The properties and uses of fluxes in molten aluminum processing

Prévia do material em texto

38 JOM • November 1998
Overview
Aluminum
Gaseous and solid fluxes play an impor-
tant role in the degassing, demagging, and
fluxing of aluminum and its alloys. Inert as
well as reactive gases, or hexachloroethane,
may be used to remove dissolved hydrogen
and sodium. Magnesium may be removed by
chlorine or an aluminum-fluoride-contain-
ing flux. Fluxes based on a KCl-NaCl mix-
ture may be used to cover and protect the
metal from oxidation. To recover aluminum
from drosses, a more reactive flux containing
cryolite or some other fluoride may be used.
In this article, the thermodynamics of alumi-
num melting and refining are analyzed in
terms of the behavior of sodium, magnesium,
and calcium. The coalescence of aluminum
drops in salt fluxes improves with fluoride
additions. With increasing MgCl2 contents
in the flux, the effects of NaF and KF addi-
tions become much less pronounced.
INTRODUCTION
For the treatment of aluminum and its
alloys, various molten-metal processing
steps are necessary in addition to melt-
ing and alloying. Historic practices, such
as fluxing, metal refining, deoxidation,
The Properties and Uses of Fluxes in
Molten Aluminum Processing
T.A. Utigard, K. Friesen, R.R. Roy, J. Lim, A. Silny, and C. Dupuis
degassing, and grain refining, are being
used in addition to newer in-line tech-
nologies, such as metal degassing, flux-
ing, and filtration.
The term fluxing is used to represent
all additives to, and treatments of, mol-
ten aluminum in which chemical com-
pounds are used. These compounds are
usually inorganic and may perform sev-
eral functions, such as degassing,
demagging, cleaning, and alloying. Flux-
ing also includes the treatment by inert
or reactive gases to remove inclusions or
gaseous impurities.
When melting aluminum scrap in open
charging well furnaces, a salt-potash-
fluoride flux is often used on the open
surface of the bath to eliminate the for-
mation of oxides and to cause the ag-
glomeration of small beads of alumi-
num, improving metal recovery. In other
cases, fluxes are used to remove oxide
build-up from furnace walls or to elimi-
nate and/or reduce oxidation. While
fluxes require energy, they are effective
in lowering the aluminum content of the
dross/mush, and the amount of alumi-
num that is skimmed from the furnace
can be substantially reduced.
FLUX COMPOSITIONS
Solid fluxes are mainly blends of chlo-
ride and fluoride salts with additives to
instill special properties. Most fluxes are
based on a mixture of KCl and NaCl,
which forms a low-temperature (665°C)
eutectic. Another common ingredient in
fluxes is NaF, which forms a ternary
eutectic with KCl and NaCl with a melt-
ing point of 607°C. A common cover flux
contains about 47.5% NaCl, 47.5% KCl,
and 5% fluoride salt. A low melting point
is important since it will improve the
fluidity of the flux.
Other cover fluxes are based on MgCl2-
KCl, which forms a low melting eutectic
at 424°C, or carnalite (MgCl2⋅KCl), which
melts at 485°C. These cover fluxes have
high fluidity and can form a thin layer on
the melt surface. However, MgCl2 is fairly
expensive, so it is primarily used in so-
dium-free fluxes for alloys containing
more than 2 wt.% magnesium. They may
also be used where it is important to
remove calcium in alloys of fairly high
magnesium content. Many ingredients
are available (Table I); these additives
affect properties such as fluidity, wet-
tability, and reactivity (Table II).
Alkali-fluoride salts act as surfactants,1
decreasing the surface tension between
the flux and the metal (Figure 1) and
between the flux and the oxides. Chlo-
ride salts, as well as AlF3 and MgF2,
exhibit this property to a much lesser
extent,1–3 because with NaF- and KF-
containing fluoride salts, the aluminum
easily picks up some sodium or potas-
sium, which are both surface-active ele-
ments.4–6 Although rarely analyzed, it is
expected that potassium has negative
effects similar to those of sodium on the
properties of the final aluminum prod-
uct. Fortunately, due to their similar
properties, the techniques used to re-
move sodium should also remove po-
tassium.
Alkali-fluoride salts have a slight (al-
though very small) solubility of oxides,
which facilitates penetration into oxide
films that contain metallic aluminum in
dross and build-up. This leads to im-
proved wettability, favoring separation
of oxide inclusions from the melt and
Table I. Characteristics of Materials Used in Fluxes
Molecular Mass Solid Density Melting Boiling
Chemical (g/mol) (g/cm3) Point (°C) Point (°C)
LiCl 43.39 2.068 605 1,325
NaCl 58.44 2.165 801 1,413
KCl 74.56 1.984 770 1,500
CaCl2 110.99 2.15 782 1,600
MgCl2 95.22 2.32 714 1,412
AlCl3 133.34 2.44 190 177.8
BaCl2 208.25 3.92 963 1,560
LiF 25.94 2.635 845 1,676
NaF 41.99 2.558 993 1,695
KF 58.1 2.48 858 1,505
CaF2 78.08 3.18 1,423 2,500
MgF2 62.31 3.18 1,261 2,239
AlF3 83.98 2.882 — 1,291*
Na3AlF6 209.94 2.9 1,010 —
LiNO3 68.94 2.38 264 600
†
NaNO3 84.99 2.261 307 380
†
KNO3 101.11 2.109 339 400
†
Li2SO4 109.94 2.221 859 high
Na2SO4 142.04 — 897 —
K2SO4 174.27 2.66 1,069 1,689
CaSO4 136.14 2.61 1,450 high
MgSO4 120.37 2.66 — 1,124
†
Li2CO3 73.89 2.11 723 1,310
Na2CO3 105.99 2.532 851 high
K2CO3 138.21 2.42 894 high
MgCO3 84.32 2.96 — 350
†
CaCO3 100.09 2.71 1339 850
* Sublimes
† Decomposes
391998 November • JOM
Table II. Properties of Selected Compounds Used in Fluxes
Chemical Element
Formula Fluidity Wettability Active Exothermic Gas Release Added
AlF3 ↑ — Yes — — —
CaCl2 ↑ — — — — —
MgCl2 ↑ — — — — —
MnCl2 ↑ — Yes — — Mn
KF ↑ — Yes — — K
NaF ↑ — Yes — — Na
NaCl ↑ — — — — —
KCl ↑ — — — — —
NaAlF3 — — Yes — — —
CaF2 ↓ ↑ — — — —
Na3AlF6 ↓ ↑ Yes — — —
Na2SiF6 ↓ ↑ Yes Yes — —
KNO3 ↑ ↑ Yes Yes N2, NOx —
C2Cl6 — — Yes — Cl2AlCl3 —
K2CO3 — — Yes Yes CO2 —
Na2CO3 — — Yes Yes CO2 —
K2TiF6 — — Yes — — Ti
KBF4 — — Yes — — B
Figure 1. Aluminum-salt interfacial tension
at 723°C.
