Introduction
The calcium sulphate waste stream of the wet phosphoric acid process, commonly known as phosphogypsum (PG), contains the majority of rare earth elements (REE) initially present in the apatite feedstock. Worldwide, around 7 billion tonnes of PG have been generated to date with a current production rate of 150–200 million tonnes per annum. In total, an estimated 21 million tonnes of REE are locked in PG waste dumps around the world at an average grade of 3 kg/t.
Due to the value of the REE contained in PG dumps, a number of processes have been developed for their recovery . The majority of these processes involve complete or partial gypsum dissolution, thus are deemed not viable due to high capital and operating costs [1]. Mintek has developed an alternative method for the recovery of REE from PG by applying Resin-in-Leach (RIL) technology [2]. However, it was found that PG from different sources was highly variable, both in terms of REE recovery as well as physical properties of the material such as viscosity and filterability of the slurry. The recovery of REE varied from 15 to 80% depending on the origin of the samples.
The need to develop a low-cost and robust process which will modify the microstructure of PG with release of locked REE was identified. Such a process can make REE recovery from various secondary sources of PG more economically attractive.
Basis for PG Modification Approaches
Solubility of gypsum in solutions of nitrates and chlorides was reported to be significantly higher than in water [3]. Uralchem developed a process where PG was re-crystallized in the presence of soluble calcium salt at concentrations of 0.075–3.75 M (in terms of Ca2+) and strong acid (pKa < 0) at a concentration of 0.2–8.0 M (in terms of H+). Subsequent to this process, up to 98% of REE was leached into solution [5]. Solvent extraction was suggested for subsequent REE recovery and processing. However, the process proposed is not attractive from an economical (high OPEX and CAPEX) and waste management point of view.
Optimum processes for REE extraction from PG matrices should fulfil two requirements: minimum use of reagents and conversion of phosphogypsum into a saleable product suitable for industrial use (which could somewhat offset the pre-treatment costs).
One of the largest consumers of natural gypsum is the construction industry. Gypsum is typically crushed, ground and heat treated until conversion into hemihydrate or anhydrate gypsum . Calcium sulphate hemihydrate, exists in two modifications i.e. α- and β-hemihydrate, which vary in physical properties and hence are used for different applications [6, 7].
Phosphogypsum is typically related to the formula CaSO4·nH2O. Depending on the value of n, the process of producing phosphoric acid is defined as a dihydrate (n = 2) process, hemihydrate (n = ½) process, or anhydrate (n = 0) process. The majority of current phosphoric acid plants world-wide are based on the dihydrate process, with a handful using the hemihydrate process, and even fewer the anhydrate installations [8].
The dihydrate type of PG cannot be used directly in the cement industry due to its high moisture and impurity content, which causes strong retardation of cement setting rates. Other types of PG (CaSO4·½H2O and CaSO4) combine with water and subsequently transform to CaSO4·2H2O during storage in tailings dumps [9]. Hence, most PG produced at phosphoric acid plants require some cleaning followed by conversion. CaSO4·2H2O can be converted into CaSO4·½H2O either via “dry” or “wet” methods.
The “dry method” is used for conversion of gypsum into β-hemihydrate (called plaster of Paris). The process consists of calcining the PG to hemihydrate and then rehydrating it with addition of limewater, Ca(OH)2. It was reported that during the dehydration process, recrystallization resulted in the release of co-crystalized P2O5 which was then stabilised by lime to inert forms [9, 10]. The Onoda Cement Company in Japan has succeeded in complete practical utilisation of PG using the “dry method” [11].
The α-product is obtained by calcining of gypsum in an autoclave using high pressure steam, i.e. “wet method” [12, 13].
It was decided to apply the above mentioned techniques used by the construction industry, i.e. the “dry” and “wet” methods, for pre-treatment of various REE-containing PG samples and evaluate their impact on subsequent recovery of REE.
Experimental
Mineralogical Techniques
X-ray Diffraction (XRD)
PG samples were pulverised for XRD analysis in order to determine the bulk phase composition. A Bruker D8 Advance diffractometer with a LynxEye detector and Fe-filtered Co Kα radiation was used. The samples were measured over a range of 3–80° 2θ, with a step size of 0.02° 2θ and counting time of 3 s per step. Identification was based on the crystal structure of crystalline phases which occur in amounts of >3 mass%. Amorphous phases were not detected. The phases were identified using Bruker Eva software.
Discrimination between α and β-hemihydrate could not be made by XRD as the XRD patterns of these two phases are quite similar, with only slight variations in the peak intensities.
Scanning Electron Microscopy (SEM)
Samples were mounted on stubs and carbon coated for analysis by secondary electron beam in order to illustrate the surface textures and crystal structure of the PG. Polished sections were prepared and carbon-coated for analysis using a Zeiss EVO MA15 SEM . Backscattered Electron (BSE) images were captured to show textural features, and phases were identified, based on chemical composition, using energy -dispersive spectrometry (EDS) .
PG Pre-treatment Methods
“Dry” Method
A certain amount of PG sample was dried in a standard laboratory oven at 90–100 °C for a period of 24 h. After drying, the sample was pulverised using a glass roller bottle to remove some lumps which formed during the dehydration process.
“Wet” Method
Schematic drawing of experimental setup
Conditions evaluated for PG re-crystallization via “wet method”
Conditions | Test 1 | Test 2a | Test 3 |
|---|---|---|---|
Pulp density (m/m), % | 5 | 2.5 | 18 |
Temperature , °C | 180 | 150 | 150 |
Steam pressure, bar (g) | 9.2 | 3.9 | 3.9 |
Overpressure, bar | 0 | 21.1 | 6.1 |
Total pressure, bar (g) | 9.2 | 25.0 | 10.0 |
Agitation speed, rpm | 300 | 300 | 300 |
Residence time, hrs | 6 | 6 | 6 |
After completion of the test, the electrical heating element of the pressure vessel was switched off and the slurry cooled by circulating water through the cooling coils. The cooling period typically ranged between 20 and 30 min. The PG slurry was then filtered and the converted PG washed with gypsum saturated water. The PG residue was allowed to dry under ambient conditions. Sub-samples of the dried solids and filtrate were then taken for chemical analysis and mineralogical characterisation .
Evaluation of temperature effect on “wet method” efficiency
Conditions | Test 4 | Test 5 | Test 6 | Test 7 |
|---|---|---|---|---|
Pulp density (m/m), % | 10 | |||
Temperature , °C | 80 | 120 | 160 | 200 |
Steam pressure/total pressure, bar (g) | 0.78 | 1.16 | 5.35 | 14.72 |
Agitation speed, rpm | 500 | |||
Residence time, hrs | 5 | |||
The remaining portion of the converted PG slurry (for each hydrothermal test) was allowed to cool down slowly in the reactor vessel to ambient temperature similar to the cooling procedure followed in the initial tests (i.e., Test 1–3). Both residues collected from each test were allowed to dry under ambient conditions and submitted for the relevant chemical and mineralogical analysis.
It should be noted that very little gypsum scaling was observed in the autoclave during temperature optimisation tests and all solid losses were recorded.
REE Recovery
REE recovery from various PG samples before and after pre-treatment was evaluated by leaching with H2SO4 solution.
100 g/L H2SO4 solution was prepared in gypsum -saturated water and used as lixiviant. A certain amount of PG solids (as-received or converted) was contacted with predetermined volumes of lixiviant (based on moisture content) in a roller bottle, for a leach residence time of 24 h under ambient conditions. In all cases, the PG pulp density was ~20% (m/m). The pH of the pulp was not controlled during the contact period. The leached PG residue solids were filtered, washed with gypsum -saturated water and dried at 50 °C overnight. PG feed samples, leach residues and filtrates were submitted for chemical analysis by ICP-OES and ICP-MS.
Results and Discussion
Mineralogical Analysis of As-Received and Pre-treated Phosphogypsum Samples
XRD analysis revealed that gypsum (CaSO4·2H2O) was the only phase identified in the PG sample collected from one of the South African dumps. The PG-converted sample via the “dry” method comprised of hemihydrate (CaSO4·0.5H2O) with traces of remnant gypsum . The wet-converted sample consisted of gypsum .
Backscattered electron cross-sectional SEM image of a as-received PG sample; b PG converted via “dry” method; c PG converted via “wet” method (condition 3 in Table 2)
The as-received sample consisted of typical gypsum with monoclinic platy structures.
The dehydrated sample (Fig. 3b) was fine-grained and much smaller in size (less than 20 μm), when compared to the as-received PG samples.
Two types of morphologies were observed in the “wet” converted PG sample, namely coarser plates/clumps of monoclinic crystals with sharp and adjoining crystal faces, typical of gypsum , as well as needle-shaped crystals typical to α-hemihydrate structure [14]. The presence of these distinctive needle-shaped crystals gave an indication that some degree of conversion of the as-received PG into α-hemihydrate had occurred, however, due to the drying procedure used (wet filtered sample was air-dried), it converted back into di-hydrate [9].
A mineralogical search of REE phases in the as-received PG sample was complicated and only on a number of occasions were distinct REE phosphate phases seen (Fig. 3a). Dehydration of the PG resulted in some release of REE locked in the PG structure, however in the case of the “wet” conversion—a change in the shape of gypsum crystals and distinct REE phases were observed. No REE were detected in the contact liquor after “wet” conversion of the PG sample.
REE Recovery from PG Samples “as is” and After Conversion
During development of the RIL process proposed by Mintek for REE recovery from PG, it was established that the maximum REE extraction which can be achieved from PG via the RIL process was similar to that obtained during leaching with 1M H2SO4, thus, direct leaching of PG with 1 M H2SO4 was used to evaluate whether modification of the gypsum resulted in improved REE recoveries [2].
REE recovery from PG samples as-received and after pre-treatment
The “dry” pre-treatment method had no effect on REE recovery when compared to the baseline recovery of 32%, however pre-treatment of PG in the autoclave resulted in a significant improvement in REE recovery to 80%, i.e. it increased from ~30–35% to up to 80%. Further development testwork was thus focused on conversion of PG by the “wet” method.
Some scale formation was observed in the autoclave during test 3 which was conducted at 18% (m/m) solids content. Therefore, during further temperature optimisation testwork, in order to avoid the scale formation, gypsum was slurried up to 10% (m/m) solids content prior to autoclaving.
REE recovery from PG samples after “wet” conversion at various temperatures
—data taken from wet conversion test 1 and 2 (see details in Table 1—Conditions evaluated for PG re-crystallization via “wet method”
Backscattered electron cross-sectional SEM image of a flash—and b slow-cooled PG samples after autoclaving at 120 °C
Variability Samples
Variability samples
Pre-treatment of PG in the autoclave resulted in an increase in the overall REE recovery in the subsequent sulphuric acid leach to ~80% regardless of sample origin and initial REE recoveries.
Way Forward
There are a number of parameters which still require optimization and verification e.g., residence time in the autoclave , impact of seeding and solids content. Moreover, there is further potential to change the approach to REE recovery from PG. Since REE phases are liberated from the PG matrix, physical methods of upgrading REE, including magnetic separation , flotation or cycloning, can become viable to recover and upgrade REE into a concentrate which will be more attractive to process further.
- Giulini process—the only known commercial process which was developed and used in two plants in Germany and one in Ireland until 1978. Depending on the impurity content of the feed PG, the process included washing or flotation or both followed by gypsum conversion in an autoclave at 120 °C and a pH 1–3. The temperature was maintained by injection of steam. Dehydration and re-crystallization occurred in the autoclave , releasing co-crystalized impurities which were dissolved in the liquid phase. The size and shape of the α-hemihydrate crystals were controlled by various additives . The slurry was withdrawn from the autoclave continuously and sent to a filter where the crystals were washed and dewatered. The main process requirements per tonne of α-hemihydrate are given in Table 3.Table 3
Consumption of process utilities per tonne of α-hemihydrate
Inputs
Units/t
Consumption
Steam
t/t
0.4–0.6
Hot water (90 °C)
m3
0.5
Process water
m3
2.0
Electric power
kWh
25
Imperial Chemical Industries Ltd. (ICI) α-hemihydrate process, which has not been used on a commercial scale, uses 2 autoclaves in series, which are operated at 150 °C, and the product crystals are separated from the liquid phase by centrifuging. Similar to the Giulini process, it makes use of crystal modifiers to control the size and shape of crystals.
Potential flowsheets for REE and α-hemihydrate recovery /production from PG
Conclusions
Highly variable and poor recoveries of rare earths elements present in phosphogypsum dumps are caused by encapsulation of REE phases in gypsum matrices. The “wet” conversion of PG in an autoclave , which is typically used in the construction industry for the production of α-hemihydrate from natural gypsum , was found to release locked REE from the gypsum matrix. This process ultimately resulted in a PG residue which is amenable to subsequent leaching and potentially physical upgrading.
This finding can possibly result in development of a viable flowsheet for REE recovery from PG dumps around the world with the simultaneous production of a product suitable for use in the local construction industry. There is still a need to optimize process further with regards to residence time, solids content, crystal modifiers and downstream processing.