Sustained Bauxite Residue Rehabilitation with Gypsum and Organic Matter 16 years after Initial Treatment

24 Bauxite residue is a high volume by-product of alumina manufacture which is 25 commonly disposed of in purpose-built bauxite residue disposal areas (BRDAs). Natural 26 waters interacting with bauxite residue are characteristically highly alkaline, and have 27 elevated concentrations of Na, Al, and other trace metals. Rehabilitation of BRDAs is 28 therefore often costly and resource/infrastructure intensive. Data is presented from 29 three neighbouring plots of bauxite residue that was deposited twenty years ago. One 30 plot was amended 16 years ago with process sand, organic matter, gypsum, and seeded 31 (fully treated), another plot was amended 16 years ago with process sand, organic 32 matter, and seeded (partially treated), and a third plot was left untreated. These surface 33 treatments lower alkalinity and salinity, and thus produce a substrate more suitable for 34 biological colonisation from seeding. The reduction of pH leads to much lower Al, V and 35 As mobility in the actively treated residue and the beneficial effects of treatment extend 36 passively 20-30 cm below the depth of the original amendment. These positive 37 rehabilitation effects are maintained after 2 decades due to the presence of an active 38 and resilient biological community. This treatment may provide a lower cost solution to 39 BRDA end of use closure plans and orphaned BRDA rehabilitation. 40 constrained audit trails regarding treatment and planting histories, can limit their 87 viability in BRDA closure plans. The objective of this study was to assess the long term 88 effects of a surface treatment to bauxite residue. Here we report the chemical and 89 mineralogical data sampled from depth profiles of bauxite residue nearly two decades 90 after initial treatment, and evaluate the ability of these treatments to provide sustained 91 rehabilitation of the substrate and associated fluid.


Introduction 41
Globally, >100 million tonnes of alumina is produced annually. 1 Producing 1 42 tonne of alumina generates 1-2 tonnes of bauxite residue (known as "red mud"). The 43 residue varies with ore type, but all are alkaline, sodic, and contain similar minerals. In 44 the Bayer process bauxite ore is digested with NaOH at high temperature and pressure 45 which results in recrystallization of iron oxides present. Silica is a common impurity, 46 which is removed from solution by precipitation of a range of characteristic Na-and Ca-47 aluminosillicate phases (e.g. sodalite and cancrinite). 2,3 These "desilication products" 48 reside predominantly in the fine fraction. Residual aluminium (oxy)hydroxide phases, 49 quartz, zircon and titanium oxides (e.g. rutile and perovskite) also occur in the 50 residues. 2,3 51 Bauxite residue has few uses (cement, iron and steel production, construction 52 materials) and most is sent to bauxite residue disposal areas (BDRAs). 4 The liquid from 53 bauxite residue is very alkaline (pH 11-13) and contains abundant sodium. 5-7 54 Subsequent dissolution of desilication products such as sodalite (Eqn 1.) and cancrinite 55 (Eqn 2), along with associated amorphous secondary phases, generates further 56 alkalinity and releases sodium in the long term. 8-10 Trace elements in bauxite, such as V 57 and As, become concentrated in the residue, and are often hosted in surface complexes 58 and secondary phases. [10][11][12][13][14] This can be environmentally problematic as Al, V, and As 59 form aqueous oxyanions in alkaline conditions which sorb poorly to sediments 15 When left untreated, bauxite residue is infiltrated by CO2 and the formation of 63 aqueous and solid carbonate consumes OH -, lowering pH. [17][18][19] The depth to which this 64 process can act within bauxite residue is controlled by the rate of in-gassing and 65 diffusion of CO2. These process can be enhanced by gypsum addition, providing excess 66 Ca 2+ for precipitation of carbonate. 20 These reactions occur rapidly at high pH and can 67 eventually buffer the pH to 7.5-8.5. 17,21 Previous work has shown that gypsum addition 68 can also decrease the mobility of trace metals and Al in bauxite residue effected 69 soils. 