Semisolid processing of metallic materials

Abstract Semisolid processing involves forming metallic alloys between the solidus and the liquidus. For the process to operate, the microstructure must consist of solid spheroids in the liquid matrix, rather than dendrites. The material then flows when it is sheared but thickens again when allowed to stand, i.e. it behaves thixotropically. This type of behaviour was first discovered by Flemings and co-workers in the 1970s and is utilised in a family of processes, some now applied commercially. Here, the current status of semisolid processing, both technologically and from a research point of view, will be reviewed.


Introduction
In the early 1970s, a PhD student at MIT in the USA was studying hot tearing of casting alloys by using a rheometer to measure the viscosity as he cooled them into the semisolid state. He found that if he was stirring the material continuously during cooling, it showed less resistance to shear than if the material was cooled into the semisolid state and then stirred. 1 The student and his supervisor identified that the microstructure of the continuously stirred materials was spheroidal (i.e. consisted of spheroids of solid in the liquid matrix), whereas the material which was cooled into the semisolid state without stirring was dendritic (Fig. 1). The material with the spheroidal microstructure in the semisolid state was thixotropic, i.e. when it is sheared, it thins and flows, and when it is allowed to stand, it thickens again. 2 This property can be exploited in semisolid processing where the material is forced to fill a die. In comparison with die casting, where the flow is turbulent, here, the flow is smooth and laminar, and hence, the resulting mechanical properties are enhanced. Consequently, parts can be made with thinner sections and hence lighter weight. As the material does not have to cool all the way from the liquid state, there is less solidification shrinkage, and therefore, parts can be made closer to net shape. The lower temperatures than die casting mean less die attrition. In addition, parts can be heat treated, and the surface finish is improved. In comparison with forging, complex shapes can be made in one shot, including undercuts, and less force is needed; hence, the machinery can be on a smaller scale.
The discovery of the thixotropic behaviour of semisolid metallic alloys has led to the development of a family of processing routes. Initially, a spheroidal microstructure has to be obtained. Sometimes, this is a prior step (as in thixoforming, thixoforging and thixocasting), and sometimes, it is integral to the forming process itself (as in rheocasting, thixomoulding and rheodiecasting).
In Fig. 2, one route to the spheroidal microstructure is shown as stirring. There are, however, a range of potential feedstock production routes, and these will be discussed further in the section on 'Routes to spheroidal feedstock'. After stirring, the material can be solidified and cut into billet. It is then reheated. If the reheated billet is placed into the shot chamber of a die casting machine and forced into the die, this is termed thixocasting, and usually, the fraction of liquid is around 60% or more. If the semisolid billet is placed between closed dies, then it is thixoforging, and the fraction solid will be quite high (around 60% or more). Thixoforming, as a term, is sometimes used to cover both thixocasting and thixoforging, or to cover a hybrid process where the billet is placed on a ram (rather than in a shot sleeve) and forced into a die. In the diagram, 'deformed billet' means that the route to a spheroidal microstructure involves deformation in the solid state. On reheating into the semisolid state, the material recrystallises, and liquid is formed around the new grains, which are fine and equiaxed. In rheocasting (and in the related routes rheodiecasting and thixomolding), the feedstock is not solidified after stirring but rather is taken straight from the semisolid state into the die casting machine.
In the following section, feedstock production routes will be discussed in more detail and in the third section, the semi-solid processing routes. The fourth section will cover industrial applications, the fifth section will cover rheology of semisolid material, the sixth section modelling and the seventh section the current status with respect to which alloys can be semisolid processed. The field has been reviewed, [4][5][6][7][8][9] and two books have been published recently. 10,11 conditions for dendrite formation may not be met. For example, constitutive supercooling will not occur if liquid is stirred to homogenise solute concentrations or if the material is held isothermally. The solid state routes to spheroidal feedstock are essentially based around the fact that deformed material will recrystallise as it is heated, and the liquid will tend to form at the recrystallised boundaries. 6

Liquid metal routes
Liquid metal routes can be divided into those which involve agitation and those which do not. Semisolid material can be vigorously stirred with a mechanical stirrer, but this tends to lead to unacceptable erosion of the stirrer and contamination. In the shear cooling roll technique, 12 liquid metal is poured between a rotating roll and a stationary cooling shoe. Passive stirring involves the liquid being forced through a static mixing device (such as a column containing glass spheres) where the geometry ensures high mixing without turbulence. [13][14][15][16] Spray forming is the atomisation of melt into a spray followed by collection of droplets on a substrate. This gives a very fine spheroidal microstructure when remelted and allows alloys, which cannot be made by other routes, to be fabricated. 17 However, it is relatively expensive. The 'new MIT process' (Fig. 3) is a hybrid of stirring and near liquidus casting. 18 The melt is held just above the liquidus, and a cooling finger, which also acts as a stirrer, is inserted. The melt is cooled to just below the liquidus, and the finger is withdrawn. This generates a cloud of nuclei. This process is very interesting from a commercial point of view. The most important agitation route to date industrially has been magnetohydrodynamic stirring (MHD). 4 High local shear is generated by rotating, linear or helicoidal electromagnetic fields (thereby avoiding contamination, gas entrapment and stirrer erosion), with stirring deep in the sump of the liquid. The alloy is filtered and degassed before treatment, and a fine spheroidal size of about 30 mm results. There is some lack of uniformity, and spheroids may retain some 'rosette' character, a factor which influences flow. The majority of material, which has been thixoformed commercially in the last 20 years, has been produced by the MHD route.
There have been significant new developments more recently, including the SEED process (swirled enthalpy equilibration device) by Alcan (Montreal, QC, Canada). This involves first the rapid extraction of a controlled quantity of heat from the molten metal charge in order to generate a semisolid slurry and then draining away the excess liquid, leaving behind a compact selfsupporting slug, which can then be handled and formed under pressure.
The 'new rheocasting (NRC) process', patented by UBE, 19,20 is an example of a route which does not involve agitation. Molten metal at near liquidus temperature is poured into a tilted crucible. Grain nucleation occurs on the side of the crucible, and the grain size is fine because a b 1 a dendritic cast microstructure and b spheroidal microstructure: alloy is A356 aluminium alloy 2 Semisolid processing routes 3 the temperature is near liquidus. To date, this has been the most important of the rheocasting type routes from a commercial point of view.
Other non-agitation routes include the direct thermal method, 21 where liquid alloy with a low superheat is poured into a cylindrical metallic mould of very low thermal mass but high thermal conductivity. Heat matching between the alloy and the mould results in a pseudoisothermal hold within the solidification range. This is a low cost method ideal for the laboratory environment but with a strong limitation on the size of the billet that can be produced.
In the cooling slope method, 22 liquid metal is poured down to a cooled slope and collected in a mould. Nucleation on the slope ensures that the spheroid size is fine. This shows considerable potential for combination with rolling ( Fig. 4), but there are issues of gas entrapment and oxide formation on the slope.
A number of manufacturers have aimed to use grain refinement to suppress dendrite growth, e.g. titanium diboride in aluminium alloys. It has proved difficult to obtain uniform grain sizes less than 100 mm, and the volume of liquid entrapped in spheroids tends to be high (entrapped liquid does not contribute to flow).
A new route has recently been developed, which is gaining considerable commercial interest. 24 It involves bringing two precursor alloys of precisely controlled composition, temperature and quantity into intimate contact and then casting the resultant alloy using a conventional casting process. This yields a component of a predetermined composition with a spheroidal microstructure.

Solid state routes
The main solid state routes are strain induced melt activated (SIMA) 25 and recrystallisation and partial melting (RAP), 26 both of which have been used commercially. These are compared in Fig. 5. Essentially, as stated earlier, the material is deformed, for example, by extrusion and then reheated into the semisolid state. Recrystallisation occurs, and the liquid forms around the grains, which are fine and equiaxed. The distinction between SIMA and RAP is that the former involves hot working (above the recrystallisation temperature) and the latter involves warm working. SIMA, therefore, requires an intermediate additional cold work step. RAP is, however, more limited in the billet size because it is difficult to introduce deformation uniformly across the section (and for successful thixoforming, a uniform spheroidal microstructure is required). The advantages of these routes are that many alloys are supplied in the extruded (or rolled) state anyway, and the spheroids tend to be more rounded, giving better flow than the 'rosette type' structures sometimes found, for example, with the MHD route.
It has become clear that some materials are readily thixoformable when reheated from the state which they are usually supplied in, e.g. high performance steels. 27 Semisolid processing routes Rheocasting In rheocasting, alloy is cooled into the semisolid state and injected into a die without intermediate solidification. The non-dendritic microstructure can be obtained by (see the section on 'Routes to spheroidal feedstock') mechanical stirring, stimulated nucleation of solid particles (as in the NRC process), electromagnetic stirring in the shot sleeve (the Hitachi process 28 ) or by subliquidus casting (the THT process). 29 The NRC process is shown in Fig. 6. The inversion of the billet causes the oxide skin to run into the biscuit and runner. With the NRC process, there is no need for specially treated thixoformable feedstock, and scrap can be readily recycled within the plant. However, there is a lack of flexibility; the crucible size and heating/cooling arrangements are specific to that volume of metal.
In subliquidus casting ( Fig. 7), 29 no processing equipment other than a die casting machine with a large shot diameter and a short stroke is required. The product emerges free of gates so no trimming is needed. The melt is chemical grain refined.

