Recovery of Palladium and Platinum Particles Suspended in the Al₂O₃–CaO–SiO₂ Slag Using Copper-Based Extractants at 1723 K

Abstract

Platinum group metals (PGMs) are precious rare metals mined only in a few regions of the world. Recycling PGMs used in automotive catalytic converters is of utmost importance to meet increasing global demand. The Rose process is the main recycling process used in Japan to recover PGMs from spent automotive catalytic converters. The recycling process involves smelting a ceramic structure, in which PGMs are supported, with Cu, Cu₂O, reductants, and fluxes such as CaO and SiO₂.

This study compared the Pd and Pt recovery capability of Cu or Cu₂O as extractants. Pd and Pt particles were simultaneously suspended in Al₂O₃–CaO–SiO₂ molten slag, and the extractant Cu or Cu₂O was added. The concentration of Pd and Pt in the slag as a function of processing time was investigated at 1723 K of the operating temperature in the Rose process, under carbon saturation. The results show that the suspended Pd and Pt particles were combined by collisions in the slag, and the recovery ratio and recovery speed of Pd and Pt suspended particles were higher when using Cu₂O than when using Cu. The Cu₂O dissolved in the slag was reduced to metallic Cu in the slag and alloyed with the suspended PGMs particles to form Cu-PGMs alloys. As a result, the particle size of the PGMs increased and sedimentation motion in the slag is promoted.

1. Introduction

Platinum group metals (PGMs) are considered the rarest elements on the planet and are classified as precious metals. The natural distribution of PGMs is highly concentrated in South Africa and Russia, accounting for more than 80% of the production from ores.^1,2) Recent predictions indicate that PGMs will be in short supply by 2030 if they are not recycled.³⁾ Moreover, because of their peculiar physical-chemical properties, PGMs cannot be replaced. For the sustainable management and efficient use of natural resources, PGMs recycling technologies are expected to become even more important in the future. Therefore, recycling PGMs is of utmost importance to meet increasing global demand.

Automotive catalytic converters account for over 60% of the PGMs global demand.⁴⁾ Since other PGMs form volatile oxides, palladium (Pd), platinum (Pt), and rhodium (Rh), supported on ceramic structures, are widely utilized and are essential for automotive catalytic converters. Currently, most recycling processes for recovering PGMs from spent automotive catalytic converters are based on pyrometallurgical processes.^5–11) In these processes, automotive catalytic converters are melted at high temperatures and fluxes, structures such as Al₂O₃ are concentrated in slag, and the PGMs are concentrated in molten metal (Cu, Fe, Pb, etc.) or these compounds (sulfide, chloride) added to separate the Al₂O₃ structure and the PGMs. Because the slag generated in the smelting process is eventually discarded, PGMs remaining in the slag are lost. Additionally, the revenue of the recycling process is significantly affected by PGMs price fluctuations rather than by other metal resources. Therefore, in recycling PGMs, a high recovery ratio and quick recovery are required.

The Rose process is the most widely used process in Japan for recycling PGMs.¹²⁾ In the Rose process, crushed automotive catalytic converters are melted in an electric furnace with fluxes of CaO, SiO₂ and Cu or Cu₂O as the extractant of PGMs at high temperatures under reducing atmosphere. The slag is separated from molten Cu, containing the concentrated PGMs. Next, oxidative smelting of the molten Cu containing the PGMs is carried out. To increase the PGMs concentration in the molten Cu, a part of Cu is separated into Cu₂O to obtain a Cu alloy with a high concentration of PGMs. The Cu₂O generated by oxidative smelting is then cooled, solidified, returned to the electric furnace, and mixed with the Al₂O₃–CaO–SiO₂ system slag of reductive smelting together with automotive catalytic converters, reduced and recovered to the alloy phase.

