Kovove Mater.

47 2009 247–252 247

The corrosion resistance of Fe3 Al-based iron aluminides

in molten glasses

A. Hotař1 *, P. Kratochvíl1 , V. Hotař2

1
Technical University of Liberec, Faculty of Engineering, Department of Materials Science,
Studentská 2, CZ 46117 Liberec, Czech Republic
2
Technical University of Liberec, Faculty of Engineering, Department of Glass Producing Machines and Robotics,
Studentská 2, CZ 46117 Liberec, Czech Republic

Received 7 August 2008, received in revised form 19 February 2009, accepted 23 March 2009

Abstract
Corrosion resistance of iron aluminides was investigated in molten soda-lime glass and
molten lead glass. Iron aluminides with different additives and comparative steel (EN X8CrNi
25-21) were tested in laboratory conditions and compared. The corrosion resistance was de-
termined using measurement of weight change and change of surface. The tinting of molten
glasses is important to describe the interaction between the molten glass and the investigated
materials. The alloys Fe25Al5Cr and Fe28Al3Cr0.02Ce have very good corrosion resistance
against both glasses. The preferential oxidation of aluminium is typical for all tested iron alu-
minides and generated alumina slows down corrosion of these alloys. The plunger for molten
glass output control was produced from Fe25Al2Cr and the application plant test of plunger
has been running in lead molten glass. Damage of the surface is typical for doped Zr iron
aluminide.

K e y w o r d s : iron aluminide (Fe3 Al type), corrosion resistance, molten soda-lime glass, mol-
ten lead glass

1. Introduction is approx. 16–19 at.%. Aluminium content of invest-

igated iron aluminides is satisfactory to originate con-
Fe3 Al-based iron aluminides have been studied as tinuous aluminium oxide layer. The temperatures for
candidates for high-temperature structural applica- the formation of high-quality protective of α-Al2 O3
tions in glass making. The main advantages of Fe3 Al- layer are 800–900 ◦C. Alloying can influence corrosion
-type iron aluminides compared to chromium-nickel resistance of iron aluminides, too. Chromium addition
steels are low material costs, low density and corro- in amount greater as 4 at.% deteriorates corrosion res-
sion resistance at high temperature [1]. On the other istance [3]. Vice-versa, cerium has positive effect on
hand, these alloys have low ductility at room temper- corrosion resistance of iron aluminides based on Fe3 Al
ature and low high-temperature strength. However, at temperatures above 1000 ◦C [4].
these disadvantages can be partly suppressed by off- The corrosion resistance of materials based on
-stochiometric composition (e.g. 28 at.% Al) and tern- Fe40–45at.%Al in molten glass was tested in the Czech
ary additives (especially chromium) in combination Republic in the early fifties of the last century, see e.g.
with grain refinement agents such as Ce. [5]. Results of corrosion resistance were very good but
Alloys based on Fe3 Al and FeAl have excel- high content of carbon was a disadvantage for using
lent oxidation resistance at high temperatures. High- in molten glass. Therefore this material was used for
-temperature corrosion resistance of alloys based on parts of furnaces, which were not in contact with mol-
Fe-Al in oxidation media is caused by creation, main- ten glass (for example burners, holders etc.).
tenance of intact and adherent layer of Al2 O3 [2]. Min- Special requirements necessary for the use of
imum aluminium content for pure binary Fe-Al alloy metals and alloys in molten glass are namely high
*Corresponding author: tel.: +420 485 353 136; e-mail address: [email protected]
248 A. Hotař et al. / Kovove Mater. 47 2009 247–252

T a b l e 1. Chemical composition of the alloys

Chemical composition
Materials
Al C Ce Zr Cr Mn Ni Fe

wt.% 14.20 0.02 – – 5.63 max. 0.45 – Bal.

Fe25Al5Cr
at.% 25.41 0.08 – – 5.23 max. 0.40 – Bal.

wt.% 16.13 0.04 0.06 – 2.85 0.46 – Bal.

Fe28Al3Cr0.02Ce
at.% 28.39 0.16 0.02 – 2.60 0.40 – Bal.

wt.% 17.39 0.05 – 0.47 3.72 0.20 – Bal.

Fe28Al3Cr0.5Zr
at.% 28.06 0.2 – 0.53 2.85 0.26 – Bal.

wt.% – max. 0.20 – – 24.00–26.00 1.50 19.00–22.00 Bal.

