1 Introduction

Aluminium and its alloys are used as lightweight materials in a broad range of industrial sectors such as automotive, aerospace, electronics, sensor applications or packaging. In general, aluminium features a good recyclability, however the globally still growing demand cannot be covered by secondary aluminium alone. A total of approximately 65 million tons of primary aluminium and approximately 11 million tons of secondary aluminium were produced worldwide in the year of 2020 [1]. Therefore, the energy intensive production of primary aluminium still plays a major role in the supply of aluminium-materials for the worldwide demand. Primary aluminium production is realized in electrolysis cells using cryolite (Na3AlF6) as the electrolyte. The typical operating temperature of such a cell ranges between 950 and 970 °C [2] and relies completely on electrical energy. Therefore, it consumes about 15 kWh per kg of pure aluminium [3] at an operating voltage of only 4 V–5 V and currents up to 600 kA [4]. Because of the costly energy carrier and the high demand on production, optimization of the cell efficiency is a crucial aspect to increase the productivity of aluminium production.

The setup of an aluminium electrolysis cell is shown in Fig. 1. The anode arms induce the direct current into the electrolyte and molten aluminium from where it passes through the cathode block and finally the collector bars. The cathode block is made of graphite and a pitch binder. To ensure a high electrical conductivity, a low carbon, high purity steel is used for the collector bars. A mechanically stable connection between collector bar and cathode block is frequently realized with a high-carbon casting iron alloy. In this study it is a grey cast iron which is casted between cathode block and collector bar. An aluminium silicate wool between the materials is used as a spacer to create a hollow space for the casting material to fill. Only this process ensures an electrical connection between cathode and collector bar.

Fig. 1
figure 1

General setup of an aluminium electrolysis cell. The enlarged part on the right shows the material compound of the cathode in a 90° rotated view

Full size image

In an effort to optimise the energy efficiency of the electrolysis cell, several investigations have been made. Investigations on possible measures on production-related energy efficiency in the aluminium industry have been performed by Haraldsson et al. [5]

Also, investigations that include the cathode of the cell, focussing on material choice, cathode design and bonding techniques of the material compound have been carried out. It was found that the main cathode material (anthracitic vs. graphitized carbon) [6, 7] and the shape and structure [8,9,10] can influence the energy efficiency of the electrolysis process in a positive way. Also, with a more regular current flow that can be realized by optimizing the structure, the conductive structure and the size of collector bar as well as the cathode block, further improvements can be made [5, 11]. Additionally, it is discussed to use electrically insulating regions between cathode block and collector bar, as well as heat proof concrete to reduce horizontal current flow [5, 8, 12]. While the connection between cathode block and electrolyte has been investigated in studies [13, 14], the solid–solid interfaces between collector bar, cast iron and cathode block have to our best knowledge only been studied in our work [15]. It was discovered that the elevated service temperatures for a period of over 5 years led to extensive diffusions processes of carbon, phase transformations and -formations within the cast iron and the collector bar. The corresponding change in microstructure for both materials can be seen in Fig. 2. Table 1 gives the difference in the chemical composition. These processes influenced the electrical conductivity at room temperature of both, cast iron and collector bar. While the electrical conductivity of the collector bar at room temperature decreased by 26%, it was surprisingly increased in the cast iron material by 52%.

Fig. 2
figure 2

Microstructure of the collector bar (a, b) and cast iron (c, d) in its initial states (a, c) and after 5 years of service in an electrolysis cell (b, d)

Full size image
Table 1 Chemical composition of collector bar (CB) and cast iron (CI) in wt% [15,16,17], the positions are corresponding to Fig. 3. The chemical composition of the collector bar after usage in the cell was measured with optical emission spectrometry. 1Measured with calibrated OES grey cast iron program
Full size table
Fig. 3
figure 3

The collector bar, connected with cast iron to the graphite cathode (a), and a cross section of a discarded collector bar after use (b). Positions from where samples have been taken are marked

Full size image

Also, for a more exact understanding of the actual service temperature of the collector bar, phase quantity analysis in combination with the measured chemical composition was used to determine those temperatures more precisely. It was discovered that the temperature ranges from 902 to 731 °C over the collector bar.

The electrical conductivity is, however, heavily affected by the temperature. Therefore, the actual influence of those discovered processes on cell efficiency is still unknown and electrical conductivity is measured from room temperature up to 950 °C in this work.