Cu and Cu₂O are used as extractants of PGMs in the Rose process. Nishijima et al.¹³⁾ reported phase equilibria data of PGMs between the Al₂O₃–CaO–SiO₂ slag system and liquid Cu, Nakamura et al.,^14,15) Chen et al.,¹⁶⁾ Baba et al.¹⁷⁾ reported the phase equilibria and solubility data of PGMs in slags of different compositions and various temperatures. To the best of the authors’ knowledge, studies reporting non-equilibrium or kinetic measurements are scarce in previous literatures. Regarding the processes of recovering PGMs from spent automotive catalytic converters using molten metal, Benson et al.¹⁸⁾ calculated the sedimentation velocity of PGMs particles and reported the recovery mechanism. Kolliopoulos et al.¹⁹⁾ also conducted experiments to recover PGMs on Cu. However, these studies used metal, and not Cu₂O, as the extractant. Except for the previous work by the authors,²⁰⁾ studies using Cu₂O as an extractant, as in the Rose process, are not reported in the literature. The previous study²⁰⁾ reported the difference in the recovery ratio and recovery speed of PGMs using Cu or Cu₂O as the extractant, demonstrating the superiority of Cu₂O. Additionally, our previous study²⁰⁾ reported experiments to recover Pd or Pt particles suspended in molten slag, without considering collisions between suspended particles.

Crucible experiments are the basis of the PGMs recycling process using Cu-based extractants. Due to the small size of Pd and Pt particles, a significant amount of time is needed for them to settle by themselves. In this work, Pd and Pt particles were suspended simultaneously in the Al₂O₃–CaO–SiO₂ slag system, at 1723 K of the operating temperature in the Rose process. The recovery ratio and recovery speed of Pd and Pt using Cu or Cu₂O as extractants were investigated. The behavior of Pd and Pt particles in the slag was studied, and the sedimentation velocity of the suspended particles was estimated.

2. Experimental Procedure

The powder sample was prepared by mixing 200 g of 35 mass%Al₂O₃–30 mass%CaO–35 mass%SiO₂. The powder sample was placed into a MgO crucible with an internal diameter of 65 mm and height of 60 mm. Then the sample was placed in an electric furnace of MoSi₂ heating elements. The temperature was set at 1723 K (±10 K) to melt the sample and maintained for 9 h, under an Ar atmosphere, to reach thermal equilibrium, and take into account the dissolution ratio of MgO. Next, the powder obtained by mixing 100 g of 35 mass%Al₂O₃–30 mass%CaO–35 mass%SiO₂ with 0.15 g or 0.3 g of Pd and Pt, respectively, in which the average diameter of Pd and Pt particles was about 0.6 µm was added from the top of the furnace through a quartz tube at room temperature (∼298 K). Immediately after the addition, a 250 mm long columnar carbon rod (∼125 g) with a diameter of 20 mm was inserted in the crucible to maintained carbon saturation inside the crucible. The sample was maintained at 1723 K (±10 K) for 3 h after the mixed powder was added, and 30 g of Cu or 33.78 g of Cu₂O (equivalent to 30 g of Cu element) with a particle diameter <53 µm was added through a quartz tube at room temperature (∼298 K). Sampling was performed immediately before and after 0.5, 1, 2, 4, 6, 8, and 12 h of adding Cu or Cu₂O. Approximately 1 g of the sample was collected from the top of the furnace by suction of the center of the molten slag using a mullite tube and a syringe, followed by water cooling. In order to confirm the natural sedimentation of Pd and Pt particles, the experiment was also conducted under the same conditions without adding Cu and Cu₂O.

After 12 h of adding Cu or Cu₂O, the carbon rod in the crucible was removed and the crucible was cooled in water. The remaining slag and alloy phases in the crucible were carefully separated. The Pd and Pt concentrations in the slag obtained by sampling from the top of crucible were quantified by inductively coupled plasma mass spectrometry (ICP-MS), and the Cu concentration in the slag and Cu, Pd, and Pt concentrations in the alloy were quantified by inductively coupled plasma optical emission spectrometry (ICP-OES). Additionally, a part of the obtained slag sample was observed, and the composition of metal particles suspended in the slag was analyzed by electron probe microanalysis (EPMA) and scanning electron microscopy with energy-dispersive X-Ray microanalysis (SEM-EDX).