EN X8CrNi25-21 (ČSN 417 255)
at.% – max. 0.92 – – 25.37–27.48 1.50 17.79–20.60 Bal.

corrosion resistance at high temperatures (to 1200 ◦C),

low tint of glass, low generation of bubbles and proper
values of mechanical strength [6] and creep properties
at high temperatures [7].
The solution of metal in molten glass has mostly
the character of oxidation [8]. The oxidation in molten
glass differs from oxidation in gaseous environment be-
cause generally protective scales are not formed on the
surface during corrosion in molten glass. The process
of corrosion can be measured as a weight loss (weight
change), while the formed oxides immediately dissolve
in the neighbouring molten glass where their concen-
tration may increase up to saturation level. This is the
process, which takes place near metallic parts inserted
into the molten glass.
The aim of the present paper is to describe corro-
sion resistance of alloys based on Fe3 Al (with differ-
ent composition) in molten soda-lime glass and molten
lead crystal. The results are compared with steel EN
X8CrNi25-21, which is now frequently used in molten
glasses.

2. Experimental procedure

The iron aluminides of different composition (Tab-

le 1) were tested for corrosion resistance and that was
compared to steel EN X8CrNi25-21. The samples were
cylinders (ø 12 mm × 18 mm long) machined from
sheets 13 mm thick, which were prepared by rolling at
1200 ◦C. The grains are elongated in the rolling direc-
tion (Fig. 1). The structure after corrosion tests (hold-
ing at 1200 ◦C) is recrystallized.
The test samples were placed in alumina cru- Fig. 1. (a) Structure of Fe28Al3Cr0.02Ce – Mastermet re-
cibles, which were prior filled with scrap glass, in agent, (b) structure of Fe28Al3Cr0.02Ce (to the right)
the furnace. All tests were carried out at 1200 ◦C. after corrosion test in molten lead glass at 1200 ◦C/72 h
The corrosion tests were performed in different mol- – Rollason’s reagent.
 A. Hotař et al. / Kovove Mater. 47 2009 247–252 249

T a b l e 2. Chemical composition of the glasses before tests; viscosities of molten glasses at 1200 ◦C

Composition (wt.%) SiO2 Al2 O3 Fe2 O3 CaO MgO K2 O Na2 O SO3 Viscosity log η (dPa)

Soda-lime glass 74.03 1.67 0.05 10.08 0.54 0.62 12.68 0.13 2.67
(transparent container glass)

Composition (wt.%) SiO2 Na2 O K2 O BaO PbO ZnO As2 O3 Er2 O3 Viscosity log η (dPa)

Lead crystal 53.0 2.5 10.8 0.7 32.0 0.7 0.2 0.018 3.28

ten glasses (Table 2) for time intervals: 24, 48, 72, 96,
168 hours. After each test the crucibles were cooled
down in furnace, subsequently carefully broken and
the glass separated from surface of alloys except for
Fe28Al3Cr0.5Zr. The corrosion resistance was determ-
ined using measurement of the weight change and
change of surface roughness. Change of chemical com-
position of glasses was measured by EDX.

3. Results and discussion

3.1. C o r r o s i o n r e s i s t a n c e o f t e s t e d
alloys

Describing of tinting by the interaction between

alloys and glasses is very important. The tinting of
both glasses after 168 hours is obvious in Fig. 2. (col-
our version on the website). The cylindrical samples
were situated (tinting the glass) in lower left corner.
Soda-lime glass is tinted more intensively than lead
crystal after interaction with alloys.
Iron aluminides tint glasses mainly to grey. Yellow-
-brown tint of soda-lime glass was also observed after
corrosion test with iron aluminides. This colour is little
intensive (glass is still transparent) and it is typical
for soda-lime glass after interaction with Fe25Al5Cr.
On the contrary, comparative steel tinted both glasses
to dark green. The dark green tint of steel was caused
especially by Cr2 O3 , which dissolves very quickly from
steels to molten glass. Only a small quantity of Cr2 O3
tints very intensively [8]. In the picture you can see
bubbles in the scrap of soda-lime glass. In preference
carbon reacted with SO3 during interaction steel with
soda-lime glass and gases (CO2 and SO2 ) generated.
The corrosion resistance was determined by meas-
urement of weight change. Weight loss is typical for
all alloys with the exception of iron aluminide alloyed
by Zr. Weight gain of iron aluminide alloyed by Zr is
due to glued-on thin layer of glass on the surface of
Fig. 2. The tinting of soda-lime glass (a) and lead crystal
specimen, therefore the corrosion cannot be quanti-
(b) after interaction with alloys, time interval 168 h; 1 – fied. Other specimens for iron aluminides have lower
EN X8CrNi25-21, 2 – Fe25Al5Cr, 3 – Fe28Al3Cr0.02Ce, 4 weight loss than comparative steel (Figs. 3 and 4). The
– Fe28Al3Cr0.5Zr. rate of solution in molten glass is very important para-
meter. The slopes of curve show that alloys Fe25Al5Cr
and Fe28Al3Cr0.02Ce have smaller rate of solution in
250 A. Hotař et al. / Kovove Mater. 47 2009 247–252