Figure 2. Standard Gibbs energy of formation
of several sulfides, oxides, chlorides, and
fluorides. The data are given at 723°C per
mole of S, O, Cl2, and F2, respectively.
13metallic aluminum from the dross. Un-
fortunately, the high melting points of
alkali-fluoride salts thicken the liquid
flux, limiting their use. Also, the dis-
posal of used fluoride-containing salts
are subject to stricter environmental
regulations than pure chloride salts.
Fluxes may contain fluoride salts, such
as cryolite (Na3AlF6), calcium fluoride
(CaF2), and sodium silicofluoride
(Na2SiF6), in amounts up to 20 percent.
The addition of oxygen-containing
compounds, such as KNO3, releases heat.
The released oxygen from the decompo-
sition of the nitrates reacts with metallic
aluminum, yielding Al2O3 and consider-
able heat. This locally increases the flu-
idity, enhancing the recovery of metallics
suspended in the oxide. In cleaning
fluxes, the reaction increases penetra-
tion of the flux into build-ups.
Certain compounds decompose into
chlorine, CO2, or metal halide gases
(AlCl3). If they are introduced beneath
the melt surface, they create bubbles that
remove hydrogen. The most notable
gas-releasing compound is hexachloro-
ethane (C2Cl6), which generates Cl2 and
gaseous AlCl3.
Compounds that react with aluminum
or its impurities can be used to add
certain elements to the melt or reduce
the concentration of others. NaF will
add traces of sodium to the melt, K2TiF6
can add titanium, and KBF4 adds boron.
To some extent, AlF3 removes Ca, Sr, Na,
and Mg, and compounds releasing chlo-
rine remove Ca, Li, Mg, and Sr.
FLUX CHARACTERIZATION
AND USES
Most secondary aluminum alloy pro-
ducers use smelter’s flux for their cover
and depend on Cl2 or Cl2-N2 for their
degassing fluxes.2,3,7–11 Fluxing is tem-
perature dependent; it must be high
enough to provide for good contact and
reactivity and achieve good physical
separation. The choice of specific com-
pounds or chemical reagents in fluxes
depends on the specific purpose(s) of
the flux.
The various constituents serve four
uses. First, they form low-melting, high-
fluiditycompounds, as is the case with
sodium chloride (NaCl)-potassium chlo-
ride (KCl) mixtures. Second, they de-
compose to generate anions, such as ni-
trates, carbonates, and sulfates, that are
capable of reacting with impurities in
the aluminum. Third, they act as fillers
to lower the cost per kilogram, provide a
matrix or carrier for active ingredients,
or adequately cover the melt. Fourth,
they absorb or agglomerate reaction
products from the fluxing action.
The uses of salt fluxes fall into five
categories: cover, cleaning, drossing, re-
fining, and wall-cleaning. Cover fluxes
prevent oxidation of the molten bath
and cause the agglomeration of metal
droplets to form larger pieces that then
sink back into the bath. Cover fluxes
(NaCl-KCl + some fluorides and CaCl2)
are designed to be used primarily with
smaller (e.g., pot or crucible) furnaces to
provide a physical barrier to oxidation
of the metal or serve as a cleanser for
alloys or scrap foundry returns. Cover
fluxes may also be used under highly
oxidizing conditions (T > 775°C), in melt-
ing fines and chips, or in making alloys
containing more than 2 wt.% magne-
sium. Although the economic value of
cover fluxes depends on each individual
case, their use on alloys that oxidize
rapidly, particularly those containing
more than 2 wt.% magnesium, is usually
cost effective.
Cleaning fluxes facilitate keeping fur-
nace or crucible walls above and below
the melt line free of build-up. Build-up
begins as a composite of metallic alumi-
num and oxide, so that it initially can be
loosened and dispersed with exother-
mic fluxes. The build-up often originates
from wet dross sticking to the furnace
walls. Since the build-up gradually in-
creases its oxide content, eventually
forming corundum, a jackhammer is
needed if it is not removed at an early
stage.
Drossing fluxes are designed to pro-
mote separation of the aluminum-oxide
dross layer from the molten metal. The
crossing fluxes are designed to react with
metallic aluminum to generate heat
2Al + KNO3 → Al2O3 + 1/2N2 + K,
Keq = 7.1 ⋅ 1060, ∆H° = –1,230 kJ
or to react with Al2O3 in the slag and
dross layer.
6Na2SiF6 + 2Al2O3 → 4Na3AlF6 +
3SiO2 + 3SiF4(gas)
 Keq = 2.3 ⋅ 1029
These crossing fluxes usually contain
compounds capable of reacting exother-
mically, giving heat and improved wet-
tability. The fluorides wet and slightly
dissolve thin-oxide films, which, with
mechanical agitation, may be broken to
release entrapped aluminum. However,
no fluoride salts can dissolve massive
Al2O3 particles.
Drossing fluxes are used to great ad-
vantage to reduce the rich metallic con-
tent of drosses that may contain up to
60–80% free metal. Considerable cost
savings result because proper fluxing
will deliver perhaps 50% metal directly
back into the melt. Drossing fluxes are
added either by weight, about 0.2–1% of
metal charged, or by a melt surface area
of 2.5 kg/m2, corresponding to a thick-
ness of about 1 mm. Too little exother-
mic combustion reduces fluxing effi-
ciency, while too much flux burns exces-
sively, creating excessive fume and loss
Ca Ba Li Sr Mg Na K
Metal
Al Si Mn Zn Fe Cu
–1,200
–1,000
–800
–600
–400
–200
0
G
ib
bs
 E
ne
rg
y 
of
 F
or
m
at
io
n 
(k
J/
m
ol
)
Fluoride
Chloride
Oxide
Sulfide
■
●
▲
■
●▼
▲
■
●
▼
▲
■
●
▲
▼
■
●
▲
▼
■
●
▲
▼
■
●
▼
■
■
●
▼
▲
■
●
▼
▲
■
●
▼▲
■
●
▼
▲
●
▲
▼
■
●▼
▲
▲
▼
✚ ✚
✚ ✚
0 1 2 3 4 5
Additive to Equimolar KCI-NaCl (wt.%)
6 7 8 9 10
400
500
600
700
800
900
1,000
In
te
rfa
ci
al
 T
en
si
on
 (m
N
/m
)
NaF
MgF2
KF
AlF3, MgCl2, & LiCl
■
■ ■ ■
■
■
■
■■
●
●
●
●
▲
▲
▲ ▼▼▼▼▼
◆◆◆◆
40 JOM • November 1998
Figure 3. Exchange equilibrium between
aluminum and different metal chlorides and
metal fluorides at 723°C based on the
reactions Al + 3 MeX = AlX3 + 3Me and Al +
1.5MeX2 = AlX3 + 1.5 Me.
Figure 4. The exchange equilibrium between
magnesium impurities in aluminum and differ-
ent metal oxides, chlorides, and fluorides at
723°C using reactions Mg + 2MeX = MgX2 +
2Me and Mg + MeX2 = MgX2 + Me.
Figure 5. The activity coefficient of MgCl2 in
NaCl-KCl melts at 723°C and 823°C.
of metallic aluminum.
Refining fluxes contain compounds
that break down and are thermodynami-
cally favorable to react with certain me-
tallic elements in the aluminum. For ex-
ample, certain chlorine-containing com-
pounds will react with molten alumi-
num containing Mg, Ca, Li, Na, and K to
form compounds that will partition to
the dross phase, where they can be re-
moved by skimming.
Wall-cleaning fluxes (e.g., Na2SiF6)
contain compounds that help remove
the oxide build-up that occurs on fur-
nace walls. These fluxes can often be
applied with a typical refractory gun-
ning device.
Most fluxing compounds are hydro-
scopic and must be stored in a dry place.
They should never be used wet because
of the danger of explosions. In addition,
this will introduce hydrogen into the
aluminum. Cover fluxes can be spread
over the melt, while crossing fluxes usu-
ally need to be mixed into the dross
layer. After a flux is used, a quiescent
time for the bath is recommended to
allow adequate settling of heavy inclu-
sions or floating out of lighter-density
fluxing salts and flux-wetted inclusions.
Optimal settling time may vary from 5–
10 minutes for a small crucible melt to 1–
2 h for a 50 tonne furnace.
An important factor is the flux’s melt-
ing and reaction temperature range. A
cover flux should be liquid at melt tem-
peratures, and drossing/exothermic
fluxes should ignite. An exothermic wall
cleaning flux is typically applied when
the walls are as hot as possible to aid
heating and softening of oxide build-
ups.12 A weekly practice consists of drain-
ing the furnace to a low level, coating the
walls with enough flux (3–6 mm) to ini-
tiate a good reaction, turning the burn-
ers on high for 10–15 minutes with the
doors closed, scraping off build-up, and
skimming debris from the melt surface.
Another method to clean walls is to
add a cleaning flux to the melt near the
walls after skimming the melt, but be-
fore tapping the furnace. During tap-
ping, the flux coats the wall as the melt
level goes down. Any build-up on the
walls reacts with the flux while the fur-
nace is recharged and is then scraped off
during the next skimming operation.
This method can be used as preventive
maintenance in melting furnaces, coun-
teracting the sticking of wet dross to the
walls of the furnace.