3. Result and Discussion

3.1 Concentration of elements in the slag over time

The effect of Cu and Cu₂O addition on the recovery of the suspended Pd and Pt in the slag was investigated at 1723 K. Figure 1 shows the back-scattering electron image and the elemental mapping of a sample in containing 0.3 g of Pd and Pt obtained by EPMA, and maintained for 3 h before the addition of the Cu-based extractants. The Pd and Pt particles formed alloys and were suspended in the slag. Tables 1 and 2 list the concentration of Cu, Pd, and Pt in the slag and the suspended metal particles in the slag, respectively, when 0.15 g of Pd and Pt were added. These results are compared with the results reported in our previous study²⁰⁾ under the same conditions, in which 0.3 g of Pd or 0.3 g of Pt were added and recovered.

Fig. 1

Back-scattering electron image and elemental mapping of suspended Pd and Pt particle in the slag (EPMA).

Table 1 Experimental results on the recovery of 0.15 g Pd and 0.15 g Pt.

Table 2 Experimental results on the recovery of 0.3 g Pd and 0.3 g Pt.

The relationship between the concentration of Cu in the slag and processing time is shown in Fig. 2. In all cases, when Cu was added, the Cu concentration in the slag increased and became nearly constant after ∼1 h. On the other hand, when Cu₂O was added, the Cu concentration in the slag increased nearly up to 8,000 mass ppm after ∼1 h, and then gradually decreased over time. These experiments were performed under the carbon saturation. Therefore, it was considered that once Cu₂O was dissolved in Al₂O₃–CaO–SiO₂ slag it was reduced by C (s) or CO (g) to metallic Cu over time.

Fig. 2

Concentration of Cu in the slag as a function of time.

The relationship between the processing time and the concentrations of Pd and Pt in the slag is shown in Figs. 3 and 4, respectively. The Pd and Pt concentrations in the slag decreased over time due to natural sedimentation without the addition of Cu and Cu₂O. Comparing the results of Cu and Cu₂O addition, at equal processing times, the concentrations of Pd and Pt in the slag were lower using Cu₂O as an extractant than with Cu. Therefore, suggesting that Cu₂O is superior to Cu as an extractant in terms of the recovery ratio and recovery speed of suspended PGMs in the presence of Pd and Pt. This tendency was consistent with our previous results in which Pd and Pt do not coexist. Regarding the amount of Pd and Pt (0.15 g and 0.3 g), results show that the concentration of Pd and Pt in the slag were lower before and after the addition of 0.3 g of the extractant. Furthermore, the changes in the concentration of Pd and Pt in the slag using 0.15 g of Pd and Pt were similar to the results of our previous study using 0.3 g of Pd or Pt. Therefore, authors consider that an increase in the number of suspended particles in the slag leads to a high frequency of collisions between the particles. Additionally, authors consider that the growth speed of the particles and the sedimentation velocity increases with higher concentration of Cu in the suspended particles.

Fig. 3

Concentration of Pd in the slag as a function of time.

Fig. 4

Concentration of Pt in the slag as a function of time.

3.2 Recovery ratio of Pd and Pt

Table 3 lists the Cu, Pd, and Pt concentrations, mass, and recovery ratios of Pd and Pt in the alloy phase after cooling in water. Because the concentrations of Al, Ca, Si, and Mg in the alloy were <0.5 mass%, the alloy phase was regarded as an alloy containing only Cu, Pd, and Pt. The recovery ratio was calculated using eq. (1)   

\begin{equation} R = \frac{m_{\text{m}}\cdot x_{\text{m}}}{m_{\text{s}}\cdot x_{\text{s}} + m_{\text{m}}\cdot x_{\text{m}}}\times 100 \end{equation} (1)

