Introduction

Copper contamination of end-of-life steel scrap has been identified as the main barrier to high-quality steel recycling.[1] Copper and steel are combined for applications in vehicles, appliances, and equipment. Conventional scrap preparation uses hammer shredding to fragment and compact these end-of-life products. Subsequent magnetic separation is not completely effective: copper wiring entwined with steel pieces and steel-encased electric motors remain, so significant quantities of copper follow into the steel melt.[2] Todate, a commercial extraction technique has not been developed, yet residual copper causes metallurgical problems—most notably surface hot shortness.[3] High-quality flat steel products have strict copper tolerances, so end-of-life steel scraps are typically recycled to make tolerant, lower-quality long products.[4] In the coming decades, new measures for copper control will be necessary to ensure that all end-of-life scraps are recycled into demanded products.[5]

An extraction process could be developed to separate copper from steel. The possible separation routes were evaluated by their potential to remove copper, and their specific energy and material input requirements by Daehn et al.[6] This analysis shows that techniques such as sulfide slagging[7] and vacuum distillation[8] have received considerable experimental investigation, but are difficult to implement in practice. However, premelt techniques, such as improved physical scrap preparation, appear to be particularly viable. Trommel[9] or ballistic sorting systems[10] separate steel pieces by size or density to yield scrap with as low as 0.15 wt pct Cu. However, 20 to 30 pct of the steel scrap stream is diverted as copper-rich. To treat heavily contaminated steel scrap and recover pure copper, preferential melting could be relatively simple. Removing external copper before it is dilute within the steel melt requires less energy. The difference in melting points between copper (1083 °C) and steel (1370 °C to 1536 °C)[11] is significant, and could be exploited to collect pure, liquid copper from solid steel. Melting is kinetically fast, and assuming the process is integrated into commercially available scrap heating systems (described by Manning and Fruehan[12]), this step would require little additional energy. Such a process would prevent additional copper from entering the steel cycle.

Preferential melting to remove copper has not been developed because there are two main practical limitations: the irregular shape of end-of-life scrap, and the adherence of liquid copper to the surface of solid steel. Liquid copper could flow into the crevices of fragmented pieces, undermining separation, as described by Katayama et al.[13] However, process design could overcome this limitation by applying mechanical force, such as shaking and agitation. Adherence behavior determines the force that must be applied to achieve separation, and thus overall feasibility. Therefore, adherence of liquid copper to the solid steel surface must be investigated first.

Previous wettability investigations are relevant to understand the adherence of liquid copper on steel. The wetting behavior of the copper-iron system is well understood, and has been studied extensively for sintering applications.[14] Liquid copper will spread over the iron substrate, penetrating grain boundaries, driven by the reduction in surface energy, and diffusing into grains until the maximum solubility of copper in austenitic iron is reached. This mechanism is characterized by Fredriksson et al.[15] Spreading is observed for most metal/metal combinations, but nonwetting is observed for most metal/oxide combinations. Oxides have a low surface energy and form weak interactions with liquid metal.[16] An oxide, or other nonmetallic layer on the steel could be utilized to minimize the adherence of liquid copper, but there is nuance to this idea.

The first study to investigate preferential melting as a means to separate copper from steel was by Brown and Block,[17] who statically held samples of copper winding on steel in oxidizing, inert, and reducing atmospheres. They observed entrapment of the liquid copper between the oxide scale and substrate in an oxidizing atmosphere, but observed direct brazing of copper to steel in a reducing atmosphere.

An oxide scale does not permanently prevent wetting, so subsequent studies investigate the influence of the character of the oxide scale. At high temperature, steel forms a complex surface scale of three oxides. Above 570 °C, FeO, wüstite, grows rapidly and has a porous structure, while more compact Fe3O4, magnetite, and Fe2O3, hematite, are present in much smaller proportions (~ 4 and 1 pct, respectively at 1000 °C[18]) at the outer surface of the scale. Hara et al.[19] varied CO/CO2(g) partial pressures to yield different scales with dominant FeO, Fe3O4, or Fe2O3 and studied the interaction of liquid copper. The work of adhesion of copper was greatest when FeO was dominant. The authors hypothesized that copper wetted due to the porous structure, but copper could be melted away from steel when compact Fe2O3 and Fe3O4 were present.

Further studies investigate the evolution of wetting behavior as the oxide scale is reduced, dissolved or physically disrupted. Bernardo et al.[20] varied the compositions of copper alloys and Fe-based substrates in inert and reducing atmospheres with the goal to improve wetting for sintering applications. Nonwetting was observed with copper on a pure Fe substrate in an argon atmosphere, but when any reductants were present in the system, including carbon in 0.5 wt pct C steel, the oxide layer was reduced to allow complete wetting within 60 seconds. Takahira et al.[21] observed “unusual” wetting behavior of liquid copper on a preoxidized iron sample in a reducing atmosphere. The liquid metal rapidly proliferated through the porous structure. Liquid copper can also dissolve iron oxide. Eustanthopolus et al.[22] show the dissolution of Fe3O4 in liquid copper is moderately favorable, to form Fe-O clusters in solution with liquid copper. These clusters develop coulombic interactions with Fe3O4 to promote adsorption at the interface. In the absence of chemical reactions, studies of similar systems show the iron oxide scale is susceptible to the physical penetration of liquid metal. Protsenko and Eustathopoulos[23] observed droplets of lead and lead-bismuth alloys (which do not form intermetallics or exhibit solubility with iron), on thin oxide layers present on Fe and steel substrates in a low oxygen partial pressure (pO2) environment. Liquid metal penetrated defects in the oxide film (pores, microcracks) and established direct contact with the solid. Shen[24] observed liquid tin wetting on iron and found that the iron oxide film was tenacious, even at low pO2, but disruption of the oxide allowed for wetting and interfacial reactions.

As can be seen above, the oxide layer is not universally protective. Wetting behavior is highly sensitive to the system conditions (temperature, atmosphere, surface roughness, composition, as reviewed by Komolafe et al.[25]), but substances can be applied to the substrate surface to control its behavior and reduce sensitivity. Fluxes are commonly applied, and act to dissolve any oxides present to promote wetting. Alternatively, to promote nonwetting, coatings may provide a barrier. Leak and Fine[26] applied a range of coatings, organic and inorganic predips, and braze inhibitors, to promote sweating of liquid copper from steel. The most successful were sodium metasilicate (Na2SiO3), which is chemically inert, and sodium sulfate (Na2SO4), which modifies the oxidation reaction of steel. The authors submerged motor armatures in a neutral molten salt medium (CaCl2 and BaCl2) and found nearly complete separation of copper from steel when the specified coatings were applied. However, multiple variables could have contributed to this result: the composition of the steel was not stated, the proportion of copper present was high (which promotes liquation), and the motors were shaken to further remove copper. The role these coatings play in achieving nonwetting is not known.

Although existing literature has examined the influence of some factors on the preferential melting of copper from solid steel, the individual and combined effects of steel carbon content, initial surface oxidation, and coatings on effective wetting behavior have not yet been systematically assessed. In this paper, we examine the influence of these parameters on the wetting behavior of liquid copper to begin defining an appropriate process window for preferential melting.

Methods

To investigate the wetting behavior of copper in different conditions, copper was melted on various steel substrates and the copper droplet was observed with a heating microscope. Subsequent microscopy of the sample cross sections revealed the character of the nonmetallic surface layer, the interaction of copper with the surface layer and the distribution of copper across the substrate. Experimental procedures—the composition and surface preparations of the various steel substrates (Section II–A), the operating conditions of the heating microscope (Section II–B) and the preparation of the samples for microscopy (Section II–C) are detailed below.

Test Matrix

Figure 1 illustrates the main variables of the wetting system, which determine the interaction at the liquid/solid interface:

Fig. 1
figure 1

The copper/steel wetting system showing the variables at the focus of this study

Full size image

  • Environment: the environment could be gaseous or liquid. Here, the wetting system is observed in a gaseous environment. To maintain an oxide layer, but limit the growth of porous FeO, an inert atmosphere (argon) was chosen as a constant. Residual oxygen was present in the furnace, which reacted with the surface of the steel substrate.

  • Steel composition: the composition of the steel varies widely in scrap, and this variability will influence iron oxidation and wetting behavior. Carbon content is the most important element because it can reduce the protective oxide layer. Unalloyed low- and medium-carbon steels, representative of the range of carbon contents encountered in practice, were chosen to understand this effect.

  • Surface condition: The surface condition of the steel scrap also varies in practice, and depends on steel composition, surface treatments, processing history and service conditions. Steel scrap may have a no initial oxide layer (such as galvanized or coated steel, where the surface layer combusts around 800 °C during preheating), or may exhibit a thick initial oxide layer of primarily magnetite from rusting during service. This range was investigated by preparing initially oxidized and initially ground sample surfaces. The surface condition may be further modified by applying a coating. Sodium metasilicate and sodium sulfate are widely available, cheap and nontoxic, so these coatings were applied to the various substrates to understand if they promote nonwetting behavior.

The test matrix is defined in Table I. Each combination of steel composition, initial surface oxidation, and coating was tested, for a total of 12 distinct tests. Two additional tests were performed as well. To confirm the observations of other researchers, and understand the interaction of copper with the typically observed three-phase oxide layer, one test was performed with low-carbon steel in air. To understand the wetting behavior of copper on a typical surface with no preparation, one test was performed on low-carbon steel in the as-received condition.

Table I The Varied Characteristics Between the Different Steel Substrates
Full size table

Commercially available steels were used for the substrates. For a representative low-carbon, unalloyed steel, CR4-grade was chosen. The CR4 steel was received in the black, cold-rolled condition as a one-mm-thick sheet, which was cut into 10 × 10 mm square pieces. Key steel was chosen to represent an unalloyed medium-carbon steel used in general engineering applications, with a nominal carbon content of 0.35 to 0.45 wt pct. Steel with yet higher carbon content may be encountered in practice, but this medium-carbon steel represents the high-end of carbon content in widespread use. 10 × 10 mm2 bar from RS Components UK was acquired, from which 1.5-mm-thick sample pieces were cut (all pieces were then ground to a 2500 grit finish—therefore no samples for medium-carbon steel were tested in the “as-received” condition). Optical emission spectroscopy was performed on the CR4 and Key steel to determine their chemical compositions beyond the nominal specifications. The compositions as determined by the average of three sparks are shown in Table II.

Table II Chemical Compositions of the Low- and Medium-Carbon Steel Substrates
Full size table

To simulate steel scrap with a thick oxide layer formed in service at room temperature, the samples were treated in a box furnace in air at 500 °C for 24 hours to form a scale of Fe3O4.[18] To simulate steel with a minimal initial oxide layer, the samples were ground with sandpaper to 2500 grit and ultrasonically cleaned in water and rinsed with ethanol immediately before testing. Iron oxide forms on the surface of iron at room temperature rapidly after exposure to the atmosphere.[27] Therefore, a thin initial oxide layer was present on these ground samples.

To apply the coatings, sodium sulfate (Na2SO4 anhydrous, from Fisher Scientific UK) and sodium metasilicate (Na2SiO3, from Sigma Aldrich UK) solutions were prepared, as described in the patent by Leak and Fine.[28] 15 g of Na2SO4/Na2SiO3 was mixed with 65 mL of water, and the solution was stirred on a hot plate for about 10 minutes at 60 °C to achieve dissolution. The steel samples were submerged into the solution and dried for about a minute under a hair dryer. The Na2SO4 coating left a white residue, while Na2SiO3 was present on the surface as a glassy gel.

Commercially available copper wiring from wires.co.uk (99.95 pct purity) was used for the copper droplet. 2.64 mm diameter bare copper wire was cut to 2.5 mm length pieces, each weighing 120 to 140 mg. All pieces were ultrasonically cleaned in water and rinsed with ethanol prior to testing to remove surface contaminants; however, a thin layer of copper oxide remained on the surface. Oxygen, either in solution with copper or in the form of copper oxide, affects the surface tension and wetting behavior of liquid copper, as characterized by Gallois and Lupis.[29] Copper wiring with a native surface oxide layer would be encountered in practice, so the condition studied here is relevant.

Droplet Observation

The aptitude of a liquid to wet a solid surface can be characterized by the shape of the liquid droplet. The sessile drop technique is a standard experimental method to measure droplet shape—most commonly Young’s contact angle, θY, formed at the contact of the liquid, vapor and solid phases,[30] as shown in Figure 1. The condition for wetting is defined as θY < 90 deg.

To observe the droplet, a Misura HSM2 heating microscope was used. The schematic is shown in Figure 2. The steel substrates were placed on the ends of two alumina sheaths, suspended within the furnace cavity. A thermocouple measured the temperature at the end of one of the alumina sheaths. The light source (a 50 W halogen spotlight bulb) illuminated the sample on one end, while a camera imaged the droplet profile through a transparent quartz screen. Argon (from Air Products BIP, N6.0 purity) flowing at a rate of 0.2 L/min was purged through the furnace cavity (200 × 18 mm diameter) and furnace inlet tube (40 × 3 mm diameter) for 60 minutes prior to testing with the sample in the chamber, and flowed at the same rate during testing. A transparent quartz screen was secured at the end of the inlet tube, with a gap of 3 mm between the screen and the tube. As the furnace cavity was not sealed, some oxygen was present in the system, which we estimate to be in the range of 10−4 to 10−3 atm.[31] The oxygen partial pressure to prevent the oxidation of iron at 1000 °C is extremely low, at 10−12 atm.[32] A piece of metallurgical coke (approximately 12 × 12 × 3 mm) was placed in the inlet tube, about 10 mm in front of the sample holder. Under flowing argon, the piece of coke had limited contact with oxygen for any substantial reaction, but CO(g) and CO2(g) may have formed,[33] the ratio of which would have had a reducing or oxidizing effect.[34] These conditions were intended to approximate an inert industrial furnace. The atmosphere during testing remained oxidizing, evidenced by the surface oxide formed on all samples and decarburization of the medium-carbon steel. Instrumentation or CO/CO2 gas mixtures can be used to directly measure and control oxygen concentration. The present apparatus was not equipped with these capabilities, so this is recommended for future analysis. The in-air test was run by flowing compressed air through the furnace, also at 0.2 l/min (the piece of metallurgical coke was removed for this test).

