Journal of Sustainable Metallurgy (2024) 10:1594–1607

https://doi.org/10.1007/s40831-024-00887-2

RESEARCH ARTICLE

Study on Inter‑electrode Process of Aluminum Electrolysis: An Insight

into Inter‑electrode Phenomena Under Current Fluctuations
Youjian Yang1 · Yonghui Yi1 · Chengping Xia2 · Jiangyu Yu1 · Qianhan Zhao1 · Fei Wang1 · Xianwei Hu1 ·
Zhaowen Wang1

Received: 25 April 2024 / Accepted: 4 July 2024 / Published online: 22 July 2024
© The Minerals, Metals & Materials Society 2024

Abstract
Currently, there are two research focuses in aluminum electrolysis industry: process control based on individual anodic
current and current modulation. These two novel technologies share the same core mechanisms: precise control of energy
balance and heat balance of aluminum electrolysis cells, which is closely linked to the changes in inter-electrode processes
when the anodic current changes. In this study, the correlation between inter-electrode characteristics, including character-
istics of the aluminum-electrolyte interface and anode-electrolyte interface, and current density as well as anode–cathode
distance during aluminum electrolysis were investigated using the scanning reference electrode method and a see-through
electrolytic cell. The obtained variation patterns of inter-electrode voltage components may serve as a reference for current
balance control and precise thermal balance management in the multi-anode aluminum electrolysis system. The see-through
lab-scale electrolytic cell was used to statistically analyze size distribution of gas bubbles released from the bottoms of three
types of anodes during aluminum electrolysis process, aiding in understanding the resistance of the gas bubble layer.

The contributing editor for this article was Hongmin Zhu.

* Youjian Yang
[email protected]
1
Key Laboratory for Ecological Metallurgy of Multimetallic
Mineral, Northeastern University, Shenyang 110819, China
2
School of Energy and Power Engineering, Xi’an Jiaotong
University, Xi’an 710049, China

Vol:.(1234567890)
Journal of Sustainable Metallurgy (2024) 10:1594–1607 1595

Graphical Abstract

Keywords Aluminum electrolysis · Inter-electrode process · Voltage distribution pattern · Bubble distribution pattern ·
Current modulation

Introduction average current carried by each feeding point has increased

from 50 kA to over 120 kA, demanding faster alumina feed-
Aluminum electrolysis industry provides a large amount ing, dissolution, and dispersion throughout the cell. Cur-
of raw material for national defense and economic devel- rently, most alumina feeding control technologies, includ-
opment, often serving as a pillar industry of a country or ing cell resistance-based control, intelligent tracking control,
region. Development of energy-saving technologies with adaptive control, and fuzzy expert control, are based on sig-
less pollution has become an increasingly important goal nals relating to whole-cell voltage and line current [5]. How-
for the industry due to its high energy consumption. Many ever, due to increased cell size, localized conditions become
researchers have focused their studies on process control more prominent, leading to greater difficulties in balancing
based on individual anodic current [1, 2] and current modu- the distribution and consumption of alumina across the cell.
lation [3, 4] to achieve this goal. Consequently, low-voltage (or local, flash) anode effects fre-
The anodic current-based process control technology quently occur in large aluminum electrolytic cells, leading
aims to achieve precise management of local conditions in to decreased current efficiency and more greenhouse gas
the electrolytic cell using current signals from individual (perfluorocarbons-PFCs) emissions [6, 7]. In recent years,
anodes. To save investment costs, over the past 30 years, technologies for measuring anodic current distribution in
the mainstream industrial pre-baked aluminum electrolysis aluminum cell have been rapidly developed, providing hard-
cells have gradually increased their current from 160 kA ware for the control system to monitor anode status across
to 400–600 kA. The scaling up of cell production capacity different areas in the cell while significantly reducing the
makes even distribution of alumina concentration in the cell size of cell control units [8, 9]. Such technologies also bring
more challenging. Although the number of feeding points opportunities for precise process control. Based on analyses
has gradually increased over the years, it hardly matches the of intensity and frequency of anodic current signals, control
rapid increase in cell production capacity. The equivalent parameters can be identified. These technologies can not
1596 Journal of Sustainable Metallurgy (2024) 10:1594–1607