where R is the recovery ratio [%], m_m is the mass of the alloy phase [g], m_s is the mass of the slag phase [g], x_m is the concentration of alloy phase [mass%], and x_s is the concentration of the slag phase [mass%]. In addition to the constituents Al₂O₃, CaO, and SiO₂, the slag contains MgO (crucible component), Cu₂O (derived from extractants) and suspended metal particles. By collecting samples over time, the slag in the crucible was reduced by ∼8 g. However, since the initial molten Al₂O₃–CaO–SiO₂ slag was 300 g, it is reasonable to assume that mass loss due to sampling has little effect on m_s. Furthermore, the concentration of Pd and Pt in the slag remaining in the MgO crucible after water cooling were roughly the same as the concentration in the slag samples collected by suction 12 h after the addition of the extractant. Additionally, a uniform concentration of Pd and Pt in the crucible was observed. The density distribution at equilibrium when particles are suspended in a liquid is given by Stokes’ law for particle sedimentation and Einstein’s theory for Brownian motion. It is well known that particles are distributed exponentially in the height direction. However, as discussed later, when the particle diameter is 1 µm, both sedimentation and Brownian motion velocities are <1 mm per hour, which is very small compared to the size of the crucible. Additionally, because of the influence of convection caused by the temperature difference in the crucible, it is considered that the suspended particles with small grain size do not reach the concentration distribution at equilibrium and exist almost uniformity in the slag. Therefore, it is assumed that the concentration of Pd and Pt in the slag is uniform, and the concentration of Pd and Pt in the slag phase are the concentration of Pd and Pt in the slag 12 h after the addition of the extractant (Tables 1 and 2). Additionally, the recovery ratio using Cu₂O as an extractant was higher than when Cu (Table 3).

Table 3 Recovery ratio of Pt or Pd in alloy phase.

3.3 Suspended metal particles in the slag

Figure 5 shows the back-scattering electron image obtained by SEM-EDX of the 0.15 g Pd and 0.15 g Pt sample suspended in the slag 1 h after the addition of Cu₂O. Metal particles of a Cu–Pd–Pt alloy with a diameter of several micrometers were observed in the slag. Furthermore, results show that the composition of these particles is nearly constant regardless of particle diameter and processing time. Figure 6 shows the relationship between the total Pd and Pt concentration in the alloy particles and the processing time. The reported concentrations are the average of five suspended particle samples. The concentration of Pd and Pt in the Cu–Pd–Pt suspended alloy decreased and the concentration of Cu increased with increasing processing time. The comparison between the extractants (Cu and Cu₂O) shows that a higher concentration of Cu in the alloy particles was obtained using Cu₂O than for Cu. Additionally, the addition of Cu₂O showed higher concentrations of Cu in the slag than those using Cu (Fig. 2). This is because the extractant Cu₂O was chemically dissolved in the slag. Subsequently, Cu₂O in the slag was reduced to metallic Cu with increasing processing time. The metallic Cu produced with this reduction formed Cu–Pd–Pt alloy particles with suspended Pd and Pt alloy particles. Moreover, more time was required to form the Cu–Pd–Pt alloy using Cu as an extractant.

Fig. 5

Back-scattering electron image obtained by (SEM-EDX) of the 0.15 g Pd and 0.15 g Pt sample suspended in the slag 1 h after the addition of Cu₂O.

Fig. 6

Concentration of Pd and Pt in alloy particles as a function of time.

3.4 Sedimentation velocity of metal particles

The relationship between the diameter of the metal particles and their sedimentation velocity in the slag was studied to understand the behavior of Pd and Pt particles using Cu₂O and Cu as extractants. According to SEM-EDX analysis, the Pd–Pt or Cu–Pd–Pt alloys in the slag showed particle diameters <10 µm. Therefore, the motion of the alloy particles in the slag can be described by sedimentation in the vertical direction (eq. (2)) and Brownian motion in the random direction (eq. (3)).²¹⁾   

\begin{equation} \text{Sedimentation motion (Stokes' law):}\ v_{t}\approx \frac{(\rho_{\text{p}} - \rho)D^{2}g}{18\mu} \end{equation} (2)