The following section presents the used materials, its microstructure and chemical composition, as well as the methods for measuring the electrical conductivity assess the numerical calculations. Section 3 presents the found results, that are discussed in the following Sect. 4. The last section summarizes the substantial findings and gives an outlook to possible future research directions in the topic of aluminium electrolysis cell energy efficiency.

2 Materials and methods

To obtain material from the cathode after service time, cross-sections have been cut from the cathode of an electrolysis cell which has been shut down after a typical service time of 5 years (Fig. 3). Samples were cut out of the cross-section to obtain cast iron and collector bar samples. The positions are also given in Fig. 3.

Samles from the collector bar in their initial condition were taken from a pristine collector bar and for samples of the cast iron in an initial state, the material was casted into a mould with a diameter of 47 mm.

Finally, all the samples for electrical measurement were machined to a square geometry of 5 mm × 5 mm and a length of 20 mm, to be able to assume a one-dimensional measuring current.

Measuring the electrical resistivity was performed using a Linseis LSR-Platform (Linseis Messgeräte GmbH, Germany, type: LSR-3) with a four terminal measurement to suppress parasitic influences. The measurements have been performed at the discrete temperatures (in °C): 30, 80, 180, 280, 380, 480, 530, 580, 630, 680, 730, 780, 830, 880, 930 and 950 for 4 times at each temperature, with a measuring current of 100 mA. After reaching each temperature a holding time of at least 10 min was introduced before measuring. To obtain the specific electrical conductivity, the reciprocal of the specific electrical resistivity was calculated. For observations between the collector bar in its initial state and after service time only the Position CB.Mid is discussed in the following sections, since all three positions showed negligible differences in the electrical properties at the range of the cell operating temperature.

Using thermodynamic data (software: Thermo-Calc, version: 2021.02.03.11.50.53, database: TCFE10) the volume fractions of occurring phases for both investigated materials with the chemical composition before and after usage in the cell for 5 years have been calculated including the temperature intervals ∆T(FCC)→(BCC), with (FCC)→(BCC) indicating the microstructural transition from a face centered cubic structure (FCC) to a body centered cubic structure (BCC) that occurs for low alloyed Fe-base materials during cooling from elevated temperature. Only the diamond phase has been rejected for these calculations to take into account the occurrence of graphite precipitates, as the grey cast iron is in a stable Fe-graphite system and the long service duration of over 5 years at elevated temperatures might also lead to graphite precipitations in the wrought collector bar. The Curie temperature of the (BCC)-phase of both materials has also been calculated using Thermo-Calc software.

3 Results

3.1 Material and temperature related electrical conductivity

While the influence of the diffusion processes on electrical conductivity at room temperature is rather high, the changed microstructure (Fig. 2) and chemical composition exerts a reduced influence at elevated temperature. The relative negative influence decreases from 17% at room temperature to 11% at the maximum tested temperature of 950 °C for the collector bar, which is at total of -0.1 1/µOhm*m (Fig. 4). The trendlines of both curves for the materials are at no point linear and reach a plateau at high temperatures.

Fig. 4
figure 4

Electrical conductivity measured over a temperature range from room temperature to 950 °C. Shown are both, collector bar in its initial condition and after "end of service life"-condition (a). The difference in electrical conductivity between both materials is decreasing over temperature (b)

Full size image

As has already been found out before, the service temperature induced changes in microstructure and chemical composition have a positive influence on the specific electrical conductivity of the cast iron [15]. These measurements now show that this has a strong effect at room temperature, resulting in an increased specific electrical conductivity by 58%. Also, the total specific electrical conductivity is reducing almost linearly with different slopes (\({\alpha }_{CI}=-6.02*{10}^{-4}; {\alpha }_{CI.Top}=-10.01*{10}^{-4}\)), until a plateau at 0.48 1/µOhm*m for the cast iron in its initial state and 0.68 1/µOhm*m for the cast iron operated for 5 years is reached at about 730 °C. Within an interval from 780 to 950 °C the positive effect has reduced from 58% at room temperature to about 41%, resulting in a total difference of 0.20–0.23 1/µOhm*m (Fig. 5). However, the absolute level of electrical conductivity at room temperature is still by about one order of magnification lower for the cast iron compared to the collector bar.

Fig. 5
figure 5

Electrical conductivity measured over a temperature range from room temperature to 950 °C. Shown are both, Cast iron in its initial condition and after "end of service life"-condition (a). As can be seen the difference between both materials is decreasing almost linearly (b)

Full size image

At lower temperature the cast iron has a negative influence on the general electrical conductivity of the cell. It is functioning as the connection material between graphite and collector bar and has a lower conductivity than the collector bar. This is the case for both conditions—the initial state and the state after 5 years of service time. However, with rising temperature the conductivity for the cast iron only drops slightly in comparison to the collector bar material. For the maximum calculated service temperature at the interface between cast iron and collector bar of 885 °C [15] the difference between the two materials in specific electrical conductivity is reduced to 0.4 1/µOhm*m with the materials being in their initial state. After the service period the cast iron has a conductivity of 0.68 1/µOhm*m and the collector bar of 0.78 1/µOhm*m at 885 °C, and therefore the difference in specific electrical conductivity is only 0.1 1/µOhm*m at 885 °C (Fig. 6).

Fig. 6
figure 6

Comparison of electrical conductivity for both CB and CI over temperature range from room temperature to 950 °C in a new condition (a) and in "end of service life"-condition (b)

Full size image

3.2 Thermodynamic calculations

Since the chemical composition did change over the operating period, also phase transformations from (BCC) to (FCC) shifted. Especially the drastic depletion of carbon in the hypereutectoid cast iron shifts these transformations to lower temperatures (Fig. 7a).

Fig. 7
figure 7

Volume fraction diagram for the cast iron (a) and collector bar (b) in its initial condition (solid lines) and in “end of service”-condition (dotted lines)

Full size image

In contrary, the collector bar in its initial state is a low carbon, high purity steel and is therefore hypoeutectoid with a broad transformation interval. After usage the carbon content increased, and the collector bar material has a hypereutectoid composition close to the eutectoid point and therefore a very narrow phase transformation interval at lower temperatures (Fig. 7b). The results of the thermodynamic calculations are summarized in Table 2.

Table 2 Transformation temperatures
Full size table

For the given materials the Curie temperature, where the matrix ceases to be ferromagnetic, has been calculated to correlate it to changes in the electrical conducitvity over temperature. Since the composition of the matrix changes continuously with temperature the Curie temperature has been calculated at the equilibrium composition for 600 °C (Table 2).

4 Discussion

From our former work [15] operating temperature for CI and CB have been calculated. Electrical conductivity for those materials used in a cell can be derived from present results.

It is commonly known that for materials with metallic bonding the electrical conductivity is decreasing with rising temperatures due to the increased vibration of the lattice. It is also known that electrical conductivity is almost always decreasing in a material when impurities, dislocations, interstitial atoms, vacancies or grain boundaries occur. The resistivity can therefore be stated by Matthiessens’s rule as \({\varvec{\rho}}={{\varvec{\rho}}}_{{\varvec{T}}}+{{\varvec{\rho}}}_{{\varvec{R}}}\), where \({{\varvec{\rho}}}_{{\varvec{R}}}\) is the residual resisitivity caused of electron scattering by impurities, dislocations, interstitial atoms, vacancies, grain boundaries and shows very little temperature dependence. \({{\varvec{\rho}}}_{{\varvec{T}}}\) is the phonon scattering and shows almost linear dependency to the temperature [17] .

For the case of the investigated materials this can be very well observed for the cast iron, both in its initial state and after usage in the electrolytic cell, where there is a linear reduction of the electrical conductivity until a plateau at around 730 °C is reached. A similar trendline of the graph can be observed for the collector bar. However, the graph for the collector bar does not show the linear dependency that is to be expected.

In our former work [15] we discovered that a massive change of the chemical composition is occurring within the investigated cathode materials during the operating time of an aluminium reduction cell, which is affecting the electrical conductivity. As can be derived from Matthiessens’s rule, the influence of the chemical composition decreases at the elevated operating temperatures, due to increased phonon scattering processes. As can be seen in Figs. 4b and 5b our experimental data shows the expected decrease. However, the electrical conductivity of the collector bar at service temperatures still decreases by up to 11% over the service life of 5 years, which can only be linked to an increased concentration of alloying elements acting as impurities. Also, the increase on electrical conductivity of the cast iron due to changes in chemical composition and phase morphology, still has a measurable effect of up to 41% increased electrical conductivity at service temperature.