only optimize the monitoring and control of local alumina Similarly, adjustments in anode–cathode distance, differ-
concentration [10, 11], but also help diagnose and track ences in electrolyte composition, and temperature changes
localized abnormalities [12]. Additionally, they can adjust also may cause corresponding changes in inter-electrode
the current load of each anode based on an anode–cathode processes, which may be explained by studying the cor-
distance (ACD) model [13, 14]. However, how anodic cur- relation between inter-electrode voltage components and
rent is distributed in the multi-anode aluminum electrolysis inter-electrode characteristics [19].
system is closely linked to inter-electrode processes, which Various methods have been developed to study inter-elec-
still needs further investigation. trode voltage components. The scanning reference electrode
Due to tiered electricity prices and carbon emission poli- method can accurately measure the potential distribution
cies, some aluminum plants have begun seeking flexible pro- between electrodes [20]. A see-through electrolytic cell has
duction strategies or using clean energy such as wind and enabled observation of characteristics of bubble released
photovoltaic power, which requires aluminum electrolysis at an anode bottom, including bubble coverage rate, bub-
cells to adapt to variable power input. Meanwhile, aluminum ble size, and bubble release frequency [21]. The power-off
electrolysis industry, as a large adjustable power load, is also method can capture the components of counter electromo-
an important aspect for the reform of the power system at tive force, including theoretical alumina decomposition volt-
the demand end. For example, the Essen aluminum plant of age and overvoltage.
TRIMET, a German company, played an important role in In this study, the scanning reference electrode method
the implementation of a demand-side power management was applied as well as a bottom-view see-through electro-
system in Germany, where the share of renewable clean lytic cell to investigate the correlation between inter-elec-
energy in its national power grid reached 40.1%, of which, trode voltage components and inter-electrode characteristics,
wind power 20.9%, photovoltaic power 7.7%, hydropower including characteristics of aluminum-electrolyte interface
3.1%, and biomass power 7.4% in 2019 [3, 4]. and anode-electrolyte interface, in two electrolytes at current
Thermal balance management is crucial for aluminum densities of 0.6–2.6 A/cm2 and inter-electrode distances of
electrolysis cells to adapt to variable power input. To man- 2–5 cm. The patterns of each voltage component changing
age thermal balance during current modulation, side heat with current density and inter-electrode distance obtained
exchangers were introduced to maintain the integrity of from the experiments can serve as a reference for future
side crust across each aluminum electrolysis cell with vari- development of anodic current control models and precise
able power input [15, 16]. The higher the frequency and thermal balance management for aluminum electrolysis.
amplitude of current modulation, the higher the accuracy
required for thermal balance control. As electrolyte between
the electrodes generates heat due to its electric resistance,
inter-electrode distance adjustment has become a key control Experiment
method for maintaining the heat balance of an aluminum
electrolysis cell. At the same time, previous studies on inter- The schematic diagram of the scanning reference electrode
electrode processes during current modulation also involves apparatus used in this study is shown in Fig. 1. A graphite
detections of low-voltage anode effects and low alumina anode and liquid aluminum cathode are placed in a graphite
concentration [17]. crucible with an inner lining of corundum. The scanning
The inter-electrode voltage in aluminum electrolysis electrode consists of a 1 mmø tungsten wire in a corundum
consists of the following components: alumina equilib- tube, which is inserted through the center of the cylindrical
rium decomposition voltage, reaction overvoltage, anode anode, as shown in Fig. 1. The tip of the exposed tungsten
concentration polarization overvoltage, cathode concen- wire is 5 mm long. During the experiment, the voltage dif-
tration polarization overvoltage, electrolyte ohmic volt- ference between the scanning electrode and the cathode is
age drop, and bubble layer voltage drop, all of which are recorded at a frequency of 4 Hz. Before each scan, the tung-
closely linked to the current applied to the anode [18]. sten tip of the scanning electrode is pre-immersed in the
Changes in inter-electrode voltage components caused liquid aluminum for 1–2 min. Due to good wettability of
by current distribution or current modulation are asso- tungsten by liquid aluminum, the tungsten electrode carry-
ciated with various inter-electrode characteristics. For ing a layer of liquid aluminum film has the same potential
example, increased current density leads to increased as the aluminum reference electrode and remains stable dur-
bubble coverage on the anode, resulting in increased volt- ing a scan. The scanning electrode is driven by a precision
age drop of a bubble layer. Changes in metal-electrolyte guide rail system and move steadily upward at a speed of
interface may lead to increased cathode overvoltage. The 0.4 mm/s. Once the tungsten tip touches the bottom of the
increase in electrolyte ohmic voltage may lead to increased anode, a scan is completed. An example of the obtained
inter-electrode temperature and electrolyte conductivity. inter-electrode voltage distribution curve is plotted in Fig. 2.
Journal of Sustainable Metallurgy (2024) 10:1594–1607 1597

The symbols and abbreviations used in this paper are listed

in Table 2.
To aid in understanding the mechanism of bubble layer
electric-resistance, a see-through electrolytic cell was used in
combination with image analysis. In electrolyte #1, statistical
analyses were conducted on the size of bubbles released from
three types of graphite anodes with different bottom designs
­ m2,
as follows: (a) flat bottom surface with an area of 50 × 22 m
(b) longitudinal slotted bottom surface with an area of 50 × 22
­mm2, and (c) flat bottom surface with an area of 100 × 50 ­mm2.
The device used has been described in previously published
literature [21], where graphite anode and graphite cathode
were installed in compartments separated by quartz partitions
in a double-chambered quartz crucible. The partitions had a
horizontal gap at the bottom to connect the two compartments,
guiding the current flow out from the bottom of the anodes.
Images of the anode bottom were captured using a camera to
Fig. 1  Apparatus for measuring inter-electrode voltage distribution
during aluminum electrolysis record transient bubble distributions, as shown in Fig. 3. The
quantity and size of released bubbles over a period of time
were statistically analyzed.

Results and Discussion

Determine the Inter‑electrode Voltage Components

The inter-electrode voltage in aluminum electrolysis consists

of several components, including alumina equilibrium decom-
position voltage, reaction overvoltage, anode concentration
polarization overvoltage, cathode concentration polarization
overvoltage, electrolyte ohmic voltage drop, and bubble layer’s
voltage drop. By examining the potential distribution curve
scanned by the reference electrode in a single test as shown
in Fig. 2, we can obtain certain information regarding these
voltage components as well as methods for calculating certain
voltage components [22].

Fig. 2  An example of inter-electrode voltage distribution curve

U = E + Eb + Ee (1)
obtained in Electrolyte #1, LACD = 4.1 cm, Jcathode = 0.31 A/cm2
E = 𝜂n + 𝜂ca + 𝜂a + 𝜂cc (2)
As different aluminum smelters operate their electrolysis
cells at different temperatures, two compositions are used Jset = I∕S (3)
for the electrolyte in the experiments as listed in Table 1.