  

\begin{equation} \text{Brownian motion (Einstein's theory):}\ \skew3\bar{X}\approx \sqrt{\frac{2\text{k}_{\text{B}}T}{3\pi\skew2\bar{D}\mu}t} \end{equation} (3)

where v_t is the terminal velocity, ρ_p and ρ are the particle and fluid densities, D is the particle diameter, g is the gravitational acceleration, μ is the fluid viscosity, $\skew3\bar{X}$ is the average moving distance, k_B is the Boltzmann constant, T is the temperature, $\bar{D}$ is the average particle diameter, and t is the elapsed time. Because the Pd–Pt alloys were formed before the addition of the extractant, it was assumed that all Pd–Pt alloy particles suspended in the slag, before the addition of the extractant, were alloys with a mass ratio of 1:1. Taking D in eq. (2), and $\bar{D}$ in eq. (3) as D, $\skew3\bar{X}$ per hour of the two types of motion in the slag were calculated. The ρ_p used in the calculations was the average density of Pd and Pt particles. The values of other physical properties used in the calculations are listed in Table 4.^22,23) Figure 7 shows the relationship between the diameter of the Pd–Pt alloy particles and the average moving distance for the two types of motion. As the diameter of the particle increases, the average moving distance of vertical sedimentation motion increases, while for Brownian motion the average moving distance decreases with increasing particle diameter.

Table 4 Values of physical properties used in the calculations.

Fig. 7

Average moving distance as a function of the Pd–Pt alloy particle diameter. (——), Sedimentation motion; (- - -), Brownian motion.

Regarding the motion of the Cu–Pd–Pt alloy in the slag, sedimentation in the vertical direction was examined (eq. (2)). Assuming that eq. (4) and eq. (5) are valid when forming an alloy with a mass ratio of Pd+Pt:Cu of x:(1 − x), the sedimentation velocity can be calculated when no combining or splitting between alloy particles occurs.   

\begin{equation} \rho_{\text{alloy}} = x \rho_{\text{Pd–Pt}} + (1 - x) \rho_{\text{Cu}} \end{equation} (4)

  

\begin{equation} m_{\text{Pd–Pt}} = xm_{\text{alloy}} \end{equation} (5)

Furthermore, the mass of the particles can be calculated by eq. (6) and eq. (7).   

\begin{equation} m_{\text{Pd–Pt}} = \frac{4}{3}\pi \left( \frac{D_{\text{Pd–Pt}}}{2} \right)^{3} \rho_{\text{Pd–Pt}} \end{equation} (6)

  

\begin{equation} m_{\text{alloy}} = \frac{4}{3}\pi \left( \frac{D_{\text{alloy}}}{2} \right)^{3}\rho_{\text{alloy}} \end{equation} (7)

Equation (8) is valid from eq. (5) and eq. (6).   

\begin{equation} m_{\text{alloy}} = \frac{\pi \rho_{\text{Pd–Pt}}D_{\text{Pd–Pt}}{}^{3}}{6x} \end{equation} (8)

Equation (9) is valid from eq. (4), eq. (7) and eq. (8).   

\begin{equation} D_{\text{alloy}} = D_{\text{Pd–Pt}}\cdot \left\{\frac{\rho_{\text{Pd–Pt}}}{x(x\rho_{\text{Pd–Pt}} + \rho_{\text{Cu}} - x\rho_{\text{Cu}})} \right\}^{\frac{1}{3}} \end{equation} (9)

Substituting the particle diameter D_alloy in eq. (9) into eq. (2), the sedimentation velocity of the Cu–Pd–Pt alloy in the slag can be calculated from eq. (10).   

\begin{align} v_{\text{t}}&\approx \frac{D_{\text{Pd–Pt}}{}^{2}g}{18\mu}\cdot (x\rho_{\text{Pd–Pt}} + \rho_{\text{Cu}} - x\rho_{\text{Cu}} - \rho_{\text{slag}}) \\ &\quad \cdot \left\{ \frac{\rho_{\text{Pd–Pt}}}{x (x\rho_{\text{Pd–Pt}} + \rho_{\text{Cu}} - x\rho_{\text{Cu}})} \right\}^{\frac{2}{3}} \end{align} (10)

When the densities of the Pd–Pt alloy and Cu (ρ_Pd–Pt, ρ_Cu), and the density and viscosity of slag (ρ_slag, μ) are constant, the sedimentation velocity is determined from the particle diameter of the Pd–Pt alloy before alloying with Cu and the composition of the alloy particles. Figure 8 shows the relationship between the sedimentation distance, the particle diameter of the initial Pd–Pt alloy, and the composition of the alloy. Results show that the sedimentation velocity increases with increasing Pd–Pt alloy particle diameter before the addition of the extractant, and with increasing concentration of Cu in the Pd–Pt alloy particles. In addition, the Pd and Pt particles alone hardly settle because the particle diameter of the Pd and Pt powders used in this experiment was ∼0.6 µm. However, alloying with Cu increased the sedimentation velocity. Furthermore, author hypothesize that the collisions between particles that were not taken into consideration in this calculation increased the particle diameter and the sedimentation velocity was sufficient for phase separation. Figure 9 shows a schematic diagram of the settling mechanism of suspended Pd and Pt particles in the slag.

Fig. 8

Relationship between the sedimentation distance and the alloy composition for initial Pd–Pt alloy particles diameter of 0.5 µm (–·–·–), 1 µm (——), and 5 µm (- - -).

Fig. 9

Schematic diagram of the settling mechanism of suspended Pd and Pt particles.

4. Conclusions

The reduction smelting process using Cu-based extractants for the recycling PGMs was investigated in this work. To understand the behavior of suspended Pd and Pt particles in the slag and Cu-based extractants efficiency, Pd and Pt particles were suspended simultaneously in the Al₂O₃–CaO–SiO₂ slag system at 1723 K. The recovery ratio and recovery speed of Pd and Pt using Cu or Cu₂O as extractants were investigated and compared. The main conclusions are as follows:

1. (1)    The recovery ratio and recovery speed of Pd and Pt suspended particles were higher using Cu₂O than when using Cu.
2. (2)    The suspended Pd and Pt particles in the slag collide and combine with each other. A Cu–Pd–Pt alloy was formed in the slag because chemically dissolved Cu₂O was reduced to metallic Cu and combined with particles. The concentration of Cu in the alloys and the particle diameter increased with increasing processing time. The concentration change was faster when Cu₂O was added. As the particle diameter increased, sedimentation motion was promoted, allowing them to move more easily to the molten Cu phase at the bottom of the crucible.
3. (3)    As the number of suspended Pd and Pt particles in the slag increases, the collision frequency between the particles also increases, allowing faster grain growth and ease of movement to the molten copper phase.

REFERENCES

• 1)   R.  Jones: JOM 56 (2004) 59–63.
• 2)   S.  Kuriki,  I.  Daigo,  Y.  Matsuno and  Y.  Adachi: J. Japan Inst. Met. Mater. 74 (2010) 801–805.
• 3)   S.  Kondo,  A.  Takeyama and  T.  Okura: Shigen-to-Sozai 122 (2006) 386–395.
• 4)   D.  Xun,  H.  Hao,  X.  Sun,  Z.  Liu and  F.  Zhao: J. Clean. Prod. 266 (2020) 121942.
• 5)   T.  Okabe,  H.  Nakada and  K.  Morita: J. Surf. Sci. Soc. Jpn. 29 (2008) 592–600.
• 6)   T.  Okabe and  K.  Nose: Mater. Cycles. Waste Manag. Res. 22 (2011) 50–57.
• 7)   Z.  Peng,  Z.  Li,  X.  Lin,  H.  Tang,  L.  Ye,  Y.  Ma,  M.  Rao,  Y.  Zhang,  G.  Li and  T.  Jiang: JOM 69 (2017) 1553–1562.
• 8)   H.  Dong,  J.  Zhao,  J.  Chen,  Y.  Wu and  B.  Li: Int. J. Miner. