Electrowinning of Aluminum —Challenges and Possibilities for Reducing the Carbon Footprint—

Abstract

Primary aluminum is produced by the Hall-Heroult process which is based on electrolysis in molten fluoride electrolyte, Na₃AlF₆-AlF₃, at ∼960 °C in which the raw material alumina is dissolved and decomposed into pure aluminum and CO₂ gas due to the use of carbon anodes. Direct CO₂ emissions are due to the anode process including perfluoro carbon (PFC) formation during anode effect. An inert anode to produce oxygen may eliminate direct CO₂ emissions including PFC gases and give possibilities to improve the cell design. CO₂ emissions from generation of electricity are the most important issue globally. Also the use of pure metals to produce alloys may significantly increase the carbon footprint due to the primary production of alloying elements. A new approach to produce alloys directly during electrolysis is proposed, and results from lab experiments show that this method may give significant reduction of carbon footprint for the production of aluminum alloys. Other sources of CO₂ emissions are production and manufacture of alumina and carbon anodes as well as loss in current efficiency for aluminum. A new process based on aluminum chloride electrolysis and recycling of CO₂ may eliminate CO₂ emissions from the production process.

1. Introduction

Production of primary aluminum metal was proposed and independently patented by Hall and Heroult in 1886. The principles of the process is unchanged since its invention, and the electrolytic Hall-Heroult process is the only industrial production route for aluminum. However, great progress has taken place over more than 100 years of development. The main improvements have been related to current efficiency, electrical energy consumption, productivity, and environmental impact.

Today the production of primary aluminum is ∼70 million metric ton.¹ China is the largest producer with more than 60 % of the global share. The primary production of aluminum has increased considerably over the past few decades due to the importance of aluminum alloys especially for transportation and building materials.

The process is based on electrolysis in a molten fluoride electrolyte, Na₃AlF₆-AlF₃ and CaF₂, at ∼960 °C in which the raw material alumina is dissolved and decomposed into pure aluminum and CO₂ gas due to the use of carbon anodes. The current efficiency with respect to aluminum can be as high as 96 % in Hall-Heroult cells.² The total cell reaction is:

  

\begin{equation} \text{Al$_{2}$O$_{3}$ (diss)} + \text{3/2 C (s)} = \text{2 Al (l)} + \text{3/2 CO$_{2}$ (g)} \end{equation} (1)

Figure 1 shows a sketch of an industrial cell.

Figure 1.

Sketch (cross-section) of a prebaked aluminum electrolysis cell.

Modern cells are operated at currents from ∼300–500 kA. The electrical energy consumption is around 12–15 kWh (kg-Al)⁻¹, which corresponds to an energy efficiency of about 50–60 %. The main losses are due to a large ohmic voltage drop in the electrolyte and high anodic overvoltage. There are considerable emissions of CO₂ from generation of electricity and the anode reaction. In addition there are emissions of CF₄ and C₂F₆ (PFC gases) during so-called anode effect. Several efforts are under way to reduce the carbon footprint of the electrolysis process. One approach is to develop inert anodes to produce oxygen to eliminate the direct emissions of CO₂ and PFC. Another idea is to produce aluminum based alloys directly during electrolysis by adding metal oxides in order to co-deposit alloying elements such as silicon, titanium and manganese. Such a method may give significant savings also in terms of carbon footprint since the current method for producing alloys is to add pure metals to liquid aluminum before casting. A new process based on aluminum chloride electrolysis is yet another approach which may reduce the carbon footprint significantly.

2. Energy and Carbon Footprint

Table 1 shows approximate values for the emissions of CO₂ equivalents in kg-CO₂ (kg-Al)⁻¹ for primary production of aluminum including production of alumina. It is clear that indirect CO₂ emissions for electricity generation is the major issue for reducing the carbon footprint.

Table 1. CO₂ emissions during primary production of aluminum.

Process step	CO2 emissions (kg-CO2 (kg-Al)−1)
Anode production	0.3
Anode reaction	1.0
Anodic formation of PFC gases (CF4, C2F6)	0.7
Electricity	1–12
Total	3–14

Excess carbon anode consumption is due to airburn and the Bodouard reaction also called carboxy burn. Also the current efficiency will affect the carbon consumption.