THE THERMODYNAMICS OF
ALUMINUM-ALLOY REFINING
BY MOLTEN SALTS
The principal metallic impurities in
molten aluminum are alkali metals
(lithium, sodium, and calcium) in very
small concentrations (<20 ppm) and
magnesium in large concentrations
(0.2–10%). Lithium, sodium, and calcium
are impurities often coming from pri-
mary aluminum production. The Gibbs
energy of formation of several sulfides,
oxides, chlorides, and fluorides is given
in Figure 2.13 As the stability of the com-
pound increases with an increasing nega-
tive value of the Gibbs energy of forma-
tion, the thermodynamic stability de-
creases, with a few exceptions, from the
fluorides down to the sulfides in the
order fluorides > chlorides > oxides >
sulfides. Among the key exceptions are
the industrially important substances
Al2O3 and MgCl2. Due to the extreme
stability of Al2O3, only a fluoride-based
inert electrolyte can be used in the Hall-
Héroult process. Because of the low sta-
bility of MgCl2 relative to the alkali chlo-
rides, an NaCl-KCl-CaCl2-based inert
electrolyte can be used for the electro-
lytic production of magnesium from
MgCl2.
The metal chlorides that have a stan-
dard Gibbs energy value more negative
than AlCl3 are more stable than AlCl3.
This means that when Cl2 is injected into
aluminum containing various metallic
elements, the chlorine will preferentially
react with these metallic impurities. The
same also applies to fluorides. Li, Na, K,
Ca, Mg, and Ba all form more stable
chlorides and fluorides than aluminum
and can, therefore, be removed by Cl2,
F2, orSF6 injection. The reaction, in the
case of magnesium, is
Mg (in Al) + Cl2 = MgCl2
∆G° = –481 kJ/mol
MgCl2 is a liquid above 712°C; it is less
dense than aluminum and tends to float
to the surface. The equilibrium constant
for reactions such as Al + 3MeX = 3Me +
AlX3, X = Cl or F, and Me = Li, Na, K and
Al + 1.5MeX2 = 1.5Me + AlX3, X = Cl or F
and Me = Ca, Mg, Ba, Sr is shown in
Figure 3 for several different metals. An
equilibrium constant much greater than
one implies that the reaction is shifted to
the right, while a value much less than
one indicates that at equilibrium the re-
action is shifted to the left. Therefore, an
alkali or alkali-earth chloride electrolyte
has no tendency to react with alumi-
num. Corresponding metal-fluoride
electrolytes are slightly more reactive. A
chloride electrolyte is, therefore, suit-
able for the refining of aluminum since it
will promote the removal of alkali/al-
kali-earth metal impurities while main-
taining high aluminum recovery during
Cl2 injection.
The removal of other impurities, such
as Zn, Si, Fe, and Cu, by chlorine or
fluorine treatment is basically impos-
sible. Similarly, if the flux used contains
compounds of such heavy metals, they
will react with the melt and contaminate
the aluminum. To remove sodium from
primary aluminum, the TAC process,
which employs the injection of AlF3 pow-
der into the metal, may be used.
AlF3 + 3Na (in metal) = Al + 3NaF
 Keq = 2.6 ⋅ 108
Since this is the inverse reaction of
those shown in Figure 3, this reaction is
highly favorable, and sodium is removed
from the aluminum.
MAGNESIUM BEHAVIOR
DURING ALUMINUM-ALLOYS
TREATMENT
Many useful aluminum alloys con-
tain magnesium in quantities of 0.1–
10%. Wrought alloys, in particular, con-
tain high levels of magnesium, and cast-
ing alloys 518 and 520 contain 8% and
10% magnesium, respectively. These
materials constitute a significant por-
tion of the scrap market and are avail-
able for recycling/remelting. Mill recy-
K Na Ba Li
Chlorides
Sr Ca
Metal
Mg Cu Mn Zn Fe Si
–40
–30
–20
–10
0
10
20
30
40
Lo
g 
(E
qu
ilib
riu
m
 C
on
st
an
t)
Fluorides●
■
■
●
●
■
●
■
●
■
●
■
●
■
●
■
●
■
●
■
●
■ ●■
BaCa Li Sr Na K
Metal Compound
Al Si Mn Zn Fe Cu
–15
–10
–5
0
5
10
15
20
25
30
Oxides
Fluorides
ChloridesL
og
 (E
qu
ilib
riu
m
 C
on
st
an
t)
■
●
▲
▲
●
■
▲
●
■
●
▲
■
●
▲
■
▲
●
■
●
■
●■
●
▲
■
▲
●
■
●
▲
■
●
▲
■
▲
▲
0.2 0.4
Mole Fraction of MgCl2
0 0.6 0.8 1
0.001
0.01
0.1
1
Ac
tiv
ity
 C
oe
ffi
ci
en
t o
f M
gC
l 2
NaCl/KCl (1/1 at 800°C)
NaCl (723–823°C)
KCl (800°C)
■■ ■
■
■
■
■
■
■
■
●■■
■
■
●
●
●
●
●
●
●
●
▲▲▲
▲
▲
▲
▲
▲
▲
▲
▼
▼▼
▼
▼
▼▼
▼
▼
▼
◆
◆
◆
◆
▲
▲
▲
411998 November • JOM
Figure 8. The percent of Al2O3 inclusions in
AA6061 during treatment with NaCl-KCl with
additions of KF.
Figure 6. The equilibrium aluminum-sodium
content versus the content of magnesium and
the MgCl2 content in the NaCl-KCl flux.
0.1
MgCl2 (wt.%)
0.01 1 10
0.01
0.1
1
10
100
So
di
um
 in
 A
l (
pp
m
)
0.1% Mg
1% Mg
5% Mg
Figure 7. The removal of sodium and calcium
from Al-4.5Mg alloys by Cl2-N2 injection for
30 minutes with a salt flux cover of KCl-MgCl2
at 740°C. Shown are parts per million of
sodium and calcium in aluminum versus MgCl2
content in the flux.
