Light Metals 2009 Edited by: Geoff Bearne

TMS (The Minerals, Metals & Materials Society), 2009

Spent Si3N4 Bonded SiC Sidelining Materials in Aluminium Electrolysis Cells

1

Zh. Wang1, E. Skybakmoen2, T. Grande1,3

Department of Materials Science and Engineering, Norwegian University of Science and Technology, N-7491Trondheim, Norway
2
SINTEF Materials and Chemistry, N-7465Trondheim, Norway
3
Corresponding author: tor.grande@material.ntnu.no
Keywords: Aluminium Electrolysis, Sidelining, Degradation, Si3N4-SiC

Abstract

Fabrication

Si3N4 bonded SiC sidelining materials with different ages in

operation have been analyzed by X-ray diffraction, electron
microscopy and chemical analysis. The chemical degradation of
the upper and the lower part of the sidelining has been
demonstrated to be substantially different. The upper part of the
sidelining has been subjected to oxidation by the pot gas and the
main degradation product found was Si2ON2. The lower part was
infiltrated by Na(g) diffusing from the carbon cathode resulting in
formation of Na2SiO3 as the main oxidation product. Chemical
reactions are proposed based on the experimental findings. The
diffusion of the degradation species into the sidelining has been
modeled by a finite element model. The degradation over time
results in the change of the thermal conductivity of the sidelining
materials.

Si3N4 bonded SiC sidelining blocks are commonly produced by

using a mixture of silicon carbide grains and finely divided
(typically < 325 mesh) elemental silicon and a binder. The block
can be formed into shapes via slip casting, vibratory casting, or by
pressing. The green blocks are fired in a furnace in an air
atmosphere at low temperature (up to 300C) and then in a
nitrogen atmosphere at temperatures below the melting point of
silicon, normally around 1400C. The nitrogen gas and silicon
metal react forming Si3N4 as a binder phase between large SiC
grains. A Si3N4 bonded SiC block typically contains 72-80 wt%
SiC and 20-28 wt% Si3N4 [6,10,11]. Dependent on the binder
used, the purity of the raw materials and the nitridation conditions,
trace of oxide impurity phases, e.g. Si2ON2, Fe2O3, Al2O3, CaO,
may also be present in addition to residual Si [14]. The porosity of
the block is around 15-16% and the density is typically around 2.7
g/cm3 [12]. Typical thickness of the blocks used in aluminium
cells is 50-100 mm.

Introduction
Si3N4 bonded SiC materials have during the last decades become
the state of the art sidelining material in aluminium electrolysis
cells due to its excellent oxidation resistance compared with
traditional carbon lining, poor electrical conductivity, and high
thermal conductivity. By using Si3N4 bonded SiC materials, the
total thickness of the sidelining and the sideledge can be reduced
and larger anodes can be used [1-4]. In modern cell designs, the
thickness of sidelining blocks can be down to 5 cm thickness.
During the operation, Si3N4 bonded SiC materials will suffer the
chemical degradation [5-8,11,13]. The development towards high
amperage cells with thinner sidelining and the demand for long
pot lifetime call for a better understanding of the chemical
stability of Si3N4 bonded SiC materials.

Si3N4 binding phase is rarely phase pure and consists of - and Si3N4 phase formed by different mechanisms [9]. The ratio
between the two phases will vary according to the fabrication
condition. -Si3N4 phase is typically dominant in the whole block.
The content of -Si3N4 phase is higher in the interior of the block
than close to the surface [13].

Experimental
Spent Sidelining Materials and Sampling
Three different spent sidelining blocks have been analyzed, which
are supplied by two producers and taken from autopsies with
different ages in operation: 270 days (Sicatec 75, LIRR, China),
604 days (Sicatec 75, LIRR, China) and 1569 days (Refrax 20,
Saint Gobain). The cross section of the spent sideling block is
shown in Figure 1.

In this paper we report on the detailed analysis of the Si3N4

bonded SiC sidelining materials recovered from the autopsies of
the pots after 270, 604 and 1569 days operation respectively.
Changes in the microstructure and mineralogical transition are
investigated and reaction mechanisms leading to degradation of
the Si3N4 bonded SiC sidelining materials are proposed. The
diffusion of the degradation gases into the sidelining block has
been simulated by a commercial software based on finite element
method. Based on the experimental finding and the simulation, the
degradation route and mechanism is proposed. The trend of the
change in thermal conductivity of the sidelining block over time is
also reported.

