Re Tten May Rint Mater Rev 2009

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Melting and remelting phenomena

M. Rettenmayr
Melting processes influence the microstructure evolution in metal alloys during casting and heat
treatments. Melting is often treated as ‘inverse solidification’, which is only appropriate in a limited
number of cases. In the present article, asymmetry between solidification and melting is reviewed
in detail. The current state of the thermodynamic description of melting under diffusion control is
outlined. Kinetic aspects that break the symmetry between solidifi-cation and melting that are
discussed are solute partitioning, solute concentration gradients and solute transport in the
involved phases. The view on nucleation of liquid and pre-melting phenomena, mostly of pure
materials, based on experimental and theoretical work, is given. Emphasis is laid on aspects of
melting with technical relevance, i.e. melting of alloys. Particu-larities of thermally and solutally
controlled melting are introduced. Mechanisms that involve both melting and solidification
simultaneously are capillary driven coarsening, temperature gradient zone melting and liquid film
migration. The impact of these processes on microstructural evolution is discussed. Open
questions concerning modelling and simulation of melting, namely, the interaction of different
melting mechanisms and the role of the solid/liquid inter-face, are identified.
Keywords: Melting, Dissolution, Solidification, Microstructure evolution, Coarsening, Liquid film migration, Temperature gradient zone melting, Dendrite
fragmentation

Introduction continuous scientific interest in melting since liquid/solid


phase transformations were studied. Different nuclea-
Solidification and melting are both phase transforma- tion behaviour of melt ‘on its own solid’ was identified
tions involving at least one liquid and one solid phase first as a reason for the asymmetry, but further aspects
respectively. As compared to solidification, research concerning structure, thermodynamics and kinetics have
work on melting documented in the literature is been pointed out over the past decades. Research work
relatively scarce. This is certainly due to the fact that is published in the literature where these aspects of
solidification is of higher interest for technical applica- asymmetry are not considered, in almost all instances for
tions, as the properties of a cast material are directly cases where solidification and not melting is the main
connected to the microstructure evolution during focus of the work. Simulation calculations for solidifica-
solidification. tion processes that include remelting should thus be
At first sight it is tempting to adapt know-how from interpreted with care.1–3
the further developed solidification theory to melting, as The influence of melting on microstructural evolution
melting and solidification could be expected to be is manifold. In large castings, the solidification structure
similar. It appears to be straightforward to treat these on a macroscopic level will generally be strongly
phase transformations symmetrically, i.e. melting as the influenced by melting, since convection can increase
inverse process of solidification. This may indeed be local temperature and composition and thus alter the
correct for liquid/solid phase transformations with low solidification time and the temperature at which
transformation rate when the assumption of local solidification occurs locally. More importantly, solidifi-
equilibrium at the interface holds and nucleation cation and melting are not necessarily separated in space
phenomena are unimportant. and time. Several processes are known that involve
However, it is a long known experimental fact that simultaneous melting and solidification in the mushy
upon cooling of liquid metals, an undercooling under zone of an alloy. Among those are dendrite arm
the equilibrium melting temperature can hardly be coarsening, temperature gradient zone melting
avoided, but that upon heating, solid metals can only (TGZM) and liquid film migration (LFM). Up to
be heated above their melting temperature while present, those processes can be described analytically,
remaining in the solid state with considerable experi- often semi-empirically, without including detailed ther-
mental effort. The lack of symmetry has caused modynamics and kinetics of melting.
In the present article, after documenting the technical
importance of melting (the section on ‘Technical impor-
FSU Jena, Materials Science and Engineering Institute, Metallic Materials tance of melting’), the known asymmetries between
Department, Löbdergraben 32, D-07743 Jena, Germany melting and solidification are elaborated; recent progress
*Corresponding author, email M.Rettenmayr@uni-jena.de in the thermodynamics of melting is summarised, and

ß 2009 Institute of Materials, Minerals and Mining and ASM International


Published by Maney for the Institute and ASM International
DOI 10.1179/174328009X392930 International Materials Reviews 2009 VOL 54 NO 1 1
Rettenmayr Melting and remelting phenomena

aspects of differences in solute partitioning and structure discussed as the main mechanism for dendrite fragmenta-
as well as phase transformation kinetics are presented (the tion which in turn leads to the columnar-to-equiaxed
section on ‘Asymmetry between solidification and melt- transition (CET)11–15 and freckle formation in Ni base
ing’). The literature on nucleation related phenomena is superalloys16 or Sn alloys.17 In these latter cases,
discussed in the section on ‘Nucleation and related secondary dendrite arms that melt off at the root are
phenomena and pre-melting’. Thermal and solutal potential nuclei for new grains.
melting are treated separately in the sections on Some technological processes that are based on
‘Thermal melting’ and ‘Solutal melting’. Finally, pro- melting need to be mentioned here, even though
cesses that involve simultaneous melting and solidifica- generally rather the technological impact than the
tion are introduced in the section on ‘Melting and melting process is in the focus of interest in the related
simultaneous solidification’. literature. Particularly there are soldering, brazing and
Melting and remelting phenomena are related in the welding as joining technologies. Additionally, surface
present work to the context that is common knowledge remelting by laser, electrical arc, electrons, etc., is used
in the scientific community working on liquid/solid as a tool to modify the surface structure and conse-
phase transformations. The focus of the present article is quently the properties of a material. As the melting
on melting of alloys and ‘real crystals’. The larger part of mechanisms are only scarcely discussed in these works
the literature on mechanisms and kinetics of melting they are not part of the present review.
deals with pure substances under laboratory conditions
or in ideal crystals in computer simulations. In
numerous instances, especially in simulations and the Asymmetry between solidification and
development of theory, the published literature con- melting
siders a state of considerable superheating, particularly In the present work, it is distinguished between thermal
close to the theoretical limit of superheating. While this and solutal phase transformations. A definition of
contributes to a basic understanding of melting pro- solutal and thermal phase transformations is not trivial,
cesses, the way to a practical application is far. Aspects as in many cases, there are aspects of both solutal and
with at least some relationship to engineering cases will thermal control. The definition is, however, straightfor-
be introduced here, the scientific literature dealing with ward when starting a phase transformation from a single
pure materials under special laboratory conditions and phase in thermodynamic equilibrium, i.e. in the absence
or the corresponding simulations are not discussed. The of supersaturations/undercoolings. In Fig. 1, a sche-
interested reader is referred to the reviews of Dash4 on matic of a part of a phase diagram is shown where the
surface melting and the recent comprehensive review on two types of phase transformation are illustrated.
melting and superheating by Mei and Lu.5
If starting with a single phase, thermal melting or
solidification are induced by a continuous increase or
Technical importance of melting decrease in temperature (more precisely in enthalpy),
In recent years, the importance of melting phenomena cases (1) in Fig. 1 (solid line for melting, dotted line for
for technical processes has become more and more solidification). It should be noted, however, that even
recognised. (Local) remelting is a phenomenon that though a change in enthalpy is the cause for the phase
occurs in a variety of casting processes and influences transformation, its kinetics might still be controlled by
the structure evolution to an important degree. Local solute diffusion in the phases. Pure thermal control
remelting can occur in non-stationary states even if in a would only be acting during rapid heating/cooling.
cast part overall the enthalpy is decreased. In fact, in his Nevertheless, in the present work, experiments that are
book ‘Castings’, J. Campell6 states verbatim: ‘convec- conducted by introducing heat into a sample and that
tion can cause remelting of large sections of a casting start single phase are treated as thermal melting, even
over an extended period of time; models that do not though solutal control might play an important role.
consider this effect are grossly erroneous’. Solutal melting or solidification is due to driving
Local remelting is not limited to ‘exotic’ cases with forces that can in principle persist under isothermal
low technical impact. Melting of considerable volume conditions, i.e. where mainly a change in composition
fractions in semi-solid or solid alloys has been identified yields a subsequent phase transformation. An example
as an important factor during a variety of processes. where in an experiment starting from the single-phase
During continuous casting, remelting can cause surface field the two-phase field in the phase diagram is entered
defects on the cast piece.7 Premature melting in recycling by one of the phases due to solute diffusion is the
of scrap steel consisting of different kinds of steel has constitutional undercooling/superheating ahead of a
been investigated in Ref. 8. In multicomponent alloys, solidifying/melting interface [cases (2) in Fig. 1, solid
diffusion of different solute elements with the respective line for melting, dotted line for solidification]. The phase
kinetic constant can lead to local melting during heat transformation behaviour in this case is likely to be
treatments.9 influenced by nucleation in the undercooled/superheated
Furthermore, local remelting can have drastic influence region.
on a solidifying micro-structure, even if the remelting Additionally, melting or solidification can be con-
volume fraction is minor. During Czochralski growth of trolled solutally, if initially both a liquid and a solid
single crystals, remelting has been observed at triple phase are present, but the composition of at least one of
points liquid/solid/gas and is a source of structural the phases is not the equi-librium composition at the
defects.10 Considering that secondary dendrite arms are interface; thus, a reaction between the phases will occur
generally thinner at the root than closer to the growth at the interface, yielding a change of interfacial
front, one can easily imagine that local remelting leads to compositions and a migration of the interface. Cases
the detachment of dendrite arms. Thus, remelting is (3) in Fig. 1 show the situation at the interface during

2 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

differences between solidification and melting.