3530 40 45 50
MgCl2 in KCl-MgCl2 Mixtures (wt.%)
55
Below Detection Limit
for Na and Ca 
at 10 ppm
60 65 70
So
di
um
 a
nd
 C
al
ci
um
 in
 M
et
al
 (p
pm
)
0
20
40
60
80
100
120
140
160
180
Ca
Na
■
■
■ ● ●
●
●
clers often need to produce new alloy
products with lower magnesium con-
tent, and secondary smelters producing
die cast and foundry ingots also need to
produce low-magnesium-content alloys
(generally less than 0.2%). Therefore,
there is a need to demag aluminum scrap
during these remelting operations.
There are three general types of de-
magging processes: chlorination, the use
of solid chlorine-containing fluxes, and
the injection of AlF3 or NaAlF4.
14–21 For a
perfect (100% efficient) reaction, 2.95 kg
of chlorine is required to remove 1.0 kg
of magnesium. Therefore, it should take
about 30 kg of Cl2 gas to remove 1 wt.%
magnesium from one tonne of alumi-
num. One problem with using chlorine
gas to remove magnesium is that the
demagging efficiency drops as the mag-
nesium content in the metal falls. This
may lead to emissions of AlCl3 and HCl
due to reactions with moisture in the air.
In such cases, the use of a thin salt flux
cover may trap the AlCl3 gas before it is
emitted into the atmosphere.
AlCl3 (g) → AlCl3
(dissolved in salt flux cover)
When Cl2 gas is injected into pure alumi-
num covered by an NaCl-KCl flux at
740°C, the capture efficiency of the salt
increases as its thickness increases up to
about 2.5 cm.22
Magnesium impurities can not be re-
moved from aluminum using an alkali-
chloride mixture alone (Figure 4). This is
illustrated by the very low values of the
equilibrium constant for exchange reac-
tions such as
2NaCl + Mg = 2Na + MgCl2
 Keq = 1.1 ⋅ 10–8
From which the sodium activity in the
metal can be determined.
 
a
a a
aNa
NaCl Mg
MgCl
= ⋅ ⋅
⋅




1 1 10 8
2
1
2
2
. –
To determine the actual sodium con-
tent in Al-Mg alloys in equilibrium with
a salt flux, the activity of the various
species involved in the reaction must be
known. The molten NaCl-KCl system is
nearly ideal, while MgCl2 shows a strong
negative deviation from the ideal. Fig-
ure 5 shows the activity coefficient of
MgCl2 in NaCl, NaCl-KCl, and KCl melts
at temperatures between 723°C and
823°C.23–25 The strong negative deviation
is caused by the formation of MgCl4
2–
complexes, which are stabilized by large
cations with low charge, such as K+. For
use in the present analysis, the activity
coefficient (γMgCl2) of MgCl2 in the equi-
molar NaCl-KCl melt is set to be 0.009.
The activity coefficient of sodium in alu-
minum is about 426, while that of mag-
nesium is 0.15. Using these activity coef-
ficients, the sodium and magnesium con-
tents in molten aluminum in equilib-
rium with an equimolar NaCl-KCl melt
(aNaCl = 0.5), are related by
 
ppm Na
wt Mg
wt MgCl
≅




0 5
2
1
2
.
.%
.%
The sodium content calculated by this
equation is shown in Figure 6 versus the
MgCl2 concentration and three contents
of magnesium in aluminum.
As seen by the large value of the equi-
librium constant (Figure 4), AlCl3 and
SiCl4 promote the removal of magne-
sium from the metal. Similarly, by add-
ing NaF and/or KF to the chloride flux,
the removal of magnesium from alumi-
num scrap is enhanced. However, this
leads to contamination of the aluminum
with sodium and/or potassium, as given
by the exchange reaction
2NaF + Mg = 2Na + MgF2
Keq = 2.6
Since MgF2 is much more stable than
MgCl2, it is expected that fluoride salt
additions will stabilize magnesium in
the flux. By adding only 1 wt.% NaF to
equimolar NaCl-KCl, the activity coeffi-
cient of MgCl2 decreases approximately
by a factor of ten.26 This means that
1 wt.% NaF additions to the salt flux
increase the sodium content by a factor
of 3.2 (≅ 101/2) above that with pure NaCl-
KCl. At higher fluoride contents, the
sodium content will increase further
since the activity coefficient of MgCl2
decreases even more. This shows that as
long as there are fluorides present in the
flux and magnesium in the metal, the
removal of sodium becomes more diffi-
cult, and higher concentrations of MgCl2
in the flux are required.
In the case of magnesium alloys in
contact with calcium compounds, the
aluminum may pick up some calcium
due to reactions such as
Mg (in Al) + CaCl2 = MgCl2 + Ca (in Al)
Keq = 6.2 ⋅ 10–9
Although the equilibrium constant is
small, since the activity coefficient of
calcium in aluminum is very low
(≅ 0.005), calcium can easily be picked
up.
 
ppm Ca
wt Mg wt CaCl
wt MgCl
≈
⋅



0 4 2
2
.
.% .%
.%
For these situations, the flux should
contain a significant amount of MgCl2 to
prevent the reactionfrom going to the
right. Such a flux may be based on
carnalite (KCl⋅MgCl2⋅6H2O), kalnite
(KCl⋅MgSO4⋅3H2O), or sylvite (KCl).
Figure 7 shows how the sodium and
calcium contents in aluminum vary with
the MgCl2 content of the cover flux used.