Table 1  Compositions, liquidus Electrolyte composition Liquidus tempera- Operating tem- Superheat, °C
temperature and operating ture, °C perature, °C
temperature of two electrolytes
used in the experiments Electrolyte #1 nNaF/nAlF3 = 2.2, 920.8 935.8 15
­CaF2 = 5%, LiF = 5%,
­Al2O3 = 3%
Electrolyte #2 nNaF/nAlF3 = 2.5, 948.5 963.5 15
­CaF2 = 5%, LiF = 3%,
­Al2O3 = 3%
1598 Journal of Sustainable Metallurgy (2024) 10:1594–1607

Table 2  Symbols and Symbol Unit Descriptions

abbreviations used in this paper
U V Total voltage range covered by the measured voltage curve
ƞn V Alumina equilibrium decomposition voltage
ƞca V Anodic concentration polarization overvoltage
ƞa V Reaction overvoltage
ƞacal1 V Calculated theoretical reaction overvoltage
ƞacal2 V Calculated reaction overvoltage including voltage drop of bubble layer
E V Counter electromotive force
Eb V Voltage drop of bubble layer
Ebcal V Equivalent voltage drop of bubble layer
Ee V Ohmic voltage drop of electrolyte
ƞcc V Cathode concentration polarization overvoltage
Lb cm Thickness of bubble layer
LACD cm Anode–cathode distance
Lcc cm Thickness of cathode transition layer
σ S/cm Electric conductivity of the electrolyte
σcc S/cm Electric conductivity of the cathode transition layer
K V/cm Slope of the segment in measured voltage curve corresponding to electrolyte
Kcc V/cm Slope of the segment in measured voltage curve corresponding to cathode
transition layer
I A Current
S cm2 Area of the anode bottom
Jset A/cm2 Anode current density set for the experiment
Janode A/cm2 Actual anode current density
Jcathode A/cm2 Cathode current density
Coverage % Bubble coverage rate at anode bottom surface
m % Current efficiency

[ ]
ln 𝜎 = 1.977 − 0.0200 × Al2 O3 − 0.0131
[ ] [ ]
× AlF3 − 0.0060 × CaF2 − 0.0106
[ ] (6)
× MgF2 − 0.0019 × [KF] + 0.0121
× [LiF] − 1204.3∕T

T is the temperature in Kelvin, and the brackets in

Eq. (6) denote the concentration of the corresponding
additives in weight percentage.
The obtained inter-electrode voltage distribution curves
show that, apart from a slight variation in slope near the
surfaces of the anode and cathode, the slope of the seg-
ment in the measured curve corresponding to the elec-
trolyte remains highly stable, indicating even current dis-
Fig. 3  A photograph of bubbles at the flat bottom anode with a sur- tribution between the electrodes, at least along the path
face area of 100 × 50 ­mm2, Electrolyte #1, current density 1.1 A/cm2 scanned by the reference electrode. Therefore, J cathode cal-
culated from the voltage distribution curve may be consid-
ered as the actual current density between the electrodes.
Jcathode = 𝜎 × K (4) In the experiments, the actual distance between the
aluminum surface and the bottom surface of the anode,
𝜎cc = Jcathode ∕KCC (5) as read from the voltage curve obtained by the scan-
ning electrode, serves as the real ­L ACD, which differs
from the set inter-electrode distance due to the bulging
Journal of Sustainable Metallurgy (2024) 10:1594–1607 1599

aluminum-electrolyte interface as the result of significant wherein, ƞ is the theoretical overvoltage, i is the current
surface tension of the liquid aluminum. density, and a and b are the Tafel parameters. Here, ƞa and
Jcathode were substituted into Eq. 12 instead [20].
Alumina Equilibrium Decomposition Voltage By fitting the values of ƞa and Jcathode into a Tafel Eq. 12,
Eqs. 13 and 14 can be obtained for Electrolyte #1 and #2,
The magnitude of ƞ n is not directly affected by ­LACD or respectively, as shown below and in Fig. 4:
current density. In this study, ƞn was calculated based on
Electrolyte #1 ∶ 𝜂acal1 = 0.745 + 0.326 × log Jcathode (13)
the operating temperature and current efficiency.
The decomposition reaction of alumina during alu-
minum electrolysis is expressed by Eq. 7 below [23]: Electrolyte #2 ∶ 𝜂acal1 = 0.498 + 0.274 × log Jcathode (14)

1∕2Al2 O3 + 3∕4C = Al + 3∕4CO2 (7)

The secondary oxidation reaction of aluminum which Anode Concentration Polarization Overvoltage
reduces the current efficiency is shown below:
The scanning reference electrode method cannot distinguish
Al + 3∕2CO2 = 1∕2Al2 O3 + 3∕2CO (8) the anode concentration polarization overvoltage and reaction
overvoltage. In this study, the anode concentration polarization
The impact of current efficiency on alumina decompo-
sition voltage is taken into account according to Pearson- overvoltage is assumed to be a constant value, ƞca = 0.036 V,
based on the literature [18].
Waddington equation as shown below:

(9) Cathode Concentration Polarization Overvoltage

( )
m = 1∕2 %CO2 + 50 %
and Ohmic Voltage Drop of Electrolyte
wherein, m denotes current efficiency.
Therefore, the overall reaction for alumina decomposi- In Fig. 2, the cathode concentration polarization overvoltage
tion can be expressed as: is shown by the instant voltage drop when the scanning probe
1∕2Al2 O3 + 3∕(4m) C = Al + (6m − 3)∕(4m) CO2 + (3 − 3m)∕(2m) CO (10)

The decomposition voltage is calculated based on ther- moves through the aluminum-electrolyte interface. It can be
modynamic data using Eq. 11 as shown below: directly read from the voltage distribution curve.
The ohmic voltage drop of the electrolyte is calculated from
ΔG⊖
T
= −nFE⊖
T (11) the following equation:
Using a common m value of 90% and corresponding
operating temperatures, the alumina decomposition volt-
age is calculated as follows: in electrolyte #1 and #2,
ηn = 1.91 V and 1.179 V, respectively. It should be noted
that the activity of alumina in the electrolyte was not taken
into consideration during the calculations.