For convenience in this paper the spent sidelining is divided into

two regions gas zone and electrolyte zone defined by the level
of the molten electrolyte in the electrolysis cell. The two regions
correspond to the two parts of sidelining above and under the
electrolyte bath level. The darker area in the electrolyte zone is
labeled infiltration area.

353

Results
Mineralogical and Chemical Analysis of Spent Sidelining
The phase composition of the blocks at the age of 270 and 604
days are similar. The main degradation products detected are
Si2ON2 in the gas zone and Na2SiO3 in the electrolyte zone, see
Figure 3. The basic conditions at the electrolyte zone induced by
Na related reactions were confirmed by a pH paper test. The
content of - Si3N4 phase is more pronounced in the gas zone than
in the electrolyte zone and in the pristine materials.

Figure 1. The cross section of the spent sidelining block at the

age of 270, 604 and 1569 days.
Samples were cut by using diamond saw from both the electrolyte
zone and the gas zone of the blocks as shown in Figure 2.

Figure 3. XRD profiles for samples from electrolyte zone, gas

zone of spent block at the age of 270 days (Sicatec 75) together
with 2 samples from pristine block. - represents -Si3N4; represents -Si3N4. Un-labeled peaks are for SiC.
For the older cell at the age of 1569 days, Si2ON2 was found
further down to the electrolyte zone near the bottom of the block
in addition to the main degradation phase Na2SiO3. In the gas
zone, chiolite (Na5Al3F14) and AlF3 were indentified besides
Si2ON2. The fluoride phases are proposed to be formed by the
condensation reaction of the pot gas NaAlF4. The presence of Si3N4 phase is more pronounced and -Si3N4 phase was hardly
detected in the gas zone. The corresponding XRD diffractograms
are shown in Figure 4.

Figure 2. Sampling position in the spent sidelining block (not in

scale).
Sample Characterization
Phase analysis of the samples was performed by using a Bruker
D8 Focus X-ray diffractometer with a LynxEye detector. Data for
phase identification were collected with a step size of 0.015 and
a count time of 0.3 s over the 2 range 10 to 80. Rietveld
refinement of the diffractograms was performed by using the
Topas software (Bruker AXS) to give quantitative phase
composition for the identified phases [13]. Total oxygen and
nitrogen contents were analyzed by LECO TC-436DR by using a
selective hot gas extraction method. Bulk density and apparent
porosity of the samples were measured according to standard
ISO5017 based on Archimedes principle. Thermal conductivity
was measured by split column method (CERAM) for samples
with diameter 50mm and length 30mm taken from the spent
blocks.

Figure 4. XRD profiles for samples from electrolyte and gas zone
of spent block at the age of 1569 days (Refrax). - represents Si3N4; - represents -Si3N4. Un-labeled peaks are for SiC.

354

Si contents for all blocks are in the range of 0.2-0.7wt% and have
no significant change with age. The amount of the degradation
phases and -Si3N4, -Si3N4 present in the spent blocks at
different ages is summarized in Table 1. The data were obtained
by quantitative Rietveld refinement.
Table 1. Rietveld refinement of the X-ray diffraction patterns.
Unit of phase amount is wt%.
1: close to the frozen bath. 4 or 5: close to the steel shell.

Figure 5. /-Si3N4 ratio as a function of Si2ON2 amount. Dotted

line is the guide to the eyes.
The chemical analysis (LECO) gives the concentration of the total
nitrogen and oxygen in the samples. The increase in oxygen
content is always accompanied by a decrease in the nitrogen
content as shown in Figure 6. The nitride binder phase is clearly
the chemically degraded part of the sidelining materials.