The results are based on experimental
tests with an Al-4.5Mg alloy doped with
sodium and calcium before the metal
was treated with an N2-10Cl2 gas mix-
ture for 30 minutes. These results are
consistent with the MgCl2 activity data
given in Figure 5. It is only after the
MgCl2 content increases to 50 wt.% that
it is possible to selectively remove the
sodium and calcium while keeping the
magnesium in the alloy.
When calcium carbonate is used (as
flux or as caulking material), two reac-
tions may cause calcium pick-up
Mg (in Al) + CaCO3 =
MgO + CO2(g) + Ca (in Al)
Keq = 7.6 ⋅ 104
2Al + 3CaCO3 =
Al2O3 + 3CO2(g) + 3Ca (in Al)
Keq= 3.7 ⋅ 10–16
1050 15
Duration (min.)
20 25 30
0.01
0.1
1
10
In
cl
us
io
ns
 (v
ol
.%
)
10 wt.% KF 5 wt.% NaF
1 wt.% KF
3 wt.% KF
5 wt.% KF
■
■
■
■
●
●
●
●
▲
▲
▲
▲
▲
▼
▼
▼
▼
◆
◆
◆
◆
●▲
◆
42 JOM • November 1998
Calcium may also be picked up by a
reaction such as
CaF2 + Mg = Ca + MgF2
Keq = 3.1 ⋅ 10–6
When using a cryolite flux for mag-
nesium alloys, the following reaction
takes place:
2Na3AlF6 + 3Mg =
6Na + 2AlF3 + 3MgF2
Keq = 2 ⋅ 10–9
leading to the pick up of sodium in the
metal.
DEGASSING ALUMINUM
ALLOYS
The simplest method to remove dis-
solved hydrogen is to hold the metal for
some time, allowing for some degas-
sing. Accelerated degassing can be
achieved by gas purging, the applica-
tion of a vacuum, tableted flux degas-
sing, or mechanical stirring.27–29 Hydro-
gen can fairly easily be removed by in-
jecting a purging gas under pressure
through a tube, pipe, lance, or porous
plug. By combining gas injection with
the use of high-speed rotors (300–
500 rpm), small gas bubbles are created.
In addition, the rotors induce metal flow,
greatly improving the kinetics of degas-
sing. Reactive gases are usually mixed
with inert gases, with concentrations of
the reactive gases up to about 20%. The
chlorine gas reacts with aluminum to
form gaseous AlCl3, and freon forms
solid AlF3. However, both chlorine and
fluorine are noxious, leading to possible
environmental problems.
A common method of degassing is to
use C2Cl6 tablets, which decompose to
form AlCl3 gas bubbles that then collect
hydrogen. The tablets may also contain
salt fluxes to help wet oxide inclusions,
thus, enabling the removal of hydrogen
associated with inclusions. To be fully
effective, the tablets should be plunged
deep into the metal and kept there until
the bubbling subsides. Although C2Cl6
provides effective degassing, the nox-
ious odor of the raw tablets creates envi-
ronmental difficulties that have forced
many foundries to discontinue their use.
After the hydrogen has been removed,
care has to be taken to prevent its content
from increasing again due to reactions
with moisture in the air.
2Al + 3H2O(moisture) →
Al2O3 + 6H (in Al)
Although the barrier of aluminum
oxide on the metal surface resists hydro-
gen pick-up, disturbances of the surface
that break the oxide barrier result in
rapid hydrogen dissolution. Further, al-
loying elements, such as magnesium,
may increase hydrogen absorption by
forming oxide products that offer re-
duced resistance to the diffusion of hy-
drogen into the melt. At temperatures
above 745°C, a complex aluminum-mag-
nesium oxide (spinel) is formed with
rapid growth potential.
INCLUSION REMOVAL
Aluminum-magnesium alloys con-
taining 12 µm sized alumina inclusions
were treated at 740°C using NaCl-KCl
salts with various amounts of NaF and
KF additions. The purpose of these tests
was to determine if salt flux could re-
duce the usage of chlorine gas during
inclusion removal. The tests were car-
ried out by placing a piece of the alloy in
a stagnant flux layer and leaving it there
for various periods of times before being
removed and analyzed microscopically.
It was found that with no fluoride salt
additions or inclusion removal took
place. As the amount of NaF and KF
added increased, the rate of inclusion
removal increased (Figure 8).30 In the
case of MgCl2-KCl based fluxes, MgCl2
seemed to inhibit, as well as delay, the
effect of the NaF and KF additions. For
50% or more MgCl2 in the base flux, no
inclusion removal was observed, even
with up to 10% KF additions. This can
be explained by the exchange reaction
MgCl2 + 2KF → MgF2 + 2KCl
Keq = 1.4 ⋅ 1011
effectively canceling the effect of KF.
Beland et al.31 have shown on a plant
scale that salt-flux injection has, indeed,
the potential to completely replace chlo-
rine for the purpose of removing inclu-
sions. At several Alcan installations, the
rotary flux injection technique is being
Table III. Qualitative Assessment of Flux Additions to NaCl-KCl in the Coalescence of Two Small Aluminum Droplets at 740 °C
Salt (5 wt.%) Coalescence Color of Flux Gas Formed Time (s) Comments
None None Clear None >900 Nothing happened
AlCl3 None Clear Small >900 No reaction after second addition and agitation
MgCl2 None Slightly cloudy Small >900 No reaction after second addition and agitation
BaCl2 None Clear Small >900 No reaction after second addition and agitation
CaCl2 Poor Clear Small >600 Second addition and agitation required
LiCl Poor Hazy, white precipitates None >600 Agitation required
MgF2 Fair Hazy Small 11 No drop spinning and agitation required
CaF2 Fair Clear, grey precipitates Small 14 No drop spinning and agitation required
AlF3 Good Grey/cloudy white precipitates Small 7 Droplets spun for about one second
LiF Good Blue fog around droplets Moderate 4 Droplets spun for less than one second
Na3AlF6 Excellent Hazy, blue fog, white precipitates Moderate <1 Droplets spun violently for 50 seconds
NaF Excellent Clear, blue fog around droplet Heavy <1 Droplets spun violently for 17 seconds
KF Excellent Clear, blue fog around droplet Moderate <1 Droplets spun violently for 5 seconds
used on a permanent basis. The flux
used is fabricated by fusing MgCl2 with
other compounds. This reduces the melt-
ing point and decreases the hydroscopic
nature of the flux.
In a different series of experiments, an
aluminum alloy (A356) with 1.3 vol.%
SiC inclusions was treated by the injec-
tion of Cl2-N2 gas mixtures.
32 It was found
that the inclusions could not be removed
by mechanical stirring or by injection of
pure nitrogen. As chlorine was added,
the inclusions started to be removed.
The rate of inclusion removal increased
with increasing Cl2 content in the gas
(20% Cl vs. 5% Cl), increasing overall gas
flow, and increasing mechanical stirring
(500 rpm vs. 200 rpm). It was proposed
that the injected Cl2 formed MgCl2 drop-
lets dispersed within the aluminum melt.
As these inclusions contacted the SiC
inclusions, they would coat the inclu-
sions with a thin film of MgCl2, changing
the wetting characteristics of the system.
In this manner, the inclusions could be
separated from the metal.
THE COALESCENCE OF
ALUMINUM DROPLETS IN
THE SALT FLUX COVER
During the bubbling of a gas into the
liquid aluminum, the use of a salt flux
decreases the tendency for oxide forma-
tion. In addition, the salt flux may also
trap some, if not all, of the gaseous AlCl3
in the bubbles as they leave the metal.
However, a disadvantage of the salt flux
is that metallic aluminum droplets may
form and remain in the salt flux, leading
to a loss in recovery. Our laboratory
work has shown that argon or N2 may
lead to significant formations of en-
trained aluminum droplets in the flux.
With the use of some Cl2 in the injected
gas22 or with the addition of fluorides to
the salt mixture33 this problem could be
managed. Table III shows how various
salt additives affected the coalescence of
two small aluminum droplets kept in an
NaCl-KCl flux.
For salt fluxes containing substantial
amounts of MgCl2, fluoride salt addi-
tions become much less potent (Table
IV). As an example, for a melt with 45%MgCl2 in KCl, a minimum addition of
431998 November • JOM
Table IV. Coalescence Times of Aluminum Droplets in MgCl 2-KCl Fluxes with NaF and
KF Additions*
 Amount of MgCl2 in KCl-MgCl2 Mixture (wt.%)
Additive (wt.%) 30 35 40 45 50 55 60 65 75
None No No No No No No No No No
3 KF 49 97 236 No No No No No No
3 NaF 38 76 184 No No No No No No
5 KF 22 41 85 No No No No No No
5 NaF 17 32 68 No No No No No No
10 KF 13 22 45 106 No No No No No
10 NaF 11 19 39 89 No No No No No
* The values listed are the average coalescence time based on three experiments. If the droplets did not coalesce after 15 minutes,
it was considered that no coalescence would take place.
10 wt.% NaF is required to promote coa-
lescence, as opposed to less than 1 wt.%
NaF in pure NaCl-KCl. The reason for
this is that MgCl2 will neutralize the
alkali fluoride salts by reactions such as
MgCl2 + 2NaF = MgF2 + 2NaCl
Keq = 2.3 ⋅ 108
CONCLUSIONS
The choice of which components to
use in a flux depends on the operating
temperature, whether the flux is to pro-
vide a molten cover, the desired reactiv-
ity, or the specific alloy chemistry. For
example, sodium-bearing, fluoride-con-
taining fluxes should not be used with
aluminum-magnesium alloys in order
to avoid sodium contamination of the
metal. When removing calcium from
high magnesium alloys, it is recom-
mended to use a flux with around
50 wt.% MgCl2. NaF, KF, and Na3AlF6
additives are useful for the coalescence
of small aluminum particles, recovery of
aluminum from a dross flux, and re-
moval of inclusions from the metal.
From a thermodynamic point of view,
metal fluorides are more stable than cor-
responding chlorides, oxides, and sul-
fides. In salts, magnesium behaves ir-
regularly. First of all, MgCl2 is not a very
stable salt as compared to alkali and
other alkali-earth chlorides. However,
MgCl2 forms MgCl4
2– complexes in chlo-
ride melts, effectively stabilizing the
magnesium chloride. On the other hand,
MgF2 is a very stable compound. There-
fore, when a fluoride salt is added to a
chloride mixture containing magnesium,
it will stabilize the magnesium in the
salt.
ACKNOWLEDGEMENTS
The financial support from the National
Science and Engineering Research Council
of Canada and Alcan International over sev-
eral years gave us the opportunity to carry
out this work. We gratefully appreciate this
support.
References
1. A. Silny and T.A. Utigard, Light Metals 1997, ed. R. Huglen
(Warrendale, PA: TMS, 1997), pp. 871–878.
2. R.D. Peterson, Recycling of Metals and Engineering Mate-
rials, eds. J.H.L. Van Linden, D.L. Stewart, and Y. Sahai
(Warrendale, PA: TMS, 1990), pp. 69–84.
3. F.K. Ho and Y. Sahai, in Ref. 2, pp. 85–103.
4. T.A. Utigard and J.M. Toguri, Metall. Trans. B, 16B (1985),
pp. 333–338.
5. T.A. Utigard, J.M. Toguri, and K. Grjotheim, Canadian
Metall. Quart., 26 (1987), pp. 129–135.
6. T.A. Utigard and J.M. Toguri, Metall. Trans. B, 18B (1987),
pp. 695–702.
7. A.H. Sully, K.K. Hardy, and T.J. Heal, J. Inst. of Metals, 82
(1952), pp. 49–58.
8. E.G. West, Trans. Inst. Weld., 3 (1940), pp. 93–97.
9. T.J. Johnston and R.D. Peterson, Recycling and Secondary
Recovery of Metals, eds. P.R. Taylor, H.Y. Sohn, and N. Jarrett
(Warrendale, PA: TMS, 1985), pp. 417–428.
10. S. Lavoie, C. Dube, and G. Dube, in Ref. 2, pp. 451–462.
11. S. Lavoie, C. Dube, and G. Dube, Light Metals 1991, ed. E.L.
Rooy (Warrendale, PA: TMS, 1991), pp. 981–985.
12. A. Flores et al., in Ref. 1, pp. 879–884.
13. T.A. Utigard, Proceedings of the International Symposium on
Extraction, Refining, and Fabrication of Light Metals (Ottawa,
Canada: CIM, 1991), pp. 353–365.
14. D.V. Neff, First International Molten Aluminum Processing
of AFS (Des Plaines, IL: AFS, 1986), pp. 341–343.
15. D.V. Neff and B.P. Cochran, Light Metals 1993, ed. S.K. Das
(Warrendale, PA: TMS, 1993), pp. 1053–1060.
16. B.L. Tiwari, JOM, 34 (7) (1982), pp. 54–58.
17. M.C. Mangalick, “Demagging in the Secondary Alumi-
num Industry,” J. of Met. , 27 (6) (1975), pp. 6–10.
18. B. Lagowski, Transactions AFS, 77 (1969), pp. 205–207.
19. P.N. Crepeau, M.L. Fenyes, and J.L. Jeanneret, Modern
Casting, 82 (7) (1992), pp. 28–30.
20. P.D. Hess, U.S. patent 3,620,716 (1971).
21. L.J. Derham, U.S. patent 3,650,730 (1972).
22.J. Lim, Master of Applied Science Thesis, University of
Toronto, Toronto, Canada (1997).
23. B.R. Davies and W.T. Thompson, Proceedings of the Inter-
national Symposium on Extraction, Refining, and Fabrication of
Light Metals (Ottawa, Canada: CIM, 1991), pp. 191–201.
24. I. Karakaya and W.T. Thompson, J. Electrochem Soc., 133
(1986), pp. 702–706.
25. D.E. Neil, H.M. Clark, and R.H. Wisdal, J. Chem. Eng.
Data, 10 (1965), pp. 21–24.
26. M. Makyta and T.A. Utigard, Scand. J of Metal., 23 (1994),
pp. 164–170.
27. D. Saha, Light Metal Age, 47 (12) (1989), pp. 17–20.
28. A.R. Anderson, “Practical Observation of Rotary Impel-
ler Degassing,” Trans. AFS (1987), pp. 533–536.
29. P.R. Whiteley, “The Role of Injection Metallurgy in Alu-
minum Industry,” The Howard Worner International Sympo-
sium on Injection in Pyrometallurgy, eds. M. Nilmani and T.
Lehner (Warrendale, PA: TMS, 1996), pp. 123–131.
30. K. Friesen, Master of Applied Science Thesis, University
of Toronto, Toronto, Canada (1996).
31. G. Beland et al., “Rotary Flux Injection: Chlorine-Free
Technique for Furnace Preparation,” Light Metals 1998, ed. B.
Welch (Warrendale, PA: TMS, 1998), pp. 843–847.
32. R. Roy, T.A. Utigard, and C. Dupuis, “Inclusion Removal
during Chlorine Fluxing of Aluminum Alloys,” Light Metals
1998, ed. B. Welch (Warrendale, PA: TMS, 1998), pp. 871–875.
33. K. Friesen et al., in Ref. 1, pp. 857–864.
ABOUT THE AUTHORS
T.A. Utigard earned his Ph.D. in metallurgy at
the University of Toronto, Canada, in 1985.
He is currently an associate professor at the
University of Toronto. Dr. Utigard is a member
of TMS.
K. Friesen earned her M.A.Sc. in metallurgy
and materials science at the University of
Toronto in 1997. She is currently a plant
metallurgist at Scepter.
R.R. Roy earned his Ph.D. in materials sci-
ence and engineering at Ohio State Univer-
sity in 1994. He is currently a research asso-
ciate at the University of Toronto.
J. Lim earned his M.A.Sc. in metallurgy and
materials science at the University of Toronto
in 1997. He is currently a Ph.D. candidate at
McMaster University.
A. Silny earned his Ph.D. in chemistry at the
Slovak Academy of Sciences in 1998.
C. Dupuis earned his M.Sc. in metallurgical
engineering at Laval University in 1997. He is
currently a senior scientist at Arvida Labora-
tories, Alcan International Ltd. Mr. Dupuis is
also a member of TMS.
For more information, contact T.A. Utigard, De-
partment of Metallurgy and Materials Science,
University of Toronto, Toronto, Canada M5S
3E4; (416) 978-3012; fax (416) 978-4155.
Coming in December . . .
• Utilizing Global Energy Resources
• The Direct Fabrication of Materials
• Radiation Effects on Corrosion
• The Zinc Coating of Steels
• Historical Metallurgy