Reaction Overvoltage

The ƞa obtained from the voltage distribution curve does

not show an obvious pattern, probably due to the gas bub-
bles released on the anode surface during the measure-
ment. Therefore, in this study the theoretical reaction over-
voltage was calculated based on the relationship between
the reaction overvoltage and current density as described
by the Tafel equation below:
𝜂 = a + b × log i (12)
Fig. 4  Tafel fitting for the calculation of ƞacal1 and ƞacal2 in Electrolyte
#1 and #2, respectively
1600 Journal of Sustainable Metallurgy (2024) 10:1594–1607

two fitted lines is mainly determined by the differences in

Ee = Jcathode × LACD ∕𝜎 (15)
alumina decomposition voltage and reaction overvoltage in
the two kinds of electrolyte used.
Voltage Drop of Bubble Layer According to Tafel Eq. (12), in the same electrolyte,
the reaction overpotential is only dependent on the current
In a test, when the reference electrode moves close to the density. In Electrolyte #1, as Jcathode increases from 0.3 A/
anode, some areas of the anode surface may be temporarily cm2 to 1.3 A/cm2, ηacal1 gradually increases from 0.577 V to
deactivated due to being covered by bubbles. As a result, the 0.782 V. In Electrolyte #2, as Jcathode increases from 0.3 A/
current is redistributed to the uncovered areas of the anode cm2 to 1.3 A/cm2, ηacal1 increases from 0.326 V to 0.551 V.
surface. This redistribution implies a change in the equipo- The difference in ηacal1 in the two electrolytes may be caused
tential lines in the electrolyte region beneath the anode. This by different electrolyte compositions and temperatures. An
phenomenon is captured by the scanning electrode and is increase in ηacal1 implies more energy is required for the elec-
manifested by changes in the potential distribution, specifi- trochemical reaction under the corresponding conditions.
cally an increase in the slope of the tested voltage distribu- The ohmic voltage drop of the electrolyte, which may be
tion curve in the region near the anode. expressed by the following equation: Ee = Jcathode × LACD/σ,
This phenomenon, caused by the partial shielding of the is directly linked to the current density, anode–cathode dis-
anode surface by bubbles, leads to a higher actual anode tance, and electrolyte’s electric conductivity. In an electro-
current density, resulting in an increase in the actual reaction lyte, Ee exhibits a linear relationship with both Jcathode and
overpotential according to Eq. (12). Therefore, the increased LACD. For instance, in Electrolyte #2, with LACD being main-
slope of the voltage distribution curve in the region near the tained at 3.87 cm, as Jcathode increases from 0.3 A/cm2 to 1.3
anode surface, as measured by the reference electrode, may A/cm2, Ee increases from 0.481 V to 2.403 V.
not be attributed to the electric resistance of the bubble layer.
Instead, it should be considered as an ‘equivalent’ voltage Counter Electromotive Force
drop caused by the increase of actual anode current density
in the bubble zone. Data in Table 3 show little correlation between the counter
Then the J anode can be obtained based on the Tafel electromotive force and LACD. From Fig. 6, it can be seen
Eq. (12), as well as the bubble coverage rate and equivalent that as Jcathode increases, E initially increases and then sta-
voltage drop of bubble layer. bilizes. It is easily understood that when current density
increases, electrochemical reactions become increasingly
𝜂acal2 = 𝜂a + Eb (16) intense. In this case, a larger reaction overvoltage is needed
to facilitate the reactions at a faster rate when E increases
(17) correspondingly. Since one of the main components of over-
( )
𝜂acal2 = a + b × log Janode
potential, ƞa, is in linear relationship with the logarithm of
Jcathode, the increasing trend of E due to the increase in ƞa is
(18)
( )
Coverage = 1 − Jcathode ∕ Janode × 100%
expected to level off as Jcathode continues to increase.

Ebcal = 𝜂acal2 − 𝜂acal1 (19) Characteristics of Aluminum‑Electrolyte Interface

The values of a and b in Eq. (17) are the same as those
This section discusses the cathode concentration polariza-
in Eqs. (13) and (14) for Electrolyte #1 and #2, respectively.
tion overvoltage, as well as the electric conductivity and
The measured and calculated data of the inter-electrode
thickness of the transition layer at the aluminum-electrolyte
voltage components are listed in Tables 3.
interface. The thicknesses of the transition layer on the liquid
aluminum surface and bubble layer on the anode surface
determine the minimum anode cathode distance in the alu-
Measured Total Inter‑electrode Voltage, Reaction
minum electrolysis process.
Overvoltage, and Ohmic Voltage Drop of Electrolyte
Cathode Concentration Polarization Overvoltage vs.
The measured results show a general trend of increasing
Current Density
total inter-electrode voltage with increasing current density,
as shown in Fig. 5. Difference in the slopes of the fitted lines
Data in Tables 3 show that the correlation between ƞcc and
for the two electrolyte compositions primarily reflect the dif-
LACD or electrolyte composition is not significant. However,
ference in the electric conductivity of the electrolytes, which
there is a clear increasing trend of ƞcc with increasing Jcathode,
is very small in this study. Difference in the intercepts of the
as shown in Fig. 7. It can be observed that ƞcc increases
Journal of Sustainable Metallurgy (2024) 10:1594–1607 1601

Table 3  Components of inter-electrode voltage

Parameters adjusted Voltage component Inter-electrode characteristic
−2
Electrolyte Jcathode/A·cm LACD/cm ƞn/V ƞacal2/V ƞca/V Ee/V ƞcc/V Ebcal/V Coverage/% σcc/S·cm−1