Fig. 2
figure 2

Schematic of the Misura HSM2 heating microscope

Full size image

The same heating profile was applied during each test. The sample was heated from 20 to 1000 °C at a rate of 50 °C/min, and from 1000 to 1150 °C at 5 °C/min. One snapshot was taken at each degree, from 1000 to 1150 °C. The contact angles of both sides of the droplet were later measured manually using ImageJ software. In some cases, the copper wetting was rapid—the thermal cycles of the medium-carbon tests were stopped early to prevent the sample from brazing to the thermocouple. After reaching the maximum temperature, the furnace was turned off, and the samples were cooled in the protective atmosphere of the furnace chamber.

Metallography

Upon cooling, samples were vacuum-cold mounted in epoxy and cross sectioned through the center. Each sample was prepared using the standard grinding and polishing procedure, to 0.3 μm with an alumina suspension. A Leica optical microscope was used to image the cross section of each sample. All samples were imaged in the unetched condition. Later, a 2 pct Nital solution was applied to medium-carbon substrates to reveal surface decarburization. Copper is easily distinguished from steel in optical microscopy. Scanning electron microscopy (Desktop Phenom G2 Pro, 5 kV working voltage in backscattered mode) was used to reveal the samples at higher resolution. Copper is distinguished as the lighter phase in these images. To determine the composition of the observed oxides and provide insight to the possible reactions resulting from the applied coatings, energy dispersive spectroscopy, EDS (Oxford instruments, x-act Penta FET Precision, 20 kV working voltage, IProbe = 50 pA) was performed at multiple points for several samples. To overcome difficulties with the nonconductive epoxy charging during EDS analysis, these samples were prepared with a gold coating and silver trace.

Results

Figure 3 shows snapshots of the copper droplet profiles on the low- and medium-carbon steel with different surface preparations at multiple temperatures upon melting, from 1060 °C to 1150 °C. The most obvious distinction is between the low- and medium-carbon substrates: the droplets on low-carbon steels are generally stable through to 1150 °C, while the droplet disperses and spreads, before the end of the heating cycle for the medium-carbon steels. The copper droplet on the coated low-carbon steels (with the exception of the oxidized-Na2SO4 sample) decreases in height and adopts an asymmetrical shape at the end of the thermal cycle. Amongst the various medium-carbon substrates, the droplet on the oxidized sample was stable for the longest, and shows the highest temperature reached before complete spreading across the substrate. The coatings on the medium-carbon substrates appear to accelerate wetting. The droplets on the Na2SO4-coated medium-carbon samples show a textured, uneven surface developing on the droplet during spreading, while copper on the Na2SiO3-coated samples appears to flow rapidly under the lifted surface layer.

Fig. 3
figure 3

Profiles of copper droplets on low-carbon and medium-carbon steels with different surface preparations from 1060 °C to 1080 °C at 5 °C increments and 1080 °C to 1150 °C at 10 °C increments

Full size image

It should be noted that there is some variability on the left-hand side of the figure, in the temperature at which copper forms a droplet. Commercially available copper wiring was used. Oxygen is the main impurity present in electrical copper, typically in the range of 250 to 400 ppm. The copper-oxygen phase diagram shows this level of oxygen decreases the melting point by 10 °C to 15 °C. A discrepancy between the temperature of the sample and the temperature measured by the thermocouple is also present, which likely explains the different melting points observed between the tests. These factors should not affect the overall comparisons of droplet evolution between the different substrates.

Figure 4 quantifies the contact angle (averaged between the left- and right-hand sides of the droplet) for the low-carbon (a) and medium-carbon (b) substrates as temperature increases (at 5 °C/min). The star-shaped marker denotes when the droplet forms, and an x-shaped marker shows when an asymmetric shape evolves (defined as when the difference between the two sides is greater than 15 deg). Again, the main trend to note is the difference between the low- and medium-carbon substrates. The contact angle decreases to zero in all of the medium-carbon cases, with the oxidized medium-carbon substrate maintaining contact angle over 90 deg at the highest temperature. The droplets for the low-carbon cases show a slight, gradual decrease in contact angle. The Na2SiO3 and Na2SO4-coated samples adopt an asymmetric shape at high temperature. The low-carbon sample treated in air shows the most rapid wetting behavior of all cases, never forming a spherical droplet, but spreading immediately over the surface.