According to Boettinger and Corriell,18 there is a
hierarchy concerning the thermodynamics of phase
transformations:
N full diffusional equilibrium: uniform chemical poten-
tial and composition of all species in all of the phases
respectively, throughout the sample
N local equilibrium: concentration gradients in (one of)
the phases, equal chemical potentials of all species at
the interface
N metastable equilibrium: one of the involved phases
cannot nucleate
N interfacial non-equilibrium: jumps in the chemical
potentials, equal compositions left and right of the
interface.
At about the same time, when this classification was
established, it was shown19 that there is a transition
between the idealised conditions of interfacial local
1 Schematic of a phase diagram for the definition of thermal equilibrium and non-equilibrium, as the partition
and solutal phase transformations; circles are for an coefficient during alloy solidification evolves from the
initial single-phase state, squares for an initial two-phase equilibrium value to unity.
state; solid lines depict melting cases, dotted lines solidi- For establishing this hierarchy qualitatively, no
fication cases; (1) phase transformation due to a continu-
explicit statement about the nature of the interface
ous change in enthalpy (thermal melting/solidification), (2)
concerning structure or width, etc., is contained. As
phase transformation due to diffusion in a single phase,
soon as a quantification of the transition from equal
e.g. constitutional superheating/undercooling ahead of an
potential to equal composition at the interface is
interface (solutal melting/solidification), (3) phase trans-
attempted, some assumptions need to be made about
formation due to a concentration gradient at the interface
the interface. For liquid/solid interfaces, the interface
(solutal melting/solidification), (4) phase transformation
thickness and the ‘interface diffusion coefficient’ were
due to solute enrichment/depletion in one of the phases
introduced,19 requiring some insight into the role of the
(solutal melting/solidification)
interface during the phase transformation. However,
more detailed insight into moving interfaces was mainly
plane front directional melting/solidification: diffusion gained in the field of solid/solid phase transformations,
in the parent phase effectuates an increase/decrease in as different interface models with finite extension were
composition of the parent phase, with the trend of established20,21 to account for effects like solute drag,
entering the two-phase field and consequently migration rearrangement of the lattice, etc.
of the interface. Before the emerging of phase field models, liquid/solid
A further well-known example for solutal melting is a phase transformations have successfully been modelled
solid being in contact with a liquid that is solutally with a mathematically sharp interface. Without ques-
enriched, i.e. the composition of the liquid is higher than tioning the advantages of a diffuse interface, it can be
the liquidus composition and in the one-phase field in stated that solid/liquid interfaces are not prone to be
the phase diagram. Under this condition, solutal melting affected by segregation of impurities or by elastic
is often called dissolution [case (4) in Fig. 1]. For strains, so that complex interfaces need not be intro-
example, at a given temperature below the freezing point duced. It is generally accepted that the sharp interface
of water pure ice will dissolve in a water/ethanol model can yield realistic results.
mixture, if the ethanol composition is higher than the A classical phase transformation model will solve a set
liquidus composition. For dissolution, the composition of heat and/or mass transport equations in all of the
in the solid phase only plays a less important role, phases and connect the phases by assumptions or
provided there is solute enrichment in the liquid melting separate models about the thermodynamic state of the
will occur for any solid composition. interface. The simplest yet widely applicable set of
conditions at the interface would be local thermody-
Thermodynamic aspects of (non-equilibrium) namic equilibrium. This implies that in a binary alloy,
solidification and melting of alloys the interface compositions will be given by the equili-
Liquid/solid phase transformations with low or moder- brium phase diagram, or, in a more general definition,
ate front velocities occur under diffusion control. It is that chemical potential of all species is equal at the
assumed that the interface is in local equilibrium, the interface in the adjacent phases.
phase transformation rate is then governed by mass In the fundamental work of Baker and Cahn,22 the
transport in the adjacent phases. As long as the local basic principles of non-equilibrium (interfacial) thermo-
equilibrium assumption holds and melt nucleation at dynamics are expounded with special emphasis on
defects is unimportant, the composition of the phases solidification. General formulations of irreversible ther-
at the interfaces can be taken from the phase diagram at modynamics23,24 are discussed and considered. It is
the relevant temperature, the phase transformation be shown under what conditions local equilibrium is lost
modelled symmetrically for solidification and melting. and the interface compositions will not be those defined
Recent progress in the formulation of thermody- by the equilibrium phase diagram any more. The
namics of migrating phase interfaces allows pointing out interface compositions decide on the driving force that

International Materials Reviews 2009 VOL 54 NO 1 3


Rettenmayr Melting and remelting phenomena

solidification, the chemical potential of the solute B will


always increase (for an equilibrium partition coefficient
keq,1), an effect that is overcompensated by a decrease
of the chemical potential of the matrix element A.
Still for the case of steady-state solidification,
equation (1) can be reformulated as20
  
DG~ 1{X L mLA {maA zX L mLB {maB
 
z X L {X a mLA {maA {mLB zmaB (2)
In this form, according to the sharp interface model, the
first two terms can be interpreted as the driving force for
the migration of the interface DGm. DGm can be used to
overcome various kinds of resistance to the migration of
the interface, e.g. pressure due to interfacial energy and
friction. The last term in equation (2) may be regarded
as the change of Gibbs energy DGt due to an adjustment
2 Tangent-to-tangent construction in a molar Gibbs free of the composition from XL to Xa by exchange of A and
energy diagram for determining the energy change (dissi- B atoms between the two phases by diffusion across the
pation) DG upon a solid/liquid phase trans-formation;22 the interface. This process is generally called trans-interface
possible composition range DXa for solidifying a-phase diffusion. Thus, the total driving force for the phase
during a liquid/solid phase transformation starting from a transformation is subdivided into
composition XL is also shown   
DG m ~ 1{X L mLA {maA zX L mLB {maB (3)
is dissipated at the interface by either interface migration  
or other processes occurring at the interface. DG t ~ X L {X a mLA {maA {mLB zmaB (4)
Baker and Cahn22 show how the change (dissipation)
In the sharp interface model, it is usually assumed that
DG in Gibbs energy is determined graphically by a
the interface migration and the trans-interface diffusion
tangent-to-tangent construction, if an infinitesimally
are two independent physical processes. Following this
thin layer with composition Xi out of a solid with
assumption, it is straightforward that each of the
composition Xa is transferred to the liquid with
processes can only occur if the driving force is positive.20
composition XL (Fig. 2). Note that the composition of
This limits the range DXa of possible compositions that
the transferred layer Xi is not necessarily that of one of
can precipitate from the liquid.25
the involved phases.
However, the simplification that the interface compo-
By the intersection points of the tangent to the liquid
sition Xi is that of one of the adjacent phases is
phase with the Gibbs energy curve of the a-phase, also a
acceptable for solidification, but not for melting.25 The
range DXa of thermodynamically possible compositions
reason is that not only in the parent phase (solid), but
of the a-phase is specified that can solidify from a
also in the product (liquid) phase, concentration
supersaturated liquid with composition XL (Fig. 2). A
gradients can lead to diffusion toward the interface
unique solution for Xa is not defined. As from
and modify the interface composition. Accepting the
thermodynamic principles, compositions can be
assumption of independent trans-interface diffusion and
excluded, but not positively and precisely defined, the
migration and re-introducing an interface composition
term ‘science of the impossible’ was coined.
Xi that is different from that in the adjacent phases,
In the formulation of the Gibbs energy dissipation equation (2) becomes for melting
upon solidification, it is assumed that at the same time as  
a layer of liquid transforms into a, it has changed its DG~ð1{X a Þ maA {mLA zX a maB {mLB
composition from XL to Xa, a change that is made  
possible by diffusion inside the liquid phase, implying z X a {X i maA {mLA {maB zmLB (5)
that there is no diffusion inside the a-phase. This latter The first two terms in equation (5) are again the driving
statement means that either diffusion in a is infinitely force for migration DGm, the last term is the trans-
slow or a precipitates with constant composition in a interface diffusion DGt.
steady-state. It also means that the composition of the Similar to the construction in Fig. 2, DGm and DGt
interface layer is not the generalised Xi (Fig. 2) any can be represented graphically25 (Fig. 3). Even though
more, but the composition Xa of the solidifying phase. the Gibbs energy dissipation during both thermal and
The transfer of atoms through the migrating interface solutal melting is given by equation (5), the different
occurs by movement of a combination of A and B atoms compositions that lead to the melting processes require a
with the new composition, Xa, from a liquid reservoir specific graphical construction for the two cases. The
into the a-phase under a change of their chemical total driving force DG is again defined by a tangent-to-
potentials mA and mB respectively, yielding a total tangent construction at composition Xi. DG is the sum of
decrease in Gibbs energy DG of the two parts DGm and DGt. For thermal melting, the
  composition Xi will always be in between the composi-
DG~ð1{X a Þ mLA {maA zX a mLB {maB (1)
tions of the adjacent phases (Fig. 3a). For solutal
Equation (1) shows that the phase transformation can melting, there are more degrees of freedom. In the case
occur and DG will decrease even if the chemical potential of melting of a solute depleted solid in contact with a
of one of the components is increasing. In fact, for rapid solute enriched liquid, Xi can be lower than the solid

4 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

3 Tangent constructions for a thermal melting and b solutal melting of a solute depleted solid in contact with a solute
enriched liquid25

composition Xa (Fig. 3b). Note that there are concen- entity governing the growth of a crystal during
tration gradients in both of the phases, implying that solidification. The similar solidification morphology of
diffusion in the two phases in the vicinity of the interface the majority of metal alloys is attributed to the similar
will yield a change from Xi to XL as one layer of solid entropy of fusion (DSf/R<1, with the gas constant R).27
with Xa melts. The asymmetry effects of atomic accommodation are
enhanced for solidification of complex crystal structures,
Solute rejection and solute absorption e.g. non-cubic structures or molecule structures. The
In alloys, solute partitioning yields different conditions non-dimensionalised entropy of fusion for such struc-
for solidification and melting. tures is larger than unity, and the tendency for faceted
For alloys with melting point depression (keq,1), crystal growth will increase with increasing entropy of
there is an excess of solute at the interface during fusion.27 Faceted growth requires additional under-
solidification. Solute is rejected into the parent phase cooling. This is a well known and experimentally shown
(liquid) by the product phase (solid). The highest solute fact and has also been shown by molecular dynamics
content in both the liquid and the solid is at the simulations (for a recent example, see Ref. 28). On an
interface. Any amount of solute that the product phase atomically smooth faceted interface, the fraction of sites
rejects can be accommodated in the liquid and is then to that are unfavourable for solidification is substantially
be transported away from the interface for the process to higher than on a rough interface. Atoms that are melting
continue (or the temperature needs to be lowered so that will on the other hand always see a rough interface and
the concentration of the solute enriched liquid reaches will still be accommodated in the melt easily.
the liquidus line again). Semiconductors are known to solidify both with faceted
During melting, on the other hand, there is solute and rough interface, depending on the growth condi-
depletion at the interface. Solute needs to be transported tions, as documented in recent studies.29,30 The same can
towards the interface for the process to continue. The be true for some intermetallic phases, as shown for the
lowest solute content is at the interface only for thermal intermetallic eutectic phase solidifying in Al–Fe–Si.31
melting from a superheated solid, for solutal melting, It is straightforward that the dynamics of the melt
there are more degrees of freedom. If solidification and structure will be different from that in the solid, leading to
melting were symmetric, during melting solute should be further aspects of asymmetry. The low range order in the
ejected from the parent phase (solid) in any amount that melt (the nearest neighbour distance, coordination num-
the product phase (liquid) requires. This is obviously not ber) is a function of temperature32 and can be permanently
the case: the amount of solute that can be released from changed by a pronounced superheating. Additionally,
the solid at the interface is certainly limited, as it needs ordering in the melt in the vicinity of the solid is an effect
to be transported to the interface by solid diffusion. that has been studied experimentally and theoretically.33
All of the mentioned effects cannot be characteristic for
Structural aspects solids. Thus, solidification and especially nucleation of a
The qualitatively different crystallographic structure solid will be affected by the structural effects in the parent
between liquid and solid, i.e. short range order v. long phase (liquid). As the parent phase for melting does not
range order, entails further aspects of asymmetry. While show such structural changes and effects, the melting
it can be assumed that an atom that is released from the process cannot be dependent on them.
solid during melting can be accommodated in the liquid
at any location along the interface, there are unfavour- Link between thermodynamics and kinetics
able locations on the solid for an atom that arrives from It is well known that in most metal alloys, mass
the liquid. This aspect of asymmetry has been worked transport by diffusion in the melt is faster by 3–4 orders
out in detail by Chalmers.26 As a result, the entropy of of magnitude than in the solid. It is one of the basic
fusion has been identified as an important characteristic aspects of asymmetry between solidification and melting