For the younger cell (270 days), the content of reaction product
phases in the electrolyte zone is much higher than in the gas zone.
For the older cell (1569 days), the oxidation in the gas zone is
more severe and Si2ON2 is found down to the electrolyte zone.
Na2SiO3 is restricted to the electrolyte zone, which means that
although Na(g) diffusing from the carbon cathode into the
sidelining in the horizontal direction, it does not penetrate much
further upwards in the vertical direction. We propose that excess
oxygen containing species diffusing downwards from the top of
the sideling condenses the Na(g) and retards further diffusion and
reaction of Na(g). Finally, it should be noted that the amount of
the degradation phase Na2SiO3 in the older cell is not higher than
that in the younger cell.
-Si3N4 phase is dominant in the binder phase of the pristine
block. The /-Si3N4 ratio is less than one. In the gas zone after
degradation by pot gas, the ratio increases due to the reduction of
the -Si3N4 phase content and increase in -Si3N4 phase. The ratio
is even revised and larger than 1 in the block at the age of 270 and
1569 days, which implies that -Si3N4 dominates in the gas zone.
This indicates -Si3N4 is more reactive and suggests the
possibility of to -Si3N4 transformation [13]. The phenomenon
is more pronounced in the cell at the age of 1569 days: /-Si3N4
ratio could reach 16.4 and -Si3N4 phase is as low as 0.5 wt%.
/-Si3N4 ratio seems to have a relationship with the amount of
Si2ON2 as shown in Figure 5. With increasing Si2ON2 amount, the
/-Si3N4 ratio increases accordingly. It implies that -Si3N4 is
preferably oxidized to form Si2ON2 and that the to -Si3N4
transformation may be related to the formation of Si2ON2. In any
case, /-Si3N4 ratio correlates well with the degree of the
oxidation by pot gas.

Figure 6. Oxygen content as a function of nitrogen content in the

spent sidelining blocks. Dotted line is the guide to the eyes.
The oxidation is becoming more severe with time as shown in
Figure 7. For the younger cell, only the area close to the bath is
oxidized, while for the older cell, the oxidation area extends
further towards the steel shell. Generally, the oxidation in
electrolyte zone is more severe than the oxidation in gas zone. In
the old cell operated for 1569 days, the oxidation in gas zone
increases significantly and the oxygen content in the two regions
are more comparable.

355

Figure 7. Oxygen content vs. distance to the bath for the spent
sidelining blocks at the age of 270, 604 and 1569 days.

Figure 9. The porosity as a function of oxygen content for the

blocks aged 270, 604 and 1569 days. Dotted line is the guide to
the eyes.

Microstructure

Thermal Conductivity

Density and Porosity After cell operation, the degradation

products (Na2SiO3 and/or Si2ON2) fill up the pores and change the
density of the block correspondingly. The porosity across the
block with different ages is shown in Figure 8. In the infiltration
area, the porosity is less than 5%. In the old cell (1569 days), the
porosity is down to around 1%. Correspondingly, the density is
increased up to 2.9 g/cm3. In the gas zone, the porosity remains
high around 9-16% and the density dose not change significantly.

Figure 10 shows the thermal conductivity as the function of

temperature for several materials. Data for the pristine materials
given by the supplier are shown for comparison.

Figure 10. Thermal conductivity for 3 samples from infiltration

area and 1 sample from gas zone as well as the values from the
suppliers specification for comparison.
Figure 8. The porosity vs. distance to the bath for the spent
sidelining blocks at the age of 270, 604 and 1569 days.

Particular attention should be paid to the thermal conductivity in

the region 550-700oC, which is supposed to be the temperature of
sidelining blocks under operation. Clearly, thermal conductivity
of spent blocks is lower than the pristine materials in this
temperature region. That is due to the formation of Na2SiO3 which
has a significant lower thermal conductivity. The reduction degree
is related to the amount of Na2SiO3, not the cell age itself.
Average Na2SiO3 contents in spent sidelining materials at the age
of 270, 604 and 1569 days are around 11.7, 4.8 and 5.6wt%
respectively, which is basically the revised sequence in the
thermal conductivity. The spent block aged 270 days with high
amount of Na2SiO3 has the lowest thermal conductivity. It is
proposed that a higher Na2SiO3 content is pointing to a lower
thermal conductivity. In gas zone, the thermal conductivity does
not decrease significantly.

The morphology of the degradation products has been previously

investigated by SEM/EPMA/WDS [13]. Na2SiO3 phase was found
dense and flake like and Si2ON2 was suggested as fiber like. That
is in agreement with that the density of the Na infiltration area
increases significantly due to filling the dense Na2SiO3 flakes in
the pores while the density and porosity of the gas zone do not
change very much because Si2ON2 grows into a less dense
microstructure.
The reduction in the porosity reflects the extent of oxidation and
the porosity correlates with the oxygen content as shown in Figure
9. The decrease in porosity may contribute to reduce the diffusion
of Na(g) into the sidelining.

356

Simulation

Na(g) at the electrolyte zone. As long as there is surplus oxidation

gases down to the electrolyte zone, Na(g) will be restricted inside
this area. Therefore, we proposed that the actual diffusion of
Na(g) in this region is governed by the reactions and less due to
reduction in the effective diffusion coefficient.