17,21 . Other approaches to decrease bauxite residue salinity and alkalinity, such as 70 treatment with acid 22 and seawater, 23 tend to only neutralise the aqueous, not the solid 71 alkalinity generating phases. Ion exchange resins, 24 and bipolar-membranes 72 electrodialysis 25 have been used to increase the longevity of treatment, yet these 73 approaches rely on continued management and the utilisation of products by an active 74 refinery. As such, common end-of-use practice is to cap BRDAs with an impermeable 75 layer, cover with topsoil, and revegetate. The costs "cap and cover" approaches are high 76 (e.g. 100k €/ha has been estimated for the BRDA in this study). However abandoning 77 BRDAs without surface cover may lead to problems with long term water infiltration 78 and dust formation. rehabilitation have also been reported elsewhere. 37,38 Yet, little is known of the 85 longevity and reliability of such surface treatments. Lack of long term data, and poorly 86 constrained audit trails regarding treatment and planting histories, can limit their 87 viability in BRDA closure plans. The objective of this study was to assess the long term 88 effects of a surface treatment to bauxite residue. Here we report the chemical and 89 mineralogical data sampled from depth profiles of bauxite residue nearly two decades 90 after initial treatment, and evaluate the ability of these treatments to provide sustained 91 rehabilitation of the substrate and associated fluid. 92

Methods 93
In September 2015 trial pits were dug to ~60 cm in a BRDA located in a 94 European Union member state with a temperate oceanic climate (average annual 95 rainfall ~1m). At this site bauxite residue was deposited into a 3m deep disposal cell in 96 1995, and subsequently treated to encourage revegetation in 1999. Therefore, 97 sampling was undertaken 20 years after deposition and 16 years after treatment. 31 98 Three plots within the BRDA were investigated. The fully treated plot was amended 99 with gypsum (3% w/w rotavated-in to a depth 30 cm), process sand (10% w/w 100 rotavated-in to a depth of 30cm), spent mushroom compost (80t Ha -1 rotavated-in to a 101 depth of 20cm), and seeded with a grassland mix (Agrostis stolonifera, Holcus lanatus, 102 Lolium perenne, Trifolium repens, and Trifolium pratense; 100 kg/ha). 31 The partially 103 treated plot was amended only with process sand, spent mushroom compost, and then 104 seeded. The third plot was left untreated. Samples of bauxite residue were collected to a 105 depth of 50 cm from the trial pits in each of three different treatment zones. Duplicate 106 sample profiles in each plot were taken from two separate clean vertical surfaces of trial 107 pits and stored in polypropylene tubes. The dual depth profiles were sampled to 108 observe and account for heterogeneity in the residue. 109 Field moist samples were stored at 5 °C before aqueous extraction for major and 110 trace metals. 10 gram subsamples were mixed with 10 mL of ultrapure water (18.5 MΩ) 111 and shaken at room temperature for seven days. The solution pH was measured using a 112 Thermo Scientific Orion ROSS Ultra electrode calibrated with 4.00, 7.00, and 10.00 113 buffers (Fisher Scientific). 1 gram field moist subsamples were mixed with 10 mL of a 114 0.1 M Na2HPO4 in 0.01 M NaOH and shaken at room temperature for 7 days for 115 phosphate extraction of metal oxyanions. Supernatant solutions from both the water 116 and phosphate extractions were filtered through 0.2 µm disposable polyethersulfone 117 filters (Sartorius) and acidified in 5% HNO3 for subsequent aqueous analysis by ICP-OES 118 (Thermo Fisher iCAP 7400 Radial ICP-OES) (see SI section S1 for further details). 119 Further 10 g field moist subsamples were also dried at 105 °C for 24 hours to 120 determine residue water content and for subsequent analysis by X-ray ray diffraction 121 (XRD; Bruker D8 Advance diffractometer, 12 min. scans, 2 to 70° 2θ), X-ray fluorescence 122 (XRF, Olympus Innovex X-5000 XRF analyser) and total carbon analysis (TC; LECO SC-123 144DR carbon analyser). The crystalline phases present were determined from XRD 124 patterns by peak fitting using EVA (version 3.0, Bruker), and semi-quantitative relative 125 proportions were calculated by Rietveld refinement using Topas (version 4.2, Bruker). 126 Total organic carbon (TOC) were measured after a 24 hour digestion in 10% HCl at 127 room temperature. Total inorganic carbon (TIC) was calculated from TC and TOC 128

measurements. 129
