Impact Milling of Printed Circuit Board Waste for Resource Recycling and Evaluation of Liberation using Heavy Medium Separation+

The authors evaluated the degree of liberation of printed circuit boards ground by impact milling to recover valuable materials. Wastes of two board types were ground with a swing-hammer-type impact mill. Heavy medium separation using tetrabromoethane was employed to separate the ground materials into the sink product, containing mainly copper, and the float product, consisting mainly of epoxy resin and glass fiber. The degree of liberation was evaluated with heavy medium separation. The sink product was larger than the float product. When processing boards whose surfaces are fully plated with copper foil we attained a degree of liberation for copper in the sink of up to 95%, which was higher than the degree of printed circuit board liberation. By contrast, the copper content of the float product was quite low at less than 1%. Thus, non-copper materials are easily liberated, and non-copper materials with few impurities can be obtained from printed circuit board wastes.


Introduction
Factory shipments of personal computers and other kinds of electronic information equipment are increasing at a blistering pace, which is indicative of the fastgrowing information society, and it is anticipated that the amounts of discarded equipment will likewise increase rapidly. The 1997 quantities have been calculated at 1,230,000 desktop computers and 430,000 laptop computers, while the weight of PCs discarded in 2002 is anticipated to be about 200,000 tons for the year.1. 2 Electronic information equipment contains many valuable materials, but a characteristic of such equipment is that they consist of about 10% printed circuit boards, which are made by copper foil wiring on epoxy reinforced with material like glass fibers, and resins such as phenol, combined in a laminated structure. Because copper accounts for about onethird to one-half, it is apparently possible to recover from electronic appliances an amount of copper equal to several percent of that used in the printed circuit 194 boards. 2 Furthermore, the electronic components mounted on circuit boards use gold, platinum, and other precious metals, which can also be recovered. 3 • 4 Major computer makers are setting up their own recovery and resource reclamation systems for discarded PCs and other items, but overall most junked equipment is disposed in landfills. 5 This gives rise to concerns about soil contamination by the lead and other hazardous substances in circuit boards, making it urgent to set up systems for resource reclamation. In view of the foregoing, the authors tried applying the shape separation technique in a process for recovering the copper from scraps produced when manufacturing printed circuit boards, and. from discarded boards. 6 · 10 In this attempt we focused on the nature of materials used in printed circuit boards, i.e., metals, consisting primarily of copper, become spherical particles when impact milled because of their great ductility and malleability, while the plastics and glass fibers making up the boards become non-spherical particles because they undergo brittle fracturing. This demonstrated that the efficient recovery of copper can be accomplished. And because the composition ratio of recovered materials is important for the reuse of copper, we also examined the recovered copper's liberatability and the dependence on milling conditions of the amount of solder contained in the recovered materials. 10 Therefore, to increase the reclamation of materials other than copper, such as plastics and glass, this report explores the liberation of plastics and glass, and the relationship between impact milling and the intermixing of plastics and glass with metals.

Experiment Samples
Our experiment used the two types of circuit boards described in Table 1 as samples. Sample 1 was glass fiber-reinforced unpatterned scrap from the printed circuit board manufacturing process. This sample therefore had no electronic components mounted. Copper content was about 60%.  Fig. 1 illustrates the experiment procedure. Samples were milled as described in previous reports. 6 · 10 They were first cut into 2-or 3-cm pieces by hand, then coarse cut to under 3 mm with a cutting mill (a model VM16 vertical cutting mill made by Orient Co., Ltd.). These pieces were milled with a swinghammer type impact mill (high-speed hammer mill 1018-LA made by Tokyo Atomizer Co., Ltd.) under appropriate milling conditions (1 mm mesh, 4,000 to 10,000 rev. min-1 ), and the product was used as the analysis sample. We used a Ro-Tap to screen the product for 10 min, and then we checked the particle size distribution. Tetrabromoethane with a density of 2,960 kg/m 3 was used to perform dense media separation, thereby separating the product into a sink product, containing mainly copper, and a float product, consisting mainly of non-copper components. 6 Concentrated nitric acid was used to dissolve only the copper from both particle groups. We then measured the amount of copper in both groups with induced-coupled plasma-mass (ICP) analysis, and evaluated the degree of liberation.  relationship between hammer speed and the median diameter X 50 (weight-based) of the sink and float products obtained with heavy medium separation of milled products. There were no appreciable differences in median diameter between samples, and just as we previously reported, 6 · 10 in all cases the float products, consisting mainly of resin and glass fibers, had smaller particle sizes than the sink products. This shows that glass fiber-reinforced epoxy resin undergoes brittle fracturing more readily than copper, and that the copper particles became large spherical particles because of their ductility and malleability. In connection with dependence on hammer speed, the graph clearly shows that the particle distribution of the milling product as a whole becomes finer as the impact increases, but distributions tended to differ somewhat between metal and resin/glass. Specifically, we observed that copper tended to become fine particles under low impact, whereas resin/glass tended to become coarse particles at high speed. 6 For the purpose of comparison the figure includes the milling results 8 for copper plate from our previous report. Although no large differences in values were discerned, we did note a tendency, when milling copper plate, for a simple value decrease in response to increasing impact. We conjecture that this is because the milling of copper that is combined with other materials differs from the milling of copper foil alone in that copper is affected by the milling characteristics of glass and plastic. Fig. 5 shows the degree of copper liberation a from sink product subjected to heavy medium separation using tetrabromoethane with a specific gravity of 2.96. 196 Here the degree of liberation a is defined by the following equation.

Liberatability
a= Wm.cu!Wm (1) Where: Wm is the particle weight recovered as sink product by heavy medium separation, and Wm,Cu is the copper content of Wm obtained by quantitative analysis.
It is evident from the definition that a is the apparent degree of liberation. In other words, the sink product conceived here is those particles with a specific gravity of 2.96 or higher, which are judged to be sink product even when nonmetals adhere to them. Accordingly, a will have a lower value than the true degree of liberation. As in Fig. 5, an 80 to 90% degree of liberation was achieved with sample 1 circuit board material for particles that were 100 11m or larger. Especially under high-impact milling, this particle size yielded a high liberation degree of at least 95%. Additionally, although it appears that the degree of liberation suddenly drops when particles are under 53 11m, it would seem this is for the following reasons: In general the finer the particles, the better the liberation, but when evaluating liberatability by means of heavy medium separation as in this instance, one cannot ignore the bias of dense media separation when separating fine particles. In other words, when separating comparatively large particles, the adherence of a slight bit of metal to nonmetals can be ignored, and the particles can be recovered as float product, but when particles are smaller than 100 11m, the adherence of even a little metal will make particle density relatively large, causing nonmetals to sink and be intermixed with the sink product. And as we shall discuss below, it is thought that because of a decline in the precision of  Fig. 6 shows the relationship between the liberation degree for each sample as a whole and milling conditions. The degree of liberation increased in both samples as hammer speed increased. With sample 1 we achieved a liberation degree of over 95%, showing that impact force is effective for liberation. On the other hand, circuit boards with resist-processed surfaces had a somewhat lower liberation degree of about 80% owing to factors such as the complexity of their laminated structure. Fig. 7 shows the relationship between float product liberatability and milling conditions. In accordance with Eq. 1, the following equation defines the parameter f3 of float product liberatability.

[3= 1-Wn.cuf Wn
Where: (2) Wn is the particle weight recovered as float product with heavy medium separation, and Wn.cu is the copper content of Wn obtained by quantitative analysis.
Just as in Eq. 1, f3 is the apparent degree of liberation, but the float product was a particle group with a particle density of under 2.96, and the resin and glass with a slight amount of metal adhering were recovered as float product. This means that f3 had a higher value than the true degree of liberation. Because the float product was a mixture of resin and glass fibers, 1-f3 here is the percentage of intermixed copper. Although liberation is facilitated more as impact force increases, f3 was 0.98 or higher in comparison with the liberatability of the copper component, and very little copper was mixed into the float product. Thus when reclaiming resin and glass, one obtains high-purity substances with little metal, which holds forth the possibility that their uses could be expanded. Fig. 8 plots the relationship between 1-f3 and particle size. Unlike the sink product, the copper intermixing rate increased as particle size grew. As noted previously, large resin and glass particles are recovered as float product even if slight amounts of metal adhere to them, indicating that liberation is inadequate. But when particles are small, those with metal adhering to them will sink, so the floating product's degree of liberation does not decrease. The reason that 1-f3 increases somewhat for fine particles is thought to be slight metal intermixing in the float product, which is because small metal particles sink with difficulty.
An evaluation of liberation and heavy medium separation used for printed circuit boards can be diagrammed as in Fig. 9. As the float product, resin and glass rather well liberated from copper are recovered, while the sink product is a mixture of well-liberated copper, as well as resin and glass to which copper has adhered, thereby giving their particles specific gravity of over 2.9. This results in the decline of apparent liberatability a, with the effect being especially apparent in fine particles.

Conclusion
In anticipation of the recovery of valuable materials from printed circuit boards, whose disposal amount is expected to increase rapidly, this paper has explored the liberation characteristics of such boards when using impact milling. We sought the apparent liberation degree of sink and float products which had undergone heavy medium separation in a dense liquid whose density was 2.96, and determined the relationship with milling conditions and other parameters. Our results produced the following conclusions.
1) Resin and glass became finer particles, and had higher liberatability, than copper. Thus even when the resin and glass are to be recycled, it is possible to recover them with very little metal mixed in.
2) Copper has a liberatability of 80% or more even from circuit boards such as those with electronic components actually mounted on them, and we demonstrated that liberatability of 90% or more is obtainable if one adjusts particle size by controlling impact force. : Degree of liberation of non-copper materials in the float product defined by Eq. 2 (-) 9) Ohya, H., S. Koyanaka, S. Endoh, H. Iwata, C. Izumikawa, H. Sasaki, and P Ditle. "Analysis of trajectory for ground printed wiring boards using shape separation," Shigen Shori Gijutsu, 44, 3-8 (1997

Hitoshi Ohya
Hitoshi Ohya received his BE and ME degrees in Chemical Engineering from Kyoto University. He has been at NIRE since 1986. He earned his doctor of engineering (DE) in 1997 from Kyshyu University for a thesis entitled "Study on shape separation of particulate materials". He works presently for environmental impact assessment of resource recycling process. E-mail: ohya@ nire.go.jp

Jae-chun Lee
Jae-chun Lee, Ph.D. is a senior researcher at Korea Institute of Geology, Mining & Materials. He stayed at NIRE as a research fellow in 1996. His main area of interest is refining of metallic and non-metallic materials by hydro metallurgical process.

Hiroyuki Iwata
Hiroyuki Iwata earned his DE degree in Mining Engineering in 1976 from Waseda University. He was the former leader of the Materials Handling and Characterization Division in NIRE. His major research interests are size reduction, characterization and separation of particulate solids. He works presently for international research cooperation program with eastern European countries.