Na(g) is proposed to diffuse from the carbon cathode into the

sidelining through the carbon ramming paste. The profile of the
Na(g) diffusion is modeled by a commercial software based on
finite element method shown in Figure 11.

Discussion
Reaction Schemes and Diffusion Route
During normal operation, the sidelining block is almost always
covered by a frozen electrolyte layer (ledge) or crust and the
sidelining is not exposed to the aluminium pad and the electrolyte
directly [20]. However, the lining materials still deteriorate
gradually over time due to Na(g) diffusion and the accompanied
reactions leading to deterioration. An overall proposed path for
the diffusion of the gaseous species into the sidelining is shown in
Figure 13.

Figure 11. Na(g) concentration profile in the carbon cathode,

ramming paste and sidelining block modeled by a commercial
software based on finite element method (D=310-11 m2/s).
The simulated Na concentration profiles are in agreement with the
appearance of the cross section of the spent block shown in Figure
12.

Figure 13. A schematic overview of proposed route of gasous

diffusion (not in scale).
Na(g) diffuses through the carbon block, ramming paste into the
electrolyte zone of the sidelining materials via pore system and
reacts with Si3N4 binder together with oxidizing gases coming
from top of the sidelining. The degradation product is Na2SiO3 by
reaction 1.1. [O] in the formula represents the oxidizing gases O2,
CO or CO2.

Figure 12. Comparison of simulated Na concentration profile and

the appearance of the cross section of a spent block.
By using combined Knudsen gas diffusion and Chapman-Enskong
model, Na(g) diffusivity in sidelining block is calculated as 710-5
m2/s. However, if with this value the diffusion distance after 604
days will be over hundred meters which is far beyond the distance
from carbon block to the steel shell and opposite to our autopsy
observation, which shows Na is restricted inside the block in the
certain time range. To simulate the real diffusion distance, the
apparent diffusivity of Na(g) should be around 310-11 m2/s, as
shown in Figure 11 and 12, which is several orders of magnitude
lower than the value estimated by the gas diffusion model. The
difference could be due to the consequence of the degradation
reactions. On one hand, the accumulated solid reaction products
will decrease the open porosity and increase the tortuosity to
decrease the effective diffusivity of Na(g). On the other hand, the
excess oxidizing gases from the top of the block will condense the

Na(g) + 1/6Si3N4(s) + 3/2[O] = 1/2Na2SiO3(s) + 1/3N2(g) (1.1)

Na(g) diffusion will slow down greatly by the comprehensive
influences of the reactions.
In the gas zone, the reactions are indeed more complicated.
Si2ON2 is the main product phase and could be formed by solidgas reaction as formula 1.2 or gas phase reaction as formula 1.3.
The gas phase reaction is suggested by the morphology of Si2ON2
particles which is fibre like [13]. Oxidation will move downwards
by time, especially with the presence of cracks on the top of the
block, and Si2ON2 can be found down to the bottom in the
electrolyte zone of the old spent block aged 1569 days.

357

degradation of the sidelining is influenced by the operation

condition of the cell.

Si3N4(s) + 3/2SiO(g) + 1/2CO(g) = 2Si2ON2 + 1/2SiC(s) (1.2)

2SiO(g) + N2(g) + CO(g) = Si2ON2(s) + CO2(g)

(1.3)

Acknowledgement

-Si3N4 to -Si3N4 phase transformation is another possible

reaction which is suggested to go through a solid-gas-solid
mechanism [13] and the transition is suggested to some degree be
related to the formation of Si2ON2 phase.

Financial support from Hydro Aluminium AS and the Research

Council of Norway through the project Thermodynamics
Applied to High Temperature Materials Technology are
gratefully acknowledged. Anne Stre is also acknowledged for the
assistance in the part of the LECO analysis.

In addition, pot gas NaAlF4(g) will condense and form chiolite

(Na5Al3F14) on the surface or even deeper inside the gas zone of
the sidelining block in case of less protection of side ledge. If no
H2O(steam) is present, NaAlF4(g) will condense to form chiolite
by direct decomposition reaction 1.4.
NaAlF4(g) = 1/5Na5Al3F14(s) + 2/5AlF3(s)

References
[1]

J. Schoennahl, E. Jorge and P. Temme, Optimization of

Si3N4 bonded SiC refractories for aluminium reduction
cells, Light Metals, 251-256, 2001.

[2]

R.P. Pawlek, SiC in Electrolysis pots: an update, Light

metals, 655-658, 2006.

[3]

Saint-Gobain Ceramics,
http://www.refractories.saintgobain.com/data
/element/node/application.

[4]

Hydro Aluminium, New sidewall lining in modern high

current hall-herault electrolysis cell, Symposium, 2003.

[5]

E. Skybakmoen, J.H. Kvello and O. Darell, Chemical

resistance of sidelining materials based on SiC and carbon
in cryolite melts a laboratory study, Light metals, 215222, 1999.

[6]

E. Skybakmoen, J.H. Kvello and O. Darell, Quality

evaluation of nitride bonded silicon carbide sidelining
materials, Light metals, 773-778, 2005.

(1.4)

If H2O (steam) is present in pot gas, chiolite will be formed from

NaAlF4(g) through reaction 1.5. That is the case that was observed
in the spent block at the age of 1569 days. Hydrogen in the
reaction can be formed by the reaction with pot gas CO and water
steam.
NaAlF4(g) + 1/10CO2(g) + 1/5H2O(g) + 1/10H2(g) = 1/5AlF3(s) +
1/5Na5Al3F14(s) + 1/10Al2O3(s) + 1/10CO(g) + 3/5HF(g)
(1.5)
Meanwhile, HF(g) has been addressed to be another degradation
source in the gas zone which will react with silica by reaction 1.6
[7] and Si3N4 binder to produce SiF4(g) by reaction 1.7.
SiO2(s) + 4HF(g) = SiF4(g) + 2H2O(g)

(1.6)

Si3N4(s) + 12HF(g) = 3SiF4(g) + H2O(g)

(1.7)

[7]

More gaseous products will be produced as illustrated in the

reactions related to HF, which will eat away the Si3N4 binder
gradually from the surface and cause the detachment of the large
SiC grains. Material loss could be observed at the top of gas zone
facing the bath.

F.B. Andersen, M. Stam, G. Dorsam and M. Spreij, Wear

of silicon nitride bonded SiC bricks in aluminium
electrolysis cell, Light Metals, 413-418, 2004.

[8]

K.R. Kvam and H.A. ye, Homogeneity and degradation

of SiC sidelinings, Ninth Internationsl Symposium on
Light Metals Production, 1997.

[9]

A.J. Moulson, Review, reaction-bonded silicon nitride:

Its formation and properties, Journal of Materials
Science, 14:1017-51, 1979.

[10]

SCP-Silicon Carbide Products.

http://www.siliconcarbideproducts.com/.

[11]

R.Etzion, J.B.Metson, N.Depree, Wear Mechanism Study

of Silicon Nitride Bonded Silicon Carbide Refractory
Materials, Light Metals, 955-959, 2008.

[12]

Simonsen AS.
http://www.simonsen.eu/PRODUCTS/sicatec-75.pdf.

[13]

Zh. Wang, E. Skybakmoen and T. Gramde, Chemical

Degradation of Si3N4-bonded SiC Sidelining Materials in
Aluminium Electrolysis Cell, J. Am. Ceram. Soc.,
submitted.

[14]

E. Skybakmoen, J.H. Kvello, O. Darell and H.

Gudbrandsen, Test and Analysis of Nitride Bonded SiC
Sidelining Materials. Typical Properties Analysed 19972007, Light metals, 943-948, 2008.

Conclusion
Autopsies of the spent blocks at different ages show generally the
same trends. The fine grained Si3N4 binder phase is the
chemically degraded part of the sidelining materials. Degradation
in the upper part (gas zone) prefers -Si3N4 resulting in the
formation of Si2ON2 by oxidizing gases (air or CO/CO2). The /Si3N4 ratio is to some degree related to the amount of Si2ON2
phase and correlates well with the degree of the oxidation by pot
gas. Chiolite can be observed in gas zone due to the condensation
reaction of NaAlF4 in case of less protection of side ledge.
Density, porosity and thermal conductivity have no significant
change in this region. In the infiltrated lower part of the sidelining
oxidation was influenced by the presence of Na(g) and the product
was Na2SiO3. The reaction products densify the block
considerably. Thermal conductivity in this part decreases due to
the formation of Na2SiO3 phase with significantly lower thermal
conductivity. Na(g) diffusivity is proposed to be governed by the
oxidation reactions. Na(g) will be retarded in the infiltration area
instead of penetrating throughout the block due to the reactions
with oxidizing species from the top of the block. Finally, the

358