A comprehensive study of the softening behaviour of two different non-heat treatable AlMn-alloys has been carried out. These alloys were a laboratory processed and an industrially processed AA3103-alloy. The primary objective of the laboratory processed alloy has been directed towards the relationship between the amount of manganese in supersaturated solid solution and the material behaviour during deformation and annealing. The focus for the industrially processed alloy was a detailed material characterisation during hot and cold rolling as well as to follow the softening behaviour of the alloy. The cold rolled microstructures were characterised with respect to subgrain size, misorientation across the subgrain boundaries, particle break-up and global texture. As far as it concerns the cold deformed microstructure it was found that the different homogenisation treatments, resulting in different amount of Mn in supersaturation, had little effect on the deformed microstructure. Detailed experimental work on the softening behaviour after cold deformation and the interaction between recrystallisation and precipitation, i.e. concurrent precipitation, has been carried out. Hardness and yield stress measurements, which defines the strength of the material, revealed that the softening behaviour was significantly slowed down in case of concurrent precipitation. It was further found that the precipitation reaction in this case occurred on the subgrain boundaries. Thus the precipitates considerably retarded the recrystallisation reaction as compared to the case when no precipitation occurred. TTT-diagrams have been constructed on the basis of hardness and conductivity measurements. From these diagrams a characteristic temperature, Tc , for the different material conditions are identified. It was found that as the annealing was carried above this temperature the microstructure consisted of a fine grained equiaxed microstructure. Below Tc the grains become elongated in the rolling direction and the average grain size became much coarser. With respect to recrystallisation texture, a very strong P-texture and in addition some ND-rotated cube texture was found in case of concurrent precipitation. This viistrong P-texture was investigated in detail, and it can be concluded that particle stimulated nucleation of recrystallisation (PSN) plays a significant role in the nucleation of these components. It was found that these texture components had a growth advantage in the early stage of annealing and that they are most probably a result of microgrowth selection, which often are related to a nucleation effect. When precipitation did not interact with recovery and recrystallisation the recrystallisation texture was either random or consisted of a weak cube texture. The softening behaviour of the different materials has been modelled by a physically based softening model, which involves both the recovery and recrystallisation reactions. Both for the AlMn-alloys and for an additional commercially pure Al-alloy, relatively good model predictions were obtained for the softening behaviour when recrystallisation occurred prior to precipitation. However, when the softening reactions were retarded by heavy precipitation of dispersoids the model, in its original form, was not capable to predict the softening behaviour. In this case an additional retarding drag, which reduces the total number of viable recrystallisation nuclei, was added. With the addition of this drag relatively good model predictions were also obtained in case of concurrent precipitation. However, the model was not capable to predict the softening behaviour of the industrially processed AlMn-alloy particularly well.

Microfluidics can be used to increase performance, reduce reagent consumption and increase throughput in chemical analysis. With the forthcoming development of more advanced microfluidic systems, the integration of actuating elements becomes essential, giving the ability to control and manipulate fluid flow as well as sample or other components. This thesis addresses miniaturisation of piezoceramic actuators, in particular important technological issues when actuators are integrated in microfluidic systems. Thick film multilayer fabrication technology for piezo­ceramics has been further developed, e.g. by introducing techniques for integration of microfabricated channel structures and via interconnects in multilayer components. New building techniques have been incorporated to allow miniaturisation of devices. A rapid prototyping technique for advanced multilayer actuators based on mechanical machining has also been developed and used in subsequent work. When interfacing the macro and the micro world in miniaturised chemical analysis systems, non-contact sample dispensing methods such as ink-jet technology are needed. Thus a piezoactuated flow-through microdispenser, suitable for high-speed on-line chemical sample handling has been investigated. A new miniaturised actuator has been developed and integrated in the microdispenser, simplifying assembly and demonstrating an improved performance of the device. With the prospect of performing automated and highly parallel analysis in reusable microarray devices, a new concept for dynamic arraying is presented. Non-contact trapping of particle or bead clusters in a microfluidic system is demonstrated utilising acoustic radiation forces in standing ultrasonic waves. The integration of piezoceramic micro­transducers has been shown to render possible localised and spatially controlled trapping of individually addressable particle clusters in micro­fluidics. The importance of the acoustic near field in miniaturised devices has been identified and utilised to give strong trapping forces. By making use of disposable chemically activated microbead arrays within a flow-through device, a flexible system is emerging with e.g. applications in proteomics.