T a b l e 3. The rate of solution in molten glasses

Glass Temperature Fe25Al5Cr Fe28Al3Cr0.02Ce EN X8CrNi25-21

( ◦C) (mm/year) (mm/year) (mm/year)

Lead crystal 1200 0.5 0.8 1.0

1200 0.9 1.2 1.9

Soda-lime glass
1350 No measurement 2.6 11.2

T a b l e 4. The roughness parameters of boundaries between alloys and glasses, DC1000 average compass dimension (mul-
tiplied by 1000), STD average standard deviation and Rt average maximum roughness

Fe25Al5Cr Fe28Al3Cr0.02Ce Fe28Al3Cr0.5Zr EN X8CrNi25-21

Glass, time period
DC1000 STD Rt DC1000 STD Rt DC1000 STD Rt DC1000 STD Rt
(–) (µm) (µm) (–) (µm) (µm) (–) (µm) (µm) (–) (µm) (µm)

Before the test 1016 0.8 3.7 1018 0.8 4.7 1017 0.7 3.2 1012 0.6 3.3
Lead crystal, 168 h 1010 1.7 7.6 1017 2.6 11.0 1057 8.4 36.2 1042 2.5 12.1
Soda-lime glass, 168 h 1020 2.0 8.3 1023 2.7 13.1 1037 10.8 46.1 1036 8.8 34.0

Fig. 3. The dependence of weight loss on time in soda-lime Fig. 4. The dependence of weight loss on time in lead crys-
glass at 1200 ◦C. tal at 1200 ◦C.

both glasses than comparative steel. The difference of

solution rate is greater at 1350 ◦C in soda-lime glass,
see Table 3.
The surface roughness of samples was measured
after the exposure to molten glass. The corroded sur-
face was observed using light optical microscopy. The
boundary curve (see Fig. 5) was described by a soft- Fig. 5. Description of boundary curve.
ware tool. For a quantification of the metal rough-
ness fractal geometry and statistic tools were used.
The dividing curve generated is described by a fractal istical description [10]. These parameters are given in
dimension (a compass dimension multiplied by 1000, Table 4.
DC1000 , for details see [9]) that expresses the degree of The values of roughness parameters show that
complexity of the interface between alloy and glass by all testing materials are less damaged after interac-
means of a single number. Also average standard de- tion in lead molten glass than in molten soda-lime
viation (STD) of all the curves and average maximum glass. Low roughness parameters of iron aluminides
roughness of all the curves (Rt ) are used for the stat- (Fe25Al5Cr, Fe28Al3Cr0.02Ce) are obvious from all
 A. Hotař et al. / Kovove Mater. 47 2009 247–252 251

