Inappropriate landfilling of residues and wastes may result in contamination of soils and groundwater. Environmental bodies in Brazil, much later than in other developed countries, have started to push industries to safely dispose their residues and wastes, as well as apply severe penalties to those companies that may postpone to tackle this issue. Nowadays the cost of storage or safe landfilling of residues is already high and it will continue to increase for Brazilian companies.

The lithium demand in the future will continue to grow, especially with the potential increasing application in batteries for electric cars. So far there is no evidence in the literature of the employment of LAR from spodumene processing in the production of PC-based materials. It is believed that this residue, as many from other industrial activities, may act as pozzolanic materials or fillers in the production of sustainable construction materials. If viable, the reuse may reduce or prevent the potential environmental impact and enable lithium producers to comply with environmental legislation and avoid penalties. The employment of LAR in the production of concrete masonry units may be also a strategic solution, as:

1. The high number of units daily produced in a prefabricated plant enable rapid consumption of the currently landfilled residue;
2. Implementation costs are expected to be negligible, possible restricted to complete or incomplete drying of the residue.
3. Reduced mechanical strength may be accepted; the ultimate properties will determine whether the units can have structural application (loadbearing concrete masonry unit) or non-loadbearing units for masonry walls.

This paper presents results of a study of a replacement for PC with lithium aluminosilicate residue in the formulations of masonry concrete blocks. The study consisted of three parts: (i) characterization of the residue; (ii) interaction with Portland cement and durability issues; (iii) production and characterization of final products.

The residue used in this study came from a plant owned and operated by Companhia Brasileira de Lítio (CBL) in the state of Minas Gerais, Brazil. After extraction of lithium from spodumede ore, a powder residue is produced; over the years it has been placed in four piles. Eight samples were collected from random locations in the four LAR existing piles (two from each pile), in order to assess the variability of the stocked material. Those samples (labeled as A1 to A8) were used for characterization.

LAR is not classified (sieved) and the size of the particles ranges from a few millimeters to 0.1 micrometers. Thus, a manual sieve analysis was carried out to determine the percentage above 0.15 mm, retained between 0.15-0.075 mm and below 0.075 mm. The fraction below 0.075 mm (75 μm) has special interest, as it could be reactive and, therefore, present pozzolanic properties. This fraction was then submitted to LASER particle size analysis (Cilas 1064), X-ray diffraction (Shimadzu XRD-7000, 2θ = 5° to 80°, at 1° per minute, step size of 0.02) and the specific gravity calculated using the Le Chatelier flask for cementitious materials.

The objective of this characterization part was to determine the effect of the interaction between the LAR and PC in the fresh properties of cement pastes and mortars. Firstly, the water content required for a normal consistency paste and setting time were determined for mixes containing 0, 10, 20, 30, 50 and 60 % wt. of LAR. An initial mix of 60/40 LAR/ PC mix was also tested for soundness (using the Le-Chatelier soundness test for cement).

Mortars containing PC, LAR and Brazilian standard quartz sand (controlled particle size distribution) were made with the same amounts of residue used for pastes (10-60 % except for 40 %). The water to cement ratio and aggregate to binder ratio were fixed to 0.48 and 3.0, respectively, as required for determination of strength and consistency (flow table test) [9].

The pozzolanic activity index (or strength activity index) was carried out according to the Brazilian standard NBR 5752 [10], which is very similar to ASTM C311 [11]. In the first, 30 % vol. of PC is replaced with the pozzolanic material (in this case, LAR). The water content is not fixed, but rather adjusted so that the reference mix (neat PC) and blended mix (70% PC + 30% pozzolan in vol.) have same consistency measured with the flow table test. The pozzolanic activity index (PAI) is the ratio between the strength of the blended mix and reference mix at 28 days, measured in cylinders of 50 × 100 mm (diameter × height). In this study, PAI was determined at 14 and 28 days of curing.

The pozzolanic reaction was also assessed using thermogravimetric analysis (TGA). Two pastes were made: (i) neat PC and (ii) 20 % wt. replacement of PC. Both pastes were kept in a desiccator in order to prevent carbonation. They were then sent to TGA analysis (10° C / min. from room temperature to 1000°C, under N₂ atmosphere) at the desired time, i.e. 7, 14 and 28 days. The amount of calcium hydroxide was then calculated for the two pastes. The reactivity of the LAR was estimated based on the reduction of the Ca(OH)₂ content with time and formation of calcium silicates, calcium aluminate and calcium sulphoaluminate hydrates.

Alkali-silica reaction (ASR) is a typical long-term expansive attack to PC-based materials when the aggregate is reactive and the amount to alkalis in the system is reasonably high, i.e. above 0.6% of Na₂O eq. [12-14]. The presence of LiO₂ is not an issue for the development of alkali-silica reaction; this oxide is rather a suppressor of that chemical attack, as the lithium-silica-gel formed does not absorb water and, therefore, does not swell [12, 15, 16]. However, mortars were tested for soundness, as the LAR also contains significant amounts of Na₂O and K₂O. ASTM 1260 [17] was used as accelerated test; 25 × 25 × 285 mm prisms were cast, cured and kept submerged into a 1 N NaOH solution at 80°C for 28 days. The high concentration of the solution and the high temperature accelerate any possible ASR and expansion of the prisms, which is monitored by comparing the length change of the samples. The ASR is more likely to take place when reactive particles are coarser [18]. Therefore, it was decided to test the susceptibility of both LAR and coarser particles of spodumene aggregates to ASR. Three different mixes were made for soundness: (i) the first mix contained 100% PC (no LAR) and fine aggregate (0.15 to 6.3 mm) obtained from the mining of spodumene; (ii) the second mix contained 50% of PC and 50% of LAR as binder. Innocuous quartz sand was the aggregate for this particular mix; (iii) the third mix was a combination of (i) and (ii), i.e. 50% PC, 50% LAR and the aggregate from the mining of spodumene. Those three mixes could determine whether ASR was likely to be formed and the effect of fine and coarser particles on the expansion.

A rapid-hardening PC (Brazilian type V) was supplied by Holcim Brazil and used in all tests listed in 2.2. This type of cement was used because it contains minor mineral additions in composition (up to 5% limestone filler).

CMU were produced at a local industry in the city of Belo Horizonte, Brazil, using a fully automatic machine. Table 1 shows the mix formulations. Five different formulations were made: a reference formulation contained no LAR and it was exactly the mix design used by the local manufacturer. Other formulations contained 20%, 30%, 40% and 50% of LAR replacing PC. The aggregate to binder ratio was equal to 20 for all mixes; the water to binder (w:b) ratio was 0.78 for all mixes, except for Mix E (containing 50% LAR), which required more mixing water during the production (w:b = 0.88). It is important to note that the LAR was neither sieved nor screened for the production tests. The idea was to use the residue as it is, not to increase the production costs. In that case, the fine particles (below 75 μm) may act as a pozzolanic material and the coarse particles as fine sand.

Twenty units of each formulation were tested in compression and the mean compressive strength and standard deviation calculated. The water absorption was measured using another 6 units of each formulation; the results were also expressed in terms of mean and standard deviation.

Two units of each formulation were cut with an electric saw (Fig. 3) and the small prims (approximately 100 mm × 50 mm × 30 mm) were submitted to the “tank test” leaching, described by the Environment Agency standard EN NEN 7375:2004 [19]. In this dynamic leaching test, the leachate is collected from 0.25 h to 64 days (8 extractions) and the cumulative leaching is plotted against the time of testing. ICP-MS was used to determine the leaching of Li, Na and K from CMU containing 20-50% of LAR.

Fig. (3). CMU after extraction of samples for leaching.

Table 2 gives the chemical composition of the LAR as provided by the manufacturer, CBL. It is possible to see that the percentage of SiO₂ is predominant, as well as the Al₂O_3, which is typical for aluminosilicates used in the production of cement and concrete [20]. It is also noticeable that the amount of alkalis, i.e. LiO₂, Na₂O and K₂O, are relatively high. Although lithium is known to mitigate the alkali silica reaction (ASR), the high amounts of Na and K oxides could represent a risk of ASR. However, the Na₂O and K₂O content (0.445 and 4.525%, respectively) are at the same level as those found in fly ash used for OPC and concrete production in countries such as France or Germany [21]. The susceptibility to ASR will be addressed in section 3.2.5.

X-ray diffraction (XRD) of the 8 samples (A1 to A8) collected from the LAR stocking piles was also carried out, in order to assess the phases present and the variability of the stocked residue. Fig. (4) shows only one XRD diffractogram, given that the eight XRD traces were nearly identical. This indicates that the LAR samples were homogeneous in terms of mineralogical composition. XRD analysis yielded a diffraction plot with distinctive peaks, implying the presence of crystalline phase solids in the LAR. The most abundant phase is an aluminosilicate of molecular formula of C_0.24Al_3.56Cl_0.76Na_2.76O₂₈Si_8.44, which occurs in nature as the silicate mineral marialite (M) [22]. Other phases in minor quantities are spodumene (S), quartz (Q) and bassanite (B).

Fig. (4). XRD difractogram of LAR (PSD < 75 µm). S = spodumene; M = marialite, Q = quartz, B = bassanite.

Manual sieving has determined the fraction of the LAR retained in a 0.15 mm sieve, between 0.075 - 0.15 mm and passing 0.075 mm. The vast majority of LAR (75.3%) has particle size below 75 microns, which is desirable to replace the PC in concrete formulations. Yet 17.5% of the material has particle size between 0.075 - 0.15 mm and 7.2% are coarser than 0.15mm; both fractions may act as a fine aggregate rather than a reactive material. Five samples of LAR with particle size below 75 μm were submitted to Laser PSD. The equipment does not detect coarser particles of LAR (in the range of mm). Fig. (5) shows the particle size distribution and cumulative distribution for one sample, as the others have nearly identical PSD curve. Overall, the mean particle size ranged from 11.91 to 13.58 μm. Results also have shown that 100% of the particles submitted to Laser PSD (i.e. diameter < 75 μm) have particle size below 30 μm. This fine fraction of LAR may demand more mixing water to maintain the same workability.

Fig. (5). Particle size distribution of LAR (PSD < 75 µm).

With regard to particle density, two samples of LAR had same results, equal to 2,50 g/cm³. This fine fraction of LAR is then less dense than PC (~ 3,14 g/cm³) [23]. This should be taken into account when preparing concrete mixtures, as the proportions of components are usually measured in volume. On the other hand, if PC is replaced with LAR in wt. percentage, attention should be paid to the fact that the volume of LAR will be greater than the replaced (PC), what may also demand more mixing water.

The characterization of CMU comprised (i) compressive strength; (ii) water absorption; leaching of alkalis (Li, K and Na). The Brazilian standard NBR 6136 “Hollow concrete blocks for structural masonry” was used as reference. This standard defines that CMU for structural masonry must have a minimum compressive strength of 4.5 MPa (lowest concrete class). In addition, the water absorption must be below 10% for each individual CMU tested, carried out using the water saturation method.

Fig. (10) shows the compressive strength of the MCU produced, as a function of the LAR content. It is possible to observe that the mean strength slightly decreased for 30% LAR content onwards. However, in general, all CMU’s may be classified as structural masonry units, as they all have mean compressive strength above 4.5 MPa and water absorption much lower than 10% (Table 5). One may indirectly assess the pozzolanic behavior of LAR by looking at the difference in strength between 14 and 28 days. In that case, mixes with 40% and 50% do not show any significant increase in strength from 14 to 28 days, which indicates that there was not sufficient PC in those mixes to promote the pozzolanic reaction.

Fig. (10). Compressive strength of CMU produced with LAR.

The MCU’s must also comply with some local legislation with regard to the leaching of alkalis. Table 6 shows the limits for Na, and Li. There is no limit for K in Brazil, but leaching of potassium was also carried out and for calculation of the concentration of Na equivalent.

Figs. (11-13) show the cumulative leaching of Na, K and Li, respectively, from the MCU produced. The cumulative concentration of Na is well below the concentration limit permitted (200 mg/l), which indicates that the leaching of this element does not represent a risk if contamination of soil and underground water. Even if the cumulative concentration of K (Fig 12) is transformed into sodium equivalent (i.e., Na_eq = Na + 0.658 × K) and added to the cumulative concentration of Na in Fig. (11), the cumulative leaching of Na would be below 200 mg/l. The cumulative leaching to Li is below 2.5 mg/l in all cases. However, the cumulative leaching of this element is generally higher for MCU’s with higher LAR content. On the contrary, the leaching of Na and K appear to be higher when the PC content is higher in the mixes, which can be observed for CMU with 20% LAR (80% PC).

Fig. (11). Leaching of Na from CMU containing LAR.

Fig. (12). Leaching of K from CMU containing LAR.

Fig. (13). Leaching of Li from CMU containing LAR.