It has been shown that the electrical conductivity decreases with rising temperature for both materials. The cast iron in both investigated conditions shows a significant change in electrical conductivity above 700 °C. In literature this change of the slope is known and often used to determine the Curie temperature [17, 18]. Calculating the Curie temperature with ThermoCalc gives a value of 720 °C for the initial composition of the cast iron and a slightly raised Curie temperature of 727 °C for the cast iron after usage. Considering the measurement accuracy, the calculations are well fitting and can be used to validate the experimental data. However, the data of the collector bar does not show a similar change of slope at the calculated Curie temperatures of 764 °C. For the CI the calculated Curie temperature is below the onset of the (BCC) → (FCC) transformation interval, therefore the change from a ferro- to a paramagnetic state occurs suddenly and creates a sudden change of slope in the experimental data for the electrical conductivity. The CB, however, due to its different chemical composition, starts to transform from a bcc matrix to a fcc matrix before the (BCC) reaches its Curie temperature. Since the fcc phase is also paramagnetic the change of the matrix from ferromagnetic to paramagnetic happens in a broader temperature interval for the CB, and this might be the reason why there is no sudden change of slope for the CB material.

Another finding of our former work was, that the diffusion of carbon into the collector bar led to microstructural changes. The matrix at service temperature is fully austenitic with isothermally formed secondary Fe3C after 5 years of service, in contrary from being mostly ferritic with little amount of austenite (Fig. 2a, b). Colvins results on the temperature dependency of high purity iron on the electrical resistivity [19] suggest, that there is a measurable anomaly of electrical conductivity for the phase change of (BCC) to (FCC) (contributing to \({{\varvec{\rho}}}_{{\varvec{R}}}\)). However, it cannot be examined in the data plots at hand, as there are not enough measurement points around this transformation in our data and the rather small effect is superimposed by the, at the service temperatures dominant, raised electron excitation \(({{\varvec{\rho}}}_{{\varvec{T}}})\).

5 Conclusions and outlook

In this work it has been shown that the diffusion related changes of microstructure and electrical properties of the cathode materials of an aluminium electrolysis cell are not only relevant at room temperatures [15] but also at elevated temperatures up to 950 °C. As suggested by Matthiessen’s rule the influence of the change in chemical composition and microstructure decreases at elevated temperatures, because electrical resistivity increases significantly with rising temperatures due to higher electron excitation. Still, in combination with the actual service temperatures of the cathode, that have been calculated before, it has been shown, that the collector bar has an 11% (-0.1 1/µOhm*m) decreased electrical conductivity, while the cast iron still has an increased conductivity by 41% (+ 0.23 1/µOhm*m) at service temperature when being operated for 5 years. Additionally, for both materials a point where the electrical conductivity does not decrease anymore with rising temperature has been observed. The shifted transformation intervals from an (FCC) to a (BCC) microstructure have only a negligible effect that has not been observed with the experimental equipment used. For the cast iron the sudden change of the slope could be attributed to the change from ferromagnetic iron to paramagnetic iron.

The substantial findings are summarized within the following statements:

  • After 5 years of service time the collector bar material has a drastically increased carbon content and shows a decrease in electrical conductivity at service temperature by 11%.

  • The cast iron, connecting the graphite and collector bar, depletes in carbon, due to diffusion processes at elevated temperature, over service time and changes microstructure, which leads to an increase of electrical conductivity at service temperature by 41%.

  • Diffusion related processes leading to microstructural and chemical composition changes affect the electrical conductivity and energy efficiency of an aluminium electrolysis cell at service temperature.

This work shows the effect of the diffusion related processes in an electrolysis cell over 5 years on the isolated materials of a cathode. However, the materials in the process work as a composite where contact resistances, quantitative proportions and geometry as well as bonding errors within the interfaces are decisive for the total electrical conductivity of the cathode. Future work needs to reflect this situation. This could be done by creating a “lab sized” compound of collector bar, cast iron and graphite, where diffusion processes can be analysed in a shorter time period.

After a full understanding of the actions within the cathode future research should aim to improve the system by performing research on a compound that incorporates diffusion barrier layers like plated copper or nickel. To support the experimental tests, the diffusion module DICTRA from Thermo-Calc software could be used to calculate the expected diffusion reactions and the efficiency of different measures to stop or slow down those processes.