Electrolyte #1 0.318 4.10 1.191 0.632 0.036 0.635 0.084 0.052 31.7 1.089
3.61 1.191 0.617 0.036 0.560 0.100 0.035 23.2 1.391
3.12 1.191 0.601 0.036 0.478 0.103 0.018 13.3 1.342
2.63 1.191 0.593 0.036 0.411 0.094 0.010 8.2 1.608
2.21 1.191 0.593 0.036 0.336 0.097 0.014 10.8 1.551
0.441 4.49 1.191 0.637 0.036 0.915 0.141 0.008 6.9 1.132
3.98 1.191 0.095 0.036 0.808 0.097 ≈0 ≈0 1.006
3.49 1.191 0.080 0.036 0.735 0.094 ≈0 ≈0 1.291
3.09 1.191 0.665 0.036 0.576 0.084 0.063 37.1 1.111
2.66 1.191 0.652 0.036 0.472 0.088 0.052 31.6 1.372
0.538 4.51 1.191 0.745 0.036 1.131 0.163 0.088 47.1 1.028
4.11 1.191 0.711 0.036 1.081 0.156 0.050 30.7 1.131
3.62 1.191 0.747 0.036 0.929 0.169 0.090 48.1 1.358
3.16 1.191 0.744 0.036 0.857 0.144 0.089 47.5 1.053
2.68 1.191 0.742 0.036 0.681 0.159 0.097 50.3 1.284
0.743 3.80 1.191 0.825 0.036 1.332 0.155 0.122 58.4 1.513
0.870 3.84 1.191 0.878 0.036 1.714 0.160 0.153 66.5 1.841
0.967 3.86 1.191 0.902 0.036 1.750 0.174 0.162 68.5 1.536
1.027 3.89 1.191 0.920 0.036 1.925 0.195 0.171 70.5 1.418
1.121 3.96 1.191 1.080 0.036 2.008 0.205 0.319 89.6 1.610
1.242 4.02 1.191 1.277 0.036 2.349 0.229 0.501 97.1 1.251
1.315 4.08 1.191 1.370 0.036 2.486 0.249 0.586 98.4 1.323
Electrolyte #2 0.318 4.08 1.179 0.373 0.036 0.576 0.075 0.016 12.4 2.339
3.55 1.179 0.326 0.036 0.481 0.081 ≈0 ≈0 1.922
3.10 1.179 0.348 0.036 0.396 0.075 0.015 11.7 2.091
2.59 1.179 0.351 0.036 0.338 0.081 0.009 7.3 1.912
2.14 1.179 0.166 0.036 0.298 0.072 ≈0 ≈0 1.768
0.441 4.64 1.179 0.466 0.036 0.858 0.106 0.121 63.9 2.839
4.17 1.179 0.407 0.036 0.780 0.109 0.041 29.1 2.997
3.69 1.179 0.416 0.036 0.670 0.106 0.093 54.3 1.783
3.10 1.179 0.369 0.036 0.523 0.031 0.043 30.1 1.563
2.77 1.179 0.369 0.036 0.511 0.100 ≈0 ≈0 1.749
0.538 4.63 1.179 0.457 0.036 1.110 0.144 0.117 62.5 1.395
4.19 1.179 0.444 0.036 0.942 0.153 0.111 60.8 1.449
3.68 1.179 0.410 0.036 0.834 0.150 0.072 45.3 1.404
3.24 1.179 0.395 0.036 0.727 0.147 0.041 29.4 1.439
2.31 1.179 0.333 0.036 0.561 0.141 ≈0 ≈0 1.139
0.549 3.55 1.179 0.479 0.036 0.802 0.138 0.368 95.5 2.798
0.743 3.81 1.179 0.485 0.036 1.196 0.156 0.104 58.3 2.163
0.870 3.85 1.179 0.482 0.036 1.415 0.163 0.126 65.4 2.334
0.967 3.87 1.179 0.469 0.036 1.601 0.209 0.115 62.1 2.108
1.027 3.91 1.179 0.435 0.036 1.758 0.213 0.129 66.2 1.949
1.121 3.95 1.179 0.551 0.036 1.959 0.225 0.370 95.6 1.863
1.242 3.99 1.179 0.382 0.036 2.207 0.244 0.401 96.6 1.768
1.315 4.05 1.179 0.363 0.036 2.403 0.250 0.411 96.9 1.600
1602 Journal of Sustainable Metallurgy (2024) 10:1594–1607

Fig. 7  The relationship between cathode concentration polarization

overvoltage and cathode current density in Electrolyte #2
Fig. 5  The relationship between total inter-electrode voltage and
cathode current density

Fig. 8  The electric conductivity of the cathode transition layer vs.

cathode current density

Fig. 6  The relationship between counter electromotive force and cath-

ode current density
approximately 1.2 S/cm and 1.8 S/cm, respectively, which
are relatively low compared to the electric conductiv-
almost linearly with the increase of Jcathode. In Electrolyte #2, ity of the pure electrolyte (calculated as 2.2–2.4 S/cm),
when Jcathode changes from 0.318 A/cm2 to 1.315 A/cm2, ƞcc suggesting that a thin layer which has a lower electric
changes from 0.081 V to 0.250 V accordingly. conductivity than the liquid electrolyte is present near
the cathodic aluminum surface. This characteristic is
Electric Conductivity of the Cathode Transition Layer consistently present in all the curves obtained. As the
at the Aluminum‑Electrolyte Interface electrolyte composition and operating temperature were
carefully controlled, the electrolyte superheat reached
Data in Table 3 show that σcc does not have a strong cor- 15 °C, raising questions about the presence of a ‘frozen
relation with LACD or Jcathode. The relationships between layer’ at the aluminum-electrolyte interface as mentioned
σcc and Jcathode in both Electrolytes #1 and #2 are shown in previous literature [24]. However, it is possible that
in Fig. 8. For both electrolyte compositions #1 and #2, a ‘cathodic transition layer’ indicated by the obtained
the average values of σcc within the measured range are curves is present, which may be caused by the significant
difference in electric conductivity between the electrolyte
Journal of Sustainable Metallurgy (2024) 10:1594–1607 1603

and aluminum, leading to redistribution of current at the

aluminum-electrolyte interface. It should be noted that
the physical presence of the cathode transition layer may
not be distinctive.