3. Inert Anode versus Carbon Anode

From the stoichiometry of the total reaction the theoretical consumption of carbon is 333 g per 1000 g Al produced. Also loss in current efficiency will increase the carbon consumption.

0.75 mol C is consumed per mol Al produced according to reaction (1). The theoretical consumption is then 333 g C per 1000 g Al produced. If the current efficiency for Al (CE_Al) is reduced, the carbon consumption will be increased proportionally, (333/CE_Al) g C per kg Al.

Under normal electrolysis conditions CO₂ is the major product of the anode process, although CO is thermodynamically favoured.

Some excess carbon consumption and CO₂ emissions are due to the carboxy reaction and airburn of the carbon anode as follows:

  

\begin{equation} \text{CO$_{2}$ (g)} + \text{C (s)} = \text{2 CO (g)} \end{equation} (2)

  

\begin{equation} \text{O$_{2}$ (g)} + \text{C (s)} = \text{CO$_{2}$ (g)} \end{equation} (3)

The real carbon consumption is more than 400 g C per kg Al.

No inert anode material for oxygen evolution has been developed successfully although a lot of research has been carried out in this area.² An inert anode will eliminate direct CO₂ emissions from the electrolysis process itself. However, the main advantage of an inert anode may be the prevention of anode effect and formation and emissions of perfluoro carbon (PFC) gases. Today CF₄ and C₂F₆ gases are formed during so-called anode effect which happens when the content of dissolved alumina decreases below a critical value so that the anode potential increases and other anode reactions than CO₂ formation will happen. PFC gases are powerful greenhouse gases with potential large impact on the climate. The aluminum producing industry is one of the largest emmitter of PFC gases. It has been shown recently that PFC gases may be formed and emitted also during normal electrolysis. This may happen in cells with several anodes which are connected in parallel; a large majority of cells has this anode design, 30–40 anodes may be present in a single cell. All the anodes have the same potential during electrolysis. However, disturbances and distribution of dissolved alumina may cause one or a few anodes to draw less current. Since the ohmic voltage drop is lower for anodes drawing less current more of the potential will be used for anode reactions. Therefore CF₄ and C₂F₆ gases may be produced since the total anode potential is unchanged regardless of current changes.³ The total cell reactions during formation of PFC gases are as follows:

  

\begin{equation} \text{4Na$_{3}$AlF$_{6}$} + \text{3C} = \text{12NaF} + \text{4Al} + \text{3CF$_{4}$}\quad E^{0} = -2.58\,\text{V} \end{equation} (4)

  

\begin{equation} \text{2Na$_{3}$AlF$_{6}$} + \text{2C} = \text{6NaF} + \text{2Al} + \text{C$_{2}$F$_{6}$}\quad E^{0} = -2.80\,\text{V} \end{equation} (5)

The emissions of PFC gases from industrial aluminum producing plants in terms of CO₂ equivalents are reported globally based on the duration of anode effects. The new knowledge of PFC formation during normal electrolysis suggests that reported values are underestimated. It is very tricky to measure the contents of PFC gases emitted from industrial cells. A better way may be to estimate the production of PFC based on measuring the current drawn by the individual anodes.

An inert anode producing oxygen gas will eliminate the direct formation of CO₂ and PFC gases. However, the reversible potential for the total electrolysis reaction will increase by about one volt without the use of carbon. This may be compensated for by optimising the design of the cell; vertical electrodes have been suggested as a way of reducing the interelectrode distance and cell voltage. A lot of research has been done in this area.⁴ Tin oxide was found to be a good candidate inert anode in the 1980’s. However, very small contents of tin in produced aluminum is detrimental for the properties. Around the turn of the century nickel ferrites became the most popular candidates, some of them also containing metallic copper and nickel. More recently metallic anodes have been emphasized with various combinations of nickel, iron and copper. In order for these anodes to be inert a lower temperature seems to be important. Therefore, new electrolyte compositions have been proposed and KF has also been added for maintaining a high enough alumina solubility. Also inert cathodes may be used to make it more convenient to optimize the design of the cell and electrodes. Titanium diboride, TiB₂, is known to be an excellent inert cathode candidate.

4. Current Efficiency and Dissolved Metals⁵

The loss in current efficiency is strongly linked to the fact that aluminum is soluble in the electrolyte. Metal solubility is a general phenomenon in molten salts.⁶ In molten cryolite based electrolytes dissolved Na must be considered in addition to dissolved Al.² A small but significant activity of sodium is established at the metal/electrolyte interface due to the following equilibrium:

  

\begin{equation} \text{Al} + \text{3 NaF} = \text{3 Na} + \text{AlF$_{3}$} \end{equation} (6)

It is known that the subvalent species AlF₂⁻ is formed as well as dissolved Na, the latter being responsible for a small contribution to electronic conductivity.² Solubility studies have been carried out in laboratory experiments, and the metal solubility is ∼0.06 wt% Al in industrial electrolyte compositions. The solubility decreases by increasing content of AlF₃ and decreasing temperature. Reliable data for the metal solubility have been published by Ødegård et al.⁷ and Wang et al.⁸

The back reaction between dissolved metals (Al and Na) and the anode product is responsible for the major loss in current efficiency, and it can be written as follows:

  

\begin{equation} \text{Al (diss)} + \text{3/2 CO$_{2}$ (g)} = \text{Al$_{2}$O$_{3}$ (diss)} + \text{3/2 CO (g)} \end{equation} (7)

The rate of the back reaction is controlled by diffusion of dissolved metals (Al and Na) through the diffusion layer near the cathode. Additions of relatively small amounts of CaF₂, LiF and MgF₂ are known to be beneficial for the current efficiency.² It is likely that the total metal solubility is reduced upon these additions. In some cases KF is added with the alumina, and the effects of KF on the operation of the electrolysis are little known.

Another phenomenon related to the dissolution of metals is called metal fog formation. So-called metal fog is a visual phenomenon which is often observed in the electrolyte near the cathode during deposition of liquid metals from molten salts. Results from laboratory experiments to study the nature of metal fog have been published.⁹ It has been established that metal fog consists of small metal droplets formed by homogeneous nucleation of metal droplets from a supersaturated solution of dissolved metal in the molten electrolyte near the cathode. Fog formation leads to reduced current efficiency for industrial electrowinning processes such as aluminum production. Fog formation may be influenced by electrolyte convection.

The current efficiency for aluminum deposition was determined by running electrolysis experiments in a laboratory cell at constant current for several hours and weighing the produced amount of aluminum and using Faraday’s law. The laboratory cell is pictured in Fig. 2. A steel cathode was used since it wets aluminum well, which is not the case for carbon. Figure 3 shows results for the current efficiency as a function of temperature in molten Na₃AlF₆-AlF₃ (CR = 2.2) with no alumina feeding at a cathodic current density of 0.9 A cm⁻². CR is the ratio of the number of moles of NaF and the number of moles of AlF₃, being 3.0 for pure cryolite. As expected the current efficiency decreases by increasing temperature because of increasing solubility of aluminum and sodium.

Figure 2.

Experimental cell for determining current efficiency.

Figure 3.

Average values for current efficiency as a function of temperature for aluminum deposition without addition of metal oxides. CR = 2.2, no alumina feeding, 0.9 A/cm², electrolysis time 4 h.

5. Alloy Production

Today aluminum alloys are produced by the following routine. The produced liquid aluminum is transported to the cast house and simple purification is done by gas treatment and holding before casting. Alloying elements of pure metals are added to the liquid aluminum prior to casting. Many of the pure metals are expensive and carry a high carbon footprint.

Fluoride electrolytes at high temperatures are know to dissolve many metal oxides. Metallic impurities more noble than aluminum tend to deposit at the liquid aluminum cathode.¹⁰ It has been shown^5,11 that cations of such impurities are reduced at the cathode at their limiting current densities (i_lim) given by the following equation which is derived from Fick’s first law:

  

\begin{equation} i_{\textit{lim}} = \mathit{nFkc} \end{equation} (8)

where k is the mass transfer coefficient and c is the concentration of the dissolved impurity element species in the bulk of the electrolyte, n is the number of electrons per mole and F is Faraday’s constant (96485 C equiv⁻¹). Studies of impurities in industrial cells have been carried out by analyses of samples of electrolyte and metal as a function of time after additions of known amounts of compounds containing impurities mainly in the form of metal oxides.⁹ The concentration of the impurity species under investigation versus time after addition can be expressed as follows:

  

\begin{equation} c = c_{o}\exp\left(-\frac{A}{V}kt\right) \end{equation} (9)

where A is the area of the active cathode, V is the volume of the electrolyte, t is the time after addition and c_o is the background concentration before the addition. Hence, a plot of ln (c) versus time should be linear if the co-deposition reaction is diffusion controlled. It was found that iron, silicon, titanium, nickel and manganese co-deposited with aluminum at their limiting current densities after additions of metal oxides in industrial cells.^5,11–13

The mass transfer coefficient may be determined from the observed relationship between concentration and time after adding the impurity compound. Typical values have been reported to be in the range from 10⁻⁵–10⁻⁶ m/s.^5,11,12

Such an approach may be used to produce aluminum based alloys by electrolysis during aluminum production by adding controlled amount of metal oxides.

Electrolysis was carried out to study the co-deposition of titanium, manganese and silicon. The metal oxides were mixed with other bath components before melting. Various concentrations were considered: 0.2 wt%, 0.6 wt%, and 1.0 wt% Ti from dissolved TiO₂ and 1, 2, 3, and 4 wt% Mn and Si from Mn₂O₃ and SiO₂. The temperature was varied from 965–980 °C.

Electrolyte samples were analyzed for the content of dissolved titanium, manganese and silicon ions, and deposited metal was analysed for alloying metal content.

Figure 4 shows the concentration of dissolved titanium as a function of time. Figure 5 shows plots of ln (c) versus time after start of electrolysis for dissolved TiO₂ according to Eq. 10. The linear fit suggests that titanium co-deposits at its limiting current density. Similar results were obtained for co-deposition of manganese and silicon.

Figure 4.

Concentration of dissolved titanium in the electrolyte as a function of time during electrolysis after addition of 0.6 wt% TiO₂.

Figure 5.

Decay of content of dissolved titanium in the electrolyte during electrolysis. 0.6 wt% Ti added before electrolysis as TiO₂, 970 °C.

The average current efficiencies of Al–Ti, Al-Si and Al-Mn alloys were determined by Faraday’s law by determining the amount of deposited metal during electrolysis. The average current efficiency for the alloy is a representation of the current efficiency of each element based on its content in the alloy. The apparent current efficiency for alloy deposition was determined. Apparent current efficiencies were used in cases where the alloy composition was unknown in order to get an estimate of the current efficiency.

The main reason for loss in current efficiency for electrodeposition of aluminum based alloys is still the back reaction between dissolved metals (aluminum and sodium) and CO₂ from the anode reaction. The metal solubility decreases with decreasing temperature, which decreases the activity of aluminum and also decreases the molar ratio of NaF and AlF₃.² The current efficiency also increases by increasing current density for aluminum deposition, mainly because the production of aluminum increases while the back reaction is independent of the current density. Therefore it is expected that the current efficiency for aluminum deposition will decrease slightly during alloy co-deposition. The current efficiency for co-deposition of alloying elements will also influence the current efficiency for the alloy. The existence of various valencies of titanium, manganese and silicon may cause cyclic red/ox reactions at the electrodes during electrolysis and is likely to be the main loss in current efficiency for deposition of the minority elements. Higher valent ions may also be reduced by dissolved metals near the cathode and lower valent ions may be oxidised by CO₂ in the bulk electrolyte.

The obtained average values for the current efficiency for Al-Ti alloys are shown in Fig. 6. These results are very promising.

Figure 6.

Average current efficiency for Al-Ti deposition as a function of temperature. 0.6 wt% Ti concentration added as TiO₂ in the range from 965 to 980 °C.

The behaviour of dissolved Mn₂O₃ in industrial aluminum producing cells was reported to be close to ideal, where manganese deposited at limiting current conditions and a high proportion of the added manganese ended up in the metal.¹³ Also in the present work the manganese behaviour was promising. Figure 7 shows results of current efficiency for deposition as a function of temperature. Figure 7 shows results of average current efficiency for deposition of Al-Mn alloy as a function of temperature when 1 wt% of Mn was added as Mn₂O₃ before electrolysis. It was found that from 8–21 wt% manganese ended up in the deposits. A significant decrease in current efficiency was observed at increasing temperature. Also, improved conversion of manganese was observed at higher temperatures probably due to faster dissolution of the oxide. It is not clear whether the possible formation of Mn (II) species may contribute to the loss in current efficiency.

Figure 7.

Average current efficiency for Al-Mn deposition as a function of temperature. 1 wt% Mn added before electrolysis as Mn₂O₃, 965–980 °C. Average current efficiency for Al-Mn deposition as a function of temperature. 1 wt% Mn added before electrolysis as Mn₂O₃, 965–980 °C.

Co-deposition of silicon was found to be challenging because of apparent problems with dissolution and subsequent precipitation of SiO₂ on the cathode. The apparent current efficiency for Al-Si deposition is expected to be lower than the current efficiency for pure aluminum deposition because dissolve silicon species consist of Si (IV). The estimated apparent current efficiencies, as shown in Fig. 8, seem much lower than expected especially at high silicon contents.

Figure 8.

Apparent current efficiency (CE) for Al-Si deposition as a function of SiO₂ addition. 1, 3 and 4 wt% Si added before electrolysis as SnO₂, 980 °C.

6. Future Possibilities - Aluminum Chloride Electrolysis and Zero Emissions Process

Aluminum chloride, AlCl₃, is an inorganic compound which sublimes at 180 °C. AlCl₃ forms low-melting molten salt electrolytes with alkali chlorides known as chloroaluminates, where NaCl-AlCl₃ mixtures have been popular. Various contents of ions of Na⁺, Cl⁻, AlCl₄⁻ and Al₂Cl₇⁻ are present depending on the composition of NaCl-AlCl₃. It may be considered an acid-base system where Al₂Cl₇⁻ is a Lewis acid and AlCl₄⁻ a Lewis base according to the following reaction:

  

\begin{equation} \text{2 AlCl$_{4}{}^{-}$} = \text{Al$_{2}$Cl$_{7}{}^{-}$} + \text{Cl$^{-}$} \end{equation} (10)

Aluminum can be deposited by reduction of both AlCl₄⁻ and Al₂Cl₇⁻. Electrodeposition of pure aluminum and alloys of aluminum with titanium, manganese and niobium has been demonstrated.¹⁴

Today primary production of aluminum takes place by the Hall-Heroult process where aluminum oxide is dissolved in a molten fluoride electrolyte and electrodecomposed into aluminum and CO₂ by the use of consumable carbon anodes. An alternative process based on electrolysis in molten chloride electrolyte with AlCl₃ as the feedstock was proposed by Alcoa in 1973. The Alcoa chloride process is in principle more sustainable than the Hall-Heroult process in terms of carbon footprint and energy efficiency. However, the development of the Alcoa process was terminated due to problems with formation of poisonous chlorinated hydrocarbons during the production of AlCl₃ by chlorination of alumina. Chloride electrolysis provides metal of high purity. The small but undesirable concentrations of sodium that characterize Hall-Heroult metal are greatly reduced in the chloride system.

Recently a new approach to improve the chloride process was proposed by the Norwegian company Hydro. The new innovative step is to eliminate CO₂ emissions by recycling CO₂ to form CO and oxygen by solid oxide electrolysis while chlorine from the electrolysis is recycled in the chlorination process. Pure oxygen is then the only emission from the process. The main steps (1–4) and reactions (11–13) are given as follows.

1. 1.    Carbo-chlorination of Al₂O₃ to produce AlCl₃
     
   \begin{align} &\text{Al$_{2}$O$_{3}$ (s)} + \text{3/2 C (s)} + \text{3 Cl$_{2}$ (g)} \notag\\ &\quad= \text{2 AlCl$_{3}$ (g)} + \text{3/2 CO$_{2}$ (g)} \end{align} (11)

2. 2.    Electrolysis in molten chloride electrolyte at 720 °C
     
   \begin{equation} \text{AlCl$_{3}$ (l)} = \text{Al} + \text{3/2 Cl$_{2}$ (g)} \end{equation} (12)

3. 3.    Recycling of CO₂ and use of CO as reducing agent
4. 4.    Solid oxide electrolysis
     
   \begin{equation} \text{CO$_{2}$ (g)} = \text{CO (g)} + \text{1/2 O$_{2}$ (g)} \end{equation} (13)

New measurements of the electrical conductivity of candidate electrolytes for the chloride process were carried out in laboratory studies. The effect of LiCl, AlCl₃, NaCl and KCl on the conductivity was studied. Some results are shown in Fig. 9.

Figure 9.

Electrical conductivity of molten chloride electrolytes.

7. Conclusions

Primary production of aluminum by molten salts electrolysis in the Hall-Heroult process suffers from high carbon footprint and low energy efficiency. This paper points at certain important issues to improve the situation. The possible introduction of inert oxygen evolving anodes will eliminate direct CO₂ emissions and undesired PFC gases formation and emissions. A new way of producing aluminum based alloys may be advantageous in terms of cost and significantly reduce the carbon footprint. Also a new approach to use aluminum chloride electrolysis with CO₂ recycling to improve the energy efficiency and CO₂ emissions is described. The short term way of reducing the carbon footprint is to switch to renewable electricity on a global scale.

CRediT Authorship Contribution Statement

Geir Martin Haarberg: Conceptualization (Lead), Investigation (Lead), Methodology (Lead), Project administration (Lead), Writing – original draft (Lead)

Conflict of Interest

The authors declare no conflict of interest in the manuscript.

Funding

Norges Forskningsråd

Footnotes

A part of this paper has been presented in 2023 Joint Symposium on Molten Salts (#4B01).

References

• 1)  https://pubs.usgs.gov/periodicals/mcs2023/mcs2023-aluminum.pdf.
• 2)  J. Thonstad, P. Fellner, G. M. Haarberg, J. Hives, H. Kvande, and Å. Sterten, Aluminium Electrolysis. Fundamentals of the Hall-Heroult Process, Aluminium-Verlag, Düsseldorf (2001).
• 3)  H. Åsheim, T. A. Aarhaug, E. Sandnes, O. S. Kjos, A. Solheim, S. Kolås, and G. M. Haarberg, Light Metals 2016, TMS, ISBN 978-1-119-22579-9, p. 551 (2016).
• 4)  S. K. Padamata, K. Singh, G. M. Haarberg, and G. Saevarsdottir, J. Electrochem. Soc., 170, 073501 (2023).
• 5)  G. M. Haarberg and P. Cui, Light Metals 2014, TMS, ISBN 978-1-11888-908-4, p. 749 (2014).
• 6)  M. A. Bredig, Mixtures of Metals with Molten Salts, In: Molten Salt Chemistry (Ed. M. Blander), Interscience, New York (1964).
• 7)  R. Ødegård, Å. Sterten, and J. Thonstad, Light Metals 1987, p. 389 (1987).
• 8)  X. Wang, R. D. Peterson, and N. E. Richards, Light Metals 1991, p. 323 (1991).
• 9)  G. M. Haarberg, Proceedings Molten Salts XIII, The Electrochemical Society Inc., Pennington, NJ, USA (2003).
• 10)  K. Grjotheim and K. Matiasovsky, Aluminium, 59, 687 (1983).
• 11)  H. Grini Johansen, J. Thonstad, and Å. Sterten, Light Metals 1977, The Metals Materials and Minerals Society, p. 253.
• 12)  T. E. Jentoftsen, G. M. Haarberg, B. P. Moxnes, A. Buen, and J. Thonstad, Proceedings, 11th Int. Aluminium Symposium (Eds. A. Solheim and G. M. Haarberg), September 2001, Trondheim-Bergen-Trondheim, Norway, p. 217 (2001).
• 13)  G. M. Haarberg, O. Awayssa, R. Meirbekova, W. Xu, and G. Saevarsdottir, ECS Trans., 109(14), 155 (2022).
• 14)  G. R. Stafford and G. M. Haarberg, Plasmas Ions, 2, 35 (1999).

Biographies

Geir Martin Haarberg (Professor, Norwegian University of Science and Technology)

Geir Martin Haarberg was born in 1955. He graduated from Norwegian Institute of Technology in September 1985, and earned PhD in electrochemistry. He worked in SINTEF research institute as a researcher from 1985 and senior scientist from 1996. He became a full professor at the Norwegian University of Science and Technology in 2000. His research interests are electrochemistry in molten salts and metal production.

Hobby: Shakespeare, fishing saba and kissu.