Professional Documents
Culture Documents
CRYSTALLIZATION
Proefschrift
door
Bedri DRINI
Doctorandus in de scheikunde
.
Dit proefschrift is goedgekeurd door de promotoren:
Prof. ir. L. Katgerman
Prof. dr. ir. P.J. Jansens
Samenstelling promotiecommissie:
Rector Magnificusvoorzitter
Prof. ir. L. Katgerman
Prof. dr. ir. P.J. Jansens
Prof. dr.-ing. O.S.L. Bruinsma
Prof. dr. ir. A.H.M. Verkooijen
Prof. dr. ir. L. Kestens
Dr. ir, D. Verdoes
Dr. ir. W. Boender
Prof. dr. ir. R. Boom
ISBN 9951-500-00-5
Publisher: DCE- Drini Consulting & Engineering
Web Site: www.dce-group.com
Voorziter
Technische Universiteit Delft, promotor
Technische Universiteit Delft, promotor
Potchefstroom University
Technische Universiteit Delft
Technische Universiteit Delft
TNO-MEP, Apeldoorn
Corus, IJmuiden
Technische Universiteit Delft (rezerv lid)
PRINDERVE T MI
[TO MY PARENTS]
Stellingen
Stellingen
Behorende bij het proefschrift
ALUMINIUM SCHROOT OPWERKING MET FRACTIONERENDE LAAG
KRISTALLISATIE
1. Met een toename van het gebruik van aluminium in de autoproductie is een
duurzaam technologisch terugwinningsproces, gebaseerd op fractionerende
kristallisatie, noodzakelijk.
2. De techniek van fractionerende kristallisatie is veel geavanceerder voor
organische stoffen dan voor metalen.
3. Zonesmelten is een relatief eenvoudige en bruikbare experimentele methode
voor het vaststellen van de mogelijkheid om gecompliceerde
legeringssystemen op te werken met fractionerende kristallisatie.
4. Moeder Natuur heeft ons het principe van fractionerende kristallisatie geleerd
door ijsbergen te creren met zoet water uit zout zeewater.
5. Opwerking van aluminium door middel van laag-groei fractionerende
kristallisatie is technisch een haalbare optie voor het ontwikkelen van een
duurzame technologie voor de terugwinning van hoogwaardig aluminium uit
schroot.
6. Post-zuiveringsprocessen, zoals zweten, besparen tijd, kosten en energie.
7. Theoretische analyse van de vallende-film kristallisatie voor aluminium
opwerking toont aan dat hetzelfde zuiveringseffect kan worden gehaald met
groeisnelheden (dus productiesnelheden) die 10 maal groter zijn dan voor het
statische proces.
8. Mensen zijn hongerig naar liefde!
Gongje Bojaxhiu alias Moeder Theresa
9. Een man zonder thuisland is als een kind zonder ouders.
10. Mensen uit ontwikkelingslanden toegang geven tot toponderwijs in de
ontwikkelde wereld is de beste bijdrage aan de ontwikkeling van de wereld.
CHAPTER I .................................................................................................................... 1
1. METAL REFINING WITH FRACTIONAL CRYSTALLIZATION: STATEOF-THE-ART AND FUTURE PROSPECTS ............................................................. 1
Abstract......................................................................................................................... 1
1.0. Introduction ........................................................................................................... 2
1.1. Theory.................................................................................................................... 3
1.2. Layer based processes ........................................................................................... 5
1.2.1. Heinzer et al. Crystallizer............................................................................... 6
1.2.2. Pechiney crystallizer....................................................................................... 7
1.3. Suspension based Processes .................................................................................. 8
1.3.1. Alcoa crystallizer............................................................................................ 8
1.3.2. Yunnan Crystallizer ...................................................................................... 10
1.4. Emerging technologies ........................................................................................ 11
1.4. A look to the future.............................................................................................. 15
1.5. Conclusions ......................................................................................................... 19
1.6. Acknowledgements ............................................................................................. 20
1.7. References ........................................................................................................... 20
CHAPTER II ................................................................................................................ 25
2. SCOPE........................................................................................................................ 25
2.1. Fractional crystallization principle and aluminium refining .............................. 25
2.2. Selection of alloy systems to be studied............................................................... 28
2.3. Research goals..................................................................................................... 30
2.4. Thesis outline....................................................................................................... 30
2.5.
References....................................................................................................... 32
5.2. Theory...................................................................................................................... 89
5.3. Experimental............................................................................................................ 91
5.4. Results and discussion ............................................................................................. 94
5.4.1 Chemical analyses ......................................................................................... 94
5.4.2. Microscopic Examinations ........................................................................... 97
5.4.3. Line scans ................................................................................................... 100
5.4.4. Inter-metallic compounds ........................................................................... 104
5.5. Conclusions ........................................................................................................... 107
5.6. Acknowledgements ............................................................................................... 107
5.7. References ............................................................................................................. 108
CHAPTER VI............................................................................................................... 111
6. FRACTIONAL LAYER CRYSTALLIZATION AND SWEATING APPLIED TO
ALUMINIUM SCRAP REFINING ............................................................................. 111
Abstract......................................................................................................................... 111
6.1. Introduction ....................................................................................................... 112
6.2 Experimental....................................................................................................... 113
6.3. Results ............................................................................................................... 115
6.3.1. Chemical analyses ...................................................................................... 115
6.3.2. Micrographs ............................................................................................... 117
6.4. Discussion.......................................................................................................... 122
6.5. Conclusions ....................................................................................................... 123
6.6. Acknowledgements ........................................................................................... 124
6.7. References ......................................................................................................... 125
CHAPTER VII............................................................................................................ 127
7. EVALUATION OF THE ALUMINIUM FALLING LIQUID FILM
CRYSTALLIZATION PROCESS STEPS .............................................................. 127
7.1. Introduction ..................................................................................................... 127
7.2. Hydrodynamics of the thin falling liquid films ............................................. 127
7.3. Heat transfer in falling liquid film ................................................................. 129
7.4. The effect of Reynolds number in diffusion boundary layer thickness ...... 130
7.5. Optimization of the aluminium falling liquid film crystallizer ................... 131
7.5.1. Optimal growth rate. .................................................................................. 131
7.5.2. Optimal crystal production rate. ................................................................ 132
7.5.3. Optimal feed temperature........................................................................... 132
CHAPTER I
This chapter was published by Drini B., Katgerman L., and Boom R: Proceedings of the ECI
Conference on Metal Separation Technologies III, June 20-24 2004, Copper Mountain, Colorado, USA,
Eds. Aune, R.E. and Keukkonen M., Helsinki University of Technology, Laboratory of Metallurgy,
Espoo, Finland; p. 34-41.
Chapter I
1.0. Introduction
The worlds metals production relies more and more on recycling of used metals.
Absolute leader in metal recycling is the gold industry for reasons of scarcity and value
per unit of weight. The steel industry, as biggest producer of metal approaching an
annual production level of 1 billion tonnes of crude steel, is an important recycler with
about 68 % recycled steel.
Aluminium, with 22 million tonnes production of row aluminium in 2003 the second
metal in terms of quantity, is increasing rapidly the fraction of recycled (secondary)
aluminium. The different aluminium alloy series, however, are not easily compatible in
terms of mixing: a beneficial element in one series is harmful for other series. Therefore
separation as well as purification of recycled aluminium alloys has received much
scientific and technological attention starting in the eighties of the last century. On top
of that the need for ultra-pure aluminium has increased, mainly for hard disc
applications in personal computers.
Melt crystallization is a separation process by which the fractional separation is effected
by directional crystallization from the melt. Fractional crystallization is applied for both
ultra purification of organic compounds [1, 2], and metals [3, 4]. Historically speaking,
three stages can be recognized in the development of fractional crystallization processes
for the purpose of refining. A first milestone was set by the invention of the Pattinson
process [5] for the extraction of silver from lead in 1833; further highlights were the
origination of the zone-melting technique in 1952 by Pfann [6], and a particular
elaboration for refining of tin as of 1975 [7]. This last example the so-called Yunnan
crystalliser is probably the most advanced application for metals at the moment. As a
continuous process and in an integrated functional design, the Yunnan crystalliser is
being used industrially (yields of 3045 tonnes/day).
In this paper one particular technique available for purification of metals, fractional
crystallization, is described and compared to the same technique used for purification of
organic compounds for the polymer industry.
Fractional
Crystallization
Suspension
Crystallization
Batch
Layer
Crystallization
Continuous
Static
Dynamic
1.1. Theory
Melt crystallization is a separation technique applied for purification of metals and
organic melts. It is widely known as fractional crystallization in crystallization worlds
and fractional solidification in the metal world.
An overview of fractional
Chapter I
crystal
Cooled
wall
melt
Crystal
layer
melt
Figure 2. Suspension (on the left) and layer (on the right) grown crystals.
Metals have a tendency of forming solid solutions. This has an adverse effect on
efficiency of purification of the metal in one crystallization stage. Therefore, if a
successful purification is to be achieved, the crystallization process has to be repeated.
This makes the process less energy efficient and also reduces the yield of the pure
crystal product. Even non solid solution forming systems are not 100 percent pure after
one crystallization stage. This is mainly due to the inclusion of the mother liquor (base
liquid metal) inside the crystal pores and adhesion of impurities on the surface of the
crystals. Therefore a further purification of crystals has to be performed. There are two
main processes applied to organic compounds for this purpose: Sweating and Washing.
Sweating is defined as a temperature induced post-purification step based on a partial
melting of crystals or crystal layers close to the melting point of the pure substance. As
a result, the impurities entrapped in the pores of crystals and adhered to their surface remelt, partially dilute with the pure crystals and drain off under the influence of gravity
forces. The temperature rises continuously, decreasing the viscosity of the entrapped
impurities and easing the draining process. The retention time of a sweating stage
applied to organic melt crystallization varies from 15 minutes for the falling film
process to several hours for static layer growth process.
Layer growth
Suspension growth
Up to100 m2 m-3
Growth rate
10-6-10-5 ms-1
10-8-10-7 ms-1
Mixing intensity
Good
Moderate
Excellent
Purification of solid
Suitable
Less suitable
No
Yes
Slurry handling
No
Yes
Mode of operation
Repetitive batch
Continuous
Design
Simple
Complex
Scale-up
Multiplication of units
Engineering
Operational attention
Normal
Above average
Maintenance intensity
Normal
Above average
solutions
There are two different washing processes. The first one is diffusion washing, where
washing liquid causes liquid/liquid diffusion of impurities from surface and pores of
crystals to the washing liquid. The second one is rinsing process, where the highly
contaminated adhered impurities at the crystal surface are washed of by pure washing
liquid. The first process has a retention time of typically 15 minutes or longer, while the
second one lasts only a few seconds. For both washing processes the washing liquid
should be overheated in order to prevent its crystallization on the crystal surface.
Chapter I
Figure 3. Layer based process for purification of silicon and germanium by fractional
crystallization [8].
Chapter I
This process is a continuous process for purifying of aluminium. The process can
integrate several stages in one crystallizer where the purified material of one stage can
serve as a feed material for the next stage and so on to desired amount of purification
stages. This can be done by adding additional cylinders parallel to each other (cascade).
Despite advantages of being a continuous and possibly multistage process, this process
has its disadvantages in that it is very sensitive to the tightness of the trough that
separates the inlet and outlet sections, lack of mixing on the crystallization zone to
provide improved conditions for mass and heat transfer, and difficulties of scaling up.
removed from the top of the crystallizer to initiate formation of the crystals in the zone
(70) and under influence of gravity they fall down to zone (72).
The bottom of the container is heated to prevent formation of incrustation and in the
same time to provide for sweating (partial melting) of the formed crystals, which
increases their purity and yield significantly compared to a previous version of this
crystallizer [11] that did not have this option. After a certain amount of crystals is
formed the crystals are pressed down by the tamper (78), which in addition to
compacting crystals to zone (72), breaks the agglomerates of crystals. The impure
aluminium mother liquor (74), high in eutectic impurities, is removed from the upper
part of the container, through the upper port (76). The remaining pure crystals are
melted and removed through the lower port (80).
While this process achieves significant purification results, it is a batch process and is
not suited for large scale operations that a recycling plant would require. Additionally, it
is a static process, which means that there is no mixing provided to enhance heat and
mass transfer, and encrustation to the walls of the crystallizer are inevitable, leading to
diminished purification efficiency.
Chapter I
10
11
Chapter I
crystals (concerning crystal growth rates and stirring) and the separation of the solid
from the liquid fraction are in most cases not fully met. Consequently, there still is room
for further progress.
A Dutch group, consisting of Corus plc. in IJmuiden, Dutch Organization for Applied
Research (known from its Dutch initials as TNO), and Delft University of Technology
are working on a fractional crystallization process for refining aluminium scrap. While
Delft is working on a layer based process, other partners in Dutch group are working in
a suspension based fractional crystallization process.
Our research group at Delft University of Technology works on a fractional
crystallization process for aluminium recycling based on static layer growth on a
vertical cooled tube with closed bottom, which is shown in figure 7 [12]. The main
emphasis in this process is paid to sweating as a post-purification process [13]. After a
crystal layer of approximately 2-3 cm is grown the cooled tube with an aluminium
crystal layer is removed from the mother liquor and is put on the sweating oven.
Cooling
air
Argon
10 cm
Crystal layer
TC
5 cm
20 cm
Sweating is based on the partial melting of the crystal layer by gentle heating close to
the melting point of the pure substance. The impure melt adhered to the surface of the
crystal layer or present in the intercellular regions of the crystal layer (see figure 8)
drains off under the influence of gravity. The temperature during the sweating process is
higher than in the crystallization stage. Although during the sweating process about 5
to10 percent of the crystals are removed as a sweat material, sweating is effective and
12
avoids an additional crystallization step in a fraction of the time (~15 min). Aluminium
with 6 wt. % silicon was used as a model alloy. This alloy was used for reasons of
simplicity of the phase diagram and relatively large temperature span of crystallization.
a)
b)
Figure 8. Optical micrographs for the AlSi6 alloy from layer growth experiments: a)
before sweating; b) after sweating (dark spots represent emptied intercellular eutectic).
The line in the picture denotes 0.1 mm [13].
The static layer crystallization experiments showed that typical growth rates of 1 to 3
m/s are optimal for successful application of sweating as a post-purification process.
The obtained results [12, 13] showed that the same purity of a crystal layer can be
achieved by applying sweating to a layer grown with 2 m/s with the layer grown at
half the speed (1 m/s) without sweating, which means that twice higher production rate
can be applied to achieve the same purification effect (figure 9).
7
6
% Si
5
4
3
2
1
0
starting alloy
1 m/s
2 m/s
2 m/s, after
sweating
Figure 9. Results for AlSi6 alloy from layer growth experiments: before sweating at
growth rate 1 and 2m/s; and after sweating of crystals grown at 2 m/s [13].
13
Chapter I
The suspension process developed by the Dutch group [14] is built in such a way that
stirring of the crystallizer, to enhance mass and heat transfer, is provided with a metallic
stirrer coated with refractory coating material. When approximately 20 to 25 % of the
aluminium is crystallized, crystals are filtered off from the mother liquor. The crystals
are much purer than from the layer based process. However, without an effective
separation process this process can not succeed to overcome problems such as low
crystal yield and removal of adhered impurities on the crystal surface and in crystal
pores.
Alcoa has also adopted the suspension based fractional crystallization process,
described in section 1.3.1, for aluminium scrap refining purposes [15]. Its performance
is in line with expectations, with understandably lower purification efficiency [15, 16],
which is known for processes that have higher alloying element concentration in the
starting material.
Figure 10. A schematic illustration of the Japanese suspension crystallizer for refining
aluminium scrap [17].
The most advanced group on research in aluminium scrap recycling is the Japanese
group [17]. They developed a suspension based process with capacity of producing
about 1000 tonne per month of purified aluminium scrap. The crystallizer is shown in
figure 10. It is similar to other processes described above with one distinction that a
compression filter is applied from the top and the crystals are packed in the bottom of
the crystallizer.
14
Zn%
Si %
1
0.8
0.6
0.4
0.2
0
Mn%
Fe%
Cu%
Initial concentration
Refined composition
Figure 11. Refining results of the radiator aluminium scrap with the Japanese
suspension crystallizer [17].
In figure 11 results of the purification efficiency of the radiator scrap are shown. It is
clearly seen that the removal efficiency of alloying elements is much lower than it is
possible from binary alloy phase diagrams. The reason for that is the presence of iron
and manganese as alloying components which form intermetalic compounds and thus
reduce the efficiency of their removal from aluminium scrap.
The main problem with suspension based processes is that there is no efficient process
of separating of the formed crystals from the mother liquor. Therefore, if a sustainable
suspension crystallization process is to succeed a better way of separating of crystals
from mother liquor has to be developed.
15
Chapter I
unit of weight, but it differs a factor of 8 per unit of volume. The latent of fusion has to
be removed on solidification and is therefore a liability for metals. However, due to the
much higher thermal conductivity for metals this liability is compensated [6].
In table 2 different parameters for organic materials and metals that can be purified by
the fractional crystallization technique are given. The data show that the main problems
for adopting technologies that are applied for organic compounds are high equipment
costs, which require special materials that can resist to the high corrosive attack of most
of the liquid metals, and difficulties of handling metal slurries. Other problems for
liquid metals could be their oxidation in contact with air, and high operational
temperatures required. The later one could pose serious problems concerning equipment
corrosion and energy efficiency. However, solutions for these problems are available in
a metallurgical process industry.
Metals
Organic compounds
300 - 1800
10-10
10-9
low
Could be a problem
20 - 400
0.1 - 0.6
200 - 400
200 - 380
0.1 1.2
1.3 3.3
1700 - 10500
<1000
Slurry handling
Difficult
Easy
Equipment corrosion
High
Low to moderate
Ease of pumping
Difficult
Easy
Equipment costs
High
Medium
-3
-1
16
then be a conceivable solution [18]. Except for the improved purification, benefits of
this concept would be that it is easily (up)-scalable with a high specific capacity.
Further, it is a continuous process, which has an advantage over batch processes
concerning energy efficiency. However, in order for this technique to succeed new
pumps should be invented that are capable of pumping metal slurries without heating it.
Corus recently patented a process for purification of metals with continuous suspension
based fractional crystallization [19, 20], which is shown schematically in figure 12. The
crystallization apparatus (1) consists of a chamber (2) in which a layer of a cooling
liquid (4) and a layer of a partially molten metal (5) to be purified are present. The
cooling liquid is drown in (arrow A) and circulated through pipe (6), where a cooling
device (8) is present, by means of pump (7) and after cooling is turned back to chamber
(2) (arrow B). The density of the cooling liquid is higher than that of the molten metal
and they should not mix with each other to avoid contamination. The molten metal is
added through inlet (10) and in contact with cooling liquid crystallization takes place.
The crystals remain for a while in suspension as a result of mixing, than fall on the top
of the cooling liquid and are transported by it towards outlet (11). In the same time the
molten metal enriched with alloying elements move counter currently towards outlet
(12). The chamber is divided in several compartments and a temperature gradient is
imposed as a result of heating up of the cooling liquid due to the contact with hotter
molten metal and dissipation of the energy of crystallization of the molten metal. The
process is continuous with the amount of the added molten metal in inlet (10) equalling
the amount of pure crystal removed in outlet (11) and mother liquor removed from
outlet (12). The same process is possible with the cooling liquid having lower density
than molten metal to accommodate removal of the crystals of metals which have lower
density of crystals than their liquid [20].
17
Chapter I
Figure 12. Schematic presentation of a cross section of the Corus crystallizer [19].
crystallization stage. Results from organic compound purification processes show that
their purifying effect for falling film [1] is even greater than for static processes.
Fractional crystallization for metals is widely used for lead, tin, and silver, and could be
applied to purify technologically important metals such as silicon, aluminium and
magnesium. Liquid silicon is a metal, and the large-scale production of solar cells based
upon silicon demands an economic purification technology that produces ultra-pure
silicon. The yield of a solar cell is strongly dependent on silicon purity. For aluminium,
there is a demand for ultra-pure aluminium (made from primary metal) for application
in the electronics industry (hard disc memories). There is also a fast growing need to
purify secondary aluminium from recycled scrap. Magnesium, as a rapidly growing
metal with applications in light-weighting design for transport and in casings for lap top
computers and mobile phones, will have to be recycled and purified because tramp
elements dissolved in magnesium strongly deteriorate the corrosion resistance of this
metal.
The future of fractional crystallization for these and possibly other metals will be bright
due to the economic, energy conservation and ecological drives.
1.5. Conclusions
Fractional crystallization technique is being used successfully for metal refining for
almost two centuries. The most advanced process for this purpose is the Yunnan
crystallizer for separating tin from lead. While existing technologies are mostly of a
limited production capacity and are mainly used to produce super-pure metals, the
increasing demand for developing this technique for recycling of metal scrap requires
processes that have much higher production capacity. Therefore research in this field is
intensifying rapidly.
Fractional crystallization technique is much more advanced for organic compounds than
for metals. This is mainly due to the better process conditions, such as low melting
temperature and lower reactor attack, of organic materials. However, a careful study of
possibilities of adopting these processes for metals will open new horizons. Prospects
are good for both modes of operation: suspension and layer crystallization. The most
important metals that could use this technique are aluminium, silicon, magnesium, lead,
and precious metals.
19
Chapter I
1.6. Acknowledgements
This research is carried out as part of the EET funded project Sustainable Technology
for the Reclamation of High-grade Aluminium from Scrap (contract number
EETK98026). Fruitful discussions, about the organic material technologies, with Prof.
P.J. Jansens of the Laboratory for Process Equipment of Delft University of Technology
are greatly appreciated.
1.7. References
Qiu, K. Sudholter, S., Kruger, J., and Yang, X.: Raffination von Metallen
durch fraktionierte Kristallisation aus der Schmelze [Historical development
and prospects: metal refining by fractional crystallization from the melt], Metall.
Vol. 49, no. 7-8 (1995) pp. 491-495.
20
8. Heinzer, H., Rath, H-J., and Schmidt, D.: Process for Purifying Solid
Substances, US Patent no. 4231755, Application: 19 May, 1978, Accepted: 04
November, 1980.
9. Boutin, R.F., Process for the Continuous Purification of Metals by Fractional
Crystallization on a Rotary Drum, US patent no.4581062, Application: 13 May,
1985, Accepted: 8 April, 1986.
10. Dawless R.K. and Graziano R.E. Fractional Crystallization Process US Patent
no. 4294612, Application: Nov. 30, 1979, Accepted: Oct 13, 1981.
11. Dawless R.K. and Jacobbs S.C., Production of Extreme Purity Aluminium,
US4273627, Application: Dec 26, 1979, Accepted Jun 16, 1981.
12. Mehmetaj, B., Bruinsma, O.S.L., Kool, W.H., Jansens, P.J., and Katgerman, L.:
Aluminium Scrap Recycling with Solid Layer Fractional Crystallization,
Proceedings of the 15th International Symposium on Industrial Crystallization
(ISIC-15),
Ed.
Chianese,
A.,
Sorrento,
Italy,
Chemical
Engineering
Ed.
Chianese,
A.,
Sorrento,
Italy,
Chemical
Engineering
Chapter I
22
CHAPTER II
2. SCOPE
Figure 2.1. Binary Alloy Phase Diagram for a Eutectic System: Schematic Enlargement of the Semi-solid
State (+L).
Chapter 2
Table 2.1. Limits on the Purification of Secondary Constituents from Aluminium by means
of Fractional Crystallization (Single-step Operation)
Element
System
Maximum Temperature
Thermodynamic Distribution
Range T (C)
Coefficient k0 ()
Si
eutectic
83
0.13
Fe
eutectic
Cu
eutectic
Mn
eutectic
Mg
eutectic
210
Cr
peritectic
Ni
eutectic
21
0.00
Zn
eutectic
279
0.87*
Ga
eutectic
633
0.20*
peritectic
Zr
peritectic
Ti
peritectic
Pb
monotectic
1.6
0.00
Sn
eutectic
432
0.00*
Cd
monotectic
11
0.05
Sb
eutectic
3.5
0.00
peritectic
Bi
monotectic
3.5
0.04
5.5
0.03
112
0.17
2.5
0.62
0.45*
* By approximation
In evaluating the performance of a fractional crystallization process, an indicator called
the distribution coefficient k is used. Relating to a mixture of the components A (here:
aluminium) and B (any other element), it is defined as:
26
Scope
k=
CB,s
.
CB,u
(1)
CB,s and CB,u are the concentrations of component B in the solidified fraction and in the
untreated material respectively. This distribution coefficient, sometimes referred to as the
effective distribution coefficient, depends on the thermodynamic properties of the system,
but also on the process kinetics. The distribution coefficient in the equilibrium condition
can be derived from the phase diagram of the concerned system and is called the
thermodynamic distribution coefficient k0.
A central question in considering the use of fractional crystallization for refining
aluminium is to what extent secondary constituents can be theoretically purged from the
base metal by this principle. Particularly, the constraints on the purification and yield are of
interest, as well as the requirements concerning such parameters as processing temperature.
An answer on this can be obtained from the phase diagrams of the concerned systems [1].
Restricting to binary alloy diagrams here, mixtures of aluminium and a single other
element theoretical limits are summarised in table 2.1.
These are grouped from top to bottom as major alloying elements, minor alloying elements
and tramp elements in aluminium.
Considerations in appraising this information are recapitulated next:
Assumption is that both fractions in the semi-solid state are separated from each other
completely; that is, without mutual attachment. Thus, the obtained distribution coefficient
constitutes a lower limit. The refining performance does essentially depend on the content
of the alloying element in the feed material, but this shows to be of minor relevance for
most of the considered binary systems. For those elements, where it is of significance,
approximations (averages) are given. Listed values are obtained by using weight
concentrations.
The refining performance does also depend on the yield P of the process, which is defined
as the quantity or flow of the product as compared to that of the unrefined input (0P1). In
the table, only the data for P=0 (no yield!) are represented.
Relating to the matter of temperature control for such processes, the maximum temperature
27
Chapter 2
range T is included in the table 2.1. It is taken as the full temperature interval of the semisolid state (i.e., TmATE in figure 2.1). Thus, it is an upper limit that may be much smaller,
depending on the actual alloying contents.
Most of the listed systems are either eutectic or monotectic. They are thus apt for refining
by means of fractional crystallization processes in which the impurities are concentrated in
the remaining melt. Some systems are peritectic, meaning that the concerned secondary
constituents end up in the opposite fraction (that is, the crystals). It can be argued that these
elements can be removed by some other treatment, like filtering the crystals from the melt.
For complicated alloy systems (like actual scrap aluminium), however, it remains
problematic that a material that is purified with respect to one category of constituents is
enriched with the other. For the major alloying elements all being eutectic the
theoretical prospects are good for iron, silicon, and copper, fair for magnesium, and
moderate for manganese. Apart from this, process control will for some cases be critical:
the available temperature interval is very small for iron and manganese.
Finally, a reservation that has to be made is that the conditions during the actual refining
operation will be principally different from the quasi-static conditions that apply for the
phase diagrams. Further, it should be noted that an extrapolation to systems with more than
two components is only indicative.
Scope
approaches are: a selection based on the wishes of Corus Group's aluminium business
units, and a selection based on generally relevant binary and ternary phase diagrams.
Binary phase systems are suitable to practice using the "tools". Ternary phase systems,
e.g. Al-Si-Fe and Al-Si-Mn, are the first systems that are relevant to industrial
applications of fractional crystallization.
Concentrations of alloying elements in the inlet stream of each alloy system have been
selected on the basis of guesses at the compositions of scrap types. The following four
systems have thus been selected :
Al - Si ;
Al - Si - Fe ;
Al - Si - Mn ; and
Al - Cu - Mg.
In the Al-Si system, there is an eutectic at about 12 mass % Si. To have a good starting
point for the process during the initial experiments, an alloy halfway between pure Al and
the eutectic, an Al-Si alloy with 6 mass % Si, has been chosen. This system is only used
to hone skills.
A ternary system that is of interest is the Al-Si-Fe system. Both Si and Fe are always
present in aluminium alloys. Two sources of Fe in aluminium scrap are foil, and parts that
have been produced with high pressure die casting[1,2]. Liquid aluminium reduces SiO2,
which comes from sand particles or refractory material. This makes that Si will be present
in molten scrap. Hence, aluminium scrap contains both Fe and Si, and the Fe content may
have to be reduced anyway. Therefore, it has been suggested to study the ternary Al-SiFe system in the project on fractional crystallization. In general, the alloys for pressure
die casting contain at least 0.6 mass % Fe to reduce sticking of the cast part to the die [1].
Food packaging is the main application for aluminium foil. An alloy that is being used for
foil production is AA8014 [3]. It contains less than 0.3 mass % Si and 1.2 to 1.6 mass %
Fe. Therefore, an alloy with 0.3 mass % Si and 1.2 mass % Fe has been selected.
29
Chapter 2
The third alloy system studied in this project is the Al - Si - Mn system, more specifically
an alloy of composition 1.5 mass % Si and 1.0 mass % Mn, because this is a typical
mixed scrap composition of braizing sheet.
The fourth system is the ternary system Al-Cu-Mg with 4 mass % Cu and 1.5 mass %
Mg. This composition represents more or less the composition of chips generated during
the machining of sheets or plates of AA2XXX alloys that are used in aircraft structures
[1,3]. These chips are not used as "raw material" for the production of wrought alloys at
the moment as they form a mixture that contains too many alloys. An avenue to close the
recycle loop for AA2XXX alloys may be the application of a purification technique like
fractional crystallization.
Studying the optimum layer growth rate for most economical way of purifying
selected aluminium alloys.
Examininning the structure of the grown crystal layers and impurity distribution
within crystal layers.
30
Scope
All the selected aluminium alloys were studied with zone melting, static layer growth,
and sweating postpurification technique. Suitable layers were obtained and studied with
different analysing techniques to unlish the impurity distribution and their local
concentration, the size of crystal cells and intercelular spaces depanding on growth rate,
etc. The obtained layers were also sweated to remove the impurities that are concentrated
mainly in the intercellular space of crystal layers and the effect on overall purification
was established.
The possibility of purifying aluminium scrap with falling film fractional crystallization
technique was tested theoreticaly and equipment for this purpose was constructed.
31
Chapter 2
2.5.
References
1. J.R. Davies et al., Metals Handbook, Tenth Edition, Volume 2 Properties and
Selection : Nonferrous Alloys and Special-Purpose Materials, ASM International
(1990)
2. D.G. Altenpohl, Aluminium: Technology, Applications, and Environment, The
Minerals, Metals & Materials Society, Warrendale, PA, USA (1998)
3. J.P. Lyle and D.A. Granger, "Aluminum Alloys", Ullmann's Encyclopaedia of
Industrial Chemistry, 5th Edition, Volume A1, Ed. W. Gerhartz, VCH
Verlagsgesellschaft mbH, Weinheim, Germany (1985)
32
CHAPTER III
3. SOLUTE REDISTRIBUTION DURING DIRECTIONAL
CRYSTALLIZATION AND INTERFACE MORPHOLOGY
3.1. Introduction
As a result of the differences in solute solubility in liquid and solid state, there is a
solute redistribution at the solid liquid interface. If the solubility of the solute in the
crystal phase is smaller than in the liquid phase, the solute is rejected at the interface
leading to solute pile up in the adjoined liquid, while if the solute solubility in the
crystal phase is higher than in the liquid the solute layer adjoined to the interface is
depleted in solute. The fact that solidus and liquidus lines of the phase diagrams do not
coincide, except for pure materials and in the exceptional cases of congruent melting,
indicates that a solid usually differs in composition from the liquid in equilibrium with
it. Therefore, crystallization processes are always accompanied with solute
redistribution which is different from the homogeneous distribution in the starting
liquid, although the total amount of solute does not change.
Throughout this chapter, except for equilibrium condition, diffusion of the solute in the
solid is not taken into account. The reason is that diffusion in liquid state is much faster
than diffusion in the solid state. In the case of metals, diffusivities in the solid state at
the melting point are by a factor of 100 to 1000 lower than in the liquid state. This is
expected, as diffusion is defined as a transport of mass from one region to another on an
atomic scale. Therefore there is much lower resistance to mass transfer in the liquid than
in the crystal phase.
In this chapter the solute redistribution for a flat solid/liquid interface is described in
sections 3.2 and 3.3. In section 3.4 the morphology of the interface is treated. Cellular
structure of crystals and solute redistribution in such morphology are presented in
sections 3.5. In section 3.6 some selected models used to describe experimental solute
redistribution are analyzed. Finally, general conclusions are presented in section 3.7.
Chapter III
k0 =
CS
CL
(3.1)
It is noted that several names are in use for k0. It is variously called as partition
coefficient, segregation coefficient, and distribution coefficient. The last term is the
Temperature
Temperature
CS
CL
Composition
a) k0 < 1
CL
CS
Composition
b) k0 >1
Figure 3.1. Schematic phase diagrams showing solidus and liquidus lines; a) k0 < 1; b)
k0 > 1.
In figure 3.1 is clearly seen that solidus and liquidus are represented with straight lines,
which is a simplification and can be true only for concentration spam. For a negative
slope, k0 is smaller than one, which means that the solute reduces the melting point of
the pure solvent or metal and crystallization leads to purer crystals than the original melt
composition (figure 3.1a). Cooling of a liquid of composition CL, of which the phase
diagram has a positive slope (figure 3.1b), results in formation of crystals with
composition CS which are enriched with impurities.
During equilibrium crystallization the concentration of solute in crystals changes
towards the overall concentration as a result of diffusion in the solid. Figure 3.2
34
schematically shows a phase diagram for which k0 < 1. When a liquid of composition
CL is cooled to a temperature T1 the first crystals will appear with composition CS1. As
crystallization proceeds towards temperature T2, the impurity concentration in the liquid
is represented by point M. The amount of formed crystals is given by the lever rule as
LM/KM. The composition of crystals is uniform and is given by point K (CS2). When
crystallization continues till the entire melt is solidified, the solute concentration of the
crystals is given by point O, and is equal to the starting concentration CL. Therefore,
true equilibrium crystallization gives a uniform concentration which is equal to that of
the starting liquid C0.
This is not often achieved due to the limited diffusion in solid or liquid. In many cases it
also not desired. In case of purification processes a concentration difference between
crystals and liquid is even imperative.
Temperature
T1
T2
T3
CS1 CS2 CL
Composition
Figure 3.2. Phase diagram and solute distribution during equilibrium crystallization.
The solute concentration in the solid part, CS, of an ingot undergoing equilibrium
crystallization, after a fraction fS is crystallized is given by equation 3.2.
CS =
k 0C0
1 + f S (k 0 1)
(3.2)
35
Chapter III
CS, C0, and k0 are expressed in mol, atom, or weight fractions, which are independent of
density.
3.3.1
Complete mixing of the liquid will never be attained since there will always be a liquid
layer in contact with the growing solid interface where solute transport is diffusion
controlled. However, when the diffusion coefficient is high or the growth rate is low,
the condition of complete mixing will be approached.
Under the condition of complete mixing, the solute concentration profile of the crystals,
crystallized directionally from a cylindrical charge, shows a profile which is defined by
the solidus line. Figure 3.3 gives the solute concentration profile for a certain fraction
crystallized. The concentration profile shows a discontinuity at the solid/liquid interface.
The ratio of the solid to liquid concentration after a fraction x is crystallized is given by
CS(x)/CL(x), which is equal to k0 (equation 3.1).
The concentration profile as a result of the crystallization is given by Scheils law [1]
and reads
C S = k 0 C 0 (1 f S ) ( k0 1)
(3.3)
36
Composition
CS(0) = k0C0
CL(x)
C0
CS(x)
SOLID
0
LIQUID
x
Fraction crystallized
37
Chapter III
CS(0) = k0C0
Composition
Initial
transient
C0
k0
C0
CL(0)
Steady state
Final
transient
CL(x)
CS
0
1
Fraction crystallized
Figure 3.4. Solute redistribution in the solid, CS, and solute concentration in the liquid
at the solid-liquid interface, CL(0). The dashed lines, CL(x) indicate the solute
concentration in the liquid ahead of the solid interface at different fractions
crystallized; k0 < 1 [2].
Tiller et al. [1] derived the following relationship between the position x in the liquid in
front of the crystal interface and the solute concentration in x, CL(x), during steady
state growth:
Gx'
C L ( x' ) = [C L (0) C 0 ]exp
+ C0
D
(3.4)
where G and D are growth rate and diffusion coefficient in the liquid respectively, and
x is distance from the solid interface. The graphic representation of equation 3.4 is
depicted in figure 3.5.
38
Composition
CL(0)
CL(x)
C0
Interface
0
Distance from interface, x
Figure 3.5. Solute redistribution ahead of the solid-liquid interface during directional
Gx
C S ( x) = C 0 (1 k 0 ) 1 exp k 0
+ k 0
D
(3.5)
Equation 3.5 describes an asymptotic rise of the solute concentration from k0C0 at x = 0,
to C0, when steady state begins (see figure 4). The length of the initial transient, XI, is
XI =
D
Gk 0
(3.6)
Equation 3.6 shows that XI increases for lower values of k0 and G. When steady state is
achieved, CS = C0 and CL(0) = (C0/k0). If we substitute these expressions in equation 4
the solute concentration in the liquid at the interface at steady state condition is given by
39
Chapter III
1 k0
Gx'
C L ( x' ) = C 0 1 +
exp
k0
D
(3.7)
Steady state will be maintained as long as there is a sufficient amount of liquid ahead of
the crystal interface for forward diffusion of the solute and G remains constant. When
the amount of liquid becomes insufficient, the solute concentration in the crystals begins
to increase above C0 in order to accommodate the excess solute, which was accumulated
in front of the interface. This stage of crystallization is called the final transient. Its
length is D/G and the final transient occupies a shorter distance than the initial transient
by a factor k0 (see figure 3.4). Characteristic values for D/G are 0.01-0.1 cm. The length
of the initial transient is dependent on the alloying system. During directional
crystallization when the liquid is stagnant, only the initial transient region becomes
purified.
40
Composition
CS(0) = k0C0
CL(x)
CL
C0
CS(x)
SOLID
LIQUID
x
Fraction crystallized
= 1 .6
1
3
1
6
(3.8)
1
2
Thickness of the boundary layer strongly depends on the intensity of stirring in the
liquid, and ranges from 0.01 mm for strong stirring to 1 mm for natural convection.
The effective distribution coefficient, keff, is now defined as the ratio of the
concentration of the solute in the solid at the interface, CS(x) (see figure 3.6), to the
concentration of the solute in the bulk of the liquid CL. It should be mentioned that keff
is considered constant over the crystallization process.
41
Chapter III
The value of keff is higher than the value of k0, and ranges from k0, for complete mixing
in the liquid, to one, for diffusion limited transport such as for the stagnant liquid (see
Section 3.3.2). The value of keff depends on the amount of convection in the melt.
The stagnant film model introduced by BPS establishes a relation between the effective
distribution coefficient and the dimensionless parameter G/D, which is used as the
normalized growth velocity. They obtained the following expression for the effective
distribution coefficient keff during steady state growth:
k eff =
C S ( x)
=
CL
k0
k 0 + (1 k 0 )e
(3.9)
G
D
G
1
ln
1 = ln 1
k
D
k0
eff
for k0<1
(3.10a)
1
ln1
k
eff
for k0>1
(3.10b)
= ln1 1
k
G
D
In this way, the /D value can be obtained experimentally by determining keff measured
from crystals grown at different growth rates under identical stirring condition and
plotting the values of ln(1/keff 1) or ln(11/keff) against the crystal growth rate. This
should give a straight line with slope -/D. The parameter keff is generally determined
from a crystallization experiment by quenching and measuring the solute concentration
in the solid near the interface and in the bulk liquid.
42
CS(0) = k0C0
Composition
CL(0)
CL(x)
C0
C0
CL
CS
0
1
Fraction crystallized
Figure 3.7. Solute redistribution in the solid (CS), solute redistribution in the liquid
(CL), and solute concentration in the liquid at the solid-liquid interface [CL(0)]. The
dashed lines, CL(x) indicate the solute concentration in the liquid ahead of the solid
interface at different fractions crystallized; k0 < 1.
Figure 3.7 shows that the solute concentration in the solid increases as a result of
increased liquid concentration, which happens due to the rejection of the solute from the
crystal interface.
43
CS(0) = k0C0
Composition
Chapter III
b
a
C0
c
d
1
Relative location in the crystal
Figure 3.8. Solute redistribution in the crystal after complete crystallization, for k0 <
1, in case of: a) complete diffusion in solid and liquid (equilibrium), b) stagnant liquid,
c) partial mixing in liquid (BPS), and d) complete liquid mixing (Scheil).
44
As a result of constitutional supercooling the local liquidus temperature, Tl, of the liquid
will be affected. Tl is related to the composition by
Tl (C0) - Tl = m(C0 Cl)
(3.11)
where Tl (C0) is the liquidus temperature corresponding to the initial alloy composition,
m is the liquidus slope in the phase diagram, C0 is the initial concentration of the alloy,
Cl is the solute concentration ahead of the advancing interface, and Tl is the liquidus
temperature ahead of the advancing interface. Liquidus temperatures determined from
equation 3.11 are indicated in figure 3.9.
C0/k0
C0
Cl
C0
x
Tl
Tl (C0)
Tq
T0
Ts (C0)
C0
Tcool
Tl
C0/k0
advancing interface during layer crystallization of a simple eutectic binary alloy [1] .
For eutectic systems the liquidus temperature of the liquid ahead of the solidifying
interface increases with increasing distance, x, from the interface. In order to determine
the stability of the growing interface, it is important to determine the temperature
imposed by the heat flow during crystallization, Tq. The temperature gradient in the
liquid, GT,l, is defined as
45
Chapter III
dTq
GT ,l =
dx x =0
(3.12)
Depending on the value of GT,l in the liquid at the solid-liquid interface, there may, or
may not exist a zone of constitutional supercooling (see figure 3.10).
Tq
Tl
Tl
Tq
dTl dTq
dx'
dx'
dTl dTq
>
dx'
dx'
T
Tcool
heat
exchanger wall
Figure 3.10. Condition for constitutional supercooling at the crystal layer interface,
GT ,l
G
<
mC 0 1 k 0
*
D
k0
(3.13)
46
where G is the growth rate, D is the diffusion coefficient, and k0 is the thermodynamic
distribution coefficient.
Experimental work followed to prove the constitutional supercooling hypothesis [5-8].
It was conducted by separating solid and liquid by decanting and inspecting the crystal
interface for signs of instability. The results were presented as a GT,l/G versus C0 plot,
which represents a straight line passing through the origin. In this way a distinction was
possible between a stable and unstable interface. The slope was compared with (-m)*(1k0)/(k0D) (see equation 3.13) and a reasonable value of D was found to comprise a
substantiation of the constitutional supercooling theory.
These experiments gave a semiquantitative merit to the constitutional supercooling
theory. However, one must notice that a precise quantitative test was beyond the scope
of these experiments. The quantities k0, m and especially D are not well known, while
GT,l is difficult to measure accurately during the crystal growth from the melt.
The constitutional supercooling principle is very useful and easy to apply even to
complex situations. Hurle [9] has given the expression that explains the condition under
which a stirred melt can be constitutionally supercooled. It reads
GT ,l
G
<
mC0 1 k0
*
*
D
k0
k0
k0 + (1 k0 )e
G
D
(3.14)
The factor, added at the right-hand side of equation 3.14, as compared with equation
3.13, is the same as in the BPS expression for keff and is always smaller or equal to 1 for
k0<1. This represents the fact that stirring enhances the interface stability.
3.4.2. Interface stability theory
The constitutional supercooling theory provides a useful estimation of the solid/liquid
interface stability in directional crystallization conditions. However, theoretically a
drawback of the constitutional supercooling criterion is that it takes into account only
the thermodynamic properties of the liquid, while ignoring the thermodynamic
properties of the crystal and the interface surface tension. Mullins and Sekerka [10-13]
studied the latter properties in their theory of interface stability.
47
Chapter III
Growth direction
0
1
2
3
the overall growth direction and expands laterally with wavelength 0 [13].
In planar crystal growth the interface is considered to be initially planar and then a small
perturbation is imagined to appear (see figure 3.11). Whether the perturbation will grow
or disappear depends on the surface energy, on the solute and thermal fields, and on the
interface kinetics. Mullins and Sekerka [11, 12] made the assumption that there is
equilibrium at the solid/liquid interface, that no convection is present, and that the
surface energy is isotropic. The instability criterion is given by
GT ,l
G
mC 0 1 k 0 s + l
H
*
*
*S
<
D
k0
2 l
2 l
(3.15)
where H is the heat of fusion, s and l are the thermal conductivities in solid and
liquid respectively, S is a dimensionless stability function, and the other parameters are
the same as in equation 13. The S function is a function of k0 and is graphically given as
a function of the dimensionless parameter A in figure 12. A is defined by
A=
k 02
G TM
*
*
1 k 0 D mC 0
(3.16)
where is the solid/liquid surface energy and TM is the melting temperature of the pure
solvent. Parameter A varies linearly with and its value is 10-3 or smaller.
48
The value of S is typically 0.8 to 0.9. For negligible small solid/liquid surface energy
S=1. Hence, the deviation of S from 1 represents the tendency of the surface free energy
to stabilize the interface. After neglecting the term H/2kl, equation 3.15 can also be
presented as [13]:
__
GT
S
GC m
(3.17)
GT =
GT , s s + GT ,l l
(3.18)
s +l
49
Chapter III
k0=0.01
k0=0.1
k0=0.3
k0=0.5
k0=0.9
k0=5
0.9
0.8
k0=1000
0.7
k0=100
k0=10
0.6
0.5
Unstable
0.4
0.3
0.2
Stable
0.1
0
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
log10A
Figure 3.13. Plot of the dimensionless stability function S versus log10A for different k0
values [13].
It can be seen from equation 3.15 that the latent heat of fusion H contributes to the
stabilizing term H/2 l , which can have an important effect in diluted solutions. Its
stabilizing effect is small for metals, but can have a dominant impact for instance for the
ice/water system [14].
a)
b)
Figure 3.13. Morphology of the growth front of an Al-Cu alloy: a) Initial perturbation
of the planar front at grain boundaries; b) Transverse section of the fully developed
cells [14, 15].
Sharp and Hellawell [14, 15] studied the interface morphology changes of an Al-Cu 2
percent alloy by a sudden acceleration of the growth rate and subsequent quenching of
the sample to observe the morphological changes in the crystal interface. They found
that the breakdown of the flat interface starts close to the grain boundaries, which act as
built-in distortions of the planar interface. This effect is clearly seen in figure 3.13a. The
interface develops further to a fully cellular structure as shown in figure 3.13b, which
shows that cells are regularly spaced.
A useful way to observe the cellular structure as it develops is by using transparent
organic compounds. Succinonitrile is the most widely used organic substance for this
purpose. The main reason for that are its plastic crystal properties, and thus, like metals,
it crystallizes with a macroscopically smooth solid/liquid interface and forms no facets.
This behavior reflects the isotropic properties of metals and a special class of molecular
crystals. A typical succinonitrile cellular interface morphology is given in figure 3.14a.
51
Chapter III
a)
b)
Succinonitrile cellular front [16]; b) Salol growing with faceted cells [17].
The cellular morphology of the transparent organic crystals of which the solid/liquid
interface develops facets is shown in figure 3.14b. Most of the non metallic crystals,
including semiconductors, develop facets due to their anisotropic crystal nature. The
conditions of formation of faceted cells are similar to the one described above for
metals.
During cellular growth the solute concentrations in the cells and in the intercellular
space differ considerably. This observation has been confirmed by numerous
researchers [18-20]. They found that the degree of segregation towards the intercellular
space can be so high that a lead rich phase was present even in a tin 0.16 % lead alloy.
52
Figure
3.15.
Concentration
profile
ahead
of
the
cellular
interface
for
Assuming a simple binary alloy like succinonitrile acetone [21] containing a eutectic
point, the maximum solute concentration in the intercellular region is equal to the
eutectic concentration (CE = C0/k0). This condition is depicted in figure 3.15.
In sections 3.2 and 3.3, the solute redistribution was considered only for directional
crystallization with a flat solid/liquid interface. Due to this assumption the equations
describing the redistribution are relatively simple. These equations are not valid to
describe the cellular growth solute redistribution. The multidimensional diffusion alone
makes it impossible to describe the redistribution in a satisfactory way. Additionally,
there are complications of solid state diffusion and radius of curvature.
Some attempts were made for describing the solute redistribution in nonplanar growth
conditions [22-27]. These microsegregation models are semi-analytical and approximate
relationships. They simplify the conditions by treating only a volume element which has
a small thickness in the growth direction and extends towards the neighboring cell wall
[22]. These are simplified models which cannot describe accurately all the processes
that go on during cellular crystal growth. Their results are in line with the experimentaly
achieved ones, but their accuracy is compromised by making numerous approximations.
53
Chapter III
C S = fC L (x)
(3.19)
Equation 3.19 is equivalent to equation 3.1 with the surface void fraction corresponding
to the equilibrium distribution coefficient. Further, Chang et al. [28] gave a modified
expression for the effective distribution coefficient given by BPS in equation 9, where
instead of the equilibrium distribution coefficient k0, a surface void fraction f parameter
is given.
k eff =
f
f + (1 f )e
G
D
(3.20)
54
1.2
1
1
ln
1
k
0.8
eff
0.6
0.4
0.2
0
0
Growth Rate
10
cm/hour
Figure 3.16. Logarithmic 1/(keff-1) versus growth rate for the organic system
As mentioned above, the two regions of linearity for the alloying system correspond
quite closely to the two distinct regions of interface morphology. This brings us to the
conclusion that a constant or average value of interface void fraction f can be given to
each of the regions of linearity. The value of f is higher than that of k0.
55
Chapter III
portion of crystals is in the second layer and the nucleation and adherence layers are
very thin. The post purification steps mentioned in figure 16 are described in the next
section.
Cooled
melt
Wall)
Step
(1)
(2)
(3)
Mechanism of
impurity
incorporation
High impurity
concentration
due to high
nucleation rate at
high supercooling
Impurities on
crystal layer due
to high crystal
growth rates
Impurities due to
adherence of
contaminated
residual melt at
the end of the
process
Operation measures to
prevent impurities in the
crystal product
-Nucleation by controlled
temperature and flow
conditions
-Controlled seeding
-Mechanically induced
nucleation by shock waves,
ultrasonic vibration, or
cavitations
Purity criteria which relates
the maximum allowable
growth rate for the
production of pure layer
Additional
purifying
measures
-sweating
-sweating
-washing
-sweating
Modeling of the layer crystallization processes is empirical [31], mainly due to the
complexity of various kinetic effects described in previous sections. During the
crystallization process the impurity segregation effect is measured by the effective
distribution coefficient (keff), which accounts for equilibrium, kinetic and phase
separation effects. Therefore, the measurement and use of the keff values must be used at
constant conditions with respect to growth rate, initial impurity concentration, etc.. In
practice the main advantage of the model described below is that it gives an expression
of the purification achievable in a single stage depending on the amount of the original
liquid remaining un-crystallized. Now we consider an initial mass of liquid m0, with
impurity concentration C0, being partially crystallized to give a reduced mass of liquid
mr, with an impurity concentration Cr. Since keff can be considered as a constant, the
56
mass and component balance for a differential balance of crystallized mass can be
integrated to give an average distribution coefficient kcr:
k cr = 1 +
log(C r / C 0 )
log(mr / m0 )
(3.21)
0.4
k_cr
0.3
0.2
0.1
0
0
0.05
0.1
0.15
0.2
%PTZ
Figure 3.18. The kcr variation as a function of the initial impurity concentration for the
57
Chapter III
Feed
liquid
One purification
stage
Phase 1:
crystallization
Phase 2:
Partial
melting/
Phase 3:
Total melting
Liquid
Product
Residue
liquid
Cooling
Gentle heating
Heating
58
The theoretical description of the sweating process is studied by Wangnick [36] for
layer based processes. The proposed equation gives the correlation for predicting the
sweating effect on the additional purification that it can achieve. The equation reads
k eff , sw = 0.028 0.277 * sw + 1.235 * 2sw
(3.22)
(3.22a)
with:
sw = k eff ,before *
and
Tsw
Teq
(3.22b)
where keff,sw is the difference between the distribution coefficients before (keff,before) and
after sweating (keff,after) of the grown crystal layer, sw is a characteristic dimensionless
number of the sweating process, Tsw is the temperature of sweating, and Teq is the
equilibrium melting temperature of the crystal.
Equation 3.22 was successfully applied to both static and dynamic layer crystallization
processes for organic compounds [36].
There are two different washing processes applied in layer crystallization processes:
rinsing and diffusion washing [36]. Rinsing is applied to remove the highly
contaminated melt that is attached to the surface of the crystal by purer liquid. The
rinsing process lasts only a few seconds. Diffusion washing causes liquid-to-liquid
diffusion of impurities from the pores of the crystal layer. To prevent the solidification
of the washing liquid onto the crystal the washing liquid is superheated. Therefore
diffusion washing is always accompanied by sweating, which further increases the
purification effect. The duration of the diffusion washing process usually is more than
15 minutes. Diffusion washing is applied to remove the impurities from the surface of
the crystal layer and in the pores of the grown layer (see figure 3.17). As a washing
liquid for the diffusion washing process one can use the starting material, which is purer
than the impurities included in the pores and the surface of the crystal. The washing
liquid can be added to the new starting material for crystallization. The purification
effect of the washing process is lower than for other purification steps. However, it is
59
Chapter III
well suited for easy removal of the impurities that are included in the outer part of the
crystal layer.
60
3.8. References
1. W. Kurz and D.J. Fisher, Fundamentals of solidification, Trans Tech
Publications, Uetikon-Zrich, 1992.
2. W.A. Tiller, K.A. Jackson, J.W. Rutter, and B. Chalmers, Acta Met., 1 (1953)
428.
3. J.A. Burton, R.C. Prim and W.P. Slichter, The Journal of Chemical Physics, 21
(1953) 1987.
4. J.W. Rutter and B. Chalmers, Can. J. Phys., 31 (1953) 15.
5. W.A. Tiller and B. Chalmers, Can. J. Phys., 34 (1956) 96.
6. D. Walton, W.A. Tiller, and J.W. Rutter, Trans. AIME, 203 (1955) 1023.
7. T.S. Plaskett and W.C Winegard, Can. J. Phys., 37 (1959) 1555.
8. G.S. Cole and W.C. Winegard, J. Inst. Met., 93 (1964) 322.
9. D.T.J. Hurle, Solid-State Electron., 3 (1961) 37.
10. W.W. Mullins and R.F. Sekerka, J. Appl. Phys., 34 (1963) 323.
11. W.W. Mullins and R.F. Sekerka, J. Appl. Phys., 35 (1964) 444.
12. R.F. Sekerka, J. Appl. Phys., 36 (1965) 264.
13. R.F. Sekerka, J. Crystal Growth, 3-4 (1968) 71.
14. R.M. Sharp and A. Hellawell, J. Crystal Growth, 6 (1970) 253.
15. R.M. Sharp and A. Hellawell, J. Crystal Growth, 6 (1970) 334.
16. A.Pocheau and M. Georgelin, J. Crystal Growth, 250 (2003) 100.
17. K.A. Jackson and J.D. Hunt, Acta Met., 13 (1963) 1212.
18. J.J. Kramer, G.F. Boling, and W.A. Tiller, Trans. AIME, 227 (1963) 364.
19. H. Biloni, G.F. Boling, and G.S. Cole, Trans. AIME, 233 (1965) 251.
20. H. Biloni, G.F. Boling, and H.A. Domian, Trans. AIME, 233 (1965) 1926.
21. B. Kauerauf, G. Zimmermann, S. Rex, M. Mathes, and F. Grote, J. Crystal
Growth, 223 (2001) 265.
22. H.D. Brody, M.C. Flemings, Trans. Met. Soc. AIME, 236 (1966) 615.
23. T.W. Clyne, W. Kurz, Metallurgical Transactions, 12A (1981) 965.
24. I. Ohnaka, Transactions ISIJ, 26 (1986) 1045.
25. C.Y. Wang, C. Beckermann, Material Science and Engineering, 171 (1993) 199.
26. H. Yoo, C.-J. Kim, Int. J. Heat Mass Transfer, 41 (1998) 4379.
27. V.R. Voller, Int. J. Heat Mass Transfer, 43 (2000) 2047.
28. C. Chang, D.Irving, and B. Kyle, AICHE-Journal, 13 (1967) 739.
61
Chapter III
29. R. Scholz, K. Wangnik, and J. Ulrich, J. Phys. D: Appl. Phys., 26 1993) B156.
30. R. Scholz, Ph.D. Thesis, University of Bremen, Germany, VDI-FortschrittsBerichte, No. 347, VDI-Verlag, Dsseldorf, Germany, 1993.
31. M. Stepanski and E. Schfer, In Couirs on:Theory and Application of Melt
Crystallization, 28th 30th September 2000, Martin Luther University HalleWitwnbwrg, Halle (Saale), Germany.
32. G.F. Arkenbout, Melt Crystallization Technology, Technomic Publishing
Company Inc., Lancaster, Basel, 1995.
33. J. Ulrich and J. Bierwirth, In Science and Technology of Crystal Growth, (J.P.
van der Erden and O.S.L. Bruinsma, eds.) Kluwer Academic Publ., Dordrecht,
The Netherlands, 1995, 245.
34. S.J. Jancic, In Industrial Crystallization 87, (J. Nnyvlt and S. Zacek, eds.),
Elsevier Science Publ., Amsterdam, The Netherlands, 1887, 57.
35. S. Rittner and R. Steiner, Chem.-Ing.-Techn., 57-2 (1985), 91.
36. K. Wangnik, Ph.D. Thesis, University of Bremen, Germany, VDI-FortschrittsBerichte, No. 355, VDI-Verlag, Dsseldorf, Germany, 1994.
62
CHAPTER IV
4. ZONE MELTING INVESTIGATION OF THE ALUMINIUM
SCRAP REFINING POTENTIAL IN FRACTIONAL
CRYSTALLIZATION1
Abstract
Recycling of aluminium is an effective way of enhancing the efficient use of energy and
limiting the burden on the environment. Purification based on fractional crystallization
appears to be a promising option for this purpose. This implies that the material is to be
processed in the semi-solid state; that is, at a temperature where a solid and a liquid
phase co-exist. In this chapter, it is shown that zone melting can serve as a perfect
indicative test for assessing the refining potential of fractional crystallization processes,
since they are both based on the same thermodynamic principle. The results of zonemelting experiments are then presented for some aluminium model alloys representative
of the scrap streams. Both chemical analyses and microscopic examination revealed that
the significant purification that takes place in it extends to more than the half of the
sample. The BPS- (developed by Burton, Prim and Slichter) and Pfan models are used
to interpret these results.
Part of this work was published by Sillekens W.H., Schade van Westrum J.A.F.M., Bruinsma O.S.L.,
Mehmetaj B., Nienoord M.; Proceedings of the Fourth International Symposium on Recycling of Metals
and Engineered Materials, Eds. Stewart Jr. D.L., Daley J.C., and Stephens R.L., TMS, Warrendale PA,
USA (2000): p. 963977.
Chapter IV
4.1. Introduction
64
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
This chapter focuses on the issue of refining aluminium scrap; that is, on treating scrap
of unfavourable chemical composition to obtain a batch containing lower levels of
alloying and contaminating elements. Zone melting is introduced as a relatively simple
technique to experimentally assess the refining potential and to predict that in fractional
crystallization. Experimental results for some aluminium model alloys will be presented
and compared with the analytical model developed by Burton, Prim and Slichter [10].
k=
Cs
Cu
(4.1)
where Cs and Cu are the concentrations of impurities in the solidified fraction and in the
original material respectively. This distribution coefficient, sometimes referred to as the
effective distribution coefficient, depends on the thermodynamic properties of the
system, but also on the process kinetics. Some sources prefer the use of the so-called
removal efficiency, , which is the complementary fraction of the distribution
coefficient k (=1k). This distribution coefficient in the equilibrium condition can be
65
Chapter IV
derived from the phase diagram of the system and is called the thermodynamic
distribution coefficient, k0.
G s
1 1
1 = 1 exp .
k k0
k d l
(4.2)
In this so-called BPS-model, G denotes the crystal growth rate and kd the mass transfer
rate. The latter represents the transfer of (accumulated) impurities away from the
growing crystal, resulting from diffusion, convection or some type of deliberate stirring.
The density of the solid and liquid are designated s and l respectively. The model
indicates that a low growth rate and a high mass transfer rate will result in a high crystal
purity.
A further indicator for the removal of secondary constituents from the primary
component is the purification ratio, R, which is defined as:
66
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
R=
Cs
Cl
(4.3)
k
Cs
= 1 (1 k ) * e l
Cu
(4.4)
67
Chapter IV
Table 4.1. Equilibrium distribution coefficients for the binary systems of the alloys
Element System
Thermodynamic
distribution coefficient
k0 ()
AlSi6
Si
eutectic
0.13
AlFe1.6Si0.4 Si
eutectic
0.11
Fe
eutectic
0.03
AlCu4Mg1.5 Cu
eutectic
0.17
Mg
eutectic
0.45
eutectic
0.13
eutectic
0.62
AlSi1.5Mn1.1 Si
Mn
Table 4.2. Calculated distribution coefficients for some binary and ternary aluminium
alloying systems[4].
Alloy
Element System
calculated distribution
coefficient k0 ()
AlSi6
Si
eutectic
0.13
AlFe1.2Si0.3
Si
eutectic
0.1
Fe
eutectic
0.03
AlCu4.4Mg1.5 Cu
eutectic
0.17
Mg
eutectic
0.45
Si
eutectic
0.17
Mn
eutectic
0.62
AlSi1.5Mn1
Finally, it should be noted that the conditions during the actual refining operation will
be principally different from the quasi-static conditions that apply for the phase
diagrams and theoretical calculations.
68
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
to obtain the same solidification conditions along the ingot. Cylindrical specimens with
diameter 10 mm and length 250 mm were machined from these blocks. These test bars
were taken from the central region of the ingots with their axis in line with the casting
direction.
Zone melting is a technique by which a specific redistribution of alloying elements can
be attained in a bar sample by repeated melting and solidification. A schematic of
sample and molten zone is presented in the insert of figure 4.2. Here, a sample of the
test material is pulled in horizontal direction through a ring-shaped heater, the position
of which is fixed. By this local heating, a confined part of the sample will melt. By
moving the sample through the heater, this molten zone travels through the sample from
head to end. This procedure means that the material is re-solidified at a controlled
crystal growth rate equal to the pulling rate of the sample.
Pulling direction
Heater
Sample
Molten zone
Stopper
Inert
gas
Induction
coil
Tube
Crucible with
sample
Guide
Pulling
wire
Weight
Linear
motor
In figure 4.2, a drawing is given of the device that has been used for the experiments.
The zone refiner consists of a quartz tube (600 mm long and 25 mm in diameter), a
water-cooled copper coil and a high-frequency inductor (power P12 kW, frequency
range 28 MHz), and an arrangement to move the tube through the coil (possible
pulling rates of 0.0420 mm/min). The alumina crucible, that contains the sample, is
placed in the tube. Design of the crucible is based on Pfann [9] and its geometry is such
that the samples are largely enclosed, preventing the molten zone from sagging. After
69
Chapter IV
sealing the tube with teflon stoppers, it is flushed with argon. A small flow of argon is
maintained during the experiment in order to prevent the formation of an aluminium
oxide layer around the aluminium melt. Figure 4.3 shows the zone melted sample. Only
the middle part of the cylindrical bars with 120 mm length was zone melted. The reason
for choosing relatively long ends is to keep the lateral heat transfer during the test more
or less constant and to minimise variations in the molten zone length, which at these
experiments was in the order of 20 mm. The AlSi6 alloy was pulled at a rate of 0.1-4
mm/min, while all other alloys were pulled at 0.1 mm/min pulling rate. The induced
heat, controlled by the power supply of the inductor, had to be adjusted to the alloy
composition such that full melting occurs in a restricted zone.
Figure 4.3. The zone-melting specimens, showing the position for chemical analyses,
and start and end of melting of the sample, dimensions are in mm.
After zone melting the chemical compositions at distinct locations within the sample
were measured and the three locations (head, middle, and end) are shown in figure 3.
These analyses were performed with optical emission spectroscopy (OES). The OESanalyses were done on cross-sections of the bars, since the variation in composition is
minimal perpendicular to the pulling direction. Concentrations were measured at the
head, the middle and the end location for each of the treated samples and are denoted
Ch, Cm and Ce respectively. Alloying contents in the untreated materials were also
measured and are designated as Cu.
The samples for microscopic analysis were cut and subsequently ground with a grinding
chapter, followed by polishing with a diamond powder (starting with 9 m and ending
up with 1 m powder size). After polishing, samples were etched by inserting in a
70
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
4.4. Results
71
Chapter IV
Table 4.3. Chemical composition of AlSi6 alloy before and after zone melting for
Element original
Head
Middle
End
Si
6.14
2.85
2.44
10.1
6.67
Si
6.14
4.83
4.33
9.54
16.67
Si
6.14
5.28
5.04
8.09
66.67
Si
6.14
6.04
6.15
7.52
Table 4.4. Chemical composition before and after zone melting (G=1.6710-6 m/s).
Alloy
Element original
Head
Middle
End
Fe
1.59
0.86
1.22
2.30
Si
0.37
0.12
0.12
0.78
Sb
0.06
0.03
0.05
0.09
Cu
4.42
1.53
3.21
Mg
1.49
0.85
1.21
2.35
Fe
0.03
0.01
0.02
0.10
Si
0.03
0.01
0.02
0.07
AlSi1.5Mn1.1 Si
1.51
0.45
0.52
3.76
1.09
0.83
0.82
1.30
AlCu4Mg1.5
Mn
10.3
72
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
rates, the cell size decreases from about 0.2 to 0.1 mm and the amount of silicon
between the cells increases when going from head position to end position. The latter
effect can be better seen in figure 4.5. In particular for the higher growth rates, the
microstructures indicate that while crystals are growing as primary aluminium
dendrites, melt is entrapped at the boundaries. This silicon-enriched melt subsequently
solidifies as an interdendritic eutectic (where the silicon appears as flakes).
In order to investigate the effect of the growth direction on the obtained crystal
morphology, samples were also cut and prepared in longitudinal direction. In figure 4.6,
the microstructures at the head position of the bar are shown for two growth rates and at
two levels of magnification. At the lower growth rate, the crystals have grown as
columns in the pulling direction, as can be deduced from the long dark lines
representing the silicon-rich intercellular areas. At the higher growth rate, this effect has
almost disappeared, although by comparing figures 4.5 and 4.6 it can be seen that the
aluminium-rich cells are still elongated in the growth direction by about a factor two.
The zone melting samples of the other alloying systems, grown at 1.67 m/s, were also
studied. The micrographs of the three ternary alloying systems, taken in longitudinal
direction in a position where 40 % of the samples had been grown, are shown in figure
4.8. In the Al-Si0.4-Fe1.6 alloy it is found that only at the beginning of the zone melted
sample a flat interface appears and that later eutectic growth takes place and that a sharp
transfer from flat to eutectic growth was observed at the position where about 40 per
cent of the sample was crystallized. The other two systems grow with flat interfaces and
no change in their structure was observed with optical microscopy.
73
Chapter IV
position
end
mid
head
1.66
16.67
66.67
Figure 4.4. Micrographs of the Al-Si6 alloy at different growth rates and at different positions in the bar after zone melting. Line segment (see
74
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
position
en
mid
head
1.66
16.67
66.67
Figure 4.5. Same micrographs as in figure 5, but with higher magnification. Line segment (see insert) corresponds with distance of 25 m.
75
Chapter IV
1.66
16.67
Figure 4.6. Micrographs of the Al-Si6 alloy at different growth rates taken in
Al-Fe1.6%-Si0.4%
Al-Cu4%-Mg1.5%
Al-Si1.5%-Mn1.1%
Figure 4.7. Optical micrographs of above cited alloying systems, growth rate 1.67
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
(Ch,g),
are given in table 4.5. Table 4.5 clearly shows that the
concentration of Si in the grains increases with increasing growth rate. However, this
increase is by far smaller than the one found from chemical analyses presented in table
4.3.
Figure 4.4 shows the line scan for AlSi6 alloy grown at pulling rate 16.67 m/s. It is
seen that the concentration of Si in the inner side of cell is stable and its average value,
given in table 4.5, is 1.02 wt %.
Table 4.5. The average values of the Si content in AlSi6, determined from EPMA line
scans for various growth rates (G) Analysis is done in the middle of the grains at the
head position (Ch,g).
G (10-6m/s) 1.67 16.67 66.67
Ch,g (wt %) 0.82 1.02
1.45
AlSi6.14
1.400
Wt %
1.200
1.000
0.800
0.600
0.400
30
35
40
45
50
55
60
Position in m
Figure 4.8. The Si content in EPMA line scan of AlSi6.14 alloy grown at 16.67 m/s.
77
Chapter IV
4.5. Discussion
CS in wt % Si
5
4
4
3
3
2
2
1
0
0
x/l
Figure 4.9. Silicon concentration after single-pass zone melting of AlSi6 alloying
system for: k0 = 0.13 (curve 1) and keff values of 0.2, 0.3, 0.4 and 0.5 (curves 2, 3, 4,
and 5, respectively).
Based on equation 4.4 the thermodynamic and experimental results of alloys studied in
this chapter are shown in figures 10-13. Pfann principle [9], which assumes no diffusion
in solid and complete diffusion in liquid, was used for calculating theoretical curves. It
78
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
is clearly seen in figure 4.10 that theory predicts that for AlSi6 only 1.5 molten zones
pass and than steady state is achieved and no purification takes place in the rest of the
sample. This is due to the fact that liquid reaches its eutectic composition and the
concentration of the metals formed is equal with the concentration of the starting
material that melts on the other side of the zone. The experimental purification of AlSi6
grown at growth rate (G) 1.67 m/s shows that it disagrees with theoretical predictions.
AlSi6
7
6
CS in wt %
5
4
3
2
1
0
0
x/l
Figure 4.10. Curves for single-pass zone melting, showing theoretical (lines) and
79
Chapter IV
AlCu4.4 alloy, which behaves as expected form theory, but with lower purification
efficiency, all the other alloys show non or very small concentration difference on
crystals along the sample. It is clearly shown that purification efficiency for most of the
alloying elements was significantly lower than theoretically predicted and therefore no
eutectic concentration was reached except for AlFe1.6 alloy. The best results compared
to theoretical predictions are achieved for Mg (figure 4.12).
AlSi0.4Fe1.6
CS in wt %
2
1.5
1
0.5
0
0
x/l
Figure 4.11. Curves for single-pass zone melting, showing theoretical (full line for Si
and dashed line for Fe) and experimental ( for Fe and for Si) solute concentration
in crystals versus distance in zone lengths (x/l) from beginning of charge, for
AlSi0.4Fe1.6 alloying systems.
CS in wt %
AlCu4.4Mg1.5
3.5
3
2.5
2
1.5
1
0.5
0
0
x/l
Figure 4.12. Curves for single-pass zone melting, showing theoretical (full line for Cu
and dashed line for Mg) and experimental ( for Mg and for Cu) solute
80
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
concentration in crystals versus distance in zone lengths (x/l) from beginning of charge,
for AlCu4.4Mg1.5 alloying systems.
Roughly 24 to 70% of the major alloying elements is removed from the material at the
head of the sample. However, the thermodynamic limits are, in most cases, and in
particular for iron, not obtained by far. A reason for this is that the tests were carried out
at relatively high growth rate. A further explanation may be that of interference between
alloying elements in case of the ternary alloys which may have a negative effect on
purification efficiency.
CS in wt %
AlSi1.5Mn1.1
1.2
1
0.8
0.6
0.4
0.2
0
0
x/l
Figure 4.13. Curves for single-pass zone melting, showing theoretical (full line for Si
and dashed line for Mn) and experimental ( for Mn and for Si) solute concentration
in crystals versus distance in zone lengths (x/l) from beginning of charge, for
AlSi1.5Mn1.1 alloying systems.
In conclusion one can say that experimental conditions applied to these alloying systems
were fare from equilibrium and therefore they disagree with theoretical predictions.
4.5.2. Comparison of results for AlSi6 alloy with non equilibrium BPS-model
The AlSi6 alloy was processed in the zone refiner at different pulling rates (i.e., growth
rates) in order to screen for the optimal operating conditions and to establish the validity
of the BPS-model. The results are plotted according to the BPS-model (equation 4.2) in
figure 4.14. As an approximation, the densities of solid and liquid aluminium are taken
equal. Clearly, the model, which predicts a straight line with a slope corresponding to
81
Chapter IV
the mass transfer coefficient kd and an intercept value k=k0 for G=0, is not correct when
taking into account the overall concentration of crystals. The slope of the curve changes
from an apparent mass transfer coefficient kd,1=810-7 m/s at low growth rates to
kd,2=310-5 m/s at higher values (say for G>510-6 m/s). Compared to this, a typical value
of 7.710-6 m/s was found for zone melting of phenol [13]. BPS model fits very well
when concentration of Si in grains is taken into account. It shows a strait line which
intercepts the y axes at value k=k0 for G=0.
ln[1/keff-1]
2
1
0
-1
-2
-3
-4
0
10
20
30
40
50
60
70
(alloy AlSi6), for average concentration in crystals () and for concentration in the
grains (). Stipulation with dashed line shows where according to prediction of BPSmodel, the line should touch the y axes at k0.
Figure 4.14 clearly indicates that the BPS-model cannot be used to accurately describe
the relation between the crystal growth rate and the distribution coefficient. Chang at al.
[14], who analyzed the crystallization of the benzene-hexadecane and used equation 4.2,
found that there were two regions of different linearity over the range of the growth
rates measured. After analysis of the interface morphology, they found out that each
region is associated with a distinguished crystal interface. The first region of linearity
was at a growth rate up to 1.2 cm per hour and corresponded to a cellular structure,
while for higher growth rates the interface was dendritic. As can be clearly seen from
the micrographic examination of the samples analyzed in this study, similar findings
were the case hear as well. Trapping of the solute in the intercellular or interdentritic
regions at the interface appears to be the main reason for such anomalous behavior of
82
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
this alloying system. BPS treated diluted solutions, assuming flat interface, which
clearly was not the case with our alloying system. Stepanski and Schfer [15] showed
that the degree of purification in layer growth process is strongly dependent on the
concentration of impurities in the system. They show the effect of the initial impurity
concentration in purification efficiency for the system Phenothiazine (PTZ) in acrylic
acid. It is clear from their obtained results that separation is significantly less effective
at higher initial impurity concentrations. This can be manly attributed to the quantitative
increase of the impurity inclusions due to the increased impurities, resulting in a crystal
layer with more entrapped and adhered liquid.
Since a proper model is essential for the design of a layer crystallization process, these
findings suggest that it is better to develop an empirical model based on the obtained
experimental results rather than relying on the theoretical models.
4.6. Conclusions
Zone melting is a relatively simple and useful experimental technique to assess the
ability to refine distinct alloying constituents from complicated alloy systems (like
industrial kinds of scrap aluminium). Regarding the obtained purification in the
experiments, results basically comply with the estimates based on the binary alloy phase
diagrams, although there are known distinct deviations from other experimentally
achieved results. Chemical analyses reveal that purity of the samples on the head and on
the middle positions are almost the same, which reveals that purification takes place in
more than the half of the original samples.
Microscopic examination of the AlSi6 alloy reveals that while aluminium cells grow,
the impurities that rich eutectic concentration are trapped between them. The
AlSi0.4Fe1.6 alloying system micrographs show a flat interface till about 40 per cent of
the sample is crystallized and later a eutectic growth resumes. The other two systems
studied in this chapter grow with flat interface and no change in their morphology was
observed by microscopy.
The BPS-model does not satisfactorily correspond with the experiments, and needs
extension to obtain a predictive tool for designing refining processes based on layerbased fractional crystallization. The microstructures of zone-refined samples indicate
83
Chapter IV
that an improved purification model should include the amount of eutectic trapped
between the cells.
4.7. Acknowledgements
Reported investigations were carried out as a part of the research project Sustainable
Technology for the Reclamation of High-grade Aluminium from Scrap, which is
undertaken in the context of the Dutch R&D stimulation programme on Economy,
Ecology and Technology (contract number EETK98026). Concerning the experiments,
the assistance of Mr C.M.S. van der Zwet Slotenmaker and Mr P.G. Commandeur at
Corus Group plc RD&T (casting of model alloys) and Mr C.C.J. Kaasschieter and Mr
M.H.F.M van Hout at TNO Institute of Industrial Technology (zone-melting tests) are
acknowledged.
84
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
4.8. References
1.
2.
3.
A.I. Kahveci and A. Unal, 2000, "Refining of a 5XXX Series Aluminum Alloy
Scrap by Alcoa Fractional Crystallization Process", Proceedings of the Fourth
International Symposium on Recycling of Metals and Engineered Materials, Ed.
D.L. Stewart Jr., J.C. Daley, and R.L. Stephens, Warrendale PA, USA: TMS, 979.
4.
5.
6.
7.
J.D. Esdaille and G.W. Walters, 1975, "Lead Purification by Crystallization and
Reflux", Proceedings of the Aus. I.M.M. conference, South Australia, 187.
8.
9.
W.G. Pfann, Zone Melting, John Wiley & Sons Inc., (2nd edition), New York, NY,
U.S.A., 1966.
10. J.A. Burton, R.C. Prim and W.P. Slichter, The Journal of Chemical Physics, 21
(1953) 1987.
11. P. Spencer, definition of thermochemical and thermophysical properties to
provide a database for the development of new light alloys, Proceedings of the
85
Chapter IV
final workshop on COST 507, Volumes 1, 2 and 3, Vaals, The Netherlands 9-11
March 1997, EUR 18499, European communities, 1998.
12. J.T. Armstrong, Quantitative elemental analysis of individual microparticles with
electron beam instruments. In Electron Probe Quantitation, Heinrich, K.F.J.,
and Newbury, D.E. (eds.), Plenum Press, (1991) 261-315.
13. P.J. Diepen, Cooling Crystallization of Organic Compounds: Processes, Purity
and Permeability, Ph.D. Thesis, Delft University of Technology, 1998.
14. C. Chang, D.Irving, and B. Kyle, AICHE-Journal, 13 (1967) 739.
15. M. Stepanski and E. Schfer, In Course on:Theory and Application of melt
Crystallization, 28th 30th September 2000, Martin Luther University HalleWitwnbwrg, Halle (Saale), Germany.
86
Zone melting investigation of the aluminium scrap refining potential in fractional crystallization
87
CHAPTER V
5. ALUMINIUM SCRAP REFINING WITH SOLID LAYER
FRACTIONAL CRYSTALLIZATION1
Abstract
Primary aluminium production is an energy intensive process. With an increase of the
aluminium use in car manufacturing a recycling technological process is imperative.
The energy use for recycling is only a fraction of the energy needed to produce primary
aluminium. In this chapter a fractional crystallization process for aluminium recycling
based on static layer growth will be described. Static layer crystallization experiments
showed that for AlSi7.2 alloying systems, typical growth rates of 1 to 2 m/s are
optimal for aluminium scrap purification. In order to achieve the preset purification
criteria for the pure product, one or more re-crystallization steps are required.
Depending on the impurity concentration, the crystallization temperature varies from
600 to 660 C. Microscopic examination of the crystal samples reveals that, while
crystals are growing as primary aluminium dendrites, impurity enriched melt is
entrapped at the boundaries. Line scans reveal that in the crystal cell area purification is
in line with thermodynamic predictions, while intercellular area is enriched with
impurities and intermetalics, which have an adverse effect in purification efficiency of
selected aluminium alloys.
Part of this work was published by Mehmetaj B., Bruinsma O.S.L., Kool W.H., Jansens P.J., and
Katgerman L.: Proceedings of the 15th International Symposium on Industrial Crystallization (ISIC-15),
Ed. Chianese, A., Sorrento, Italy, Chemical Engineering Transactions, Volume 1, 2002, p.879 884.
Chapter V
5.1. Introduction
The manufacturing industries are focusing more and more on processes that address an
increased environmental awareness on the part of the public and stricter regulations for
chemical disposal and pollutant emission by the governments. As a result new
environmental friendly and energy efficient technologies are emerging.
Such an
scrap recycling. The zone melting results of the alloys studied in this project were
published by Sillekens et al. [6], while a thermodynamic study of the alloys investigated
in this project is provided by Boender et al. [7]. Similar projects were carried out in
Japan [8] and in the USA [9].
In this chapter a layer based fractional crystallization process is described. A
background on application of fractional crystallization in metallurgy is given in section
5.2. In section 5.3 the experimental procedure and equipment description is provided.
The chemical analyses, EPMA line scans, and micrographic description of the studied
alloys grown at different growth rates are provided in section 5.4.
5.2. Theory
In evaluating the performance of a fractional crystallization process, an indicator called
the effective distribution coefficient, keff, is used. Relating to a mixture of the
components A (here: aluminium) and B (any other element), it is defined as:
k eff =
CS
.
CL
(5.1)
CS and CL are the concentrations of component B in the solidified fraction and in the
liquid material respectively. The effective distribution coefficient depends on the
thermodynamic properties of the system and on the process kinetics. The distribution
coefficient in the equilibrium condition can be derived from the phase diagram of the
concerned system and is called the thermodynamic distribution coefficient, k0.
A central question when considering the use of fractional crystallization for refining
aluminium is the extent to which secondary constituents can be purged from the base
metal. The theoretical limits for some binary Al alloys are summarised in table 5.1.
In table 5.2 calculated distribution coefficients and T values in case of studied binary
and ternary systems are presented based on the work of Boender et al. [7]. The
calculations were performed by using programs ChemSage and ThermoCalc using data
of the COST507 project [10]. Thermodynamic studies of the alloying systems studied in
this chapter revealed that for all but one alloying system, there is no difference between
the distribution coefficients of binary and ternary systems [7]. Only for AlSi1.5Mn1 a
significant decrease of silicon removal efficiency is found.
89
Chapter V
System
Si
eutectic
Fe
eutectic
Cu
eutectic
Mn
eutectic
Mg
eutectic
Maximum temperature
Thermodynamic
range T (C)
distribution coefficient k0 ()
83
0.13
5.5
0.03
112
0.17
2.5
0.62
210
0.45
Table 5.2. Calculated distribution coefficients for some ternary alloys and temperature
ranges of the semi-solid region for the critical alloying component. [7].
Alloy
calculated
distribution
coefficient k0 ()
AlSi6
Si
eutectic 40
0.13
AlFe1.2Si0.3
Si
eutectic
0.1
Fe
eutectic 3
0.03
eutectic 99
0.17
Mg
eutectic
0.45
Si
eutectic
0.17
Mn
eutectic 2
0.62
AlCu4.4Mg1.5 Cu
AlSi1.5Mn1
With reference to temperature control for such processes, the maximum temperature
range T is included in the table 5.2. It is taken as the temperature interval of the semisolid state in which the purification of the alloying system occurs (i.e., TlTE), where Tl
is liquid temperature of the given alloying system (in table 5.1 it is liquid temperature of
pure aluminium), and TE is eutectic temperature of a given binary system. The limiting
90
factor in ternary alloys is reaching the eutectic composition (table 5.2), after which no
further purification will take place during solidification.
All of the listed systems are eutectic. They are appropriate for refining by means of
fractional crystallization. The theoretical prospects are good for iron, silicon, and
copper, fair for magnesium, and moderate for manganese. Besides this, process control
will be critical for some cases: the available temperature interval is very small for iron
and manganese.
5.3. Experimental
The static layer crystallizer consists of a Nabertherm K4/10 tilting furnace and a
Eurotherm 2604 temperature controller. The furnace has a 4-liter crucible of ISOgraphite, with the maximum chamber temperature of 1273 K, designed for melting
aluminium. Thermocouples are made of NiCr-Ni, type K. A schematic drawing of the
static layer crystallizer is depicted in figure 5.1 and the equipment for this process is
given in figure 5.2.
The temperature controller is designed in such a way that during the melting stage the
temperature is controlled using the thermocouples attached on the outside wall of the
crucible, and during the crystallization stage the temperature is controlled with the
thermocouple inserted in the aluminium melt. Both stages are PID controlled, with
automatic switch from melting to crystallization PID mode. Eurotherm 2604
incorporates a self-correcting input circuit to preserve the instrument calibration
accuracy. Temperature configuration is achieved either via the front panel interface or
by using the Eurotherm iTools configuration package, which runs under the Windows
2000 operating systems.
The alloy of aluminium with 7.2 percent silicon (AlSi7 corresponding to its nominal
composition according to the EN AW system) is prepared by addition of a certain
amount of ALSi50 to the technically pure primary aluminium melt. The ternary
aluminium alloys described in this chapter were prepared by adding of a pre-calculated
amount of master alloys to the commercially pure aluminium melt. The cold finger is
inserted to the prepared alloy and the temperature is lowered to the melting temperature
of the alloy by vigorous air blowing to the cold finger. After the melt temperature is left
to stabilize for some time the cooling rate is adjusted by airflow to achieve the desired
91
Chapter V
growth rate. The growth rate is determined by dividing the crystal layer thickness by the
time it took to grow it.
Argon
5 cm
TC
10 cm
Crystal layer
Cooling
air
20 cm
Cold
finge
Crystallization
oven
Crystal
Temperature layer
controller
Figure 5.2. Equipment of static layer crystallizer for aluminium scrap refining.
The samples for chemical analyses were cut to include all the layers of the crystal. The
AlSi7.2, AlFe1.3Si0.3, AlCu4Mg1.4, and AlSi1.5Mn1 samples were dissolved in a
mixture of 10 ml water, 10 ml HCl, and 6 ml HF. A closed Teflon vessel was used to
92
dissolve the samples in a microwave oven. Chemical analyses of the above mentioned
alloys, except for Mn, were carried out using atomic absorption spectroscopic technique
(Perkin Elmer 2380(A)). The Mn analysis of the alloying system AlSi1.5Mn1 were
performed by Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES)
technique (Perkin Elmer plasma 40). The overall error of chemical analyses is 3 %.
The samples for microscopic analyses were cut and subsequently grinded with grinding
chapter fallowed by polishing with diamond powder (starting with 9 m and ending up
with 1 m powder size). The micrographs were taken in a Neophot 30 inverted stage
optical microscope equipped with a Polaroid digital camera model DMC1 with DC
input of 6 V, 3 A, connected to a PC.
The Electron Probe Micro Analaysis (EPMA) were performed with a JEOL JXA 8900R
microprobe using an electron beam with energy of 15 keV and beam current of 20 nA
employing Wavelength Dispersive Spectrometry (WDS). The composition at each
analysis location of the sample was determined using the X-ray intensities of the
constituent elements after background correction relative to the corresponding
intensities of reference materials. The points of analysis were located along a line with
increments of 4 m and involved the elements Al, Cu, Mg, Fe, Mn, and Si. The thus
obtained intensity ratios were processed with a matrix correction program CITZAF
[11]. The counting errors and detection limits are given in table 5.3 for each spectral
line used in the quantitative analysis.
Table 5.3. The counting error (one sigma) and detection limits for the measured X-ray
lines.
Element X-ray Line
(p.p.m.)
Al
K (1.487)
0.05
10
Cu
K (8.046)
0.05
51
Mg
K (1.254)
0.02
11
Mn
K (5.898)
0.05
37
Si
K (1.740)
0.02
16
Fe
K (6.403)
0.06
40
93
Chapter V
concentrations of the aluminium scrap streams in the Netherlands. The obtained crystal
layers were analysed with chemical analyses, EPMA line scans, and their micrographic
examination was carried out.
5.4.1 Chemical analyses
The results of the chemical analyses of the layer growth experiments for AlSi7.2 alloy,
carried out at different growth rates (G), presented as overall distribution coefficient
(koa= Ccr/C0), which is defined as a concentration of the impurities in the crystal layer
(Ccr) versus impurity concentration of the starting alloy (C0), and average distribution
coefficient (kcr, equation 5.2), are shown in figure 5.3. The koa expression gives the
overall partition of impurities and does not take into account the changing concentration
of impurities in the liquid. In order to account for that, we consider an initial mass of
liquid m0, with impurity concentration C0, being partially crystallized to give a reduced
mass of liquid mr, with an impurity concentration Cr. Since effective distribution
coefficient (keff) can be considered as a constant, the mass and component balance for a
differential balance of crystallized mass can be integrated to give an average distribution
coefficient kcr [1]:
k cr = 1 +
log(C r / C 0 )
log(mr / m0 )
(5.2)
94
AlSi7.2
0.65
0.6
0.55
0.5
0.45
0.4
0.9
1.1
1.3
1.5
1.7
1.9
2.1
G in m/s
Figure 5.3. Overall distribution coefficients (koa, ) and average distribution coefficient
(kcr, ) versus growth rate (G) for AlSi7.2 alloy.
In practice the main advantage of the model described by equation 5.2 is that it gives an
expression of the purification achievable in a single stage depending on the amount of
the original liquid remaining un-crystallized. It also gives a more accurate value to
compare with theoretical predictions as given by thermodynamic partition coefficient
(k0).
Considering that about 40 % of the original material was crystallized the overall
impurity removal efficiency ( = 1 koa) of the layer growth process for AlSi7.2
alloying system is significant and varies from 37 % for the G = 2 m/s to50 % For G=1
m/s. In order to compare with theoretical possibilities we take the average distribution
coefficient for G = 1 m/s, which shows a removal efficiency of 57 % and compare it
with the thermodynamic possibilities. The impurity removal efficiency for this growth
rate is significantly smaller than thermodynamically possible (87 %). The reason for this
is that the tests were carried out at finite growth rates. Chemical analyses of the alloy
grown at 16 m/s revealed that no purification happened.
The impurity removal efficiency for the alloy grown at 1.7 m/s by zone melting
technique [6] is between the one observed for the same alloy grown at 1 m/s and 1.7
m/s by layer growth process. However, comparing the purification efficiency of the
layer growth process and zone melting technique for the 1.7 m/s growth rates, the
purification efficiency of zone melting is about 5 % better. There are two main possible
95
Chapter V
explanations for that: (i) incorporated impurities due to mother liquor adhering at the
end of the crystallization process were not removed effectively (no additional washing
was applied); and (ii) Zone melting experiments were carried out by using induction
heating, which besides heating provides a significant mixing effect.
Results of the chemical analyses of the layer growth experiments of the AlFe1.3Si0.3,
AlCu4Mg1.4, and AlSi1.5Mn1, presented in weight percent, overall partition
coefficient, and average partition coefficient, are shown in table 5.4.
Table 5.4. Chemical composition of the samples before and after layer growth.
Alloy
AlSi0.3Fe1.2
AlCu4Mg1.5
AlSi1.5Mn1
Element
Untreated
Crystals
C0/w%
Ccr/w%
koa
kcr
Si
0.33
0.18
0.54
0.31
Fe
1.33
0.71
0.53
0.23
Cu
4.02
2.68
0.66
0.35
Mg
1.39
0.74
0.53
0.33
Si
1.46
0.75
0.51
0.48
Mn
1.02
0.86
0.84
0.7
The applied growth rate (G) for all the alloying systems presented in table 5.2 was G=2
m/s. The impurity removal efficiency is significantly smaller than thermodynamically
possible. The overall purification efficiency of the iron and silicon on AlFe1.3Si0.3 are
almost the same, which based on the binary phase diagram is not expected. The
purification efficiency of iron was negatively effected by presence of silicon, while the
purification efficiency of silicon was positively effected by presence of iron if a
purification efficiency of the binary alloy is taken into account. However, from the
average distribution coefficient it is shown that partition of iron is higher than that of
silicon. In this case 40 % of the original material was crystallized and concentration of
iron in remaining liquid was almost 2 wt%, which is higher than eutectic concentration
of AlFe binary alloy (1.9 wt %). This shows that no more than 40 % of the original
material of this system can be crystallized in layer crystallization process at this growth
rate.
A high purification efficiency of magnesium on AlCu4Mg1.4 alloying system was
observed. This can be attributed to two possible effects: (i) evaporation of the
96
97
Chapter V
the highest growth rate. At the highest growth rate, this effect is no longer visible. The
cell size decreases with increasing growth rate. For the two middle growth rates, the cell
size decreases in going from about 0.4 to 0.3 mm and the amount of silicon between the
cells increases. Looking in comparison with the zone melting results the cells are twice
as large, for the same growth rate. However the amount of the inclusions between the
cells is much higher for the layer grown samples.
a)
b)
c)
d)
Figure 5.4. Optical micrographs for AlSi7 alloy from layer growth experiments: a)
grown at 1m/s seen in the direction of growth; b), c) and d) cross-sections of the
samples grown at respectively 1.6, 2, and 16 m/s.
The microstructure of the first three micrographs reveals that impurities are mainly
concentrated between the cells. Due to the low temperature gradients for the aluminium
layer, these impurities are entrapped as a liquid metal between the cells at the time when
crystallization stage is finished and may be removed by sweating or washing postpurification processes. If all the intercellular material would be removed by sweating,
98
AlFe1.3Si0.3
AlCu4Mg1.4
AlSi1.5Mn1
Figure 5.5. Optical Micrographs for layer grown samples of the above cited alloying
systems, growth rate 2 m/s (longitudinal sections on the left; cross-sections on the
right). The line in the picture denotes 0.1 mm.
99
Chapter V
100
AlSi7.2
4
Si (wt-%)
3
2
1
0
0
100
200
300
400
500
600
700
position in m
Figure 5.6. The Si content in EPMA line scans of AlSi7.2 alloy cells grown at growth
rate 2 m/s. The measurement was taken at a marked line in figure above the line scan.
The line scans for both Si and Fe of AlFe1.3Si0.3 alloying system presented in figure
5.7 are within the thermodynamic predictions. The significant oscillations of the
concentration lines, particularly for Fe, are due to the counting error for these elements
(see table 5.3). The concentration of Fe in crystals is lower than the counting error and
therefore shows an indication of concentration rather than an accurate concentration
line. In contrast to line scans, the overall purification efficiency of the iron and silicon
on AlFe1.3Si0.3, as shown in table 5.4, are almost the same. The Purification efficiency
of iron is negatively affected by intermetalic compound formations that were found out
to be present on the intercellular regions and will be discussed in the next section.
101
Chapter V
AlFe1.3Si0.3
wt %
Iron
Silicium
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0
100
200
300
400
500
600
Position in m
700
800
900
1000
Figure 5.7. The Si and Fe content in EPMA line scans of AlFe1.3Si0.3 alloy cells grown
at growth rate 2 m/s. The measurement was taken at a marked line in figure above the
line scan.
In figure 5.8 the line scans for Mg and Cu of the AlCu4Mg1.4 alloying system are
presented. The purification efficiency of Cu in crystals is as expected from
thermodynamic predictions. It is notable the high purification efficiency of Mg on
AlCu4Mg1.4 alloying system. Similar findings for this alloy were reported for overall
purification efficiency presented in section 5.4.1 and the same reasoning for increased
purification efficiency could be applied.
102
AlCu4Mg1.4
wt %
Copper
Magnesium
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
500
400
300
200
Position in m
100
Figure 5.8. The Cu and Mg content in EPMA line scans of AlCu4Mg1.4 alloy cells
grown at growth rate 2 m/s. The measurement was taken at a marked line in figure
above the line scan.
A remarkable result is that Mn concentration on cells is almost the same like overall
concentration on the crystals. This is due to the higher content of crystal cells in
comparison with intercellular material and lower concentration of Mn in the
intercellular area compared to other alloying elements studied in this work. A peak on
Si content is present in figure 5.9, which could be due to the invisible grain boundary
presence.
103
Chapter V
AlSi1.5Mn1
wt %
Manganese
Silicium
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0
50
100
150
200
250
Position in m
Figure 5.9. The Si and Mn content in EPMA line scans of AlSi1.5Mn1 alloy cells grown
at growth rate 2 m/s. The measurement was taken at a marked line in figure above the
line scan.
In conclusion the line scan EPMA analyses reveal that significant purification in cells is
achieved. However, if an effective removal of the impurities from crystals is to be
achieved effective sweating process has to be performed in order to remove the impurity
rich intercellular material.
5.4.4. Inter-metallic compounds
Thermodynamic studies done for the alloying systems studied in this project, which
were carried out by our project partners at Corus and TNO [7], revealed that for ternary
alloys several intermetalic compound formations are possible. Intermetalic compounds
have an adverse effect on purification efficiency of the alloying systems. In this study
104
-AlFeSi. However, from our extensive measurements they were not found to be present
on our samples. Formation of the intermetalic compound Al13Fe4 is the main cause for
decreasing the purification efficiency for iron for this alloying system and therefore
describes the reason of departure from thermodynamically predicted purification of iron
from this alloying system. In principle the intercellular impurities could be removed by
105
Chapter V
post purification processes such as sweating, which would have very positive effect on
increasing purification efficiency for this alloying system.
Table 5.5. The chemical composition of the intermetalic compounds found in the
intercellular region of AlFe1.3Si0.3 alloying system.
Al wt %
Si wt %
Fe wt % Compound
Comment
60.04
0.88
39.07
Al13Fe4
Calculated [7]
60.06
1.23
38.71
Al13Fe4
69.99
0.77
29.24
83.65
0.58
15.51
Cu
Mg
Compound
Comment
Calculated [7]
38.1
44.8
17.1
AlCuMg-S
62.1
27.5
10.4
46.1
53.9
AlCu-
Calculated [7]
49.9
49.1
1.0
90 % AlCu-
The intermetalics appearing in intercellular space of the AlSi1.5Mn1 were analysed and
results are presented in table 5.7. Based on calculations the presence of Al6Mn
intermetalic compound is expected and AlSi eutectic. However, Due to contamination
of the starting material the presence of Fe is measured, which complicates calculation of
the right percentage of intermetalics detected from our measurements.
106
Table 5.7. The chemical composition of the intermetalic compounds found in the
intercellular region of AlCu4Mg1.4 alloying system.
Al
Si
Fe
Mn
Compound
Comment
25.3
Al6Mn
Calculated [7]
5.5. Conclusions
Refining aluminium by means of layer growth fractional crystallization seems a
technically viable option in developing a sustainable technology for recycling of the
high-grade aluminium from scrap.
Chemical analyses reveal that significant purification is achieved. However, if an
effective removal of the alloying elements from aluminium scrap is to be achieved, a
repetition of the crystallization process has to be performed.
Microscopic examinations of the crystal samples, cut at different directions of crystal
growth, reveal that while crystals are growing as primary aluminium cells, an impurity
enriched melt is entrapped in the intercellular space.
Line scans reveal that concentration of impurities within cells is in line with
thermodynamic predictions while impurity rich area is detected in the intercellular
space.
In the intercellular space a number of intermetalic compounds were found to be present.
This leads to decreased purification efficiency of the layer grown crystals.
Application of the post-purification processes, such as washing or sweating, may
increase the purification efficiency significantly.
5.6. Acknowledgements
This research is carried out as a part of the research project Sustainable Technology for
the Reclamation of High-grade Aluminium from Scrap that is undertaken in the context
of the Dutch R&D stimulation programme on Economy, Ecology and Technology
(contract number EETK98026). For the experimental work, the assistance of Mr J.J.
Jansen of the Material Sciences Department is gratefully acknowledged.
107
Chapter V
5.7. References
1. Arkenbout G.J., 1995, Melt Crystallization Technology. Lancaster PA USA:
Technomic Publishing Company Inc.
2. Jansens P.J. & Matsuoka M., 2000, Melt Crystallization, in: Encyclopedia of
Separation Science, Eds. Wilson I.D., Adlard E.R., Cook M., and Poole C.F.,
Academic Press, San Diego, New York London, Sidney, Tokyo, 966-975.
3. Sillekens W.H., Verdoes D., and Schade van Westrum J.A.F.M., 1999,
"Refining Aluminium Scrap by means of Fractional Crystallization: Technical
Feasibility", Proceedings of the Fourth ASM International Conference and
Exhibition on the Recycling of Metals, ASM Europe, 105-114.
4. Esdaille J.D. and Walters G.W., 1975, "Lead Purification by Crystallization and
Reflux", Proceedings of the Aus. I.M.M. conference, South Australia, 187-195.
5. Pfann W.G., 1966, Zone Melting. John Wiley & Sons Inc., New York, NY,
U.S.A., 2nd edition.
6. Sillekens W.H., Schade van Westrum J.A.F.M., Bruinsma O.S.L., Mehmetaj B.,
Nienoord M.; 2000, "Refining Aluminium Scrap by means of Fractional
Crystallization: Basic Experimental Investigations"; Proceedings of the Fourth
International Symposium on Recycling of Metals and Engineered Materials, Ed.
D.L. Stewart Jr., J.C. Daley, and R.L. Stephens, Warrendale PA, USA, 963977.
7. Boender W., Waringa C.J., Krielaart G.P., Folkertsma A. and Verdoes D., 2002,
Refining Aluminium Scrap With Fractional Crystallization: Assessing Its
Feasibility With Thermodynamics, Proceedings of the TMS Annual Meeting &
Exhibition, Light Metals 2002. Ed.Schneider W., Seattle, Washington, USA,
TMS,1061-1068.
8. Sotome T. and Ohtaki M., 1998, "Application of Fractional Crystallization for
Refining of Molten Aluminium Scrap", Conference Proceedings of ICAA 6, vol.
1, Ed. T. Sato et al. Tokyo, Japan: The Japan Institute of Light Metals, 351-356.
9. Kahveci A.I. and Unal A., 2000, "Refining of a 5XXX Series Aluminum Alloy
Scrap by Alcoa Fractional Crystallization Process", Proceedings of the Fourth
International Symposium on Recycling of Metals and Engineered Materials, Ed.
D.L. Stewart Jr., J.C. Daley, and R.L. Stephens, Warrendale PA, USA: TMS,
979-991.
108
109
Chapter V
110
CHAPTER VI
6. FRACTIONAL LAYER CRYSTALLIZATION AND SWEATING
APPLIED TO ALUMINIUM SCRAP REFINING
Abstract
In this chapter1 a fractional crystallization process for aluminium recycling based on
static layer growth will be described. Fractional layer crystallization is a separation
process by which the fractional separation is effected by directional crystallization from
the melt. As such it is an alternative to suspension crystallization. The main emphasis
will be paid to sweating as a post-purification process. Sweating is based on the partial
melting of the crystal layer by gentle heating close to the melting point of the pure
substance. The temperature during the sweating process is higher than in the
crystallization stage. The sweating process removes only the impurities entrapped in the
intercellular space, while impurities in the cells are not affected. Although during the
sweating process about 10 - 20 % of the crystals are removed, sweating is effective and
avoids an additional crystallization step in a fraction of the time.
Part of this work was published by Mehmetaj B., Kool W.H., Jansens P.J., and Katgerman L.:
Proceedings of the 7th International Conference on Semi-solid Processing of Alloys and Compounds (7th
S2P), Eds. Tsutsui Y., Kiuchi M., and Ichikawa K., National Institute of Advanced Sci. & Techn.- Japan
Society for Technology of Plasticity, Tsukuba, Japan, 24-28 September 2002, p. 611 - 616.
Chapter VI
6.1. Introduction
A common practice in aluminium recycling is the re-melting of aluminium scrap and
reuse for secondary alloys. In cases, where the alloying contents are higher than
according to the specification for a specific alloy, the molten scrap is diluted with
primary aluminium till the required chemical composition is achieved. With the
continuous increase of the aluminium consumption the amount of aluminium scrap that
needs to be recycled cannot be processed in this way in future. Therefore, closed loop
recycling is an imperative.
Melt crystallization is a separation process by which the fractional separation is effected
by directional crystallization from the melt. Fractional crystallization is applied for both
ultra purification of organic compounds [1,2] and metals [3]. Historically, the first
application of fractional crystallization for metals was the invention of the Pattison
process for extraction of silver from lead in 1829 [4]. It was followed by the
development of the zone melting technique in the 1950s by Pfann [5].
Fractional Crystallization
Suspension
Crystallization
Batch
Layer
Crystallization
Continuous
Static
Dynamic
Postpurification
Sweating
Diffusion
washing
Rinsing
112
wall and at the end of the experiment are easily separated from the remaining liquid
(mother liquor), which is enriched with the impurities.
After growth of the layer sweating is used as a post-purification process. Sweating is
based on partial melting of the crystal layer by gentle heating close to the melting point
of the pure substance. Then, the impure melt adhered to the surface of the crystal layer
or present in the intercellular regions of the crystal layer drains off under the influence
of gravity. In this study both layer growth and the effect of sweating on the further
purification are described. To that purpose layers are grown onto a cold tube (cold
finger).
6.2 Experimental
A binary aluminium alloy with 6 wt % silicon is prepared by addition of an amount of
AlSi50 master alloy to the commercially pure aluminium melt. The ternary
AlCu4Mg1.5 alloy is prepared by adding of a pre-calculated amount of master alloys of
copper and magnesium to the commercially pure aluminium melt. The cold finger is
inserted in the melt and the temperature is lowered to the melting temperature of the
alloy by vigorous air blowing to the inside walls of the cold finger. The temperature is
stabilized for short period and then the cooling rate is adjusted by the airflow to achieve
the desired growth rate. The growth rate is determined by dividing the crystal layer
thickness by the growth time. After the crystallization stage the cold finger is removed
from the oven and is put in a preheated oven, which is used for the sweating stage.
Sweating is done at a constant temperature for about 10 15 minutes. The sweating
stage observed initially is called sweating A, and after some shaking of the cold finger is
called sweating B. Sweating B is carried out in order to remove the sweated liquid that
is entrapped due to oxide layer that is formed in the aluminium crystal layer during the
movement of cold finger with crystal layer from crystallization to sweating oven. The
weight of the sweated material is compared with the original weight of the grown layer
to obtain the percentage of the material lost as a result of sweating. The temperature
profile in the experiment for AlSi6 alloy is shown in figure 6.2. Similar curve pattern,
but with different temperature values, is true for AlCu4Mg1.4 alloy. The temperature
curve for the AlFe1.3Si0.3 and AlSi1.5Mn1 was much narrower. A more detailed
description of the experimental set-up is provided in [6] and is depicted in figure 6.3.
113
Chapter VI
Temperature/C
680
660
IV
640
II
III
620
600
0
0.5
1.5
2.5
3.5
4.5
Time/h
Figure 6.2. Temperature profile during the experiment with AlSi6: I: cooling interval;
II: layer growth; III: removal of the cold finger to preheated sweating oven; and IV:
sweating of crystals at constant temperature.
Temperature
controller
Sweating
oven
Crystallization
oven
Cold
finger
Figure 6.3. Equipment for Solid layer crystallization and sweating process.
The samples for chemical analyses were cut to include all the layers of the crystal. The
samples were dissolved in a mixture of 10 ml water, 10 ml HCl, and 6 ml HF. A closed
Teflon vessel was used to dissolve the samples in a microwave oven. Chemical analyses
of the alloys studied in this chapter were carried out using atomic absorption
spectroscopic technique (Perkin Elmer 2380(A)). Each sample was measured 3 times
and the results given in this chapter are average of these three measurements. The
overall error of chemical analyses is 3 %.
114
Optical microscopy of the grown layers before and after sweating were taken in a
Neophot 30 inverted stage optical microscope equipped with a Polaroid digital camera
model DMC1 with DC input of 6 V, 3 A, connected to a PC.
The Electron Probe Micro Analaysis (EPMA) were performed with a JEOL JXA 8900R
microprobe using an electron beam with energy of 15 keV and beam current of 20 nA
employing Wavelength Dispersive Spectrometry (WDS). The composition at each
location of the sample was determined after background correction. The obtained
intensity ratios were processed with a matrix correction program CITZAF [7].
6.3. Results
In order to establish the potential of refining aluminium scrap by means of fractional
crystallization and sweating processes, AlSi6.2, AlFe1.3Si0.3, AlCu4Mg1.4, and
AlSi1.5Mn1 were selected. The first alloy was included as it represents a rather basic
and comprehensible model system. Other alloying systems were chosen as they
represent concentrations of the aluminium scrap streams in the Netherlands. The
obtained crystal layers of AlSi6.2 and AlCu4Mg1.4, before and after sweating were
analysed with chemical analyses, EPMA line scans, and their micrographic examination
was carried out. Sweating of the other two alloying systems; AlFe1.3Si0.3 and
AlSi1.5Mn1 was not successful. The main reasons for failure of sweating of these
alloying systems were: non ideal experimental conditions (sweating was carried out in
another oven and not in the same oven where crystallization took place), and the
temperature window for sweating of these alloying systems is to narrow.
115
Chapter VI
The results of the chemical analyses of the layers grown before and after sweating for
AlSi6.2 alloying system are shown in table 6.1. Si concentration in the starting alloy
(C0) was 6.2 wt%. Si concentration in the grown layer (Ccr) is much lower and reduces
further after sweating. Si concentration in the sweated liquid (Csw) is high and even
higher than in the starting alloy. The overall distribution coefficient (koa = Ccr/C0) is also
given in table 6.1. It becomes lower during layer growth and sweating and it clearly
indicates that during layer growth purification has taken place and during the sweating
stage additional purification occurred.
Table 6.1: Chemical composition of the samples of ALSi6 alloy before layer growth,
after layer growth, and after two sweating stages A and B.
Starting
After
After
After
alloy
Layer Growth
Sweating A
Sweating B
6.20
3.90
3.40
3.08
6.70
6.80
koa
0.63
0.55
0.50
Layer growth rate under the conditions applied for alloying system AlCu4Mg1.4 was 2
m/s. The results of the chemical analyses of the layers grown before and after sweating
are shown in table 6.2. The starting alloy was AlCu4Mg1.4. Although from mass
balance of starting master alloys it was expected that magnesium content of starting
alloy would be 1.51 per cent, due to its loss as a result of evaporation and dross
formation its content is significantly lower (see table 6.2). Copper and magnesium
concentrations in the grown layer (Ccr) are much lower and reduces further after
sweating, while their concentration in the sweated liquid (Csw) is high and even higher
than in the starting alloy. The overall distribution coefficient (koa) is also given in table
6.2. It becomes lower during layer growth and sweating and it clearly indicates that
during layer growth purification has taken place and during the sweating stage
additional purification occurred for this alloying system.
The impurity removal efficiency during the layer growth is significant for the applied
growth rate of 2 m/s (see table 6.2). For comparison of impurity removal efficiency
with theoretical possibilities for all alloying systems reader is referred to chapter 5. The
purification effect of the sweating process concerning copper is similar to that for AlSi6
116
AlCu4Mg1.5
Element
Untreated
Crystals
After Sweating
C0/w%
Ccr/w%
Ccrsw (wt %)
Csw (wt %)
Cu
3.90
2.83
2.20
4.13
Mg
1.39
0.97
0.83
1.37
koa
Cu
0.72
0.56
koa
Mg
0.70
0.60
The purification efficiency of both crystal growth stage and sweating stage are relatively
high. However, if a further purification of the aluminium scrap has to be achieved lower
growth rates as well as repetition of the process have to be applied, in which the purified
layer is subjected to a subsequent process of layer growth and sweating, at a higher
temperature, which corresponds with the grade of the purification reached in the first
step.
6.3.2. Micrographs
In figure 6.4 cross-sections of the grown crystals of AlSi6 alloy before and after
sweating stages A and B are shown. These micrographs of the crystals clearly show that
with increasing amount of sweating the intercellular eutectic material is removed and
that the porosity of the crystals is increased significantly, i.e. the intercellular eutectic
material is removed.
117
Chapter VI
a)
b)
c)
Figure 6.4. Optical micrographs for the AlSi6 alloy from layer growth experiments: a)
before sweating; b) after sweating stage A; and c) after sweating stage B. The line in
the picture denotes 0.1 mm.
In figure 6.5 cross-sections of the grown crystals of AlCu4Mg1.4 before and after
sweating stage A are shown. These micrographs of the crystals clearly show that with
increasing amount of sweating the intercellular eutectic material is significantly
removed and that the porosity of the crystals is increased significantly.
118
a)
b)
Figure 6.5. Optical micrographs for the AlCu4.4Mg1.4 alloy from layer growth
experiments: a) before sweating; b) after sweating stage A. The line in the picture
denotes 0.1 mm.
The samples before and after sweating were analyzed using Electron Probe MicroAnalysis (EPMA) in order to determine the content of Si, Cu, and Mg in large grains
along a line and especially near the grain boundary. The points of analysis were located
along the line with increments of 4 m and involved the elements Al, Si, Cu, and Mg.
wt. % Si
4
3
2
1
0
0
200
400
600
Position (m)
Figure 6.6. The Si content in EPMA line scan of AlSi6, grown at growth rate of 2 m/s,
before sweating, and position of the line scan (above).
119
Chapter VI
Si (wt- %)
3
2
1
0
0
200
400
600
800
1000
Position (m)
Figure 6.7. The Si content in EPMA line scan of AlSi6, grown at growth rate of 2 m/s,
after sweating, and position of line scan (above).
The Si composition profiles, given in figure 6.6, show that in a region of about 50 m
near the grain boundary the Si content gradually increases. In the middle of the large
grains the Si content is about 1 wt % and does not change after sweating (see figure
6.7). This shows that the sweating process has an effect on the removal of the
intercellular eutectic material, while at the same time the Si content in the grains is not
affected.
The Cu and Mg compositions profiles, given in figure 6.8, show that in the intercellular
region a highly impure material is present. The concentration of Cu is significantly
higher than that of Mg, which is shown on secondary axes. This explains the reason for
lower purification efficiency of Mg by applying sweating described in section 6.3.1.
120
wt %
10
60
50
40
30
20
10
0
8
Aluminium
Copper
Magnesium
6
4
2
0
1
Postion (m)
Figure 6.8. The Al, Cu and Mg content in EPMA line scans of AlCu4Mg1.4 alloy
intercellular material of crystals grown at growth rate of 2 m/s, and position of line
scan (above, line where white spots appear).
Copper
Magnesium
2,0
wt %
1,5
1,0
0,5
0,0
0
100
200
300
400
500
Position ( m )
Figure 6.9. The Cu and Mg content in EPMA line scans of AlCu4Mg1.4 alloy cells after
sweating, grown at growth rate 2 m/s, and position of line scan (above).
121
Chapter VI
In the middle of the large grains, as shown in figure 6.9, the Cu and Mg content is the
same with that reported in chapter V for the un-sweated sample, and does not change
after sweating. This shows that the sweating process has an effect on the removal of the
intercellular eutectic material, while at the same time the impurity content in the grains
is not affected. This is in line with the findings for AlSi6 alloy explained above.
6.4. Discussion
The impurity removal efficiency ( = 1 keff) for the AlSi6 alloy during the layer
growth is significant (37 %) for the applied growth rate of 2 m/s. The impurity
removal efficiency is significantly smaller than that thermodynamically possible (87 %).
The reason is that the tests were carried out at a relatively high growth rate. Microscopic
examination shows that while cells are grown impurities are enclosed in the intercellular
space as eutectic. The Si concentration in the cell interiors is close to that expected from
the phase diagram at the layer growth temperature (about 0.8 wt%). Similar results were
achieved also with the ternary alloy AlCu4.4Mg1.5.
Attempts to perform sweating to other two alloying system studied in this project,
AlFe1.6Si0.3 and AlMn1Si1.5 were not successful. The reasons are as follows: the
temperature frame of melting the crystal mass were very narrow and the temperature
control in the sweating oven was not accurate enough to perform a gentle heating in
order for the intercellular impurities to drain off. Hence, the intercellular impurities
were melted together with the pure crystals form cells. One has to mention that sweating
of these alloying systems is possible if better temperature control is applied, i.e.
sweating of crystals in the same oven where they grow.
From the micrographs of the studied alloys it is found that approximately 10 % of the
crystallized layer consists of eutectic intercellular material, which suggests that by
sweating maximum 10 % of the layer could be effectively removed. However, in
practice more material is removed by sweating. The reason is that besides entrapped
eutectic a significant amount of the crystals is melted, which opens the way for the
impurities to drain off, resulting in an amount of material sweated out, which is about
122
twice as high as the amount of intercellular eutectic. However, due to some locked
impurity pockets not all the impurities could be removed by sweating process.
Sweating as a post-purification process reaches the same purity of crystals as was
observed during layer growth (without sweating) with 1 m/s growth rate [6].
Therefore, in order to obtain certain purification, one may grow a layer with the
required purity. The alternative strategy is to grow such a layer at faster growth rate,
which involves a higher impurity level, and to apply sweating which reduces the
impurity level to that required. It is found that the later strategy is better, since only half
of the time is needed. The composition of the sweated liquid is not significantly higher
than the original alloy and is returned in the process.
For an effective purification three strategies may be considered. The first is only layer
crystallization under static conditions; the second is a dynamic crystallization process,
which as a result of mixing improves mass and energy transfer increasing this way
purification efficiency; and the third is layer crystallization at a higher growth rat with
static or dynamic layer crystallization process, followed by a post-purification step such
as sweating. The later saves costs and energy.
The morphology of the crystal layers are affected by the applied growth rate and get
more stable with decreasing growth rate [6], resulting in a decreased amount of
intercellular eutectic material and an increased cell size, which leads to less liquid
pockets in the crystal layer and more open channels. As a result of this, it is anticipated
that by applying lower growth rates the effect of sweating in purification of the crystal
layer would increase and the amount of the crystal loss would decrease significantly. A
significant purification of the crystal layer is achieved in the process of layer growth and
sweating. However, if a further purification of the aluminium scrap has to be achieved
lower growth rates as well as repetition of the process have to be applied, in which the
purified layer is subjected to a subsequent process of layer growth and sweating, at a
higher temperature, which corresponds with the grade of the purification reached in the
first step.
6.5. Conclusions
123
Chapter VI
layer growth purification has taken place and during the sweating stage additional
purification occurs.
Sweating process removes only impurities entrapped in the intercellular space, while
impurities entrapped inside the cells are not affected. With the increasing amount of
sweating the intercellular eutectic material is removed, while the impurity content
within the grains is not affected. The amount of material sweated out is higher than the
amount of the eutectic present, which means that also some crystals of the layer are
drained off.
For an effective purification three strategies may be considered. The first is only layer
crystallization under static conditions; the second is a dynamic crystallization process,
which as a result of mixing improves mass and energy transfer increasing this way
purification efficiency; and the third is layer crystallization at a higher growth rat with
static or dynamic layer crystallization process, followed by a post-purification step such
as sweating. The later saves costs and energy.
6.6. Acknowledgements
This research is carried out as part of the EET funded project Sustainable Technology
for the Reclamation of High-grade Aluminium from Scrap (contract number
EETK98026). Fruitful discussions at the Technical Project Group Meetings with the
EET partners and in particular with Dr. D. Verdoes (TNO-MEP, Apeldoorn) and Dr. P.
de Vries (Corus RD&T, IJmuiden) are gratefully appreciated. For the experimental
work, the assistance of Mr J.J. Jansen of the Material Sciences Department is gratefully
acknowledged.
124
6.7. References
125
Chapter VI
126
CHAPTER VII
7.1. Introduction
Falling film crystallization process proved to have several advantages over the static
crystallizer for organic compounds. As a result of the flowing film the liquid becomes
turbulent. The turbulence has positive effect on enhancing both energy and mass
transfer of the crystallizer. Due to this effect falling film crystallizer for organic
compounds achieves the same purification efficiency, as for static crystallizer, with four
to five times higher growth rate [1].
Although there are considerable differences between organic compounds and metals
behaviour during fractional crystallization the basic processes are similar. The main
problems for adopting this process for metals is high operation temperatures and
corrosiveness of metal melt to equipment. However, mass and heat transfer of metals
are higher and makes production rates much faster.
In this chapter an evaluation of the falling film process for aluminium scrap refining as
well as equipment design and building will be described. This will be done to see the
potential of following research in this direction in the future.
Chapter VII
Re =
As it is seen in the expression above, the Reynolds number expresses the ratio between
the inertia to the viscous forces.
Based on the value of the Re the regime of flow in thin films is divided as follows:
3. Turbulent flow
Re2 < Re
Different authors use different values of the Reynolds number to describe the transfer
from one regime of flow to the other. For review of the flow of liquids in thin films,
where a table of authors and critical Re numbers for transition from laminar to turbulent
flow regime is given by Fulford [2]. However, generally accepted values for transition
from one flow regime to another are: Re1 from 2 for falling films in vertical wall to 10
for inclined walls; Re2 is considered to be between 250 and 500, with the Re of 400
widely used in falling film fractional crystallization processes.
According to Brauer [3] the mean thickness of a falling laminar liquid film is expressed
in the form;
3 2
=
g sin
1/ 3
Re1 / 3
128
Film thickness in mm
200
400
600
800
1000
Re
Figure 7.1. The thickness of the aluminium falling liquid film on the vertical plate, as a
The film thickness of the aluminium falling film in vertical plate depending on Re
number for laminar flow regime is plotted in figure 7.1.
0.88
Nu = Re Pr *
1
+
1
.
317
Pr
129
Chapter VII
D
*
D =
Nu * m *
0.344
This expression shows that the diffusion layer thickness decreases proportionally with
the film thickness , which in tern decreases with decreasing kinematic viscosity and
increasing the slope of the falling film. Figure 4 shows the diffusion boundary layer as
a function of Reynolds number for vertical liquid aluminium falling film. The diffusion
coefficient of most of the liquid metals is within the range of 1-10 * 10-9 m2/s.
Therefore a value of 5 * 10-9 was taken as a representative for the metals present in
aluminium alloys that will be studied in this project, and was used in calculations for
figure 7.2.
diff. layer
0,035
0,03
0,025
0,02
0,015
0,01
0,005
0
200
400
600
800
1000
Re
Figure 7.2. The diffusion boundary layer of an aluminium falling liquid film over a
The optimum criteria for crystal growth rate can be found from the BPS-model
described in the chapter III of this thesis:
k eff =
k0
k 0 + (1 k 0 ) exp G *
( )
and
G* =
G c
*
kd l
131
Chapter VII
The optimal dimensionless crystal growth rate determines the crystal production rate.
mc = c * Gopt * A = c * G * *
*A
Where,
This expression shows that the decrease of the diffusion layer thickness, D, increases
crystal productivity. Diffusion layer thickness, D, decreases with increasing mass feed
rate (see figure 4). Therefore an optimum feed mass rate that decreases diffusion layer
thickness, D, is important for the design of the falling film crystallizer.
The temperature of the aluminium feed should be high enough to prevent interface
stability. In the same time a very high feed temperature causes problems for the uniform
layer thickness. For organic systems a feed temperature of 5-10 K above the melting
point of the feed system is found to be optimal [Mayer, 1973].
132
Re =
mf
w l
Where,
At low feed rates the interface stability of the crystal surface is not stable. For organic
compounds a minimum Re = 50 is as a criteria to avoid this effect [7, 8]. As shown in
figure 4 the diffusion layer thickness decreases with increasing Re. However, at Re >
1000 the decrease is negligible. Hence, the reasonable upper value of Reynolds number
is 1000.
7.5.5. Optimum crystal growth rate for aluminium falling film process.
For this purpose the following data will be taken into account:
D = 5 * 10-9 m2/s
Re = 1000
G* = 0.1
133
Chapter VII
Gopt = G *
= 0 .1 *
5 *10 9 m 2 s 1
= 3.73 *10 5 ms 1
5
1.34 *10
From this approximate calculation can be seen that the optimum crystal growth rate in
an aluminium falling film crystallization process can be in the range of 10-5 m/s, which
is 10 times as fast as for most of the organic systems.
134
Al melt
at Tm + 5 K
Cooling
Plate
10 dm3/h
aluminium
5 cm
Falling
Al thin film
Tm
30 cm
Residue
Aluminium melt flow rate versus Reynolds number for the newly designed falling film
crystallizer is presented in figure 7.4. For an optimum enhancement of the mass transfer
a typical flow rates with Re = 1000 is considered to be an optimal in melt crystallization
world. It clearly shows that flows of 10 litters per hour would be optimal for the newly
designed aluminium falling film crystallizer.
135
Chapter VII
Al flow
12
liters/hour
10
8
6
4
2
0
0
200
400
600
800
1000
Re
Figure 7.4. Aluminium melt flow rate versus Reynolds number for the newly designed
136
137
Chapter VII
At the moment the molten aluminium flow is in contact with the cooling plate the
heating of the plate is replaced by controlled cooling. This cooling is done by
compressed dry air, flowing through small pipes and than spread by a filter element of
sinter metal. The cooling air control valve is proportionally controlled from the
computer. On the top of the top-reservoir extra pressure is applied to provide constant
flow of aluminium melt through the valve.
Figure 7.6. The control scheme of the aluminium falling film crystallizer
138
139
Chapter VII
7.9. References
1. J. Ulrich, in Theory and Application of melt Crystallization, 28th 30th
September 2000, Martin Luther University Halle Witwnbwrg, Halle, Germany
2. G.D. Fulford, in Advances in Chemical Engineering, Volume 5, Academic
Press, New York 1964.
3. H. Brauer, VDI Forschungsheft, 457 (1956) 22.
4. D.R. Poirier and G.H. Geiger, Transport phenomena in materials processing,
TMS, Warrendale, 1994.
5. H. Schlichting, Boundary-layer theory, McGraw-Hill, New York, 1955.
6. Ossipov, Appl. Math. Modelling, 23 (1999) 419.
7. G.J. Arkenbout, Melt Crystallization Technology, Technomic Publishing
Company, Lancaster, PA, U.S.A., 1995.
8. M.U. Mayer, ber die gerichtete fraktionierte Kristallisation aus dem
Rieselfilm, Ph.D Thesis, ETH Zrich, 1973.
9. B. Drini, W.H. Kool, P.J. Jansens, and L. Katgerman: Method of Recovering a
Metal from a Mixture, World patent no. PCT/ WO/2004/085686, 07 of October
2004.
10. B. Drini, W.H. Kool, P.J. Jansens, and L. Katgerman: Werkwijze voor het
winnen van een metaal uit een mengsel, patent no. NL1023009C, 1 December
2004.
140
141
Summary
from the liquid fraction are in most cases not fully met. Consequently, there still is room for
further progress.
Fractional crystallization technique is much more advanced for organic compounds than for
metals. This is mainly due to the better process conditions, such as low melting temperature
and lower reactor attack, of organic materials. However, a careful study of possibilities of
adopting these processes for metals will open new horizons. Prospects are good for both
modes of operation: suspension and layer crystallization. The most important metals that
could use this technique are aluminium, silicon, magnesium, lead, and precious metals.
Zone melting is a relatively simple and useful experimental technique to assess the ability
to refine distinct alloying constituents from complicated alloy systems (like industrial kinds
of scrap aluminium). Regarding the obtained purification in the experiments, results
basically comply with the estimates based on the binary alloy phase diagrams, although
there are known distinct deviations from other experimentally achieved results. Chemical
analyses reveal that purity of the samples on the head and on the middle positions are
almost the same, which reveals that purification takes place in more than the half of the
original samples.
Microscopic examination of the AlSi6 alloy reveals that while aluminium cells grow, the
impurities that rich eutectic concentration are trapped between them. The AlSi0.4Fe1.6
alloying system micrographs show a flat interface till about 40 per cent of the sample is
crystallized and later a eutectic growth resumes. The other two systems studied in this
thesis (AlCu4Mg1.5 and AlSi1.5Mn1.1) grow with flat interface and no change in their
morphology was observed by microscopy.
In chapter 5 a fractional crystallization process for aluminium recycling based on static
layer growth is described. Static layer crystallization experiments showed that for AlSi7.2
alloying systems, typical growth rates of 1 to 2 m/s are optimal for aluminium scrap
purification. In order to achieve the preset purification criteria for the pure product, one or
more re-crystallizations steps are required. Depending on the impurity concentration, the
crystallization temperature varies from 600 to 660 C.
142
143
Summary
this effect falling film crystallizer for organic compounds achieves the same purification
efficiency, as for static crystallizer, with four to five times higher growth rate.
Theoretical analysis of the falling film process for aluminium scrap refining reveals that the
optimum growth rate of such crystallizer is within the order of 10-5 m/s, which is about 10
times faster than optimum growth rate for static process. This increase in the optimum
growth rate is attributed to the improved mass and heat transfer due to the turbulence that
appears as a result of the melt flow down the heat exchanger wall. The higher growth rates
compared to static crystallizer make this process suitable for a process that requires high
production rates, which with aluminium scrap recycling is the case. Further (experimental)
studies are needed, whereby several technical challenges associated with falling film
crystallization of aluminium are addressed, before the techno-economic potential of this
new technology can be established unambiguously.
144
SAMENVATTING
Primaire aluminiumproductie is een energie-intensief proces. Met een toename van
het
gebruik
van
aluminium
in
de
autoproductie
is
een
technologisch
Duurzame
Technologie
voor
het
Terugwinnen
van
Hoogwaardig Aluminium uit schroot, dat is ondernomen in het kader van het
Nederlandse
Onderzoeksstimulatieprogramma
voor
Economie,
Ecologie
en
Samenvatting
de productiecapaciteit beperkt is. Ook wordt in de meeste gevallen nog niet volledig
voldaan aan de verschillende voorwaarden aan de vorming van de kristallen (wat
betreft de kristalgroeisnelheid en het roeren) en de scheiding van de vaste- en de
vloeistoffractie. Als gevolg hiervan is er nog ruimte voor verdere vooruitgang.
De fractionerende kristallisatie techniek is veel geavanceerder voor organische stoffen
dan voor metalen. Dit is vooral het gevolg van de voordeliger procescondities, zoals
het lagere smeltpunt en de mindere reactorcorrosie, van organische materialen. Toch
zal een nauwgezet onderzoek naar de mogelijkheden om deze processen te gebruiken
voor metalen, nieuwe horizonten openen. De vooruitzichten zijn goed voor de beide
procesmethoden: suspensie- en laagkristallisatie. De belangrijkste metalen waarvoor
deze techniek gebruikt zou kunnen worden zijn aluminium, silicium, magnesium, lood
en kostbare metalen.
Zonesmelten is een relatief eenvoudige en bruikbare experimentele techniek om de
mogelijkheid vast te stellen om bepaalde legeringsbestanddelen te raffineren uit
gecompliceerde legeringen (zoals industrile aluminium afvalsoorten). Wat betreft de
behaalde zuiverheid in de experimenten komen de resultaten ongeveer overeen met de
schattingen gebaseerd op binaire fasendiagrammen van legeringen, hoewel er
bepaalde afwijkingen zijn van andere experimenteel behaalde resultaten. Chemische
analyses tonen aan dat de zuiverheden van de monsters aan de kop en in het midden
bijna hetzelfde zijn, hetgeen aantoont dat zuivering plaats vindt in meer dan de helft
van de originele monsters.
Microscopisch onderzoek van de AlSi6 legering toont aan dat tijdens de groei van de
individuele aluminium cellen ertussen vloeistof wordt ingebouwd die verrijkt is qua
onzuiverheden. De microscoopfotos van de AlSi0.4Fe1.6 legering laten een plat
oppervlak zien totdat ongeveer 40 procent van het monster gekristalliseerd is, daarna
treedt eutectische groei op. De andere twee systemen die bestudeerd zijn in deze
dissertatie (AlCu4Mg1.5 en AlSi1.5Mn1.1) groeien met een plat oppervlak en er werd
geen verandering in morfologie geobserveerd met de microscoop.
In hoofdstuk 5 is een een fractionerende kristallisatie proces beschreven voor
aluminium terugwinning gebaseerd op statische laag-groei. Statische laag146
vereiste
productzuiverheden
te
behalen
zijn
of
meerdere
doen
toenemen.
Gedurende
de
laag-groei
heeft
opzuivering
Samenvatting
Bedri DRINI
148
PRMBLEDHJE
Prodhimi i aluminit primar sht nj proces q krkon shum energji. Me rritjen e sasis s
prdorimit t aluminit pr prodhimin e automjeteve, zhvillimi i nj procesi teknologjik pr
riciklimin e aluminit sht i domosdoshm. Energjia e nevojshme pr riciklimin e aluminit
sht vetm nj pjes e vogl e energjis q do t duhej pr ta prodhuar aluminin nga xehja.
Pr t kontribuar n zhvillimin e nj procesi t riciklimit t aluminit nj projekt i prbashkt
i hulumtues i Universitetit Teknik t Delftit me partnerve Corus plc nga IJmuideni, TNOMEP nga Apeldorni dhe TNO Instituti i Teknologjis Industriale nga Eindhoven sht
zhvilluar si pjes e projektit hulumtues Teknologji e Qndrueshme pr Riciklimin e
Aluminit nga Mbeturinat , I cili sht ndrmarr n kuadr t programit hulumtues
holandez, i cili stimulon programet pr Ekonomi, Ekologji dhe Teknologji (kontrata me
numr EETK98026). Prderisa partneret ton e kan hulumtuar procesin e kristalizimit
fraksional
Prmbledhje
150
kristaleve.
Aplikimi i teknikave t post-pastrimit, si jan djersitja e kristaleve dhe larja e tyre, e rrisin
n mnyr t konsiderueshme. Gjat kristalizimit pastrimi i aluminit ndodh kurse gjat
djersitjes se kristaleve ato pastrohen edhe me shum. Me rritjen e sasis se kristalit t
djersitur rritet edhe pastrtia e tyre. Me rritjen e sasis s djersitur t shtress rritet sasia e
mbetjes eutektike qe sht e vendosur n mes t kristaleve primare, kurse papastrtit q
gjinden n kristalet primare nuk largohen. Djersitja bazohet n shkrirjen parciale t shtress
kristalore duke e ngrohur ngadal deri t pika e shkrirjes s kristaleve. Temperatura gjat
djersitjes sht m e lart sesa gjat kristalizimit. Edhe pse gjat djersitjes 10 20 % e
kristaleve humbin, procesi i djersitjes sht efektiv dhe mnjanon nj kristalizim shtes t
nevojshm pr pastrimin e skrapit n nj kohe shumfish m t shkurt. Sasia e mass
kristalore t humbur gjat djersitjes sht me e madhe sesa sasia e mbetjes eutektike q
gjendet n mes t kristaleve primare, q do t thot se edhe nj pjes e kristaleve primare
shkrihet. Pr nj pastrim efikas mund t konsiderohen dy mnyra. Njra sht t rrisim
shtresn kristalore ngadal, kurse tjetra t rrisim at me shpejt dhe pastaj ta djersitim at.
Metoda e dyt kursen koh dhe para.
151
Prmbledhje
Bedri DRINI
152
ACKNOWLEDGEMENTS
I would like to express my gratitude to number of people who have contributed to the
completion of this book.
First of all I would like to thank Prof. Ir. Laurens Katgerman and Prof. Dr. ing. Dolf
Bruinsma for giving me the opportunity to carry out this research at TU Delft.
Special thanks go to my promoters Prof. Ir. Laurens Katgerman and prof. Dr. ir. Peter
Jansens for their guidance and support during my research. Last year my stay outside the
Netherlands has made too difficult to complete this work. Due to Prf. Dr. ir. Peter Jansenss
persistence and patience, and his assistants Leslie van Leeuwens help in completing all
the administrative procedures, it was possible to finalize this work.
I am grateful to Prof. Dr. ing. Dolf Bruinsma for his support and guidance during the first
year of my research work. He taught me the fundamentals of fractional crystallization and
was always ready to discuss all the problems I was facing during that initial period of my
work.
I am very grateful to the project technical group people: Dr. ir. Dirk Verdoes and Ir.
Michil Niewenoord of TNO-MEP in Apeldoorn; Dr. ir. Wim Boender and Dr. ir. Paul de
Vries from Corus Ijmuiden, Dr. ir. Wim Sillekens, jan Schade van Westrum and Mark van
Haut form TNO Industry. Their support during the TPG meetings was very useful and in
the same time it helped me understand the issues of suspension based process.
I am also grateful to people with whom I have discussed issues considered in this thesis and
with whom I have written couple of articles: Prof. Dr. ir. Rob Boom, Dr. Ir Pim Kool and
others.
I am thankful to my colleagues and friends from my group: Dr. Dmitri Eskin, Dr. Lidy
Appachitei-Fratila, Dr. Iulian Appachitei, Dr. Jan Zuiderma, Dr. Volodia Bilovol, Suyitno,
and others for making a friendly atmosphere during coffee brakes and their support with all
scientific and personal consultations during my stay in the Netherlands. I am also grateful
to the help given by researchers at the API department for useful discussions during
monthly meetings.
Acknowledgments
My experimental work would not be completed successfully if there were not the talented
group of GST technical group: Jan van Etten, Tjerd Tobi, Jan-Peter Boomsma. Special
thanks go to Jaqcue Jansen for his mastership on dealing with molten aluminium and great
help with my experimental setup.
Last but not least my thanks go to my family and friends for inspiration and support during
my research at TU Delft. I thank my wonderful wife, Marika, for her selfless support and
encouragement. My daughters Teut and Liza were my endless source of joy and
inspiration.
154
PUBLICATIONS
Publicistic articles:
1. Bedri Mehmetaj; Hazards from polluted water, (in Albanian), Daily Rilindja,
Prishtin - Zrich - Tiran, 10 June 1996.
2. Bedri Mehmetaj; QSL- A new method of extraction of lead, (in Albanian),
Daily Rilindja, Prishtin Zrich - Frankfurt-Tiran, 20 Jun 1998.
3. Bedri Drini; Air pollution shortens your live, (in Albanian), Daily Bota Sot,
Prishtin - Zrich, 27 Oktober 2002.
Reports and Studies:
1. Verdoes, D, Nienoord, M, Boender, W, Landskroon, JPS, Vries, PA de,
Drini, B
28
pp.
http://www.senternovem.nl/mmfiles/EETK98026_tcm24-
166126.pdf.
2. Drini, B. and Dushi, M. Study for prospects of metal industry development
in Kosovo, financed by the Ministry of Trade and Industry, NKI Drini
Consulting and Engineering, Prishtina, Kosova, 2005, pp 52.
Journal and Conference Papers
1. Sillekens W.H., Schade van Westrum J.A.F.M., Bruinsma O.S.L., Mehmetaj B.,
Nienoord M.; "Refining
Aluminium Scrap
by
means
of
Fractional
Publications
Structure, Redox and Photophysical properties of Polypyridyl Ruthenium(II)
Complexes Containing Carboxylate Substituted Pyridyltriazoles; European
Journal of Inorganic Chemistry, (2002), p. 1765 1771
4. Mehmetaj B., O.S.L. Bruinsma, W.H. Kool, P.J. Jansens, L. Katgerman
Aluminium scrap recycling with solid layer fractional crystallization,
Proceedings of the 15th International Symposium on Industrial Crystallisation
(ISIC-15),
Ed.
Chianese,
A.,
Sorrento,
Italy,
Chemical
Engineering
156
CURRICULUM VITAE
CURRICULUM VITAE
Name:
E-mail:
bedri.drini@dce-group.com
Date of birth:
February. 7. 1970
Education:
Work Experience:
1997 1999
Other activities:
157