Thickness of the Cathode Transition Layer

at the Aluminum‑Electrolyte Interface

Statistical analyses were conducted on the thickness of the

cathode transition layer at the aluminum-electrolyte inter-
face. The results indicate that Lcc does not have a strong
correlation with LACD or Jcathode. As shown in Fig. 9, Lcc is
less than 0.8 cm.

Characteristics of Anode‑Electrolyte Interface

Anode Bubble Coverage Rate Fig. 10  The relationship between the anode bubble coverage rate and
cathode current density
Figure 10 shows the relationship between the anode bub-
ble coverage and Jcathode. It can be observed that as Jcathode
increases, the anode bubble coverage exhibits a significant Voltage Drop of Anode Bubble Layer vs. Current Density
increasing trend. This is because increased current density
leads to faster electrochemical reactions with more gas gen- Figure 12 depicts the relationship between Ebcal and Jcathode.
eration on the anode surface, resulting in higher bubble cov- It is evident that as anode bubble coverage increase, Ebcal
erage rate. initially increases and then stabilizes (the stabilization corre-
The increase in anode bubble coverage on the anode sur- sponds to a maximum bubble coverage rate of no more than
face significantly reduces the effective working area of the 100%) with the increase of Jcathode. When Jcathode changes
anode, leading to a substantial increase in the actual current from 0.3 A/cm2 to 1.1 A/cm2, Ebcal changes from 0.01 V to
density on the effective working surface area. As a result, 0.13 ~ 0.17 V.
Ebcal significantly increases as illustrated in Fig. 11, where
Ebcal shows a clear positive correlation with the anode bub-
ble coverage.

Fig. 9  Thickness of the cathode transition layer vs. cathode current Fig. 11  The relationship between the equivalent voltage drop of bub-
density ble layer and anode bubble coverage rate
1604 Journal of Sustainable Metallurgy (2024) 10:1594–1607

When the aluminum cathode comes into contact with C ­ O2

generated by the anode, it will get re-oxidized in accordance
with Eq. (8), which is one of the main causes for loss of cur-
rent efficiency in aluminum electrolysis. The smaller LACD
is, the more likely the cathodic aluminum will react with
the anodic C­ O2, resulting in more severe re-oxidizations of
aluminum [25].
In order to achieve lower energy consumption in alu-
minum electrolysis, modern technologies tend to operate at
low LACD. This study reveals that Lb and Lcc are approxi-
mately 0.5 cm and 0.8 cm, respectively, and do not vary
significantly with changes in LACD or current density. These
findings indicate that reducing LACD to 1.3 cm could lead to
significant re-oxidizations of aluminum. To reduce the pos-
sible contact between the suspended tiny aluminum droplets
(or dissolved aluminum) near the cathode with dynamically
Fig. 12  Equivalent voltage drop of bubble layer vs. cathode current released anodic gas bubbles, an additional 1 cm is added to
density LACD. In this study, complete inter-electrode voltage curves
could be obtained when LACD was set to 2.3 cm or more.
In industrial aluminum electrolysis cells, the cathodic alu-
minum fluctuates significantly under the effect of the elec-
tromagnetic field, leading to uneven working surface of each
anode. Therefore, another additional 1 cm of LACD is often
added to ensure production targets can be met. This is con-
sistent with the practice of some aluminum smelters control-
ling LACD at 3–3.5 cm. If further reduction in LACD is to be
achieved, the following factors may be taken into account:
use of slotted anodes (to reduce bubble size and electrolyte
disturbance), improvement in anode quality (increased elec-
trode’s working surface area), and cathode design optimiza-
tion (reduced fluctuation of liquid aluminum surface), and
others.

Size Distribution of Gas Bubbles Released

from the Anode

Fig. 13  Scatter plot of anode bubble layer thickness with cathode cur-
Earlier research [21] has examined the distribution patterns
rent density of bubbles adhered on the anode bottom surface and their
impact on the anode working condition. This section primar-
ily focuses on the principle of released bubbles and presents
Thickness of the Anode Bubble Layer the effects of anode geometry on the size distribution of the
released bubbles.
It is worth noting that the scanning reference electrode It has been observed that the variation in the amount of
method does not provide data of the actual bubble thick- released anode gas is closely linked to the alumina distribu-
ness. This method directly measures inter-electrode volt- tion pattern and, consequently, the anode deformation pat-
age distribution, which can be considered as distribution of tern. The rapid flow of the electrolyte, agitated by the buoy-
equipotential lines. Herein, the term ‘bubble layer thickness’ ancy forces of the released bubbles, indicates a fast mass
refers to the thickness of the zone affected by the bubbles. transfer rate of the fed alumina. This allows for quick replen-
Figure 13 shows a scatter plot of Lb obtained from all ishment of the consumed alumina at the anode-electrolyte
experiments conducted in Electrolyte #1 and #2. It can interface, leading to a rapid release of the anode gas and low
be seen that except for one outlier, Lb remains lower than incidence rate of anode problems [26].
0.5 cm when different cathode current densities are applied. Thorne [27] analyzed different bubble behavior on vari-
ous industrial and graphite anode samples. She discovered
Journal of Sustainable Metallurgy (2024) 10:1594–1607 1605