results. Austenitic steel (EN X8CrNi25-21) has sim- ten glass at present because alloys Fe25Al5Cr and
ilar values as Fe28Al3Cr0.02Ce after interaction in Fe28Al3Cr0.02Ce have very good corrosion resistance.
lead molten glass but surface of this steel is more dam- The plungers for molten glass output control from
aged after interaction in soda-lime molten glass. In the glass furnace were casted from iron aluminide base on
case of Fe28Al3Cr0.5Zr, the damaged surface layer is Fe3 Al. Final shapes of plungers were produced by in-
very thick. Probably this is caused by selective sur- vestment casting to save cost production. The as-cast
face corrosion. The detrimental effect of Zr on oxid- plunger from Fe25Al2Cr is cyclically heated up from
ation resistance of iron aluminides is realized also in room temperature to working temperature. The work-
[11]. ing temperature of molten glass is 900–1050 ◦C and
temperature of furnace atmosphere is 1130–1150 ◦C.
3.2. C o r r o s i o n m e c h a n i s m o f i r o n The plunger is now used without damage for 79 cycles
aluminides (i.e. 312 hours at service temperature) under these
conditions.
Diffusion of aluminium from samples and preferen-
tial oxidation of aluminium are typical for all tested 4. Conclusions
iron aluminides. The reaction of Al with SiO2 causes a
formation of Al2 O3 and Si in both glasses. Al2 O3 layer 1. Alloy Fe25Al5Cr has better corrosion resistance
is formed at the glass/alloy interface, which is partially against both molten glasses (soda-lime molten glass
protected against further solution of iron aluminides. and lead molten glass) than the steel EN X8CrNi25-
In addition alumina, which is concentrated near the -21. The alloy Fe25Al5Cr tinted less intensively both
surface of samples, increases viscosity and also slows glasses and the rate of solution and damage of surface
down corrosion of these alloys. were the lowest.
The generated Si small particles cause grey tint of 2. Alloy Fe28Al3Cr0.02Ce offers better resistance
soda-lime glass. If a reducing effect of aluminium is also to soda-lime glass than EN X8CrNi25-21. It dis-
lower, aluminium reacts with SO3 and Fe2 O3 and the solves slower and the surface is smoother. On the other
yellow-brown tint of soda-lime glass generates. Small hand, the corrosion resistance of iron aluminide with
bubbles in soda-lime glass are caused by the reaction Ce to lead molten glass is comparable to steel EN
Al with SO3 , too. X8CrNi25-21.
In the case of a lead crystal, Al reacts especially 3. The alloy Fe28Al3Cr0.5Zr has different results of
with PbO and the reaction causes a formation of corrosion resistance. Thin layer of glass is glued on the
Al2 O3 and Pb. The reaction of aluminium with SiO2 samples of iron aluminides with Zr therefore the cor-
also takes place at the same time. Small particles of rosion cannot be quantified. Large damage of samples
Pb and Si formed grey area around the sample. surface is typical for this type of iron aluminide.
Higher corrosion rates of all tested alloys in soda- 4. The preferential oxidation of aluminium is typ-
-lime glass are caused by different chemical compos- ical for all tested iron aluminides. Generated alumina
ition, especially by the content of SO3 (Table 2) be- formed a layer at the glass/alloy interface, which par-
cause corrosion resistance of majority of metals de- tially protected iron aluminides against further solu-
creases with increasing content of SO3 [8]. Viscosity tion. In addition, alumina around samples increases
of glasses also has influence on rates of alloy solution viscosity and it also slows down corrosion of these al-
because diffusion and glass flow are slowed down with loys.
increasing viscosity. But lower viscosity of lead crystal 5. Tested iron aluminides have sufficient content
than that of soda-lime glass does not increase solution of aluminium for good corrosion resistance in molten
rate of alloys in lead crystal too much (see Table 2). glass. Small differences among corrosion rates of tested
All tested iron aluminides have sufficient content iron aluminides are probably caused by alloying ele-
of aluminium for good corrosion resistance in molten ments (Cr, Zr, Ce).
glass. Differences of aluminium content in iron alumin- 6. The good laboratory results stimulated the ap-
ides are small and lower content of Al can cause only plication plant tests of iron aluminide based on Fe3 Al
shorter oxidation lifetime of iron aluminides in mol- run in lead molten glass. Iron aluminides (especially
ten glasses. Therefore alloying elements (Cr, Zr, Ce) Fe25Al5Cr) could replace heat resistant steels for ap-
probably have an effect on corrosion rate. For under- plications in soda-lime molten glass to 1200 ◦C. Feeder
standing of the effect of alloying elements on corrosion plungers, parts of gatherer and elements for mechan-
mechanism of iron aluminides in tested molten glasses, ical homogenizing of molten glass are the prospective
further studies and analysis are necessary. applications.
Acknowledgements
3.3. A p p l i c a t i o n p l a n t t e s t s
The authors would like to thank A. Smrček for help
The application plant tests also run in lead mol- with describing of corrosion mechanism.
252 A. Hotař et al. / Kovove Mater. 47 2009 247–252

This work was supported by the Research Plan of Min- [6] HOTAŘ, A.—KRATOCHVÍL, P.: Kovove Mater., 40,
istry of Education MSM 4674788501. 2002, p. 45 (in Czech).
[7] KRATOCHVÍL, P.—HANUS, P.—HAKL, J.—VLA-
SÁK, T.: Kovove Mater., 42, 2004, p. 73.
References [8] KIRSCH, R.: Metals in Glassmaking. Amsterdam, El-
sevier Science Ltd. 1993.
[1] McKAMEY, C. G.—DeVAN, J. H.—TORTORELLI, [9] MANDELBROT, B. B.: The Fractal Geometry of
P. F.—SIKKA, V. K. J.: Mater Res, 6, 1991, p. 1779. Nature. 2nd ed. New York, Freeman WH and Co.
[2] TORTORELLI, P. F.—NATESAN, K.: Mat. Sci. 1982.
Eng., A258, 1998, p. 115. [10] HOTAŘ, V.—NOVOTNÝ, F.: In.: Proceedings of 11th
[3] LEE, D. B.—KIM, G. Y.—KIM, G. J.: Mat. Sci. Eng., International Conference on Fracture. Ed.: Carpinteri,
A339, 2003, p. 109. A. Turin, CCI Centro Congressi Internazionale s.r.l.
[4] XINQUAN, Y.—YANGSHAN, S.: Mat. Sci. Eng., 2005, p. 588.
A363, 2003, p. 30. [11] STEIN, F.—PALM, M.—SAUTHOFF, G.: Interme-
[5] UXA, V.: Sklář a keramik, 6, 1956, p. 225 (in Czech). tallics, 13, 2005, p. 1275.