Acid soluble inorganic and organic substances (AIC and AOC) were determined 130 in 12 samples after extraction with 2 M HCl (1 g soil in 5 mL of 2 M HCl for 3 days at 4 131 °C). The extractant was then separated by centrifugation at 8000 g for 10 min, pH 132 neutralised by drop-wise addition of 2 M NaOH, evaporated to dryness; and finally the 133 resulting solid dissolved in ultra-pure water at 1 g·L −1 . 39 Total carbon and total 134 inorganic carbon in the extractant was determined using a Shimadzu total organic 135 carbon analyser 5050A (LOD 4 µg kg -1 ). 136 Separate samples of bauxite residue were collected from beneath the exposed 137 vertical surface of each trial pit using a clean spatula, and sealed in sterile 138 polypropylene centrifuge tubes. These samples for DNA analysis were refrigerated 139 within 4 hours and frozen within 48hrs. DNA was isolated from 0.5 g of each sample 140

Sampling observations 160
Both the fully treated and partially treated sites were vegetated with a variety of 161 perennial grasses (Holcus lanatus), trifoliate clovers (Trifolium pratense), and occasional 162 small shrubs (Salix spp.; Fig. S1), as has been described previously. 31 The untreated plot 163 was largely unvegetated with one or two areas of stunted grasses (Fig. S1). The root

Substrate characteristics 171
The pH of the untreated residue was 10.2 at the surface and steadily increased to 172 12.0 at a depth of 50 cm ( Fig. 1; SI Table S2). The pH of the treated plots were notably 173 and significantly lower (p < 0.001; Table S3). The fully treated residue was pH 7.6 at the 174 surface, and increased steadily to a value of 9.6 at a depth of 50 cm. The pH value of the 175 partially treated residue was 7.6 at the surface, increased steadily to a value of 10.8 at a 176 depth of 50 cm, and was not significantly different from the fully treated residue (p > 177 0.05; Fig 2; Table S2-3). 178 The amount of sodium available to aqueous extraction of the untreated bauxite 179 residue was ~900 mg kg -1 of bauxite residue, and with exception of concentrations at 180 the surface and at 50 cm there was little variation with depth ( Fig. 1, Table S2). The 181 amount of Na that could be extracted from the fully treated and partially treated 182 samples demonstrated no trend with depth and were not significantly different from 183 each other (p > 0.05; Table S3). Fully and partially treated residue contained 184 concentrations approximately 10-15 % of those extracted from the untreated residue at 185 the same depth (p < 0.001; Fig. 1; Table S2-3). The concentration of silicon available to 186 aqueous extraction in the untreated bauxite residue was 5 mg kg -1 , and apart from the 187 measured concentration from 50 cm there was minimal variation with depth ( Fig. 1, 188 Table S2). Si concentrations extracted from fully treated and partially treated bauxite 189 residue were ~4 mg kg -1 below 5 cm, and ~ 13 mg kg -1 above 5cm, there was no 190 significant difference between fully, partially, or untreated residue (p > 0.05; Fig. 1; 191 Table S2-3). Calcium concentrations from the aqueous extraction of untreated bauxite 192 residue ranged from 3 mg kg -1 at the surface to below the limit of detection at 50 cm 193 (0.11 mg kg -1 ) ( Fig. 1, Table S2). In contrast Ca concentrations from fully treated and 194 partially treated samples were significantly different to the untreated residue (p < 195 0.001; Table S3), 143 mg kg -1 at the surface decreasing to ~10 mg kg -1 at 20 cm, with 196 further slight concentration decrease to ~2 mg kg -1 at 50 cm with no significant 197 difference between treatments (p > 0.005; Fig. 1; Table S2-3). 198 The aluminium concentration available to aqueous extraction in untreated 199 bauxite residue was ~10 mg kg -1 at the surface which increases steadily with depth to 200 ~65 mg kg -1 at 50 cm (Fig 2. Table S2). Conversely, Al concentrations available in fully 201 and partially treated samples were significantly different (p < 0.001, Table S3) and near 202 the detection limit (0.09 mg kg -1 ) at all depths, apart from at 30-50 cm where Al 203 concentrations were 1-10 mg kg -1 (Fig 2. Table S2). There was no significant difference 204 between treatments (p > 0.05, Table S3). The amount of vanadium available to aqueous 205 extraction from untreated bauxite residue was ~5 mg kg -1 and did not vary greatly with 206 depth (Fig 2. Table S2). Aqueous extractable V in fully treated and partially treated 207 samples were near detection limit at the surface (0.03 mg kg -1 ) and increased gradually 208 with depth to maximum concentrations of 3.9 mg kg -1 at 50 cm, significantly different 209 from untreated residue (p < 0.001, Table S3) but not significantly different between 210 fully and partially treated residue (p > 0.05; Fig 2; Table S2-3). Aqueous available 211 arsenic concentrations from untreated bauxite residue were highest at the surface (0.3 212 mg kg -1 ) and decrease with depth to 0.9 mg kg -1 at 50 cm depth (Fig 2. Table S2). With 213 the exception of one sample, all measurements of aqueous extractable As from fully 214 treated and partially treated bauxite residue were below detection limit (0.045 mg kg -1 ) 215 and significantly different from the untreated residue (p < 0.001; Fig 2; Table S2-3). 216 Extraction at high pH using disodium phosphate demonstrated substantial 217 concentrations of Al, V, and As were available in all bauxite residue treatments. 218 Phosphate extractable Al concentrations from all treatments are generally all 25-50 mg 219 kg -1 at all depths (no significant differences between treatments; p > 0.05; Table S2-3). V 220 concentrations from the phosphate extraction of untreated bauxite residue range from 221 30-75 mg kg -1 at the surface to 30 mg kg -1 at 50 cm depth (Fig 2. Table S2). Phosphate 222 available V from fully treated and partially treated samples was lowest at the surface 223 (~15 mg kg -1 ) and increases with depth to ~75 mg kg -1 at 50 cm, but with no significant 224 differences between untreated, fully treated, or partially treated residue (p > 0.05; Fig 2. 225 Table S2-3). Arsenic concentrations extracted from untreated bauxite residue at high 226 pH with phosphate are ~2.5 mg kg -1 at the surface and decrease to < 1 mg kg -1 at 50 cm 227 (Fig 2. Table S2). Phosphate extractable As from fully treated and partially treated 228 samples increase with depth from ~1 mg kg -1 at the surface to ~2.5 mg kg -1 at 50 cm 229 (Fig 2. Table S2). Phosphate extractable As from fully treated and partially treated 230 residue were significantly different (p < 0.05), though neither were significantly 231 different from the untreated residue (p > 0.05; Table S3). 232 The water content of the residue (weight of water as % of dry weight) at both the 233 fully and partially treated sites was over 50% near the surface, exhibited a minimum of 234 ~30 % at approximately 10 cm, and then increased to between 35 and 45 % at depths 235 below 20 cm (Table S2). In contrast the water content in the untreated profile was 35% 236 near to the surface, exhibited a maximum value of ~50 % at 10cm, and then decreased 237 slightly to 40 % at depths below 30 cm. Water in the untreated residue was significantly 238 different to fully treated residue (p < 0.001), but not significantly different from 239 partially treated residue (p > 0.05; Table S3) 240 The bulk mineralogy of bauxite residue from all plots were largely similar and 241 consist of 40-45% iron oxy-hydroxides, 20-30% aluminium oxy-hydroxides, 20-30% 242 titanium oxides, and 10-15% feldspathoids (Table 1, Table S4). At the untreated bauxite 243 residue plot there were no differences in the relative proportions of each phase with 244 depth. Variations in the relative proportions of phases within the residue as a function 245 of depth and treatment were within the range of uncertainty of Rietveld refinement (5 246 %). The alkali generating feldspathoid and desilication product cancrinite was present 247 at all depths in all treatment sites (Table 1, Table S4). There was little difference in the 248 bulk elemental composition measured by XRF with either depth or treatment (Table  249 S5). Fe, Al, Ca, Si and Ti were the most abundant oxides in present each site (36 ±3, 250 10±2, 15 ±2, 5 ±1 and 4 ±1 wt. % respectively). Carbon was most concentrated in the 251 top 10 cm of the fully treated profile (Fig. 3), where TOC was approximately 2.5% and 252 TIC was 1.5%. Below 10 cm there was no discernible difference in carbon content 253 between the fully treated and untreated profiles. Samples of untreated bauxite had less 254 than 0.5% TOC and TIC at all depths. Acid extractable inorganic carbon (AIC) and 255 organic carbon (AOC) was only detectable in the top 10 cm of the fully treated and 256 untreated bauxite residue, and was below or at the limit of detection (<4 µg kg -1 ) in all 257 other samples (Table S2). 