Fig. 4
figure 4

Droplet contact angle over temperature for the various preparations of low-carbon substrates (a) and medium-carbon substrates (b). The heating rate in all cases is 5 °C/min

Full size image

The cross sections of the solidified copper on the substrates are compared side-by-side in Figure 5. The oxidized, ground, and as-received low-carbon steel samples show a spherical droplet with limited spreading. The reduced height and asymmetric shape of the droplets on the coated low-carbon samples is explained by some copper flow between the surface scale and the substrate. The exception is the oxidized-Na2SO4 sample, which shows no copper flow between the surface layer and the substrate and negligible diffusion of copper into the substrate. The medium-carbon substrates show copper spread sporadically over the top and bottom surfaces. In all cases, various pores, surface cracking and internal cavities of the copper droplet can be seen. This is a common observation for pure copper and high copper alloys, which are intrinsically susceptible to porosity.[35] The specific volume of copper decreases by 5 pct in the liquid–solid transition. Commercial copper casting uses a controlled, fine dispersion of oxygen to offset this decrease in volume. The sample cross sections exhibit cracks and pores from uncontrolled shrinkage and gas formation during solidification. Oxygen was present in liquid copper from the atmosphere and dissolution of the iron oxide. During cooling, oxygen solubility decreases. Excess oxygen can also react with hydrogen, resulting in water vapor porosity.[36]

Fig. 5
figure 5

Cross sections of solidified copper on all substrate preparations. All samples were cooled in the furnace chamber after the thermal cycle. All low-carbon samples reached a maximum temperature of 1150 °C, but to prevent brazing of liquid copper to the alumina sheaths, the thermal cycle was stopped early for most medium-carbon samples. The maximum temperature reached for the medium-carbon samples is shown in the figure

Full size image

Discussion

The mechanisms controlling the observed wetting behavior must be determined. As discussed in the introduction, diffusion of liquid copper into the steel substrate, as well as chemical dissolution, reduction, and mechanical disruption of the oxide are known to promote wetting of copper to steel. Examining the varied substrates at higher magnification reveals evidence of these mechanisms. The implications for process development are then discussed.

Liquid Copper Wetting on Steel in Air

Figure 6 shows how copper distributes within the typical three-oxide-layered scale that forms on low-carbon steel when air is circulated through the furnace. The majority of the scale would be wüstite at high temperature, which transforms partially to magnetite upon cooling. Copper penetrates the pores in the wüstite, while the majority of copper appears to be present at a seam close to the substrate. This seam exhibits nonuniform features of oxides and pores. It appears iron oxide was delaminated from the surface by the flowing liquid copper, which exposed the surface to oxygen to allow further reaction. This result confirms that to enable copper separation, scale formation must be limited.

Fig. 6
figure 6

Iron oxide scale with liquid copper entrapment when low-carbon steel was treated in air

Full size image

Liquid Copper Wetting on Low-Carbon Substrates

Figure 7 shows micrographs of the main features of the copper droplets on the as-received and initially ground and oxidized low-carbon substrates. The main phenomena shown in these images are iron oxide dissolution and oxide precipitation within and on the surface of the copper droplet, copper diffusion into the substrate, and liquid copper flowing outwards in the gap between the oxide scale and substrate. The initially oxidized sample shows more oxide dissolution (and iron oxide precipitation) than the initially ground sample. EDS analysis reveals that iron oxide precipitated as a seam in the copper droplet. A thin oxide formed on the surface of the copper droplet (EDS analysis shows the composition of this oxide varies and contains both iron and copper). For all samples, the iron oxide scale shows two layers distinguishable by color. The surface layer is a higher-oxygen oxide, magnetite or hematite. When contact of the scale with the substrate is disrupted, the transport of Fe ions is disrupted as well and the scale oxidizes further. The transformation of the iron oxide scale after a gap forms between the scale and substrate is described by Kim and Lee.[37] Thus, the proportion of the higher-oxygen oxide layer is an indication of when the scale was disrupted. The flow of liquid copper disrupted scale contact, especially for the as-received and ground conditions. Copper wets to the iron substrate beneath the surface scale, but overall a stable droplet with limited outward copper flow between the oxide layer and substrate can be achieved on uncoated low-carbon steel whether there is a significant oxide layer initially present, or if the surface is ground or in the as-received condition. The oxygen in the furnace atmosphere reacts to form an adherent iron oxide layer on the initially ground sample. The oxygen also increases the surface tension of liquid copper and reacts to form surface oxides to impede droplet flow.