which, in turn, is related to the impurity level and structure

of the anode material, including the degree of isotropy and
carbon crystallite size. However, in Thorne’s study, the effect
of anode wettability on polarization was not considered.
Figure 14 shows the size distribution of bubbles released
from the bottom of three types of anodes with different bot-
tom geometries as follows: (a) flat surface with an area of
50 × 22 ­mm2, (b) slotted surface with an area of 50 × 22
­mm2; and (c) flat surface with an area of 100 × 50 ­mm2.
For type (a) anode, the bubble size distribution is shown
in Fig. 14a. It can be seen that bubbles of 0 to 100 m ­ m2
account for 40% of the total bubbles released during the
measured period, while bubbles of 300 to ­500mm2 account
for approximately 50% of the total. Moreover, the latter
larger bubbles cover 30%–50% of the anode bottom surface
area before being released, leading to highly regular voltage
fluctuations in the cell.
In contrast, the bubbles released from the bottom of type
(b) slotted anode are mainly of 0–100 m ­ m2, as shown in
Fig. 14b. The slot significantly reduces the minimum width
of the anode from 22 to 10 mm, making bubble release eas-
ier by providing easier access to anode edge.
After the flat surface of the anode bottom is further
increased to 100 × 50 ­mm2, bubbles of 0–100 ­mm2 account
for 45% of the total released bubbles, those of 100–1000
­mm2 approximately 45%, and those larger than 1000 ­mm2
the rest 10%, as shown in Fig. 14c.
It can be observed that the size of bubbles released from
the anode bottom is closely related to the minimum width
of the anode surface, or say the critical area of the bubble.
The critical area of the bubble on type (a) and (c) anodes are
484 ­mm2 and 2500 ­mm2, respectively. Due to the presence
of the slot, the critical area of the bubble on type (b) anode
is 100 ­mm2. Subsequently, in Fig. 14, very few bubbles in
the three types of anodes exceeded these critical sizes. It
can be deduced that, under the same current density, anode
with smaller bubble sizes imply faster renewal of bubbles
on the anode surface, which not only ensures great electro-
lyte circulation on the anode surface, but also keeps anode
bubble coverage rate low, leading to a lower voltage drop of
bubble layer.

Conclusion

This study employed the scanning reference electrode

Fig. 14  Size distribution of bubbles released from the bottom of three method to measure voltage distribution between the anode
types of anodes and cathode during aluminum electrolysis process. The
inter-electrode voltage components were analyzed, and the
correlations between inter-electrode characteristics, includ-
that the resistance of the bubble layer and the oscillation ing characteristics of the aluminum-electrolyte interface
of cell voltage caused by bubble evolution are significantly and anode-electrolyte interface, and current density were
affected by the wettability of the anode by the electrolyte, discussed. It was found that as the current applied to the
1606 Journal of Sustainable Metallurgy (2024) 10:1594–1607