258 DNA mass isolated per gram of sample demonstrated a strong vertical gradient 259 and significant difference between the treated (fully treated and partially treated) and 260 untreated sites (Fig. 3, Table S6). DNA was concentrated in the top 12 cm of the fully 261  The alpha diversity indices for each sample are shown in The geochemistry of 20 year old untreated bauxite residue 289 Fresh bauxite residue is highly alkaline (pH 10-13), highly sodic (abundant 290 mobile Na), contains abundant solid phase alkalinity (e.g. desalination products; 2-51%) 291 and can also can contain trace metals above threshold intervention levels. 10,12,26,27,48-52 292 The desilication products in fresh residue tend to have higher proportion of sodalite to 293 cancrinite 10 however, with age sodalite can transform into cancrinite. 53 Initially the high 294 pH and sodium contents are due to remnant NaOH from the Bayer Process. Previous 295 work has shown that repeated replacement of pore water decreases the mass of fresh 296 bauxite residue but does not alter final pH, Na + , Al(OH)4 -, CO3 2-, or OHconcentrations 8 297 due to the dissolution desilication products, and associated amorphous phases (Eqn 1, 298 2). When left untreated, the pH of bauxite residue is controlled by the balance between 299 CO2 infiltration from the atmosphere, and OHproduction through desilication product 300 dissolution. 301 20 years after deposition the measured pH of the untreated bauxite residue 302 ranges from pH 10 at the surface to pH 12 at 50 cm. XRD analysis indicates that 303 cancrinite was the primary desilication product present (Table S4). At the surface, CO2 304 in-gassing, in combination with cancrinite dissolution, and associated amorphous Fe, Al, 305 and Si phase solubility, buffers the pore fluids to approximately pH 10. Atmospheric CO2 306 in-gassing appears to extend ~20 cm from the surface (Fig. 1). Below 20 cm the bauxite 307 residue appears to be isolated from the atmosphere and dissolution of cancrinite results 308 in higher pH (≥ 11.5; Fig. 1). Cancrinite dissolution also controls long term Na 309 availability (Eqn. 2), and results in aqueous available Na concentrations of ~900 mg kg -1 310 in untreated bauxite residue after 20 years. However, dissolution of cancrinite appears 311 to be incongruent at high pH. Cancrinite dissolution should produce equimolar 312 concentrations of Na, Si, and Al, (Eqn. 2) but the measured concentrations are far from 313 stoichiometric (Fig. S3). Aqueous extractable Na concentrations from untreated samples 314 are 100 to 400 times higher in concentration than extractable Si and 10 to 150 times the 315 Al concentration, indicating a preferential retention of Si and Al in the solid phase. 316 This preferential retention of Al and Si in the solid phase is probably controlled 317 by the precipitation of amorphous and crystalline secondary phases. At the highest pH 318 measured, Al concentrations are close to equilibrium with gibbsite (Al(OH)3) (Fig S3). 319 The measured Al concentrations decrease as the pH decreases from 12 to 10, but 320 exceeds concentrations in equilibrium with gibbsite. Over this pH range, Si 321 concentrations are much lower than those expected for SiO2(am) equilibrium, suggesting 322 an alternative solubility limiting phase. At high pH, with high Na concentrations, Al and 323 Si can co-precipitate in amorphous cation-bridged alumino-silicate gels, 54 which may 324 explain the low concentrations observed. 325 Sustained alkalinity generation throughout untreated bauxite residue is a 326 concern because it may be associated with increased mobility of potentially toxic 327 metal(oid) oxyanions such as Al, V, and As. Both V and As are reported to be present in 328 bauxite residues primarily in the 5+ oxidation state as vanadate and arsenate species 329 10,12 , and are found as surface adsorbed species (V can also be associated with 330 neoformed hydrogarnet phases such as Katoite). 12 Conversely, Al availability is usually 331 controlled by the solubility of Al (oxy)hydroxide phases, which typically have much 332 higher solubility at high pH (see discussion above). 55 In alkaline phosphate extractions 333 both OHand phosphate ions compete strongly for available sorption sites and promote 334 the mobility of metal oxyanions. 14,20 The results of these extractions, therefore, 335 demonstrate that there is abundant V and As adsorbed to bauxite residue (Fig. 2). In the 336 untreated samples, where pH > 10, As and V sorb poorly to mineral surfaces, 14-16,21,56-58 337 which is why only 10 and 15 % of the phosphate extractable As and V respectively were 338 extractable water this fraction will be mobile in residue pore waters. 339 In summary, the bauxite residue from the untreated plot retains many of the 340 characteristics of the fresh bauxite residue 20 years after deposition: high pH, a sizeable 341 quantity of desilication products (particularly cancrinite), abundant available Na, high 342 Al, V, and As concentrations, low organic carbon concentrations. Thus, untreated, it is an 343 environment that is not conducive to spontaneous macro-or microorganism 344 colonization through translocation. 345 346 Treated bauxite residue 347 16 years after bauxite residue treatment with process sand, organic matter and 348 gypsum significant pH reduction (2 units) was observed over a depth that extends at 349 least 30 cm below the actively treated surface layer ( Fig. 1; Table S2). Aqueous sodium 350 concentrations were an order of magnitude lower in the treated plots than untreated 351 plot at all depths ( Fig. 1; Table S2), and the availability of aluminium, vanadium, and 352 arsenic were all lower in treated than untreated bauxite residue ( Fig. 2; Table S2). 353 These observations demonstrate that positive treatment effects observed in the short 354 term are sustained, such as: improved permeability, particle aggregation, and drainage; 355 pH neutralisation; decreased Na, Al, and Fe availability. 28,29 In natural soils, organic 356 matter plays a key role in controlling particle aggregation, 59-61 and the application of addition decrease the rate of OHand Na + production from the dissolution of cancrinite 386 and associated secondary phases (Fig. 1). 387 Aqueous extracted aluminium concentrations from partially and fully treated 388 bauxite residue plotted as a function of pH (Fig. S3) fall on a line parallel to, but in 389 between, the solubility lines of gibbsite and Al(OH)3 (am). This is different to the trend 390 observed for the untreated samples at higher pH, suggesting a different solubility 391 controlling phase. Between pH 8 and 10 formation of dawsonite (NaAlCO3(OH)2) and an 392 amorphous precursor to boehmite have been observed in bauxite residue 393 treatment. 12,72 and may be the solubility controlling phases at this site. The phosphate 394 extraction shows that there is abundant extractable Al, V, and As in both the partially 395 and fully treated bauxite residue ( Fig. 2; Table S2). However the aqueous extractions 396 showed that nearer to neutral pH Al is secured in secondary phases, and the majority of 397 V and As is sorbed to mineral surfaces 14-16,21,56-58 making Al, V, and As, much less 398 available to aqueous solution (Fig. 2). Surface treatment with process sand, gypsum, and organic matter is a stable, 446 reliable, and safe solution to bauxite residue rehabilitation. Bauxite residue pH is 447 neutralised, Na + is less available, and metal oxyanions (Al, V, and As) are less mobile. 448 The beneficial effects of treatment are long term and extend 20-30 cm beyond the depth 449 of application. The formation a passively treated zone, which is ≥ 20% of the total 450 disposal cell depth, is sufficient to separate the surface environments from the 451 potentially highly alkaline, sodium rich, and trace metal containing residue at depth. 452 The presence of alkalinity generating phases in both treated plots highlights the 453 importance of maintaining a strong biologically active surface layer. Were this layer to 454 be removed or substantially disrupted, and its supply of acid neutralising molecules 455 lost, the system would likely return to a high pH steady state, with high Na, Al, V, and As   Concentrations of Al, V, and As in solution following aqueous and phosphate (PO4) 748 extractions from fully treated, partially treated, and untreated bauxite residue as a 749 function of depth. Note the change in x-axis scale for aqueous and phosphate extracted 750 V and As. The dotted line represents the limit of detection for each element. 751 752  Table 1. 758 Semi-quantitative percentage of crystalline phases present in bauxite residue as a function of treatment and average across depth, fitted 759 using Rietveld refinement. Uncertainty on the Rietveld refinement is approximately 5 %. Full details are available in Table S2. 760 761