International Materials Reviews 2009 VOL 54 NO 1 5


Rettenmayr Melting and remelting phenomena

that the melt with the faster mass transport (that can be
further accelerated by buoyancy) is the parent phase for
solidification, but the product phase for melting. For
solidification, the slow diffusion in the product phase
(solid) yields that equation (1) (that is actually derived
neglecting the flux Ja in the solid) is still a good
approximation for the Gibbs energy change at the
interface, even if concentration gradients in the solid are
present; the interfacial compositions and the interface
velocity are hardly influenced by processes in the solid.
On the other hand, in the case of melting, slow diffusion
in the parent phase (solid) will move the interface
composition of the product phase (liquid) towards or
into the two-phase field.25 This can change the
concentration gradient in the liquid (that governs the
melting process) and thus influence the interface velocity
drastically.
The kinetics of a phase transformation process is
linked via equaiton (5) with the thermodynamics.
The term Xa2Xi in equation (5) can be defined
considering the mass balance as25
X a {X L
X a {X i ~ (6)
1{J L =J a
where JL and Ja are the fluxes of B in the liquid and a-
phase respectively. If the diffusion in the parent phase a
is negligible, the trans-interface term DGt becomes zero,
there will be no adjustment of composition by trans-
interface diffusion.25 For melting, (almost) all of the
available driving force can be spent for the migration
of the interface; for solidification, a part of the
available driving force is dissipated by trans-interface
diffusion.
By equation (6), thermodynamics and kinetics are
coupled, the thermodynamic state of the interface is no
longer ‘science of the impossible’, but fully defined.
Considering the thermodynamic constructions in
Fig. 3, schematic representations of concentration pro-
4 Schematic composition profiles for thermal and solutal
files in the liquid and solid phase for both thermal and
solid/liquid phase transformations: a thermal solidifica-
solutal liquid/solid phase transformations can be pro-
tion, b thermal melting (situation corresponding to
vided (Fig. 4). In all parts of Fig. 4, there is the solid
Fig. 3a) and c solutal melting (dissolution of solute
phase on the left, the liquid phase on the right, the
depleted solid in a solute enriched melt, corresponding
interface is moving in the direction of the arrow, i.e.
to the initial conditions as shown in Fig. 3b
solidifies towards the right and melts towards the left.
For solidification, both thermally and solutally con-
trolled (rapid) processes yield interface compositions may enter the two-phase field in the phase diagram an
inside the two-phase field in the phase diagram, i.e. Xa thus be lower than XeL. In this case, the composition of
above the solidus composition Xea and XL below the the solid will be in the one-phase field in the phase
liquidus composition XeL, this latter condition implying diagram below Xea.
that the driving force must be some undercooling of the
liquid (Fig. 4a). Nucleation and related phenomena and
For melting, there are more degrees of freedom. pre-melting
Supersaturation (superheat) of the solid will be the
driving force for thermal melting (Fig. 4b, correspond- It has long been known that solids even in the form of
ing to Fig. 3a). Generally, thermal melting will occur high purity, low defect single crystals exhibit little kinetic
with increasing temperature and thus decreasing inter- inhibition for melting. This is believed to result form the
face compositions. Solutal melting can either occur absence of a nucleation barrier. Not only is superheating
during directional melting in a steady-state with uniform avoided by the easy nucleation of liquid, but the melting
concentration in the melt (not shown in Fig. 4), or if temperature at the surface is even somewhat depressed.
solute enriched liquid is in contact with a solid that is This effect is referred to as pre-melting.
saturated with solute or solute depleted (initial situation A heuristic explanation for pre-melting states that
as in Fig. 3b). However, the solute enriched liquid will atoms at the surface of a solid crystal have a lower
be diluted at the interface by the dissolving solid, leading coordination number and are thus thermally less stable
to a concentration profile as shown in Fig. 4c.25 For than bulk atoms. Thus, the surface of a crystal can
solutal melting, the interface composition in the liquid transform to a (thin) liquid layer already below the

6 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

equilibrium melting temperature Tm. Besides this melting of Al in the transmission electron microscope on
‘mechanistic’ view, the unstable surface is sometimes a hot stage,44 no evidence was found for pre-melting
explained in textbooks34 by a balance of interfacial below Tm. It should be noted that in this and other
energies: the sum of solid/liquid and liquid/vapour work, superheating near a grain boundary was not
energies is supposed to be smaller than the solid/vapour observed.
surface energy, also entailing the formation of a liquid Various models of pre-melting at grain boundaries45
layer at temperatures slightly below Tm. It is worth have been proposed. Computer simulations by molecu-
noting that not all materials feature a combination of lar dynamics,46,47 Monte-Carlo48,49 and phase field50
interfacial energies that lead to pre-melting.35 methods suggest grain boundary melting below the
The origin of the idea that a solid may have a liquid melting point. In highly symmetric tilt boundaries, some
layer on its surface stems from observations of superheating of a grain boundary could be achieved in a
Faraday36 on ice as early as 1859. Further experimental simulation.51 It is pertinent to note that computer
evidence was provided when preparation techniques simulations dealing with grain boundaries only give
were available to generate well defined surfaces of high hints on the behaviour of real grain boundaries, as
purity crystals and when measurement techniques were segregated impurities will change their properties
available to detect the ultra-thin liquid layers at the drastically. Owing to the common melting point
onset of melting. Direct evidence of surface melting on depression upon alloying, it is highly probable that
metals is documented for Pb37 and for Al.38 A summary pre-melting is more pronounced in real crystals (see the
of further experimental findings can be found in Ref. 39. HRTEM and Auger spectroscopy studies in Ref. 52).
It is worth noting that surface melting is crystal face Pre-melting can also occur at line defects (disloca-
dependent and thus not observed for all crystal surface tions).53–55 Detailed observations of melting behaviour
orientations. It appears that close packed {111}-surfaces within bulk colloidal crystals56 illustrate that there is a
of fcc metals do not pre-melt. Less densely packed ‘hierarchy’ of defects as far as the nucleation of liquid is
{110}-surfaces show complete surface melting,5 whereas concerned. With increasing temperature, pre-melting
intermediate packing densities lead to incomplete sur- starts at grain boundaries, then affects dislocations and
face melting. finally partial dislocations. In molecular dynamics
From the theoretical side, it is nowadays assumed that simulations, the role of different defects on melting57,58
specific modes of vibrations, fluctuations and changes of has also been found, even though the occurrence of pre-
electronic structures at a free surface are connected with melting has not been confirmed.59,60
an excess free energy that leads to pre-melting.40 It is From the experiments and the corresponding theore-
generally accepted that various processes occur at a tical interpretation on pre-melting, it can be concluded:
crystal surface as it is approaching its melting tempera- N pre-melting occurs on loosely packed surfaces,
ture from far below. At low temperatures, there is no arbitrary surfaces do not necessarily form a liquid
surface liquid layer. As temperature increases, above a layer below the melting temperature
critical temperature, a disordered surface region forms N in experiments that are not especially designed for
on the crystal with properties in between those of solid studying pre-melting, melting starts close to or at the
and liquid. Before melting, the surface region grows in melting temperature; the liquid layer on a surface is
thickness until pre-roughening, roughening and wetting not relevant for practical purposes
are initiated. Wetting is an early stage of surface melting N surface or interfacial strains can influence local
and equivalent to the wetting as described by the energy remelting processes (TGZM, LFM, see below); it
balance above. remains unquantified to what degree this influence is
The thickness of the liquid surface layer increases significant
significantly as Tm is approached. Close to Tm some 20 N easy nucleation of a liquid layer is not restricted to
atomic layers have been detected to be liquid. At Tm the free surfaces, internal two-dimensional defects like
thickness of the liquid layer diverges. For the ‘equili- grain boundaries or interfaces can act similarly.
brium thickness’ of the liquid layer, a power law
temperature dependence has been deduced.41 Thermal melting
For a surface strained in a direction parallel to the
surface, there is an additional free energy contribution Slow heating of pure substances and alloys
to the surface energy that modifies the formation of a Literature describing melting due to (slow) heating is
liquid surface layer at temperatures near Tm.42 The scarce. In the few instances, melting is treated symme-
surface strain enhances the free energy of the solid, but trically to solidification, which may be justified for the
leaves the liquid unaffected, thus favouring melting majority of cases. Since melting is controlled by the
independently of the nature and direction of the strain. kinetics in the liquid, convection may play an important
However, the onset temperature for surface melting is role. Hence, due to the difficulties in quantifying the
also influenced by the strain, and under certain effects of convection precisely, it is difficult to draw
conditions, increased up to Tm. In this case, incomplete conclusions on details of the thermodynamic state of the
surface melting may be observed or surface melting may interface. Purely diffusion controlled melting experi-
be entirely suppressed for a material where a strain free ments conducted under microgravity conditions with
surface exhibits complete surface melting. pure transparent materials yield close agreement with
A practically important question is to what degree conventional phase transformation theory. The melting
pre-melting and nucleation of liquid at an internal processes occurs for flat interfaces at Tm, the melting
surface or interface are comparable to a free surface. kinetics is governed by the heat flow in the liquid
Pre-melting at grain boundaries has been observed in Al phase.61 For curved interfaces, the melting temperature
4 K below Tm.43 However, in in situ observation of needs to be corrected for the melting point depression

International Materials Reviews 2009 VOL 54 NO 1 7


Rettenmayr Melting and remelting phenomena

50 K s21 (Ref. 67) demonstrate that melting can occur


partitionless at the T0 temperature. One of the peculia-
rities of the chosen alloy system is that the solid
diffusion of Bi in Sn close to the melting temperature
is two orders of magnitudes slower than that of the
majority of metal alloys. Also, due to the small size of
the powder particles, less defects are to be expected in
the particles than in a bulk sample. An endothermal
signal is detected near the T0 temperature for various
heating rates. A deflection of the heating curve close to
5 Concentration profiles in a sample after homogenisa- the T0 temperature is found at heating rates as low as
tion in the two-phase region and after subsequent heat- 0?1 K s21. It is concluded that the solid is super-heated
ing (left) and corresponding phase diagram (right): the until partitionless melting occurs at T0.
solid at the higher temperature is superheated Rapid heating of Nb–Ti wires (1?3 mm diameter) has
(hatched) at all locations64 been performed with heating rates between 100 and
10 000 K s21.68 Before collapsing, the Nb–Ti wires reach
due to the Gibbs–Thomson effect.62 Melting of the liquid fractions of 0?4–0?9. The temperature is determined
strongly anisotropic pivalic acid shows a peculiarity: a by simultaneous measurement of the radiance temperature
locally enhanced melting rate is due to steep gradients in and the normal spectral emissivity. The authors propose a
the mean curvature, a thermo-capillary heat flux in the model with diffusion control and local equilibrium at the
solid phase is induced.63 interface. By this model, the heating curves of material
In alloys, the situation during slow heating is more with different grain sizes are well reproduced. Maximum
complex, since it is hard to characterise to what degree melting front rates are estimated to 2 mm s21.
superheating in the solid can be achieved in a practical In conclusion of the rapid heating experiments of
case. Raising the temperature in a semi-solid alloy with alloys, it can be stated that melt nucleation in alloys will
uniform concentration in both the liquid and solid phase occur independently of the heating rate close to the
will inevitably lead to a constitutionally superheated solidus temperature at the free surface and internal
solid (Fig. 5). Such types of profiles have been predicted interfaces. With rapid heating, a finite amount of
in Refs. 1 and 3. Note that if there is uniform superheating can be achieved: direct measurement of
composition in the liquid due to the faster diffusion, the superheating appears to be problematic, and the
the solid fraction is larger than the equilibrium solid documented values are rather estimates than precise
fraction predicted by the lever rule, while during values. Different bulk melting temperatures with respect
solidification, the lever rule predicts the maximum to solidus T0 and liquids temperature have been found.
possible solid fraction. It will be discussed below that Thus, there is different melting kinetics depending on the
more mechanisms than just diffusion are available to alloy system. The dependency on the alloy system
reduce superheat in the solid, especially if other appears to be more prominent for melting than for
interfaces are in the vicinity of the considered interface. solidification. Reasons for the different behaviour of
The difficulty of achieving a superheated solid experi- different alloy systems include
mentally is presented in the section on ‘Theory on (i) the defect structure of the solid crystals
melting and superheating of pure solids’. (ii) the structure of the solid/liquid interface
(iii) the viscosity of the melt
Rapid heating of alloys (iv) the kinetics in the solid.
A few experiments are documented in the literature
where it is attempted to superheat a solid alloy by rapid Theory on melting and superheating of pure
heating. solids
Bulk samples of Sb–Bi alloys were heated with heating In addition to the different nucleation behaviour of
rates up to 30 000 K s21 by an electrical current.65,66 A liquid as compared to a crystalline solid, further criteria
plateau in the heating curve is found above the solidus have been introduced in the literature that add further
temperature, but below the T0 temperature. The T0 aspects of asymmetry between solidification and melting
temperature is defined as the temperature at which solid regarding the migration of the interface.
and liquid phase possess the same Gibbs free energy, The most quoted early concept of a melting mechan-
allowing for a partitionless phase transformation. The ism that is different from the classical nucleation/growth
estimated melting front velocity from a grain boundary to mechanism of solidification is that of Lindemann,69,
the centre is given as 2–7 mm s21 for superheatings of 40– even though the idea has been introduced in the
60 K. The sample temperature is measured with a thermo- literature earlier.70 It is supposed that at the melting
couple: averaging effects of the thermocouple are not point a surface instability of the oscillating atoms
discussed. The author shows that there can be repreciptia- occurs, i.e. that the atoms acquire enough energy and
tion of solid phase with equilibrium composition, i.e. a free volume ‘to slip past one another’ (the root mean
melting/resolidification mechanism that assists the process square displacement of the atoms reaches a critical
of reducing supersaturation: a liquid film migrates from fraction of the interatomic distance), or the crystal
the grain boundary through the supersaturated grain; ’shakes itself to pieces’. The Lindemann criterion has
behind the film the equilibrium concentration resolidifies been further elaborated in the following decades5 and is
to form a non-supersaturated crystal. still occasionally used. However, thermodynamically the
Heating experiments using small Sn–Bi particles with Lindemann criterion is incomplete, since the free energy
5–20 mm diameter with heating rates between 0?1 and of the liquid is not taken into account.

8 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

In 1939, Born71 introduced his concept of the Experiments on superheating


‘rigidity catastrophe’, a shear instability introduced Whether superheating of crystals, particularly metal
through the elastic constants that also shows that the crystals, can be achieved experimentally has been
temperature range for which the crystal can be stable is questioned for many decades. In fact, still in the 1960s,
limited. The melting point is thus defined as the authors state that ‘superheating of bulk solid metals
temperature at which the rigidity of a crystal becomes above their melting point is not observed’.86
zero. Similar to the Lindemann criterion, the Born However, already in the 1930s, it was concluded from
criterion is a one-phase theory. It was also challenged observations of Ga irradiated with a light source that
by observations of elastic moduli that failed to vanish superheating of solid Ga by 0?1 K is possible.87 Later
at Tm.72–75 on, several pure substances that form viscous melts
Frank76 introduced in 1939 more of a thermodynamic (different modifications of SiO2,88 P2O5,89 Na2O.SiO2,90
approach by comparing melting with a solid state order/ Na2S2O5.5H2O,91 As2O392) were reported to be observed
disorder transition. He postulates critical temperatures in a superheated state.
above (and below) Tm that limit the range of existence of Superheating of ice has been described in Ref. 93. It is
the superheated (or undercooled) state. These tempera- shown that superheating of ice of up to 0?4 K is possible
tures are not symmetric to Tm, which is assumed to by dielectric heating: the ice is used as the dielectric in a
explain that superheating is more difficult to attain than capacitor, liquid water nucleates at different locations
undercooling. (‘Tyndall figures’) in the bulk of the solid ice. The
The concept of upper limits for the stability of a superheating is on the one hand directly measured by a
superheated solid has subsequently be extended. The thermocouple; on the other hand, it also becomes evident
hierarchy of stability limits includes the isochoric indirectly through the ongoing growth of the liquid nuclei
(temperature at which liquid and superheated solid have after the external heating is stopped. Further evidence
the same volume),77 the isenthalpic (equal enthalpy)78 and quantification of Tyndall figures caused by light
and eventually the isentropic (equal entropy) condi- radiation into the bulk is documented in Ref. 94. The
tions.78 Subsequently it has been postulated, that before growth of the liquid regions follows conventional growth
(i.e. at temperatures below) one of the above mentioned kinetics and thermodynamics, i.e. it can be described with
instabilities would occur, thermally activated nucleation the aid of transport of latent heat, interfacial tension, and
of liquid inside the superheated crystal occurs.79 the anisotropic interfacial kinetics of ice.
The role of vacancies for the lattice instability is Superheatings and melting rates have also been
discussed in different contexts. Gorecki80 systematically quantified carefully in Ga single crystals.95 The melting
studied the possible role of vacancies for melting. He kinetics of thin Ga rods in a capillary (0?5 mm diameter)
showed that in metals the equilibrium concentration of was studied as a function of the crystallographic
vacancies at the melting point reaches a value of 0?37% orientation. Superheatings of 0?1 K were measured
at the onset of melting, independently of the crystal through the change of electrical resistivity. A quantita-
structure (fcc, bcc and hcp) of the metal. In a subsequent tive relationship between melting rate and superheating
paper,81 further correlations of the property changes due is given. It is distinguished between dislocation-free and
to vacancies and the melting temperature are pointed dislocation-assisted conditions.
out. These correlations include latent heat of melting, An important step in the achievement of higher super-
volume change upon melting and the change of electrical heatings over an extended period of time is described in
resistivity. Ref. 96, following an idea of Turnbull97 to coat the
An extension of the view of melting as being vacancy surface of a crystal in order to prevent contact of solid
induced is presented in Ref. 82. The authors study the with vapour. Ag spheres of 120–160 mm diameter were
role of vacancy diffusion rather than vacancy concen- produced and coated with Au of 11 mm thickness by
tration. They find that close to Tm a vacancy migrate to electrodeposition. The Ag spheres were heated to 25 K
such an extent that it interacts with several lattice above Tm for 1 min. Some of the Ag spheres had melted
positions at the same time (‘is decomposed by neigh- at Tm, which can be easily identified by the Au from the
bouring atoms’), leading to lattice disordering. Recently, coating that is dissolved in the Ag. Other spheres
it has been shown82 that at the equilibrium melting remained solid, as shown by the diffusional Au profile in
temperature, the vacancy concentration close to a free the Ag in the vicinity of the Au layer.
surface is approximately three times higher than in the Subsequently, substantial superheatings have been
bulk. This phenomenon is interpreted as a hint for the achieved with smaller and smaller particles down to the
role of vacancies for pre-melting. size in the nanometre range. For free particles, a size
Molecular dynamics simulations suggest that both dependent melting point depression is expected and also
heterogeneous and homogeneous melting processes are found experimentally in numerous cases. Embedded
mediated by atoms with defective coordination. Their particles can exhibit superheating, an effect that can be
concentration in the crystalline regions close to the more pronounced for non-spherical particles than for
surface at the equilibrium melting point was found to be spherical ones. The crucial role of interfaces and strains
approximately the same as in the bulk system at the limit in superheating experiments of nanoparticles has been
of superheating.84,85 demonstrated in Ref. 98 where freezing temperatures
Further defect controlled melting mechanisms are above Tm have been observed upon resolidification of
discussed in Ref. 5. In a context of studying liquid/solid melted particles. Ample literature is available in the field
phase transformations in ‘real crystals’, results on of superheated nanoparticles. The interested reader is
melting close to the theoretical limit of superheating referred to the review in Ref. 5.
are not applicable. Work on the number of atoms in a For documentation of ultrafast heating by shock wave
liquid nucleus is thus not discussed here. loading (Fe, Mo, V, Ta, W) and intense laser irradiation

International Materials Reviews 2009 VOL 54 NO 1 9


Rettenmayr Melting and remelting phenomena

(Al, Pb) with superheatings up to the order of 0?5Tm, see system (succinonitrile–acetone) is documented in
Ref. 99. Ref. 64. The authors focus on the morphology of the
liquid regions. They observe nucleation first at grain
boundaries, spherical nuclei in the solid ahead of the
Solutal melting melting front and dendrite-like melting morphology.
Constitutional superheating during directional Concluding the section on superheating of solid
melting experiments alloys, the following can be stated.
Constitutional undercooling and the stability of a planar Superheating a solid alloy is certainly more easily
solidifying solid/liquid interface were a prominent possible than for a pure substance. It appears that
research topic since the 1950s. Later on, a series of superheating is facilitated if the melt is viscous. Thus,
papers was published attempting to transfer the progress results obtained from organic alloys should not be
in this field both theoretically and experimentally to directly applied on metal alloys. The nucleation events
melting. Experiments on melting thin layers of K–Na of liquid in the solid bulk ahead of the melting front64,100
alloys100 showed that melting starts at grain boundaries show that the conditions for nucleation from a super-
(presumably due to impurities) and that there are heated solid are satisfied, but the amount of super-
independent melt nuclei ahead of the melting front, heating or supersaturation can so far only be estimated.
suggesting that a superheated region in the solid should The interpretation of the directional melting experi-
exist. ments with Sn–Sb and Sn–Bi alloys102,103 suffers from
Stability analysis similar to the Mullins-Sekerka the low spatial resolution of the concentration measure-
analysis for solidification is developed in Ref. 101. It ments and the only vaguely known position of the
interface at the time of the quench. As a consequence,
predicts a constitutional superheating criterion for
the interface composition is not measured, but con-
melting, similar to the constitutional undercooling
cluded on.
criterion for solidification. The comparison with the
For quantifying superheating unambiguously, it
experiment is not fully meaningful, as the theory is
would be desirable to have precise information about
applicable to steady state, but the experiments with the
both local temperature and local composition in the
K–Na alloy are with transient melting.
solid. Experiments with organic alloys provide readily
Directional melting experiments are documented for
the necessary temperature information, with metal
Sn–Sb alloys in Ref. 102 and for Sn–Bi alloys in
alloys (post mortem) good composition information.
Ref. 103. The microstructures are carefully analysed
At present, information about the spatial distribution of
‘post-mortem’ after quenching and compared with EDX
both temperature and composition in one and the same
measurements. The authors conclude on constitutional
system is not available.
supercooling ahead of the melting interface of up to
3?3 K over a length of up to 0?3 mm for the Sn–Sb alloy. Dissolution in a solute-rich melt
It is stated that it is impossible to attain temperature Dissolution of a solute depleted solid in a solute
gradients experimentally that render the melting front enriched melt is prone to be governed by convection.
stable (in the order of 104 K mm21 for melting rates of Hence, it is difficult to specify the experimental
0?2 mm s21), and the breakdown of the front as well as conditions directly at the interface, and quantification
melt nuclei ahead of the front are observed for the faster of experimental results is tedious.
melting rates and lower temperature gradients. For the In Ref. 106, melting experiments with transparent
Sn–Bi alloys, the superheating is not confirmed; over the systems and with Pb–Sn and Sn–Pb alloys are docu-
length of the solid that is predicted by theory to be mented. Depending on the initial conditions of the
superheated, (former) liquid droplets that are assumed experiments, the authors observe convection and diffu-
to ‘relieve the superheating’ are found. sion controlled dissolution. Convection assisted melting
The stability analysis for melting is refined in occurs with a rate that is approximately two orders of
Ref. 104. The authors point out that constitutional magnitude faster than diffusion controlled melting.
superheating predicts that a melting interface would be Gradients or interfacial conditions are not quantified
less stable than a solidifying interface, but that stability in detail. Literature concerned with the effects of
analysis predicts that it would be more stable. The thermosolutal convection on dissolution rates relevant
reason is that morphological instability is governed by to oceanographic and geological contexts is quoted, e.g.
diffusion in the liquid for both solidification and the melting of icebergs and reactions in magma
melting, but that constitutional superheating is governed chambers.107,108 It is found that the mixing enthalpy
by solid diffusion for melting and liquid diffusion for upon dissolution of a solid exerts a certain influence on
solidification. This is another point demonstrating the the melting rate and needs to be considered. Detailed
asymmetry between solidification and melting. simulation calculations considering convection and
Directional melting experiments with organic succi- mixing enthalpy are carried out in Ref. 109. The
nonitrile–water alloys are presented in Ref. 105. simulation results are compared with experiments on
Diffusion control for melting and local equilibrium at melting of ice and dissolution of Si in Fe-rich melts.
the interface is assumed, and consistency between Small solid particles (r<10 mm) of transparent
observed melting rates and a solution of the diffusion organic alloys that float in hot melt are observed in
equation is found. Composition distributions are not Ref. 110. The melting process is assumed to occur in an
assessed or discussed, but for some cases, the authors adiabatic environment, i.e. with solute or temperature
assume a superheated solid. They suggest that their gradients around the melting particles. The melting
experimental procedure can be used for determining the kinetics of the particles is assumed to be due to the
liquid diffusion coefficient. Another series of experi- kinetics of solute or heat in the liquid phase. The authors
ments on directional melting with a transparent alloy distinguish between alloys with high and low solubility

10 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

6 Melting front velocities of succinonitrile particles as a


function of temperature110

of the solvent. For alloys with high solubility, the onset


of thermally controlled melting is above the liquidus 7 Former melting front of pure Al that was in contact
temperature, but below the melting temperature of the with Al–Mg, and simulated concentration distribution of
solvent. A sharp transition between lower solutally and Mg in the melt and shape of the front114
higher thermally controlled dissolution and melting
rates respectively, is found (Fig. 5), confirming theore- local equilibrium after an instantaneous disturbance are
tical results in Ref. 111. obtained.
In a series of papers,112–114 solutal melting is studied First calculations based on the thermodynamics25 for
systematically. A pure metal is brought into contact with initial conditions leading to melting120 indicate that the
a solute rich binary alloy at a temperature where the interface properties need to be different for solidification
pure metal is still solid and the binary alloy is liquid with and melting in order to find a consistent solution.
a solute composition distinctly above the liquidus Calculations for steady-state directional melting121 yield
composition. Thus, both parts of the sample were in interface compositions that are influenced by friction, if
their one-phase field in the phase diagram respectively. during melting, the same interface mobility is assumed
A high reliability of the experimentally determined front as for earlier solidification calculations. Molecular
velocity has been achieved despite the convection dynamics and kinetic Monte-Carlo calculations, mostly
control of melting. on solidification, confirm the existence of different
In the dissolution experiments,112 a surprisingly low characteristics of interfaces.122 The dependence of
composition at a formerly melting interface was mea- the interface mobility in pure Ni123 and pure Mg124 on
sured. However, the reliability of the measurement of the the growth direction also demonstrates the impact of the
low composition is limited, as discussed in Ref. 115. In interface on growth.
Ref. 114, the possibility of an interface composition in the As a result of the solutal melting experiments and the
liquid that is lower than the equilibrium composition, i.e. subsequent development, it can be stated:
inside the two-phase field in the phase diagram, is N concentration gradients in the liquid must be con-
introduced. Good agreement between experimental and sidered when describing the kinetics of a melting
simulated shapes of the melting front is achieved (Fig. 7). process
In Ref. 115, simulation calculations based on an extre- N structure and properties of a solid/liquid interface
mum principle confirm qualitatively the possibility of appear to be different for solidification and melting
interface compositions in the liquid inside the two-phase N under diffusion control, non-equilibrium interfaces
field for the case of high interface mobilities. A high will approach equilibrium within microseconds,
mobility is certainly characteristic for liquid/solid inter- under convection control non-equilibrium interfaces
faces. The quantitative disagreement about the time span may exist longer.
to reach local equilibrium in Refs. 114 and 115 may be
attributed to the fact that the case for an instantaneous Melting and simultaneous solidification
(short) disturbance of local equilibrium is treated in
Ref. 115, while the disturbance persists over a longer time Dendrite arm coarsening and dendrite
span in Ref. 114. fragmentation
As a result of the solutal melting experiments112–114 Dendrite arm coarsening is an effect that is driven by the
and the subsequent discussion,116,117 the thermody- reduction of surface energy, involving a remelting/
namics of Baker and Cahn was further elaborated, resolidification mechanism. In the literature, coarsening
yielding the possibility for the quantification of the is generally described in terms of a change of secondary
interfacial compositions in the adjacent phases, as dendrite arm spacing. For this, several morphological
described above and in Ref. 25. Quantification of the changes have been proposed (Fig. 8). A secondary arm
interfacial compositions based on Ref. 25 has been can melt back radially,125 axially,126 melt off at its
attempted for rapid solidification.118,119 In Ref. 119, the root,126,127 or two secondary arms can coalesce.128 By
same initial conditions as in Ref. 115 are treated and definition, a coarsening process does not involve any
similar results concerning the time span to re-adjust change of the phase fractions. The melting of secondary

International Materials Reviews 2009 VOL 54 NO 1 11


Rettenmayr Melting and remelting phenomena

observed in a transparent alloy under forced convection


conditions, with the conclusion that ‘shear forces cannot
be completely discounted’ in Ref. 139. In solidification
experiments from undercooled droplets in levita-
tion,140,141 different causes for fragmentation in Ni–Cu
alloys at low and high undercoolings are assumed: while
constitutional melting is effective at low undercoolings,
mechanical shearing is suggested to govern the frag-
mentation process at high undercoolings. Other authors
assert that only remelting can be the cause for dendrite
fragmentation.142
Melting at the dendrite root leading to fragmentation
8 Coarsening mechanisms as proposed by different is prevalent if the solidification front velocity is
authors decelerated.133,142–144 While recalescence is assumed to
be responsible for the fragmentation in Ref. 133, no
arms in the context of coarsening must thus be evidence is found for thermal recalescence in Ref. 144. It
accompanied by simultaneous solidification at regions is thus concluded that an increase in solute composition
with low curvature convex shape and/or concave in the mushy zone causes the detachment of dendrite
shape. arms.
In real dendritic microstructures, all of the above Calculations considering fluid flow around a dendrite
mentioned mechanisms have been observed, even revealed that under non-forced convective conditions,
though the melting off at the roots126 is rarely seen fragmentation is unlikely to occur. This has been shown
during steady state dendritic growth. It should be noted in an analytical model145 and simulation calcula-
that in the majority of cases, coarsening only occurs tions.14,146 Under rapid solidification conditions, the
once the secondary arms are overgrown by their forces of strong fluid flow on fine structures may be
neighbours, or the growth is stopped by the neighbour- sufficient for at least mechanical deformation of dendrite
ing dendrite(s). Once the large turnover of energy due to arms.147 By mechanical deformation (bending), a
solidification comes to an end, the weaker contributions dendrite arm would reduce its cross-section toward the
of interfacial energy become effective, and the over- flowing melt and thus avoid to be sheared off. Further
grown dendrite arms start melting back, generally at the conditions for mechanical deformation have been
same time radially and axially. identified,148 and experimental evidence for bent den-
Coarsening increases the diffusion distances during drites is given in Cu–Sn and Cu–O alloys. Subsequently,
solidification in the liquid and solid phases. This might it has been shown149 that misorientations in dendrites
lead to a less homogeneous microstructure, since the due to bending are far from being a rare incidence.
homogenising effect of back diffusion is reduced. On the In recent experiments in Al–Cu that were documented
other hand, the remelting/resolidification mechanism of with high resolution synchrotron radiation,150 fragmen-
coarsening leads to the opposite, homogenising effect, tation could be observed in situ. It is shown how solute
since the compositions of the remelting/resolidifying flow into the mushy zone, solute advection and solute
phase fractions are slightly different. Generally, the net rejection due to local growth lead to solute pile-up at the
effect of coarsening is a homogenising one. There is also roots of the secondary dendrite arms. This solute pile-up
an influence on the solidification path and potentially on alters the local compositional balance at the solid/liquid
the type of secondary phases that crystallise toward the interface at the root, causing the root to remelt, which in
end of solidification.129 Various work is concerned with turn results in further branch detachment.
these aspects of coarsening in binary130 and multi- In conclusion, it appears that melting phenomena are
component alloys.131,132 the most prominent and most frequently observed cause
Both melting of secondary dendrite arms at the root for dendrite fragmentation. However, being aware of
and mechanical breakdown have been discussed as cause this, dendrite fragmentation is still hard to be quantified,
for dendrite fragmentation and thus for the columnar- as several causes for melting can be named that might
to-equiaxed transition in castings and for microstruc- occur on different time scales. All of these mechanisms
tural defects like freckles or stray crystals. have been evidenced experimentally. Particularly, there
In early experiments on fragmentation, it has been are
found that local remelting is the most likely cause for (i) capillary driven adjustment of local
fragmentation.133 The authors find evidence that radii14,133,151
excludes mechanical shear forces as major cause: when (ii) solute rejection due to local solidification in the
they start mechanical stirring in the liquid phase of their vicinity of the dendrite arm roots150
transparent system, there is a delay between the onset of (iii) flow of solute rich melt into the mushy zone150
stirring and the generation of dendrite fragments. In (iv) flow of melt from hot regions toward the mushy
further studies on transparent alloys,134,135 local melting zone
was confirmed as fragmentation mechanism. However, (v) recalescence14,133
later on, some work was published where at least in (vi) back diffusion in the solid dendrite in multi-
certain melt, velocity ranges a correlation between melt component alloys that changes the local solidus
flow and fragmentation was found.136,137 Other authors temperature.9
assume a stronger influence of melt flow and state that Fully understanding and predicting the interplay of the
grain refinement is not controlled by the breaking of six mechanisms mentioned above will require sophisti-
dendrites by shear forces alone.138 Fragmentation is cated simulation calculations.

12 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

liquidus line. TGZM is not an important mechanism for


the coarsening of dendritic structures for technical
casting conditions,163 while it may be significant for a
combination of a high temperature gradient and a low
solidification velocity.
Besides the effects on dendrite arms and liquid
droplets, TGZM can contribute to the formation of
macroscopic concentration gradients in a mushy zone in
a (steep) temperature gradient. This mechanism is first
discussed in Ref. 164. The transport of solute atoms
out of the mushy zone in the liquid is described in
Ref. 165.
Direct observation of TGZM by synchrotron radia-
tion is reported for Al–Ni166 and for Sn–Bi.167 Samples
are partially melted in a (steep) temperature gradient,
and resolidification of the mushy zone to form solid
phase with a macroscopic concentration gradient is
tracked. In some cases, detachment of secondary
dendrite arms is observed.167 The evolution of the
macroscopic concentration gradient is characterised for
the purpose of defining good initial conditions for
subsequent directional solidification.168 It is shown that
the boundary condition at the outer boundary of the
mushy zone toward the liquid is dependent on convec-
9 Schematic diagram of the solute distribution in a den- tion in the melt. The evolution inside the mushy zone
drite arm along the line AB during solidification in a yields similar concentration profiles for experiments on
temperature gradient161 earth and in space. After similar melting experiments in
a temperature gradient,169 post-mortem analysis of the
Temperature gradient zone melting concentration distribution and estimation of the rate of
the kinetic processes, the amount of solute that diffuses
In the presence of steep temperature gradients, a
out of the mushy zone lead to the conclusion that
melting/resolidification mechanism can occur involving
TGZM and liquid diffusion cannot be the only physical
either a liquid droplet migrating in the solid phase, or a
processes that lead to the macroscopic gradients.
sequence of dendrite arms moving in the liquid phase, in
either case toward the hot zone of the sample. Not Liquid film migration
considering thermotransport, the kinetics of TGZM is The movement of a thin liquid zone in a mostly solid
controlled by diffusion in the liquid phase: solute atoms structure is being referred to as LFM. It is generally
diffuse in the liquid from colder regions with high accepted that grain boundaries are favourable sites for
concentration toward regions with higher temperature at the nucleation of liquid. Considering that melting starts
lower concentrations. The consequence is a depletion of by the formation of a liquid layer between the grains, it
solute atoms at the colder side (leading to local is expected that at the interface in both the liquid and the
solidification) and an enrichment at the warmer side solid phase the equilibrium compositions as given by the
(leading to local remelting). As opposed to coarsening, phase diagram are found (Fig. 10). If the local geometry
the process is not driven by the reduction of surface is perfectly symmetric, the liquid layer should be
area, but by solutal effects that are established due to the stationary at constant temperature. If the symmetry is
adjustment of equilibrium concentrations at the inter- broken, it starts to migrate. Reasons for the broken
face (Fig. 9). symmetry might be fluctuations, concentration differ-
TGZM was observed in arctic ice152 in 1926. For the ences across the liquid layer, coherency strains or
case of migration of dendrite arms TGZM was first anisotropic interfacial energy.
described by Pfann,153 for the case of the migration of It is generally taken as evidence for LFM if a grain
liquid droplets by Tiller.154 boundary is growing against the curvature pressure, or if
TGZM of dendrite arms leads to an asymetric (saw- there are regions positive and negative curvature along
tooth like) concentration distribution in the dendrite one and the same grain boundary (Fig. 11). Triple
arms, as e.g. measured in unidirectionally cast copper junctions are occasionally mentioned as a cause of
alloys,155 in investigations by different authors in curved grain boundaries, as the triple junctions are
steels156–160 or in Al alloys.161 In organic alloys, it has assumed to be less mobile than the ordinary straight or
been reported that secondary dendrite arms migrate slightly curved boundary. This is not in accordance with
along the primary stem towards the dendrite tip up to a grain growth theory where it is assumed that grain
distance of four dendrite arm spacings.162 growth is controlled by triple junctions at low tempera-
An equation is given for the migrating velocity of the ture only.
geometric centre of the dendrite arm vm as162 It is worth noting that the liquid film is highly mobile,
DGT as enthalpic effects are not involved in the process: the
vm ~ (7) latent heat needed for melting the solid at one side of the
mL X L (1{keq )
film is provided by solidification at the other side.
where D is the diffusion coefficient in the liquid, GT is Different mechanisms are discussed in the literature
the temperature gradient and mL is the slope of the that lead to a migration of the liquid film. Initially,

International Materials Reviews 2009 VOL 54 NO 1 13


Rettenmayr Melting and remelting phenomena

10 Concentration distribution at a liquid layer between two grains moving to the right, suggested with coherency strains
as driving force171 (left), suggested with supersaturation as driving force180 (right)

coherency strains in the solid phase on one side of the In several works, it is assumed that different
film were assumed to be the driving force for LFM. coherency strains in two adjacent grains are due to
However, an estimation170 leads to the conclusion that composition differences across the grain boundary. The
solely the reduction of coherency strains is not sufficient higher solute content in the dissolving grain entails a
to drive the liquid layer. It is pointed out that coherency higher coherency strain that is reduced by the liquid film
strains can only be responsible for the film migration at and reprecipitation in the growing grain with a different
the initial state, and that the anisotropy of interfacial composition. Similar effects as LFM are diffusion
energy can be symmetry breaking.171 induced grain boundary migration (DIGM),175 chemi-
Impurities appear to influence the migration rate cally induced grain boundary migration (CIGM)176 or
significantly.172 An Inhomogeneous distribution of Fe discontinuous precipitation.177 It is assumed that they
impurities in Al–Cu–Mg alloys is assumed to be the are similarly based on different coherency strains in the
cause for a reduction of the migration rate and for adjacent grains due to a varying composition.
the detachment of liquid droplets that remain behind the It is, however, likely that chemical driving forces
liquid film. The authors in Ref. 172 speculate about should be the stronger ones as compared to interfacial
effects of Fe on the coherency strain. Recent analytical pressure or coherency strains close to the melting point.
models for LFM173,174 also assume coherency strains as In alloy systems with finite solubility of the solute, for
driving force and include the anisotropy of the inter- thermodynamic reasons, there must be a temperature
facial energy. dependence of the solubility. In a precisely known
solidus temperature as a function of solute composition
in the phase diagram the solid can thus be slightly
superheated, so that melting occurs only at the grain
boundaries and not in the interior of the grains. This has
been used in Ni alloys to generate migrating liquid
films.178 The liquid film grows in thickness if there is a
superheated region around the liquid film,179 since the
excess solute of the solid phase will remain in the liquid
as the liquid film passes by.
Processes similar to LFM, but certainly driven by the
reduction of supersaturation, have been identified
during melting and resolidification in a (steep) tempera-
ture gradient.169,180 Different initial microstructures
(single crystal, equiaxed fine grained and columnar) give
insight in the role of LFM. LFM occurs both laterally,
i.e. isothermally perpendicular to the direction of the
heat flow, and longitudinally along the temperature
11 Resolidified microstructure in Al–Cu with curved grain gradient. In the longitudinal direction, the temperature
boundary after LFM (right arrow) and former position of gradient breaks the symmetry at the migrating film.
grain boundary (diffuse ‘ghost boundary’, left arrow) LFM along the temperature gradient occurs at a higher

14 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

velocity than laterally, leading to elongated grains. particularly defect controlled melting and the melting/
Mean interface velocities in the order of 1 mm s21 have resolidification mechanism as in TGZM or LFM
been determined. N further development of theories of melting and
In conclusion of LFM, it can be stated that the interfaces as well as sophisticated simulation calcula-
melting/resolidification mechanism will assist the reduc- tions are needed to predict microstructure evolution
tion of superheat/supersaturation at a solid/liquid inter- directly, as in phase field models, if local remelting is
face, if other solid/liquid interface are in the vicinity. A involved.
variety of symmetry breaking effects can cause a liquid Open problems with technical relevance are effects in
layer at a grain boundary to start migrating. For multicomponent alloys, and detailed studies of the
modelling, the driving forces need to be identified interaction of temperature field, composition field and
unambiguously. It is unlikely that there will be a capillarity that lead to fragmentation.
‘general LFM model’ that describes the migration of a Non-equilibrium melting can at present only be
liquid film for different boundary conditions. simulated for small deviations from local equilibrium.
Further development of irreversible thermodynamics
concerning larger potential jumps as during joining
Conclusion and outlook processes with dissimilar materials (soldering, brazing) is
Melting and remelting phenomena are frequently necessary to find consistent results. All continuum
occurring in a variety of processes and exert a strong models known to the author assume a linear relationship
influence on the microstructural evolution in numerous between driving force and process rate, a relationship
cases. The technical relevance of melting is more and that has been confirmed only for small driving forces.
more acknowledged, but until present, mostly scientific No attempts are known to the author to model defect
questions are the subject of research. Most of the controlled melting at very low superheatings, i.e. in a
research work deals with pure substances or binary practical context. Atomistic modelling for further
alloys, leaving numerous basic technical questions quantification of the influence of line defects (disloca-
unanswered. tions) and area defects (internal interfaces, stacking
Melting and solidification are not symmetric in several faults, etc.) could be a first step; for practical purposes, a
structural, thermodynamic and kinetic aspects. This continuum theory considering defects would be highly
becomes visible mostly for rapid processes under a loss desirable. The theoretical development also needs to be
of local equilibrium. For slow melting front velocities sustained by experimental progress.
and diffusion control, diffusion in the solid is generally Detailed studies with molecular dynamics or kinetic
sufficiently fast to warrant local equilibrium at the Monte-Carlo simulations are required concerning the
interface. However, even in this case, symmetric treat- structure of the interface. It appears that the variability
ment of melting and solidification can lead to erroneous of the interface is a question that has so far not be
results, as super-heating in the solid has been predicted addressed sufficiently. It appears to be impossible to get
in casting simulations, but is unlikely to be observed in further insight into the structure of a moving interface
reality. The kinetics of the process of reducing super- through direct observation in experiments. Atomistic
heating in a solid alloy is not yet fully understood. This modelling or precise knowledge of the consequences of
concerns not only the melting/resolidification mechan- varying interfaces structures is necessary.
ism as in LFM, but also defect controlled melting at a Documented research on melting started more than a
liquid/solid interface. century ago with the fundamental question why super-
The mobility of the solid/liquid interface appears to be heatings are generally not observed during melting. The
variable, depending on the alloy system and the focus of research soon shifted to extreme cases and
direction of migration of the interface. For solidifica- conditions, e.g. suggesting the different ‘catastrophes’ to
tion, there is a dependence on the crystallographic be responsible for the asymmetry of melting and
direction of migration. Furthermore, there is the solidification. While through the research on extreme
possibility that for a melting interface both thickness conditions certainly substantial progress has been made
and structure are different from a solidifying interface. in the basic understanding of melting, there are still
When modelling the thermodynamics of the interface plenty of open questions concerning the details of the
during a phase transformation, the difference of kinetic first fundamental question and the quantification of
coefficients in the adjacent phases needs special con- melting in practical cases.
sideration. It has been shown that a fast kinetics (high
kinetic coefficients) especially in the product phase can References
change the interface compositions and thus concentra-
1. M. Rappaz and V. Voller: Metall. Mater. Trans. A, 1990, 21A,
tion gradients at the interface and the kinetics of the 749–753.
phase transformation process. Further clarification is 2. J. Campbell: in ‘Modeling of casting, welding and advanced
expected from continuum models based on the interface solidification processes’, (ed. M. Cross and J. Campbell), 907–914;
kinetics25 and extremum principle models. 1995, Warrendale PA, TMS.
3. H. Combeau, J.-M. Drezet, A. Mo and M. Rappaz: Metall.
When modelling remelting during solidification in the Mater. Trans. A, 1997, 27A, 2314–2327.
simulation of casting or microstructure evolution in 4. J. G. Dash: Contemp. Phys., 1989, 30, 89–100.
general, one should consider that: 5. Q. S. Mei and K. Lu: Prog. Mater. Sci., 2007, 52, 1175–1262.
N melting is controlled by the kinetics in the product 6. J. Campbell: ‘Castings’; 1991, Oxford, Butterworth-Heinemann.
7. E. E. Emley: Int. Met. Rev., 1976, 21, 75–115.
(liquid) phase
N superheating is unlikely to be observed in real
8. J. Szekely, Y. K. Chuang and J. W. Hlinka: Metall. Trans., 1972,
3, 2825–2833.
crystals; in addition to solute diffusion, there are 9. B. Hallstedt: in ‘The SGTE casebook: thermodynamics at work’,
more mechanisms that assist to reduce superheating, (ed. K. Hack), 2nd edn; 2008, Cambridge, Woodhead Publishing.

International Materials Reviews 2009 VOL 54 NO 1 15


Rettenmayr Melting and remelting phenomena

10. N. van den Bogaert and F. Dupret: J. Cryst. Growth, 1996, 166, 56. A. M. Alsayed, L. M. F. Islam, I. J. Zhang, L. P. J. Collings and
446–451. A. G. Yodh: Science, 2005, 309, 1207–1210.
11. A. M. Mullis and R. F. Cochrane: J. Appl. Phys., 1997, 82, 3783– 57. J. F. Lutsko, D. Wolf, S. R. Phillpot and S. Yip: Phys. Rev. B,
3790. 1989, 40, 2841–2855.
12. M. Schwarz, A. Karma, K. Eckler and D. Herlach: Phys. Rev. 58. S. R. Phillpot, J. F. Lutsko and D. Wolf: Solid State Commun.,
Lett., 1994, 73, 1380–1383. 1989, 70, 265–268.
13. M. A. Martorano, C. Beckermann and C.-A. Gandin: Metall. 59. T. Nguyen, P. S. Ho, T. Kwok, C. Nitta and S. Yip: Phys. Rev. B,
Mater. Trans. A, 2003, 34A, 1657–1674. 1992, 46B, 6050–6060.
14. R. H. Mathiesen, L. Arnberg, P. Bleuet and A. Somogyi: Metall. 60. S. R. Phillpot, J. F. Lutsko, D. Wolf and S. Yip: Phys. Rev. B,
Mater. Trans. A, 2006, 37A, 2515–2524. 1989, 40B, 2831–2840.
15. S. C. Flood and J. D. Hunt: ‘ASM metals handbook’, Vol. 15, 61. M .E. Glicksman, A. Lupulescu and M. B. Koss: J. Thermophys.
‘Casting’, 30; 1990, Metals Park, OH, ASM. Heat Transfer, 2005, 17, 69–76.
16. M. C. Schneider, J. P. Gu, C. Beckermann, W. J. Boettinger and 62. M. E. Glicksman, A. Lupulescu and M. B. Koss: J. Cryst. Growth,
U. Kattner: Metall. Mater. Trans. A, 1997, 28A, 1517–1531. 2005, 276, 459–465.
17. H. Yasuda, I. Ohnaka, K. Kawasaki, A. Sugiyama, T. Ohmichi, J. 63. M. E. Glicksman, A. Lupulescu and M. B. Koss: in ‘Free
Iwane and K. Umetani: J. Cryst. Growth, 2005, 262, 645–652. boundary problems’, 219–230; 2007, Basel, Birkhäuser.
18. W. J. Boettinger and S. R. Corriell: in ‘Science and technology of 64. D. Benielli, N. Bergeon, H. Jamgotchian, B. Billia and P. Voge:
the undercooled melt’, (ed. P. R. Sahm et al.); 1986, Dordrecht, Phys. Rev. E, 2002, 65E, 051604
Martinus Nijhoff Publications. 65. Y. S. Koucharenko: Izv. Akad. Nauk SSSR Ser. Met., 1973, 6,
19. M. Aziz and M. Kaplan: Acta Metall., 1988, 36, 2335–2347. 116.
20. M. Hillert: Acta Mater., 1999, 47, 4481–4505. 66. Y. S. Koucharenko: Fiz. Metal. Metaloved., 1975, 39, 815–820.
21. J. Svoboda, J. Vala, E. Gamsjäger and F. D. Fischer: Acta Mater., 67. W. P. Allen, H. J. Fecht and J. H. Perepezko: Scr. Metall., 1989,
2006, 54, 3953–3960. 23, 643–648.
22. Baker, J. Cahn: in ‘Solidification’, 23–58; 1971, Metals Park OH, 68. D. Basak, W. J. Boettinger, D. Josell, S. R. Coriell, J. L. McClure,
ASM. S. Krishnan and A. Cezairliyan: Acta Mater., 1999, 47, 3174–
23. L. Onsager: Phys. Rev., 1931, 37, 405–426. 3158.
24. L. Prigogine: ‘Thermodynamics of irreversible processes’; 1961, 69. F. A. Lindemann: Z. Phys., 1910, 11, 609.
New York, Interscience Publishers. 70. W. Sutherland: Philos. Mag., 1891, 32, 42.
25. M. Hillert and M. Rettenmayr: Acta Mater., 2003, 51, 2803–2809. 71. M. Born: J. Chem. Phys., 1939, 7, 591.
26. B. Chalmers: ‘Principles of solidification’, 1964, New York, J. 72. O. D. Slagle and H. A. McKinstry: J. Appl. Phys., 1967, 38, 437–
Wiley. 445.
27. K. A. Jackson: ‘Liquid metals and solidification’, 174; 1958, 73. O. D. Slagle and H. A. McKinstry: J. Appl. Phys., 1967, 38, 446–
Cleveland, OH, ASM. 450.
28. F. Delogu: Nanotechnology, 2008, 19, 175703. 74. O. D. Slagle and H. A. McKinstry: J. Appl. Phys., 1967, 38, 451–
29. K. Fujiwara, K. Maeda, N. Usami, G. Sazaki, Y. Nose, A. 458.
Nomura, T. Shishido and K. Nakajima: Acta Mater., 2008, 56, 75. L. Hunter and S. Siegel: Phys. Rev., 1942, 61, 84–90.
2663–2668.
76. F. G. Frank: Proc. R. Soc. Lond. A, 1939, 170A, 182–189.
30. A. Deal, E. Balikci and R. Abbaschian: Metall. Trans. A, 2007,
77. J. L. Tallon: Nature, 1989, 342, 658–660.
38A, 100–115.
78. H. J. Fecht and W. L. Johnson: Nature, 1989, 334, 50–51.
31. B. Dutta and M. Rettenmayr: Mater. Sci Technol., 2002, 18,
79. Z. H. Jin, P. Gumbsch, K. Lu and E. Ma: Phys. Rev. Lett., 2001,
1428–1434.
87, 1–4.
32. F. Q. Zu, X. F. Li, L. J. Guo, H. Yang, X. B. Qin and Z. G. Zhu:
80. T. Gorecki:, Z. Metallkd, 1974, 65, 426–431.
Phys. Lett. A, 2004, 324A, 472–478.
81. T. Gorecki: Z. Metallkd, 1976, 67, 269–273.
33. W. D. Kaplan and Y. Kauffmann: Annu. Rev. Mater. Res., 2006,
82. L. W. Wang, L. Zhang and K. Lu: Philos. Mag. Lett., 2005, 85,
36, 1–48.
213–219.
34. D. A. Porter and K. E. Easterling: ‘Phase transformations in
83. Q. S. Mei and K. Lu: Philos. Mag. Lett., 2008, 88, 203–211.
metal and alloys’, 2nd edn; 1992, London, Chapman and Hall.
35. B. Pluis, D. Frenkel and J. F. van der Veen: Surf. Sci., 1990, 239, 84. F. Delogu: Mater. Sci. Eng. A, 2006, A416, 33–39.
282–300 and references therein. 85. F. Delogu: J. Phys. Chem. B, 2006, 110B, 12645–12652.
36. M. Faraday: Proc. R Soc. Lond., 1859/1860, 10, 440. 86. M. Fine: ‘Introduction to phase transformations in condensed
37. J. W. M. Frenken, P. M. J. Maree and J. F. van der Veen: Phys. systems’; 1964, London, Macmillan.
Rev. B, 1986, 34B, 7506–7516. 87. M. Volmer and O. Schmidt: Z. Phys. Chem., 1937, 35B, 467–480.
38. A. W. D. van der Gon, R. J. Smith, J. M. Gay, D. J. O’Connor 88. N. G. Ainslie, J. D. Mackenzie and D. Turnbull: J. Phys. Chem.,
and J. F. van der Veen: Surf. Sci., 1990, 227, 143–149. 1961, 65, 1718–1724.
39. D. W. Oxtoby: Nature, 1990, 347, 725–730. 89. R. L. Cormia, J. D. Mackenzie and D. Turnbull: J. Appl. Phys.,
40. M. C. Desjonqueres and D. Spanjaard: ‘Concepts in surface 1963, 34, 2239–2244.
physics’; 1996, Berlin, Springer. 90. G. S. Meiling and D. R. Uhlmann: Phys. Chem. Glasses, 1967, 8,
41. M. B. Baker and J. G. Dash: J. Cryst. Growth, 1989, 97, 770–776. 62–68.
42. U. Tartaglino amd E. Tosatti: Surf. Sci., 2003, 532, 623–627. 91. I. Gutzow, A. Razpopov and R. Kaischew: Phys. Status Solidi,
43. W. I. Pumphrey and J. V. Lyons: Nature, 1949, 163, 960–961. 1970, 1, 159.
44. T. E. Hsieh and R. W. Baluffi: Acta Metall., 1989, 37, 1637–1644. 92. I. Karutz and I. N. Stranski: Z. Anorg. Allgem. Chem., 1957, 292,
45. R. Lipowsky: Phys. Rev. Lett., 1986, 57, 2876–2876. 330–342.
46. M. Tang, W. C. Carter and R. M. Cannon: J. Mater. Sci., 2006, 93. M. Käss and S. Mangun: Z. Kristallogr., 1961, 116, 354–370.
23, 7691–7695. 94. S. Takeya: Appl. Phys. Lett., 2006, 88, 074103.
47. G. Ciccotti, M. Guillope and V. Pontikis: Phys. Rev. B, 1983, 27B, 95. G. J. Abbaschian and S. F. Ravitz: J. Cryst. Growth, 1975, 28, 16–
5576–5585. 20.
48. G. Besol and O. G. Mouritsen: Phys. Rev. B, 1994, 50B, 6573– 96. J. Daeges, H. Gleiter and J. Perepezko: Phys. Lett. A, 1986, 119A,
6576. 79–82.
49. J. Q. Broughton and G. H. Gilmer: Phys. Rev. Lett., 1986, 56, 97. D. Turnbull: J. Chem. Phys., 1952, 20, 411–424.
2692–2695. 98. H. Rösner, P. Scheer, J. Weissmüller and G. Wilde: Philos. Mag.
50. A. E. Lobkovsky and J. A. Warren: Physica D, 2002, 164D, 202– Lett., 2003, 83, 511–523.
212. 99. S. N. Luo and T. J. Ahrens: Appl. Phys. Lett., 2003, 82, 1836–
51. W. Fan and X. G. Gong: Phys. Rev. B, 2005, 72B, 064121. 1838.
52. J. Luo, V. K. Gupta, D. H. Yoon and H. M. Meyer: Appl. Phys. 100. D. P. Woodruff and A. J. Forty: Philos. Mag., 1967, 15, 985–993.
Lett., 2005, 87, 231902. 101. D. P. Woodruff: Philos. Mag., 1968, 17, 283–294.
53. S. F. Edwards and M. Warner: Philos. Mag., 1979, 40, 257–278. 102. J. P. Verhoeven and E. D. Gibson: J. Cryst. Growth, 1971, 11, 29–
54. L. Burakovsky, D. L. Preston and R. R. Silbar: Phys. Rev. B, 38.
2000, 61B, 15011–15018. 103. J. P. Verhoeven and E. D. Gibson: J. Cryst. Growth, 1971, 11, 39–
55. L. Burakovsky, D. L. Preston and R. R. Silbar: J. Appl. Phys., 49.
2000, 88, 6294–6301. 104. H. S. Chen and K. A. Jackson: J. Cryst. Growth, 1971, 8, 184–190.

16 International Materials Reviews 2009 VOL 54 NO 1


Rettenmayr Melting and remelting phenomena

105. X. Wan, Q. Han and J. D. Hunt: Metall. Mater. Trans. A, 1998, 142. A. Hellawell, S. Liu and S. Z. Lu: J. Met., 1997, 3, 18–20.
29A, 751–755. 143. H. Yasuda, I. Ohnaka, K. Kawasaki, A. Sugiyama, T. Ohmichi,
106. D. Fang and A. Hellawell: J. Cryst. Growth, 1988, 92, 364–370. J. Iwane and K. Umetani: J. Cryst. Growth, 2005, 262, 645–652.
107. H. E. Huppert and J. S. Turner: J. Fluid Mech., 1980, 100, 367– 144. S. Liu, S. Z. Lu and A. Hellawell: J. Cryst. Growth, 2002, 234,
384. 740–750.
108. R. C. Kerr: J. Fluid Mech., 1994, 280, 255–285, 287–302. 145. J. Pilling and A. Hellawell: Metall. Mater. Trans. A, 1996, 27A,
109. H. F. Hu and S. A. Argyropoulos: Metall. Mater. Trans. B, 1997, 229–232.
28B, 135–148. 146. B. Billia, N. Bergeon, H. Nguyen–Thi, H. Jamagotchian, J.
110. Q. Han and A. Hellawell: Metall. Mater. Trans. B, 1997, 28B, Gastaldi and G. Grange: Phys. Rev. Lett., 2004, 93, 126105.
169–173. 147. A. M. Mullis, D. J. Walker, S. E. Battersby and R. F. Cochrane:
111. A. W. Woods: J. Fluid Mech., 1992, 239, 429–448. Mater. Sci. Eng. A, 2001, A304, 245–249.
112. M. Rettenmayr, O. Warkentin and H. E. Exner: Z. Metallkd, 148. K. Dragnewski, A. M. Mullis, D. Walker and R. F. Cochrane:
1997, 88, 617–619. Acta Mater., 2002, 50, 3743–4745.
113. B. Dutta and M. Rettenmayr: Metall. Mater. Trans. A, 2000, 31A, 149. R. D. Doherty: Scr. Mater., 2003, 49, 1219–1222.
2713–2720. 150. D. Ruvalcaba, R. H. Mathiesen, D. G. Eskin, L. Arnberg and L.
114. M. Rettenmayr, O. Warkentin, M. Rappaz and H. E. Exner: Acta Katgerman: Acta Mater., 2007, 55, 4287–4292.
Mater., 2001, 49, 2499–2510. 151. E. R. Rubinstein: ‘Dendritic growth kinetics and structure’, PhD
115. E. Gamsjäger, J. Svoboda, F. D. Fischer and M. Rettenmayr: thesis, RPI, Troy, NY, USA, 1989 (cited in Ref. 139).
Acta Mater., 2007, 55, 2599–2607. 152. W. G. Whitman: Am. J. Sci., 1926, 211, 126–132.
116. M. Hillert and J. Agren: Scr. Mater., 2002, 46, 455–457. 153. W. G. Pfann : Trans. AIME, 1955, 203, 961–964.
117. M. Rettenmayr: Scr. Mater., 2003, 48, 315–319. 154. W. A. Tiller: J. Appl. Phys., 1963, 34, 2757–2767.
118. M. Buchmann and M. Rettenmayr: Scr. Mater., 2007, 57, 169– 155. F. Weinberg and E. Teghtsoonian: Metall. Trans., 1972, 3, 93–
172. 111.
119. M. Buchmann and M. Rettenmayr: Scr. Mater., 2008, 58, 106– 156. A. Türkeli and D. H. Kirkwood: in ‘Solidification and gravity’,
109. (ed. A. Roósz and M. Rettenmayr), 149–156; 1996, Zürich,
120. M. Buchmann and M. Rettenmayr: Int. J. Mater. Res., 2008, 99, Transtec Publications.
613–617. 157. H. Frederiksson: Metall. Trans., 1972, 3, 2989–2997.
121. M. Buchmann and M. Rettenmayr: J. Cryst. Growth, to be 158. L. Schmidt and H. Fredriksson: Ironmaking Steelmaking, 1975,
published. 1, 61–67.
122. K. A. Jackson: Interf. Sci., 2002, 10, 159–169. 159. B. A. Rickinson and D. H. Kirkwood: in ‘Solidification and
123. D. Y. Sun, M. Asta and J. J. Hoyt: Phys. Rev. B, 2004, 69B, casting of metals’, 44–52; 1979, London, The Metals Society.
024108. 160. E. Schürmann, S. Baumgartl, L. Nedeljkovic and M. Tripkovic:
124. Z. G. Xia, D. Y. Sun, M. Asta and J. J. Hoyt: Phys. Rev. B, 2007, Steel Res., 1987, 58, 498–502.
75B, 012103. 161. D. J. Allen and J. D. Hunt: Metall. Trans. A, 1976, 7A, 767–770.
125. T. Z. Kattamis, J. C. Coughlin and M. C. Flemings: Trans. 162. F. Y. Xie, T. Kraft, Y. Zo, C. H. Moon and Y. A. Chang, Acta
AIME, 1967, 239, 1504–1511. Mater., 1999, 47, 489–500.
126. M. Kahlweit: Scr. Metall., 1968, 2, 251–254. 163. E. Halder and H. E. Exner: Acta Metall., 1988, 36, 1665–1668.
127. A. Papapetrou: Z. Kristallogr., 1935, 92, 89–130. 164. T. A. Lograsso and A. Hellawell: J. Cryst. Growth, 1984, 66, 531–
128. K. H. Chien and T. Z. Kattamis: Z. Metallkd, 1970, 61, 475–479. 540.
129. M. Rappaz and M. Rettenmayr: Curr. Opin. Solid State Mater. 165. J. Beech: J. Cryst. Growth, 1984, 67, 385–390.
Sci., 1998, 3, 275–282. 166. H. Nguyen-Thi, G. Reinhart, A. Buffet, T. Schenk, N.
130. A. Roósz, E. Halder and H. E. Exner: Mater. Sci. Technol., 1986, Mangelinck-Noel, H. Jung, N. Bergeon, B. Billia, J. Hartwig
2, 1149–1155. and J. Baruchel: J. Cryst. Growth, 2008, 310, 2906–2914.
131. R. Beaverstock: in ‘Solidification processing 1997: proceedings of 167. B. Li, H. Brody and A. Kazimirov: Metall. Mater. Trans. A, 2006,
the 4th Decennial Conference on Solidification processing’, (ed. J. 37A, 1039–1044.
Beech and H. Jones), 321–324; 1997, Sheffield, University of 168. H. Nguyen-Thi, B. Drevet, J. M. Debierre, D. Camel, Y. Dabo
Sheffield. and B. Billia: J. Cryst. Growth, 2003, 253, 539–548.
132. B. Dutta and M. Rettenmayr: Mater. Sci. Eng. A, 2000, A283, 169. U. Bösenberg, M. Buchmann and M. Rettenmayr: J. Cryst.
218–224. Growth, 2007, 304, 281–286.
133. K. A. Jackson, J. D. Hunt, D. R. Uhlmann and T. P. Seward: 170. M. Kuo and R. A. Fournelle: Acta Metall. Mater., 1991, 39,
Trans. TMS AIME, 1966, 236, 149–158. 2835–2845.
134. M. S. Christenson and F. P. Incropera: Int. J. Heat Mass 171. J. S. Kirkaldy: Acta Mater., 1998, 46, 5127–5133.
Transfer, 1989, 30, 47–68. 172. W. A. G. MacPhee, G. B. Schaffer and J. Drennan: Acta Mater.,
135. M. S. Christenson, W. B. Bennon and F. P. Incropera: Int. J. Heat 2003, 51, 3701–3712.
Mass Transfer, 1989, 30, 69–79. 173. E. A. Brener and D. E. Temkin: Phys. Rev. Lett., 2005, 94, 1–4.
136. D. M. Smith, J. A. Eady, L. M. Hogan and D. W. Irwin: Metall. 174. D. E. Temkin: Acta Mater., 2005, 53, 2733–2738.
Trans. A, 1991, 22A, 575–584. 175. J. R. Michael and D. B. Williams: in ‘Interface migration and
137. Q. T. Fang and M. J. Bruno: in ‘Light metals 1991’, (ed. E. L. control of microstructures’, (ed. C. S. Pande et al.), 73–81; 1986,
Rooy), 1856; 1991, Warrendale, PA, TMS. Metals Park, OH, ASM.
138. C. Vives: Int. J. Heat Mass Transfer., 1990, 33, 2585–2598. 176. M. Hillert and G. R. Purdy: Acta Metall., 1978, 26, 333–340.
139. C. J. Paradies, R. N. Smith and M. E. Glicksman: Metall. Mater. 177. M. Hillert: Metall. Trans., 1972, 3, 2729–2741.
Trans. A, 1997, 28A, 875–883. 178. Y. J. Baik and D. N. Yoon: Acta Metall., 1985, 33, 1911–1917.
140. J. F. Li, Y. C. Liu, L. Lu, G. C. Yang and Y. H. Zhou, J. Cryst. 179. G. R. Purdy: in ‘Phase transformation in materials’, Vol. 5, (ed. P.
Growth, 1998, 192, 462–470. Haasen), 305–338; 1999, Weinheim, VCH-Verlag.
141. D. Herlach, K. Eckler, A. Karma and J. Schwarz: Mater. Sci. 180. M. Buchmann and M. Rettenmayr: J. Cryst. Growth, 2005, 284,
Eng. A, 2001, A304, 20–25. 544–553.

International Materials Reviews 2009 VOL 54 NO 1 17

You might also like