electrolytic cell increased, the electrolyte’s ohmic voltage and challenges in reudusing these emissions. Proceedings of
drop generally increased linearly, leading to a basically 37th international ICSOBA conference, pp. 735–758
8. Potocnik V, Arkhipov A, Ahli N, Alzaroonil A (2017) Measure-
linear increase in the total cell voltage. However, increase ment of DC busbar currents in aluminium smelters. Proceedings
of reaction overvoltage and voltage drop of bubble layer of 35th international ICSOBA conference, pp 1113–1128
gradually leveled off at higher current densities. This paper 9. Wong C, Shi J, Bao J, Welch B, Skyllas-Kazacos M, Jassim A,
also explored the relationship between size distribution Mahmoud M, Nikandrov K (2023) A smart individual anode
current measurement system and its applications. Light Metals.
of released bubbles and minimum width of anode bottom https://​doi.​org/​10.​1007/​978-3-​031-​22532-1_6
based on bubble sizes measured in an electrolysis cell with 10. Sun Y, Gui W, Chen X, Xie Y, Xie S, Zou Z (2023) A dynamic
a see-through bottom. The research results provide valuable spatial distributed information clustering method for aluminum
insights into current distribution mechanism in multi-anode electrolysis cell. Eng Appl Artif Intell 126:106793. https://​doi.​
org/​10.​1016/j.​engap​pai.​2023.​106793
electrolysis systems and variations in aluminum processes 11. Yao Y, Cheung C, Bao J, Skyllas-Kazacos M, Welch B, Akhme-
during current modulation. tov S (2017) Estimation of spatial alumina concentration in
an aluminum reduction cell using a multilevel state observer.
Acknowledgements The authors would like to acknowledge the AIChE J 63(7):2806–2818. https://​doi.​org/​10.​1002/​aic.​15656
financial support from Liaoning Natural Science Foundation (Grant 12. Lundby E, Rasheed A, Gravdahl J, Halvorsen I (2023) Sparse
No. 2022-MS-121), the Fundamental Research Funds for the Central deep neural networks for modeling aluminum electrolysis
Universities (Grant No. N2225045), and the National Natural Science dynamics. Appl Soft Comput 134:109989. https://​doi.​org/​10.​
Foundation of China (Grant No. 52341402). 48550/​arXiv.​2209.​05832
13. Hua J, Beckstein P, Manger E, Kolas S, Jensen O, Marholm S
Funding This work was funded by Liaoning Natural Science Founda- (2024) Numerical modeling of anode changes and their effect
tion, 2022-MS-121, Youjian Yang, the Fundamental Research Funds on current distribution and magnetohydrodynamic behavior of
for the Central Universities, N2225045, Xianwei Hu, the National an aluminium reduction cell. Light Metals. https://​doi.​org/​10.​
Natural Science Foundation of China, 52341402, Zhaowen Wang. 1007/​978-3-​031-​50308-5_​61
14. Dion L, Kiss L, Poncsak S, Lagace C (2018) Simulator of non-
Declarations homogenous alumina and current distribution in an aluminum
electrolysis cell to predict low-voltage anode effects. Metall
Conflict of interest The authors declare that there is no conflict of in- and Mater Trans B 49(2):737–755. https://​d oi.​o rg/​1 0.​1 007/​
terest regarding the publication of this paper. s11663-​018-​1174-2
15. Eisma D, Patel P (2016) Challenges in power modulation.
Essent Read Light Metals. https://​doi.​org/​10.​1002/​97811​18647​
851.​ch101
References 16. Taylor M, Etzion R, Lavoie P, Tang J (2014) Energy balance
regulation and flexible production: a new frontier for aluminum
1. Meng Y, Tie J, Li C, Zhao R, Jiang H, Peng X, Xiao H, Liu D, smelting. Metall Mater Trans E 1:292–302. https://​doi.​org/​10.​
Lei J (2024) Accurate measurement of anode current in alu- 1007/​s40553-​014-​0029-2
minum electrolysis: from ideal to reality. Light Metals. https://​ 17. Cheung C, Menictas C, Bao J, Skyllas-Kazacos M, Welch B
doi.​org/​10.​1007/​978-3-​031-​50308-5_​75 (2013) Frequency response analysis of anode current signals
2. Yang S, Zou Z, Li J, Zhang H (2016) Online anode cur- as a diagnosis aid for detecting approaching anode effects in
rent signal in aluminum reduction cells: measurements and aluminum smelting cells. Light Metals. https://​doi.​org/​10.​1007/​
prospects. JOM 68(2):623–634. https:// ​ d oi. ​ o rg/ ​ 1 0. ​ 1 007/​ 978-3-​319-​65136-1_​151
s11837-​015-​1738-4 18. Kvande H, Haupin W (2000) Cell voltage in aluminum elec-
3. Vogel L, Schenuit C, Jian L (2019) Industrial demand side flex- trolysis: a practical approach. JOM 52(2):31–37. https://​doi.​org/​
ibility in China. Deutsche Energie-Agentur GmbH (dena). Ger- 10.​1007/​s11837-​000-​0044-x
man Energy Agency. https://​www.​seman​ticsc​holar.​org/​paper/​ 19. Yang Y, Gao B, Wang Z, Shi Z, Hu X (2016) Study on the inter-
Indus​trial-​Demand-​Side-​Flexi​bility-​in-​China/​bc2db​ae8fb​e1669​ electrode process of aluminum electrolysis. Metall Mater Trans
9503e​93318​74d4b​9d549​caf53#​citing-​papers. Accessed 1 Aug B 47B:621–629. https://​doi.​org/​10.​1007/​s11663-​015-​0508-6
2019 20. Haupin W (1971) A scanning reference electrode for voltage
4. Depree N, Dussel R, Patel P, Reek T (2016) The ‘virtual bat- contours in aluminum smelting cells. J Metals 23(10):46–49.
tery’—operating an aluminium smelter with flexible energy https://​doi.​org/​10.​1007/​BF033​55737
input. Light Metals. https://​doi.​org/​10.​1007/​978-3-​319-​48251- 21. Yang Y, Zhang Y, Yu J, Wang Z, Shi Z (2021) Study on the
4_​96 inter-electrode process of aluminum electrolysis(II)—digital
5. Kvande H, Moxnes B, Skaar J, Solli P (2016) Pseudo resistance analysis of the anode gas distribution patterns on the anode
curves for aluminium cell control—alumina dissolution and cell surface using a see-through cell. Appl Sci 11:7702. https://​doi.​
dynamics. Essent Read Light Metals Alum Reduct Technol. org/​10.​3390/​app11​167702
https://​doi.​org/​10.​1007/​978-3-​319-​48156-2_​113 22. Hives J, Thonstad J, Sterten A, Fellner P (1996) Electrical
6. Batista E, Dando N, Menegazzo N, Espinoza-Nava L (2016) conductivity of molten cryolite-based mixtures obtained with
Substainable reduction of anode effect and low voltage PFC a tube-type cell made of pyrolytic boron nitride. Metall Mater
emissions. Light Metals. https://​d oi.​o rg/​1 0.​1 007/​9 78-3-​3 19-​ Trans B 27B:255–261. https://​doi.​org/​10.​1007/​BF029​15051
48251-4_​89 23. Pearson T, Waddington J (1947) Electrode reactions in the alu-
7. Wong D, Welch B, Nunez P, Dion L, Spirin A (2019) Latest minium reduction cell. Discuss Faraday Soc 1:307–320. https://​
progress in IPCC methodology for estimating the extent of PFC doi.​org/​10.​1039/​DF947​01003​07
greenhouse gases co-evolved in the aluminium reduction cell 24. Solheim A, Giskeodegard N, Holt N (2016) Sideledge facing
metal in aluminium electrolysis cells: freezing and melting in
Journal of Sustainable Metallurgy (2024) 10:1594–1607 1607

the presence of a bath film. Light Metals. https://​doi.​org/​10.​ Publisher's Note Springer Nature remains neutral with regard to
1007/​978-3-​319-​48251-4_​55 jurisdictional claims in published maps and institutional affiliations.
25. Solheim A (2024) Limits for the current efficiency in Hall-
Heroult cells. Light Metals. https://​doi.​org/​10.​1007/​978-3-​031-​ Springer Nature or its licensor (e.g. a society or other partner) holds
50308-5_​60 exclusive rights to this article under a publishing agreement with the
26. Majid N, Taylor M, Chen J, Young B (2015) Aluminium process author(s) or other rightsholder(s); author self-archiving of the accepted
fault detection and diagnosis. Adv Mater Sci Eng. https://​doi.​ manuscript version of this article is solely governed by the terms of
org/​10.​1155/​2015/​682786 such publishing agreement and applicable law.
27. Thorne R, Sommerseth C, Ratvik A, Rorvik S, Sandnes E, Los-
sius L, Linga H, Svensson A (2015) Bubble evolution and anode
surface properties in aluminium electrolysis. J Electrochem Soc
162:104–114. https://​doi.​org/​10.​1149/2.​03215​08JES