Cold chamber high pressure die casting, (HPDC), is an important commercial process for the production of complex near net shape aluminium and magnesium alloy castings. The work presented in the thesis was aimed at investigating the microstructure formation in this type of casting. The solidification characteristics related to the process and the alloys control the formation of grains and defects. This again has a significant impact on the mechanical properties of the castings. The investigations were carried out mainly using the AM60 magnesium alloy and the A356 aluminium alloy. Two different casting arrangements were used: the cold chamber HPDC and the gravity die casting methods, which allowed for different flow and solidification conditions. The microstructures in the castings were investigated using optical microscopy, image analysis, scanning electron microscopy, electron back scatter diffraction measurements and electron probe microanalysis. In the HPDC experiments, the shot sleeve solidification conditions were investigated primarily by changing the melt superheat on pouring. This significantly affected the microstructures in the castings. The fraction of externally solidified crystals (ESCs) was consistently found to be largest near the gate in both the AM60 and the A356 die castings. This was attributed to the inherent shot sleeve solidification conditions and the flow set up by the plunger movement. When the superheat was increased, a lower fraction of ESCs was found in the castings. Furthermore, a high superheat gave ESCs with branched dendritic/elongated trunk morphology whilst a low superheat generated coarser and more globular ESCs, both in the AM60 and the A356 castings. The ESCs typically segregated towards the central region of the cross sections at further distances from the gate in the die castings. When a thin layer of thermal insulating coating was applied on the shot sleeve wall in the production of AM60 die castings, it nearly removed all ESCs in the castings. Using an A356 alloy, (and no shot sleeve coating), with no Ti in solution gave a significantly lower fraction of ESCs, whereas AlTi5B1 grain refiner additions induced an increase in the fraction of ESCs and a significantly finer grain size in the castings. The formation of globular ESCs was enhanced when AlTi5B1 grain refiner was added to the A356 alloy. In controlled laboratory gravity die casting experiments, typical HPDC microstructures were created by pouring semi-solid metal into a steel die: The ESCs were found to segregate/migrate to the central region during flow, until a maximum packing, (fraction of ESCs of ~35-40%), was reached. The extent of segregation is determined by the fraction of ESCs, and the die temperature affects the position of the ESCs. The segregation of ESCs was explained to occur during flow as a result of lift forces. The formation of banded defects has also been studied: the position of the bands was affected by the die temperature and the fraction of ESCs. Based on the nature of the bands and their occurrence, a new theory on the formation of defect bands was proposed: During flow the solid distribution from the die wall consists of three regions: 1) a solid fraction gradient at the wall; 2) a low solid fraction region which carries (3) a network of ESCs. A critical fraction solid exists where the deformation rate exceeds the interdendritic flow rate. When the induced stress exceeds the network strength, deformation can occur by slip, followed by liquid flow. The liquid flow is caused by solidification shrinkage, hydrostatic pressure on the interior ESC network, and gaps forming which draw in liquid.

In the large family of carbon-based coatings there are members capable of providing a rare and very desirable combination of high wear resistance and low friction. The field of application for this kind of coatings is steadily increasing which, naturally, give raise to new questions and demands. Coatings of this type can be produced in a number of different ways, but the parameter influencing their properties most is the chemical composition. To facilitate investigations of the compositional impact on the synthesis and properties of coatings, a way to alloy magnetron-sputtered thin films was developed in this thesis. It does not involve the use of reactive gases or additional material sources; instead metallic foils are attached onto the magnetron target surface and thus sputtered alongside the target material. This co-sputtering route was later used to synthesize carbon coatings alloyed with Ta, Zr, W and Al in various amounts and configurations. It was shown that the co-sputtering method could be used to alloy coatings in a very simple and straightforward manner, with excellent possibilities of controlling the amount of alloying elements. The process temperature could be kept as low as 70 °C. Carbon coatings alloyed with transition metals (MeC:C) displayed polycrystalline nanocomposite structures with 5 nm metal carbide crystallites in a matrix of near-amorphous carbon. Alloying with Ta resulted in a radical 80 % reduction in friction coefficient during dry sliding in air, from 0.22 for pure carbon to 0.04 for TaC:C. This was found to be due to facilitated graphitization and the formation of lubricating surface oxides. Al additions to TaC:C resulted in a transformation of the TaC phase to a metastable, previously unreported Ta1-χAlχC. The coefficient of friction remained unaltered, but the oxidation rate compared to TaC:C was significantly reduced due to the formation of AlTaO4 instead of Ta2O5.

Demands on lower fuel consumption, reduced pollution, increased operating times, etc. force the automotive industry to constantly improve the performance of critical machine elements. In this development various carbon based coatings have proven very promising, mainly because of their low friction and high wear resistance in dry sliding contacts. The contact conditions can be very different in various machine element applications, e.g. both rolling and sliding contacts. Additionally, most contacts are usually lubricated. Hence, other properties of the coating may be required in order to obtain low friction and wear, as for instance a beneficial running-in ability. In lubricated contacts the very high wear resistance of carbon coatings can cause fatigue damage resulting in delamination of the coating, especially when deposited on rough substrates. In rolling contacts the fatigue damage can be reduced simply by using smoother surfaces, but the thickness of both the coating and the interlayer also has a strong influence on fatigue damage. In lubricated sliding contact tests it was found that the running-in ability could be improved by modifying the coating and/or by using an appropriate surface preparation prior to coating deposition. An increased Cr-content in the top-layer of the carbon coating reduced the friction due to the formation of a stable tribo-layer on the uncoated counter body. An even further reduced friction could be obtained by employing a fine wet-blasting of the substrate giving sharp surface asperities. The friction reduction is thought to be a result of a transition towards a higher degree of full film lubrication, due to a very fine smoothening process of both the coating and the counter body during the running-in process.

This thesis work consists of two parts, each with a different motivation. Part II is the main part and was partly conducted in industry, at ScanWafer ASA’s plant no.2 in Glomfjord. The large growth in the Photo Voltaic industry necessitates a dedicated feedstock for this industry, a socalled Solar Grade (SoG) feedstock, since the currently used feedstock rejects from the electronic industry can not cover the demand. Part I of this work was motivated by this urge for a SoG- feedstock. It was a cooperation with the Sintef Materials and Chemistry group, where the aim was to study the kinetics of the removal reactions for dissolved carbon and boron in a silicon melt by oxidative gas treatment. The main focus was on carbon, since boron may be removed by other means. A plasma arc was employed in combination with inductive heating. The project was, however, closed after only two experiments. The main observations from these two experiments were a significant boron removal, and the formation of a silica layer on the melt surface when the oxygen content in the gas was increased from 2 to 4 vol%. This silica layer inhibited further reactions. Multi-crystalline (mc) silicon produced by directional solidification constitutes a large part of the solar cell market today. Other techniques are emerging/developing and to keep its position in the market it is important to stay competitive. Therefore increasing the knowledge on the material produced is necessary. Gaining knowledge also on phenomenas occurring during the crystallisation process can give a better process control. Part II of this work was motivated by the industry reporting high inclusion contents in certain areas of the material. The aim of the work was to increase the knowledge of inclusion formation in this system. The experimental work was divided into three different parts; 1) Inclusion study 2) Extraction of melt samples during crystallisation, these were to be analysed for carbon- and nitrogen. Giving thus information of the contents in the liquid phase during soldification. 3) Fourier Transform Infrared Spectroscopy (FTIR)-measurements of the substitutional carbon contents in wafers taken from similar height positions as the melt samples. Giving thus information of the dissolved carbon content in the solid phase. The inclusion study showed that the large inclusions found in this material are β-SiC and β-Si3N4. They appear in particularly high quantities in the top-cuts. The nitrides grow into larger networks, while the carbide particles tend to grow on the nitrides. The latter seem to act as nucleating centers for carbide precipitation. The main part of inclusions in the topcuts lie in the size range from 100- 1000 µm in diameter when measured by the Coulter laser diffraction method. A method for sampling of the melt during crystallisation under reduced pressure was developed, giving thus the possibility of indicating the bulk concentration in the melt of carbon and nitrogen. The initial carbon concentration was measured to ~30 and 40 ppm mass when recycled material was employed in the charge and ~ 20 ppm mass when no recycled material was added. Since the melt temperature at this initial stage is ~1500 °C these carbon levels are below the solubility limit. The carbon profiles increase with increasing fraction solidified. For two profiles there is a tendency of decreasing contents at high fraction solidified. For nitrogen the initial contents were 10, 12 and 44 ppm mass. The nitrogen contents tend to decrease with increasing fraction solidified. The surface temperature also decreases with increasing fraction solidified. Indicating that the melt is saturated with nitrogen already at the initial stage. The proposed mechanism of formation is by dissolution of coating particles, giving a saturated melt, where β-Si3N4 precipitates when cooling. Supporting this mechanism are the findings of smaller nitride particles at low fraction solidified, that the precipitated phase are β-particles, and the decreasing nitrogen contents with increasing fraction solidified. The carbon profile for the solid phase goes through a maximum value appearing at a fraction solidified from 0.4 to 0.7. The profiles flatten out after the peak and attains a value of ~ 8 ppma. This drop in carbon content is associated with a precipitation of silicon carbide. It is suggested that the precipitation of silicon carbide occurs after a build-up of carbon in the solute boundary layer. FTIR-measurements for substitutional carbon and interstitial oxygen were initiated at the institute as a part of the work. A round robin test was conducted, with the Energy Research Centre of the Netherlands (ECN) and the University of Milano-Bicocci (UniMiB) as the participants. The measurements were controlled against Secondary Ion Mass Spectrometer analyses. For oxygen the results showed a good correspondence between the FTIR-measurements and the SIMS. For carbon the SIMS-measurements were significantly lower than the FTIR-measurements. This is probably due to the low resistivity of the samples (~1 Ω cm), giving free carrier absorption and an overestimation of the carbon content.

Ceramic membranes made from mixed oxygen-ionic and electronic conducting perovskite oxides can selectively separate oxygen from air at elevated temperatures. These membranes have several potential applications that require a continuous supply of oxygen. For example, they may be an alternative for cryogenic production of oxygen or alternative electrode materials in solid oxide fuel cells. Of particular significance is the partial oxidation of methane to syngas (CO + H2). By combining air separation and partial oxidation of natural gas into a single step, the need for expensive oxygen production by cryogenic means may be eliminated. Combined with existing processes for gas-to-liquid production such as Fisher-Tropsch and methanol synthesis, the MIEC membrane technology represents a very attractive route for conversion of natural gas to liquid fuels. The research in this field was initially concerned with the search for materials with the optimum oxygen flux. Today, the long term stability of the membranes is probably the main issue. The membranes have to be stable under operating conditions, which include mechanical stability and chemically compatibility with other materials like sealing and support materials. However, the current understanding of the long term chemical and mechanical reliability is poor and this is one of the major challenges for solid state ionic research. The aim of this work has been to investigate the mechanical properties and the chemical stability of La0.5Sr0.5Fe1-xCoxO3-δ (x = 0, 0.5, 1) materials when they are exposed to thermal and chemical gradients. The chemically induced stresses due to reduction of the valence state of the transition metals are of particular importance with respect to the mechanical stability. In paper I, the oxygen non-stoichiometry, investigated by thermogravimetrical analysis, and thermal end chemical expansion, studied by dilatometry and high temperature X-ray diffraction, of La0.5Sr0.5Fe1-xCoxO3-δ materials are reported. The oxygen deficiency was observed to increase with decreasing partial pressure of oxygen and increasing temperature corresponding to expectations and previous reports. At ambient temperature the thermal expansion coefficient of the materials were in the range 15- 18·10-6 K-1. Above a certain temperature thermal reduction of the material take place, and the thermal expansion coefficient due to chemical expansion raise to 16-36·10-6 K-1. The chemical expansion εc, defined as the linear expansion due to a change in partial pressure of oxygen at constant temperature, reached a maximum in the range 0.036-0.039 for the materials studied at 800ºC. The change in ionic radii of the transition metals is the main contribution to the chemical expansion. The crystal structure of the perovskite materials were shown to be slightly rhombohedral at ambient temperatures and a transition to cubic phase were observed above 300ºC. This non-linear thermal expansion behavior is a major challenge for the applications of the mixed conductor materials. La0.5Sr0.5Fe1-xCoxO3-δ membranes in an oxygen partial pressure gradient will have different oxygen deficiency on either side of the membrane. The increasing oxygen deficiency is accompanied by a volume expansion as shown in paper I, and this will lead to chemically induced stresses. These stresses and the failure that might follow can be prevented by creep of the materials. Creep is also important due to dimensional stability. In paper II, the steady-state creep performance under compression of La0.5Sr0.5Fe1-xCoxO3-δ (x = 0.5, 1) as a function of temperature, atmosphere, load and two different grain sizes is reported. The stress exponent found for the materials was close to unity and an unusual low inverse grain size exponent close to one was found for one of the materials. The activation energy of the two materials was not equal and the influence of secondary phases on the creep was discussed. The obtained creep behavior and microstructural investigation after measurements point to a diffusion related mechanism for the creep. Higher creep rates are found under reducing conditions and this suggest that creep relaxation of mechanical or chemical induced stresses may enhance the mechanical stability of oxygen permeable membranes. In Paper III, the mechanical properties of La0.5Sr0.5Fe1-xCoxO3-δ (x = 0.5, 0.75, 1) were investigated by several methods. Fracture strength was measured by four-point bending, fracture toughness was measured by SENB and SEVNB methods and finally Young’s modulus were investigated by four-point bending and resonant ultrasound spectroscopy. Four-point bending showed a non-linear ferroelastic behavior at ambient temperature due to rhombohedral crystal structure. Above the ferroelastic to paraelastic transition temperature the materials showed elastic behavior, however, at temperatures from about 800ºC a non-elastic respond was observed due to creep. The measured fracture strength and fracture toughness were observed to increase with increasing temperature, which was attributed to frozen-in stress gradients in the materials during cooling due to different oxygen stoichiometry. These stress gradients caused the low fracture strength and fracture toughness at ambient temperature. At higher temperatures, the stresses are assumed to relax resulting in a higher strength and fracture toughness. At high temperature, the non-linear respond made systematic errors in the calculated strength and fracture toughness. The Young’s modulus was measured from four-point bending and by resonant ultrasound spectroscopy for two of the materials. These data obtained by these two different methods were not in good agreement, which demonstrate the difficulty to obtain reliable data for the Young’s modulus of such materials by four-point bending. The presented findings have demonstrated the importance of understanding ferroelasticity and chemically induced stresses in order to comprehend the mechanical properties of such mixed valence state perovskite materials. A high oxygen flux is required in order to realize the oxygen permeable membrane technology. At the same the chemical stability of the materials in a pO2 gradient must be good for a sufficient long period of time. The oxygen flux performance and the long term stability of La0.5Sr0.5Fe1-xCoxO3-δ (x = 0, 0.5, 1) are the topics of Paper IV and V. Oxygen fluxes through the membranes are found as a function of oxygengradient and temperature in a oxygen permeation cell using air and inert gas on each side. The oxygen flux was observed to increase with decreasing pO2 on the secondary side until the surface exchange became rate limiting and the fluxes reach a constant value. By further increase of the pO2 gradient, the flux seemed to decrease and this was attributed to the pO2 dependence of the surface exchange coefficient. The apparent activation energy of the oxygen permeation was in good accordance with previous investigation of similar materials. After about 5 week of exposure in an oxygen gradient at about 1150°C, the membranes were carefully examined by electron microscopy for evidence for kinetic demixing and decomposition. Dependent of the overall composition of the membrane, different secondary phases were formed at the primary surface of the membrane. For the cobalt containing materials, isolated grains or clusters of grains of cobalt oxide were formed. In case of the La0.5Sr0.5FeO3-δ membrane, a dense and about 20 µm thick layer of the secondary phase SrFe12O19 was formed at the primary side. The overall (La+Sr)/(Fe+Co) ratio was also seen to influence on the phase formed at the primary side. Kinetic demixing was also demonstrated in all the membranes although the metal concentration profiles were not drastically changed from the initial concentrations. The formation of secondary phases was reflected in the (La+Sr)/(Fe+Co) ratio across the membrane. The largest deviation from the nominal stoichiometry was seen close to the surfaces indicating steeper chemical gradients close to the surfaces. These phenomena may strongly limit the long term stability of thinner membranes e. g. films on a porous substrate.

No description available.

The overall objective of this work has been to develop aluminium alloys, which after hot and cold deformation are able to withstand high temperatures without recrystallising. This has been done by investigating aluminium alloys with various additions of hafnium, scandium and zirconium, with a main focus on Hf and to which extent it may partly substitute or replace Zr and/or Sc as a dispersoid forming elements in these alloys. What is the effect of hafnium, alone and in combination with Zr and/or Sc and how do hafnium containing alloys perform? It is shown that hafnium may alter or modify the casting structure, though, not to the better as it can form TCGs in combinations with Zr and/or Sc. This is not advantageous neither as far as it concerns grain refining nor precipitation of dispersoids. When precipitation of binary Al-Hf is compared to Al-(Hf)-(Zr) alloys, hafnium shows even slower precipitation than in Al-(Hf)-(Zr) alloys and also much slower and with a poorer spatial distribution of dispersoids than in Al-Sc or Al-Hf-Sc alloys. As a consequence, it may be concluded that binary aluminium-hafnium alloys are of limited interest as they display a poor recrystallisation resistance when no other alloying elements are added. However, when hafnium is added together with scandium and/or zirconium, precipitation may actually improve both in Al-Hf-Zr alloys and in scandium containing alloys like Al-Hf-Sc and Al-Hf-Sc-Zr. Hafnium can still not completely replace neither Zr nor Sc due to its poor precipitation properties. However, Al-Hf-Sc-Zr alloys show a better dispersoid distribution (number density and volume fraction) than what was observed in Al-Sc-Zr alloys after extrusion. The Al-Hf-Sc-Zr alloys also show extreme high temperature properties, by withstanding recrystallisation at high temperatures and long annealing times. After severe cold deformation, the extruded profiles of both the Al-Sc-Zr alloy and the Al-Hf-Sc-Zr alloy displayed a remarkable recrystallisation resistance. The reason why the Al-Zr-Sc- and the Al-Hf-Zr-Sc alloys behave so well has been investigated by detailed 3D Atom Probe investigations. Due to the homogeneous precipitation of Al3Sc dispersoids and the retarding effect from Hf and/or Zr containing shells, thus limiting the coarsening of these dispersoids, these combinations have been shown successful. As we have seen when Hf and Zr are added in combination equal or improved recrystallisation properties can be obtained. This is specially the case when both these elements are added together with scandium. Since Hf and Zr are extremely difficult (and thus costly) to separate, Al-Zr master alloys used in industry today can in the future probably contain more Hf, lowering the cost of master alloys. The work presented in this thesis have hopefully added some new insight and a better understanding of the effects of adding various dispersoid forming elements to aluminium, alone and in various combinations, which may be useful for industry today and a basis for further alloy development.

When remelting aluminium scrap, metal losses due to dross generation is a common problem. Reduction of these losses will give substantial economic and environmental benefits. Dross is generated when aluminium metal oxidizes and films of oxide envelope molten metal. When a cold metal object is immersed in a melt, the heat of the melt around this is transferred so rapidly into the object that a shell of melt often solidifies to the surface of the object. When scrap with low bulk density is charged to a melt, solidification of melt on the cold scrap prevents melt from entering the cavities in the bulk of the scrap, and the bulk density remains low. Thus the scrap tends to float on the melt surface. Submersion of this scrap is important to avoid oxidation and subsequent dross generation. One solution to this is to roll scrap to a strip and feed it into the melt. This system has been examined by studying feeding of a continuous, thin aluminium plate into molten aluminium. Also, the effect of lacquer was considered, as well as feeding the plate into a launder with melt flowing along the surface of the plate. An analytical, one-dimensional, steady-state model has been developed to describe the melting and the melting mechanisms. It is based on a shell solidifying on the plate surface and a gap introducing a thermal resistance 1/hg between the shell and the plate. The thermal resistance 1/hl of the boundary layer of the melt is included. Depending on these resistances, the initial temperature of the plate and the melt temperature, a shell will form, and the plate will penetrate a distance P into the melt before it melts away. An experimental apparatus was designed and constructed to feed aluminium plate from a coil into a melt bath at a specified velocity. The plate could be withdrawn rapidly to “freeze” the situation like it was below the melt surface. The penetration depth P of the plate could be measured and shell formation observed. More than 200 experiments were performed, and by comparing the penetration depth at different feeding velocities and melt temperatures to model predictions, the two heat-transfer coefficients hl and hg could be determined by curve fitting. They agree reasonably well with values found in the literature and calculated from boundary-layer theory. In a few experiments, the plate feeding was recorded on video tape, and the cross section of some plates was studied in a microscope. Feeding of somewhat thicker plates was also tried. This gave valuable background information for comparing the experiments to the model. We believe that snap-off of the plate due to low mechanical strength around the melting temperature may affect the measurement of the penetration depth of the plate. Attempts were also made to measure the temperature in the plate by attaching thermocouples to its surface. The obtained temperature profiles in the plate were compared to the model predictions, but the method needs improvement. A criterion for formation of a shell is formulated and tested against experimental observations. Qualitative agreement is achieved. Even if there is no shell formation, it seems that there will be an air film with thermal resistance 1/hg. This indicates that the melting rate will be independent of whether a shell is formed or not. Two additional models with only one heat-transfer coefficient are also developed in order to challenge the main model. From this analysis it is found that the use of two heat-transfer coefficients is necessary to describe the system. The model should be of direct interest when feeding rolled scrap into molten aluminium. Improvement of the model can be attained by reconsidering the assumptions made, but then numerical methods must undoubtedly be applied. These new models should include the snap-off mechanism.