Rheomoulding
Rheomoulding is allied to polymer injection moulding. It employs either a single screw 30,31 or a twin screw. 32,33 Liquid metal is fed into a barrel where it is cooled while being mechanically stirred by a rotating screw. The semisolid is then injected into a die cavity. This allows the continuous production of large quantities of components, and no specially produced feedstock material is needed. Workers at Brunel University are in commercial trials of what they call the rheodiecasting process 34

Thixomoulding
Thixomolding is licensed by a firm called Thixomat (Ann Arbor, MI, USA). Many firms in Japan, the USA and the Far East use the process to produce components from magnesium alloys. Magnesium alloy pellets are fed into a continuously rotating screw. The process is therefore similar to rheodiecasting, but the starting material is solid pellets rather than liquid. The energy generated by shearing contributes to heating the pellets into the semisolid state (but some additional heating is applied). The screw action produces the spheroidal microstructure, and the material is fed into a die. One of the advantages of the process is that it allows forming with a range of solid fractions from 5 to 60% (although typically it tends to be relatively low).

Thixoprocesses
In thixocasting, the material is solid initially, and the alloy has been treated in such a way that when it is reheated into the semisolid state, a spheroidal microstructure is obtained. The liquid content is relatively high (.50 vol.-%). Reheating can be in a radiant furnace or an induction heating furnace. The high liquid fractions mean that the billet tends to collapse and is therefore tipped into the shot sleeve from a container. This process has been used commercially for making fuel rails for cars.
Thixoforging 35,36 involves heating the feedstock into the semisolid state and placing it between die halves. The parts of the die are then brought together. Material is used efficiently because of the lack of runners, gates and press discard. The liquid volume fraction is generally less than 50 vol.-% so that the billet can be handled. The process has not been used commercially, perhaps because of the difficulty in obtaining repeatability.
In thixoforming, 17 the billet is heated as a vertical cylinder (in contrast with thixocasting) and then injected into the die, either horizontally or vertically. The liquid content is between 30 and 50 vol.-%. A range of automotive and other components have been produced by this route in industry. Billets are heated on a carousel with induction. Process control is demanding, but cycle times are very comparable with those in die casting and indeed may be faster because full solidification is not required.

Industrial practice and applications Process issues
For the rheocasting routes, heating of the feedstock billet is not required. However, for the thixoprocesses, the general method is to use induction heating in a carousel arrangement so that a billet is always ready for processing. The induction heating sequence may be stepwise in order to obtain as uniform a temperature profile across the billet as possible (given that induction heating works through the skin effect). Process control is demanding because the spheroidal microstructure must be as uniform as possible across the cross-section with a consistent liquid/solid fraction. The viscous slug must retain its shape until the shear force is applied. Experience shows that it is difficult to obtain satisfactory thixoprocessing unless the equipment has real time and closed loop control. For example, feedback must occur so that, as the ram makes its stroke, the velocity is constantly monitored and controlled to given values. Otherwise, as the ram forces material into the die and encounters increasing resistance, it will slow down. In practice, successful semisolid processing is based around optimised profiles for pressure and velocity versus time.

Examples of commercial components
Components made by thixoforming because of the near net shape requirement include computer heat sinks ( Fig. 9), industrial bolt connectors, racing bike swing arms, computer disk drive motor bases and mountain bike suspension parts. For other parts, it is the capability to achieve enhanced mechanical properties which is significant, e.g. motorcycle chassis frame arms and mountain bike forks. In some cases, thixoforming can produce a component which is pressure tight where die casting cannot, for example, intricately shaped gas control valves. One of the problems with die casting is the incorporation of oxide films because of the turbulent flow; this does not occur with semisolid processing, where the flow into the die is smooth and laminar. Other examples of leak tight, pressurised applications include master brake cylinders, acrylonitrile butadiene styrene system valves, missile connector supports and airbag canisters. All these are made from aluminium alloys, usually A356 or A357 in the heat treated condition.
Many components for mobile phones, laptops and automotive applications are made from magnesium alloys by thixomoulding (Figs. 10 and 11).
Experimental determination of parameters for semisolid processing Background rheology Figure 12 shows different potential types of fluid behaviour. In a Newtonian fluid, the shear stress is proportional to the shear rate, and the constant of proportionality is the viscosity g. Thixotropic fluids are non-Newtonian, i.e. the shear stress is not proportional to the shear rate. The viscosity is then termed the apparent viscosity and is dependent on shear rate, pressure, temperature and time. Some non-linear fluids also show viscoelasticity, i.e. they store some of the mechanical energy as elastic energy. Thixotropic materials are generally thought not to store energy elastically and show no elastic recovery when the stress is removed. However, Favier and Atkinson 37 have recently shown that to accurately model the initial stages of the rapid compression response of thixotropic billets, an assumption of elastic behaviour by the interconnected solid skeleton before deformation is required. If a fluid exhibits a yield stress and then gives a linear relationship between shear stress and shear rate, it is termed a Bingham material. The Herschel-Bulkley model is where behaviour is non-linear after yield.
There is dispute over whether thixotropic semisolid alloys display yield and whether they should be modelled as such. Barnes 38 concluded, in a review of thixotropy, which did not include semisolid metallic alloy materials as such, that the presence of a yield stress as reported by some workers for thixotropic materials is probably due to the limitations of their experimental apparatus in not being able to measure shear stresses at very low shear rates. Koke and Modigell 39 have, however, used a shear stress controlled rheometer to measure yield stress directly on Sn-15Pb. They distinguish between a static yield stress where the fluid is at rest before the application of a shear stress and a dynamic yield stress where the fluid is being continuously sheared. The yield stress increases with rest time before deformation because of the increasing degree of agglomeration of the spheroids of solid. In terms of modelling semisolid alloy die fill, the use of a yield stress may be appropriate because a vertical billet does not collapse under its own weight unless the liquid fraction is too high. In addition, in rapid compression experiments (to be described later), an initial peak in the load versus displacement curve is detected.
The Ostwald de Waele relationship (t~k : c n ) is used to describe fluids, which do not have a yield point and where there is a power law relationship between the shear stress t and the shear rate. If the exponent n is 1, this reduces to the expression for a Newtonian fluid with the constant k equal to the viscosity g. A shear thinning material (whose viscosity decreases as the shear rate increases) would have a value of n of less than 1, and a shear thickening material would have n greater than 1.
Thixotropic materials are essentially shear thinning but also thicken again when allowed to rest (i.e. all thixotropic materials are shear thinning but not all shear thinning fluids are thixotropic).
For a thixotropic material at rest, when a step increase in shear rate is imposed, the shear stress will peak and then gradually decrease until it reaches an equilibrium value for the shear rate over time (Fig. 13). The higher the shear rate after the step, the lower the equilibrium viscosity. The peak viscosity encountered will increase with increasing rest time before it recovers back to the equilibrium viscosity of the shear rate specified.
In semisolid metallic alloys (as in other thixotropic materials such as clays, mousses, emulsions and flocs), the thixotropic behaviour is associated with agglomeration and disagglomeration. When the material is allowed to stand, or stirred at a low shear rate, spheroids link together to form agglomerates (Fig. 14). With a step change up in shear rate, the agglomerates are broken down, and the viscosity decreases towards a new equilibrium level.
The higher the shear rate (faster stirring), then the lower the viscosity. The higher the fraction solid, the higher the viscosity. If the shear rate is changed, then over a period of time, the viscosity will reach a new equilibrium value (steady state). The viscosity in the steady state depends on the balance between agglomeration and disagglomeration and also on particle morphology; pure spheres give the lowest viscosity. Liquid entrapped within the spheres does not contribute to shear flow.

Transient versus steady state behaviour
In semisolid processing, the injection into the die takes less than 1 s. If the material is initially at rest (as in thixoforming), in that fraction of a second, the viscosity is changing by orders of magnitude from 10 6 to 10 3 Pa s (equivalent to that of heavy machine oil) (the viscosity change will be less in rheocasting type processes where there has been insufficient time in a rest state for the agglomeration to occur to a great extent). For modelling, we need to use computational fluid dynamics (or for high solid fractions, a 'soft forging' type of approach), but the issue is essentially where to derive the material parameters from. Several workers have carried out rheological After a shear rate increase, there is a characteristic time for the slurry to achieve the steady state. Quaak 3 deduced that, in fact, there are at least two processes occurring (Fig. 15), and therefore, the breakdown should be represented with a double exponential. At the instant after a jump up in shear rate, Quaak argues that the structure is the same ('isostructure'). Very quickly, deagglomeration occurs, and then over a slower time scale, the fragments spheroidise and coarsen by diffusion. It is the very fast process which is most relevant to transient behaviour in thixoforming, and hence, we must design experiments to investigate that.

Rheometry
A rheometer can be used to carry out rapid changes in shear rate. 3,39,40,[45][46][47][48][49][50][51] The advantage of a rheometer is that the shear rate in the gap between the bob and the cup can essentially be treated as uniform (but note some concerns which are now emerging 52 ). A typical experiment would involve stirring the alloy (usually Sn-15Pb as a model because few rheometers will operate at semisolid aluminium temperatures), while cooling from above the liquidus point to the required semisolid temperature (to within 1uC) to obtain a non-dendritic microstructure. Stirring continues until an apparent steady state is obtained with the specimen protected from oxidation by passing inert gas over the material. Shear rate jumps for several different rest times, and initial and final shear rates, can be obtained. Typical results are shown in Fig. 16. 40 The challenge experimentally is to collect data very rapidly during the shear rate jump (1 Hz in this case) and to separate out the behaviour of the rheometer from the behaviour of the material. It is important to correct for the inertia of the measuring head (by subtracting from the data the results for the same test but in air) and to allow for any electronic switching that occurs during the shear rate change by only using data from after when 90% of the final shear rate is achieved.
Breakdown times can be obtained by fitting an exponential to the data obtained in the second after 90% of the specified final shear rate has been achieved. The longer the rest time before the shear rate jump, the lower (i.e. slower) the breakdown time due to increased solid particle sizes and degree of agglomeration. These increases would impede the movement of particles when the shear stress is imposed. The peak viscosity increases with rest time because of the greater degree of agglomeration. The 'first steady state viscosity' increases with rest time. The 'first steady state viscosity' is a pseudo steady state between the fast disagglomeration process and the slower diffusion process (Fig. 15). It is not a full steady state. The breakdown time for the fast disagglomeration process is about 0?15 s. 40 The ease with which the particles flow past each other depends on the particle size, the fraction liquid and the degree of agglomeration. The breakdown time is independent of the initial shear rate but controlled by the final specified shear rate. Breakdown is faster for higher final shear rates and faster than recovery.
The challenge now for rheometry is to obtain data for steels, 53 despite the difficulties with operating rheometers at these higher temperatures.

Rapid compression experiments
The difficulty with rheometers is that there are few that can operate above about 50% solid (because the torque that is required is too high). However, thixoprocessing often takes place with around 60%. One possibility is then to use the thixoformer itself (or a hydraulic press) to carry out rapid compression experiments. [54][55][56][57] The shear rates and temperatures are then akin to those in the process itself. However, in contrast with a rheometer, the shear rate varies from point to point within the volume of the material. There are two ways of approaching this. One is to assume an average shear rate applies across the material volume. The other is to model the compression process with computational fluid dynamics (CFD), which, in itself, requires the material parameters one is aiming to measure. Those material parameters then have to be obtained by an iterative process, fitting the CFD results to the experiment. Figure 17 shows a typical signal response to rapid compression. 57 The peak load is thought to represent the resistance to flow of the skeletal structure connecting agglomerates. There is rapid breakdown under shear (within 10 ms). The peak decreases with increasing temperature (and hence fraction liquid). It increases with increasing soaking time before compression as the skeletal structure is built up.
The load versus displacement response can be obtained by using a load cell arrangement. 57 Recent experiments have shown that the load can also be obtained from data from the ram operation without the need for a load cell. 58 Typical results obtained using the averaging approach (and assuming Newtonian behaviour in the analysis, a shortcoming) are shown in Fig. 18. Shear thinning occurs with an increase in shear rate. With increasing temperature, the viscosity 15 Fast and slow processes after either sudden shear rate increase or sudden decrease 3 16 Shear rate jumps from 0 to 100 s 21 after different rest times for Sn-15Pb (fraction solid: 0?36) 40 decreases. The CFD type of approach has not been fully developed and is discussed further below because it is inherently a modelling approach.

Modelling
There have been a variety of approaches to modelling. 9 Illustrative highlights are given here. CFD approaches include finite difference, both one-and two-phase, and finite element, again both one-and two-phase. Micromacromodelling is also under investigation. 59 The power of modelling is illustrated in Fig. 19.
In the interrupted filling with the original die design, there are defects where the material is flowing back on itself, with the flow front having been disrupted by the bolt holes and the central boss. The simulation mimics this well. An improved die design widens the ingate, removing the constriction typical of a die casting ingate. The flow front is then relatively smooth. In practice, the improved die design gave excellent results.
Much of the modelling work in the literature is based around the model of Brown and co-workers, 60-62 where a structure coefficient l is introduced, which varies between 0 for a structure which is fully broken down (disagglomerated) and 1 for a structure which is fully built up (agglomerated). The flow resistance is due to the hydrodynamic flow of agglomerates and the deformation of solid particles within agglomerates. The rate of change in l is a balance between agglomeration and disagglomeration, with the latter dependent on shear rate because the rupture of particle-particle bonds is shear induced.

Finite difference one-phase modelling
Modigell and Koke 50 have used a one-phase finite difference model (based on FLOW3D from Flow Science Inc.) to examine flow around an obstacle. Figure 20 illustrates the startling difference in behaviour between a Newtonian fluid and a thixotropic one.
It is important to check that the model can predict not only the flow front but also the rheometer results. 63 Results from FLOW3D are shown in Fig. 21 using an ADI solver developed by Flow Science to cope with the fact that in thixoforming, the viscosity and shear rate change by orders of magnitude over very short distances and times. A number of workers have used the MAGMAsoft package with the thixotropic module (finite difference, one-phase) and obtained reasonable agreement between interrupted filling tests and the simulation. 64,65 Finite element one-phase modelling Alexandrou and co-workers 52,66-70 have carried out important work with a finite element one-phase model, with a continuous Bingham law to avoid the discontinuity at the yield stress. The work has explored the relative importance of the inertial, yield stress and viscous flow effects.
Alexandrou and co-workers have been able to predict the regimes under which various flow patterns occur (Figs. 22 and 23) and, in particular, the conditions where defects such as the 'toothpaste instability' are found.
Orgéas et al. 71 use a power law cutoff model in the PROCAST package (Fig. 24). Shear thinning only occurs if a cutoff value is exceeded. This can be set at different values in different parts of the component. The behaviour is, therefore, ratchet type, i.e. an increase in the shear rate beyond the largest shear rate experienced so far will lead to decrease in viscosity. Otherwise, the viscosity is unchanged. Orgéas et al. also observe that, under certain conditions, flow leads to the concentration of liquid in veins in the structure.

Micro-macromodelling
Favier et al. 59 proposed a constitutive equation accounting for the mechanical role of four phases within the material: the solid globules, the solid bonds, the entrapped liquid and the free liquid. An internal variable which modifies the liquid-solid spatial distribution captures the evolution of the microstructure with the shear rate. The model was originally established in a viscoplastic framework and has been shown to predict the characteristics of the transient response for shear rate jumps for solid fraction lower than 0?5. 59 It is now being extended to include elastic behaviour because the initial sharp rise to a peak in rapid compression tests is not consistent with viscoplastic behaviour. 37 In the latest results, this sharp rise can be modelled (Fig. 27).

Experimental validation
Apparatus for filming die filling has been reported. 81,82 One of the challenges which has been overcome is the need for an effective arrangement with the window, which is also inexpensive, for use with aluminium alloys, where the high temperatures are problematic. There are advantages to observing the flow in situ. Interrupted filling tests have disadvantages as the material can continue to travel because of inertia, after the ram is stopped. Results are shown in Fig. 28. Note that with a broader obstacle, the flow fronts meet when the die is more full.
Overall  Alloys for semisolid processing The range of alloys available for semisolid processing and the research to widen the range of applicability have been reviewed. 83 More recently, there has been significant work on steel alloys for thixoforming. 84,85 Virtually all the existing commercial activity is with aluminium casting alloys, particularly A356 and A357, and with magnesium alloys (in the case of thixomoulding). Typical properties for aluminium alloys from Vforge (Lakewood, CO, USA; one of the main suppliers of semisolid processed aluminium alloys) are 241 MPa yield strength for A356 in the T6 condition (with 10-12% elongation) and 262 MPa for A357, again in the T6 condition, (elongation: 8-10%).
There is extensive interest in semisolid processing higher strength aluminium alloys, which are normally wrought, and higher temperature materials, such as copper based alloys, cast irons and steels. In addition, there is increasing interest in composites. Some results for the normally wrought Al alloys are shown in Fig. 29. The results here show that it is ductility which is the major challenge; the yield strengths are close to target (although the target is exceeded with the high strength casting alloy 201) and, for ductility, with 6082. The difficulty here is separating out the processing defects, which are inherent to the composition of the alloy and those which are due to laboratory scale methods of making spheroidal microstructure material. For example, some of these results have been obtained with material from a cooling slope, and there are questions as to whether oxide has been incorporated as the material has flowed down the slope. However, wrought composition alloys are inherently vulnerable to hot tearing, and, at least in part, the ductility problems are associated with that. This is because the freezing range tends to be relatively wide with a long 'tail' during the final stages of solidification. There has been extensive discussion in the literature about the design of high strength alloys specifically for semisolid processing, 83,87-95 with consideration of the freezing range and the rate of liquid formation within the semisolid processing range. Thermodynamic prediction is potentially a useful tool. There is no commercial production of high strength alloys yet. There has been some exploration of copper alloys, particularly in the context of motor squirrel cages (where the alloying element can, in fact, be oxygen, to obtain a freezing range while maintaining a relatively high electrical conductivity. However, currently, there is not known to be any commercial application. Considerable attention is being focused on steels 84,85 and has included the development of steel compositions specifically designed for thixoforming (Fig. 30).
The modified alloy gives a relatively low melting temperature and temperature for 50% liquid (a critical consideration for thixoforming) in comparison with the non-modified compositions. The slopes at 50% liquid and at the end of solidification are also relatively low. The latter is a key criterion for avoiding hot tearing according to Kazakov. 89 The highly alloyed steels (such as tool steels) are the most amenable and show good results for short runs. However, to some extent, the lack of availability of commercial equipment suitable for such production (e.g. with the facility to avoid oxidation during the high temperature heating) has limited the exploration of long run production issues. Tool materials need development. These issues have been recently reviewed. 96 Other high temperature materials under investigation include ductile irons, 97 stellites 98 and high performance steels. 27 Metal matrix composites are also of interest. 17

Conclusions
The current status of semisolid processing has been reviewed. The discovery of thixotropic behaviour in semisolid alloys with spheroidal rather than dendritic microstructures has been described. The terminology has been introduced, and the origin of the behaviour was identified. Routes to feedstock material, and the continuing fertility of ideas in this area, have been described. The distinctions between process routes have been highlighted, and the continuing developments in improving routes so that the economics become increasingly competitive are of note. Some examples of industrial applications have been given. Background rheology has been summarised so that efforts to obtain experimental parameters for modelling can be understood. The variety of models has been discussed, and future challenges have been identified. Finally, the range of alloys which are currently semisolid processed is described. There are challenges in widening the range of applicability, but significant progress has been made. The technology is gaining increasing industrial application.
Ward, P. Kapranos, C. M. Sellars and D. H. Kirkwood. In addition, a range of industrial collaborators, the EU and the UK DTI and EPSRC funding bodies, are acknowledged. Collaborators in the EU COST Consortium COST 541 'Thixosteel' are also thanked for stimulating discussions, in particular, A. Rassili, J.