Fig. 7
figure 7

Optical micrographs of the copper droplet on low-carbon steel substrates in the as-received, oxidized, and initially ground conditions (a). SEM images in (b) show the copper/oxide/steel interface. SEM images of the oxide layer are shown in (c)

Full size image

Figure 8 shows the low-carbon samples with an applied Na2SO4 coating. There is a significant difference in wetting behavior between the Na2SO4-coated initially oxidized and ground samples. Decohesion of the oxide with the substrate is observed in the initially ground sample, with copper flowing in this gap, while the scale on the oxidized-Na2SO4 sample appears to be the most resistant against disruption from the liquid copper. The difference between the two samples may be the reaction progression. Na2SO4 deposits on iron at high temperature cause an accelerated oxidation reaction known as “hot corrosion.” Iron reacts with Na2SO4 to form iron sulfide and oxides. Shi[38] shows that Na2SO4-Na2O eutectic melt is formed, which maintains contact between the Na2SO4 deposit and substrate to progress the reaction. Iron oxide dissolves in the melt and precipitates as discreet particles at the melt/solid interface. Once all the sulfate ions are reduced, only solid Na2O and iron oxide are left. EDS analysis confirms that the scale of the Na2SO4-coated initially oxidized sample consists of iron oxide at all points tested. With significant iron oxide initially present, the reaction may have progressed to completion, such that the scale was solid and protective when the liquid copper melted. However, hot corrosion may have been in the propagation stage for the initially ground sample. The molten deposit destroys the protective reaction-product oxide scale,[39] and the scale may have fractured at the molten layer, a weak point. Overall, it appears that attaining a resilient, nonwetting scale with the Na2SO4 deposit depends on the initial oxidation condition. If the reaction is incomplete, the deposit may weaken the scale with a liquid reaction product.

Fig. 8
figure 8

Optical micrographs of the copper droplet on initially ground and initially oxidized low-carbon steel substrates with an applied Na2SO4 coating (a). SEM images in (b) show the copper/oxide/steel interface. SEM images of the oxide layer are shown in (c)

Full size image

The cross sections of the initially ground and oxidized surfaces coated with Na2SiO3 are shown in Figure 9. The coating appears to have caused decohesion between the iron oxide and steel substrate, which then provided a pathway for copper flow. Na2SiO3, also known as waterglass, is used in a wide range of high-temperature adhesion and repair applications. The melting point of Na2SiO3 is 1089 °C.[40] During heating, excess water evaporates such that waterglass cures[41] in a polymerization process,[42] leaving silicate compounds with glass-like properties. The material can flow to fill cracks and act as a sealant. In this case, it appears the applied Na2SiO3 flowed through the oxide layer and the silicate compounds collected in the wüstite pores, shown as the darker phase at the base of the oxide scale. EDS analysis confirms that the scale consists of iron oxide and another phase, in which Na, Si and O present, but could not be stoichiometrically identified. SiO2 may have reacted with iron oxide and form fayalite, Fe2SiO4. Surface roughness at the interface is also evident, a feature of fayalite formation.[43] Overall, the formation of a silicon-rich liquid phase may have compromised the connection between the oxide layer and the substrate, allowing copper flow. Silicates may have also introduced additional stress and volume mismatch between the oxide and substrate. Lastly, the vaporization of water may have also contributed to the lack of adhesion. Matsumoto et al.[44] studied the adhesive performance of waterglass and found the adhered film thickness decreased as temperature increased up to 1000 °C, which the authors attributed to the formation of a vapor film. In any case, Na2SiO3 did not act to seal the connection between the oxides in substrate. The mechanisms of iron oxide dissolution and copper oxide precipitation at the base of the droplet are not apparent.

Fig. 9
figure 9

Optical micrographs of the copper droplet on initially ground and initially oxidized low-carbon steel substrates with an applied Na2SiO3 coating (a). SEM images in (b) show the copper/oxide/steel interface. SEM images of the oxide layer are shown in (c)

Full size image

Liquid Copper Wetting on Medium-Carbon Substrates

The images revealing the main characteristics of the medium carbon initially ground and initially oxidized cross sections are shown in Figure 10. Copper established a connection with the steel substrate and penetrated grain boundaries by flowing between the oxide layer and substrate. The oxidized sample exhibited the slowest spreading of all the medium-carbon substrates, and evidence of iron oxide dissolution in copper and precipitation of a seam of iron oxide, similar to the low-carbon cases, can be seen. The spread of copper is sporadic, and pores and formation of an oxide on the surface of copper can be seen. The oxidation of the liquid copper surface during spreading may have contributed to the sporadic distribution, as copper oxide would act to intervene and prevent liquid flow. The iron oxide scale on the uncoated medium-carbon samples is much thinner than the oxide scale observed in the uncoated low-carbon samples.

Fig. 10
figure 10

Optical micrographs of the copper spread on medium-carbon steel substrates in the initially ground and initially oxidized condition (a). SEM images in (b) show the copper/oxide/steel interface. SEM images of the oxide layer are shown in (c)

Full size image

The extensive wetting in the medium-carbon cases cannot be attributed only to the lack of adherence between the oxide and substrate, as decohesion was observed for coated low-carbon substrates, but the droplet remained intact. It is likely that the formation of CO(g) caused bulging and blistering of the scale to accelerate wetting. As reviewed by Chen et al.,[45] all steel types are susceptible to blistering above 800 °C due to internal stresses during growth as iron ions diffuse outward and vacancies diffuse inward, as well as from the differing crystal structures and thermal expansion coefficients between the scale and substrate. However, higher-carbon steels are more susceptible to blistering due to the accumulation of gases. Carbon reacts with oxygen in wüstite to form CO(g), which fills pores and voids in the scale, or escapes through cracks and ruptures. CO(g) can also act as an additional reductant to produce CO2(g). Figure 11 shows the initially oxidized medium-carbon sample upon etching with Nital to reveal ferrite (α-Fe, the white phase) and cementite (Fe3C, the dark phase). The decreased presence of cementite close to the surface confirms significant decarburization. The etched original medium-carbon steel material is shown in image (c) for comparison.

Fig. 11
figure 11

The initially oxidized medium-carbon substrate with a 2 pct Nital etching to reveal decarburization, with higher magnification images shown in (a) and (b). The etched original medium-carbon substrate is shown in image (c)

Full size image

The scale on the medium-carbon steel also appears to be intrinsically more irregular and less adherent. Figure 12 shows the initial oxide scales at several points formed on the low- and medium-carbon steels after the oxidizing treatment, before testing. The oxide scale on low-carbon steel was thick and adherent, while the scale on the medium-carbon steel was thinner, and in places detached or exhibited pores at the interface. Carbon is known to cause a more erratic oxidation reaction,[46] but the steels also had different processing histories. Medium-carbon steels are typically in the form of bars, and may be hot rolled or annealed. Low-carbon steels are more commonly cold-rolled into sheets. More adherent and thicker scales form on cold-worked surfaces, while the scale adhesion on annealed/electropolished surface is poor, resulting in an irregular scale structure.[47]

Fig. 12
figure 12

Initial oxide scales for medium- and low-carbon steel samples

Full size image

The cross sections of the medium-carbon substrates with the Na2SO4 coating are shown in Figure 13. It appears that the reaction for a solid, protective scale was incomplete for the Na2SO4-coated samples. The blistering of the scale would have disrupted contact between the substrate and Na2SO4 coating to terminate the hot corrosion reaction. EDS analysis shows that the majority of the scale is iron oxide, but an occluded darker phase contains significant quantities of Fe, Na, O, as well as Si (which would have oxidized from the Si-containing steel substrate and may have formed fayalite). It appears that Na2SO4 and Na2S may have reacted with copper upon melting. Voids in the scale close to the interface are apparent. Carbon from the steel, or CO(g), may have reduced the Na2SO4 deposit to form Na2S and CO(g) or CO2(g), respectively.

Fig. 13
figure 13

Optical micrographs of the copper spread on medium-carbon steel substrates in the initially ground and initially oxidized condition with the Na2SO4 coating applied (a). SEM images in (b) show the copper/oxide/steel interface. SEM images of the oxide layer are shown in (c)

Full size image
$$ {\text{Na}}_{2} {\text{SO}}_{4} + 4{\text{C}} \to {\text{Na}}_{2} {\text{S}} + 4{\text{CO}}\left( {\text{g}} \right) $$

The reaction of copper with Na2S, shown below, is well known to be favorable.[48] Precipitates are distributed throughout the solidified copper, which EDS analysis confirms is copper sulfide.

$$ {\text{Na}}_{2} {\text{S}} + 2{\text{Cu}} \to {\text{Cu}}_{2} {\text{S}} + 2{\text{Na}}\left( {\text{g}} \right) $$

The droplet profiles of the medium-carbon Na2SO4-coated samples in Figure 3 show a textured reaction product enveloping the surface of the copper droplet as it spreads.

Figure 14 shows the features of the Na2SiO3 coated samples. Again, it appears silicates are present as a darker phase in the oxide, but this likely aided in decohesion. It appears the scale lifted and provided a pathway for copper flow. The spreading of liquid copper was most rapid for Na2SiO3-coated medium-carbon samples.

Fig. 14
figure 14

Optical micrographs of the copper spread on medium-carbon steel substrates in the initially ground and initially oxidized condition with the Na2SiO3 coating applied (a). SEM images in (b) show the copper/oxide/steel interface. SEM (optical for the oxidized with Na2SiO3 sample due to issues with sample preparation) images of the oxide layer are shown in (c)

Full size image

Overall, the coatings do not appear to be especially useful. The improved liquation aided by Na2SO4 and Na2SiO3 as described by Leak and Fine[26,28] is not substantiated here. The difference in environments between this study and the study by Leak and Fine may explain this discrepancy. The molten salts employed in their tests, BaCl2 and CaCl2, may have dissolved the iron oxide scale present, such that the Na2SO4 and Na2SiO3 barriers were critical to prevent brazing. In a gaseous environment, the coatings weaken adherence between the scale and substrate to provide a pathway for copper flow.

Process Development

These results provide a basis to begin defining a preferential melting process. As discussed in the introduction, copper external to steel requires less energy to remove compared to copper dilute in the steel melt. Upon melting steel scrap, copper and iron are completely miscible, but the solubility of liquid copper in solid austenite is limited. Indeed, the limited solubility of liquid copper in austenite as iron oxidizes during hot rolling is the principle behind the surface hot shortness problem. The proposed preferential melting process seeks to exploit the same principle, but during scrap heating. Wetting must be minimized, so an adherent nonmetallic surface layer must be maintained for the duration of the treatment. Copper may wet to the steel substrate once the surface layer is penetrated or compromised, but the results here show a stable liquid copper droplet can be achieved on an uncoated low-carbon steel in an inert, or lightly oxidizing, atmosphere. Further work required to develop this idea is discussed below. Directions for exploring a wider range of conditions are discussed as well.

For energy-efficiency and industrial feasibility, the heat for preferentially melting copper must be retained to then melt the steel scrap, so the process conditions must be compatible with the conventional EAF route. The conditions of low-carbon content and a lightly oxidizing atmosphere to achieve nonwetting do not present any significant scale-up obstacles. However, further work is required to determine the levels of carbon concentration and oxygen partial pressure, which cause liquid copper to transition from nonwetting to wetting behavior. Low-carbon sheet steel is used in cars, goods, and equipment, which makes up the shredded scrap contaminated by copper. However, these steels may contain up to 0.25 wt pct C. Decarburization can be prevented by adjusting the furnace atmosphere, specifically by increasing the carbon potential. The atmospheric conditions required to heat treat a range of carbon steels while avoiding decarburization are specified,[31,34] so knowledge and expertise in downstream processing could be applied to scrap heating. Similarly, the oxygen potential of the atmosphere must be further investigated to determine the boundary between achieving a protective, thin oxide layer on both the droplet and substrate, and problematic entrapment of liquid in the scale.

Knowledge on scaling behavior during steel cooling and hot rolling could be applied. Scale formation and adhesion has been studied extensively.[49] Increased H2O or CO2 in the atmosphere is known to increase the plasticity of the scale and improve adhesion.[45] Precipitating a magnetite seam at the scale/steel interface with an appropriate heating profile is also shown to improve adhesion,[50] and there are many coatings to control the iron scale proposed in patents.[51,52]

The other key part of the process is applying a mechanical force to separate liquid copper from the steel surface. Dynamic flow behavior of a liquid metal on a solid substrate can generally be inferred from the evolution of a static droplet.[22] However, the goal of the process would be to minimize contact and avoid phenomena such as dissolution of the oxide layer and penetration into pores observed in these static tests. The time-scale of these behaviors and how liquid copper flows over the scale must be determined with further experimental testing.

A liquid environment, such as a chemically inert molten salt bath, might be especially helpful to aid in the separation of liquid copper from steel. The bath composition could be optimized to promote coalescence of the liquid copper, increase the thermal efficiency of heating, and decrease the weight of the scrap by 25 to 60 pct due to buoyancy,[53] which would help with shaking and agitation. The salt bath could also aid in controlling substances covering end-of-life steel scrap such as, zinc, paints, and tin coatings. The combustion of these substances during current EAF steelmaking is problematic, causing dust and pollution,[54] and alternative processes for preheating scrap while simultaneously removing or treating coatings are being developed.[55,56] Molten salt baths are currently investigated at the pilot scale as a more efficient and clean method to treat municipal waste.[57] Such setups are scalable, as molten salts are also widely used in metallurgy, for aluminum melting and steel heat treating. The interaction of liquid copper with the steel surface in an inert molten salt must be explored.

Thus, future study is required to investigate promising directions for process development. Overall, a viable process must be designed through an iterative consideration of scientific and practical aspects: converging an understanding of the wetting behavior of copper on steel, with the industrial constraints of treating heterogeneous end-of-life scrap.

Conclusions

This study establishes the influence of steel carbon content, initial surface oxidation, and coatings on the wetting behavior of liquid copper to steel. The key parameter to affect rapid wetting appears to be carbon content. Blistering of the oxide scale provided a pathway for extensive copper flow between the scale and substrate. The Na2SO4 and Na2SiO3 coatings appeared to weaken the connection between iron and its oxide to allow copper flow, to supersede the typical iron oxide dissolution observed in the noncoated samples. The exception was the oxidized-Na2SO4 surface on low-carbon steel—the resultant scale was fully intact at the interface and prevented liquid copper penetration. The applicability of the coatings is questionable, but a stable droplet was observed on the uncoated low-carbon steel regardless of the initial oxidation condition, which is promising for future process development. With a renewed urgency to control copper in steel recycling, previously investigated techniques must be reconsidered. For this technique and others, it is important to test the limits for controlling tramp elements.