Oral Presentation International Solvent Extraction Conference 2025

Removal of arsenic from sulfuric acid solutions: over 40 years of continuous improvement to an industrial process (122716)

Joris Roosen 1 , Rayco Lommelen 2 , Koen Binnemans 2 , Jan Luyten 1
  1. Corporate Research & Development, Umicore, Olen, Antwerpen, Belgium
  2. Department of Chemistry, KU Leuven, Leuven, Belgium

Introduction

In 1976, De Schepper and Van Peteghem (Metallurgie Hoboken-Overpelt, now Umicore) filed a patent on the solvent extraction (SX) of arsenic from sulfuric acid solutions from the bleeding of copper electrorefining or copper electrowinning. [1] The process was industrially commissioned at Umicore in 1980. Since then, the installation has operated continuously under various regimes for more than 40 years. This contribution aims to share industrial experiences and insights over this period, demonstrating that continuous improvements are necessary to maintain the relevance of an industrial operation.

 

Initial development

The objective of the process is to remove arsenic from aqueous acidic sulfate streams by contacting it with an organic phase comprising an organophosphorus compound represented by the formula (RO)3P=O, wherein R can be an unsubstituted or substituted alkyl, aryl or aralkyl group. [2] In this matrix, arsenic is expected to be present in its undissociated form (As(III) as H3AsO3 and As(V) as H3AsO4). Therefore, the solvating extractant tributyl phosphate (TBP) was selected for batch screening experiments. These experiments revealed suitable extraction conditions for H3AsO4 and selectivity between H3AsO4 and H2SO4. While As(V) demonstrates a favorable distribution ratio, As(III) is poorly extracted. Therefore, it is necessary to oxidize the feed solution, for instance, using H2O2 prior to the SX process.

 

Degradation of the organic phase

Over time, a degradation pathway of TBP towards dibutyl phosphate (HDBP) and monobutyl phosphate (H2MBP) has been observed. These degradation products exhibit acid-base behavior and pH dependent solubility. The solubility of HDBP is approximately 18 g/L at 20 °C in pure water, and decreases with increasing acid concentrations.

Although the rate of this degradation reaction is moderate, a more significant issue appeared to be the formation of precipitates in the mixer-settlers of the stripping section. The presence of Fe(III) led to the formation of a white precipitate in the stripping section, impairing phase separation and reducing performance of the SX installation. The solubility product of Fe(DBP)3 is influenced by acid concentration and temperature and is around 10-12. [3]

A potential mitigation strategy involves an alkaline cleaning step, wherein the solvent is mixed with an aqueous NaOH solution. The conjugate base of HDBP exhibits high solubility in the NaOH solution, making it effective for cleaning TBP from HDBP. However, the resulting NaOH solution presents challenges for treatment due to significant organic load and traces of metals such as arsenic. The DBP ion could be quantitatively removed from this NaOH stream through acidification to 100 g/L H2SO4, under which conditions the DBP ion is protonated again, forming an insoluble organic layer that can be easily removed. This alkaline cleaning procedure is implemented in the industrial plant.

The presence of HDBP (and H2MBP) is typically analyzed using GC-MS. However, due to the requirement for highly skilled analytical personnel, a simpler pretreatment and titration method has been developed for quick assessment of the HDBP concentration in the organic solvent. This method has demonstrated sufficient accuracy to determine the appropriate timing for regeneration.

 

Thermodynamic modelling

With several experimental data points available, creating an appropriate equilibrium model for this extraction system was considered necessary to simulate the performance of the SX system under various conditions. Therefore, a thermodynamic model has been developed to describe the (competitive) extraction of H2SO4 and H3AsO4 from NiSO4 solutions. A thermodynamic modelling approach was selected over an empirical (correlative) model. OLI, with its Mixed Solvent Electrolyte (MSE) framework, was chosen as the software tool for this purpose. The derived model parameters are detailed in [4]. The model's accuracy falls within the range of experimental errors, allowing to predict plant performance based on factors such as incoming feed conditions.

 

Further improvements

Once an installation is commissioned, its capacity is determined by the design. However, boundary conditions often change over time. To evaluate new opportunities, batch and continuous chemical tests, along with decantation tests in a geometrically equivalent scale model, have been used to assess the feasibility of new conditions. Recently, model calculations have become an additional tool to explore potential operating windows and opportunities. These calculations and experiments are intended to identify and justify the necessary investments.

  1. [1] A. De Schepper, A. Van Peteghem, Metallurgie Hoboken-Overpelt, Treatment of solutions containing impure metals (1976), patent number 4061564.
  2. [2] F. Rondas, J. Scoyer and C. Geenen, Solvent extraction of arsenic with TBP - the influence of high iron concentration on the extraction behaviour of arsenic (1995), Proceedings of the Copper 95 - Cobre 95 International Conference, p. 325.
  3. [3] W. Davis Jr. and D. O. Rester, Kinetics of precipitation of ferric dibutyl phosphate from aqueous solutions (1968), Journal of Inorganic and Nuclear Chemistry, vol. 30, p. 3317.
  4. [4] R. Lommelen and K. Binnemans, Molecular thermodynamic model for solvent extraction of mineral acids by tri-n-butyl phosphate (2023), Separation and Purification Technology, vol. 313, p. 123475.
  • Abstract category selection: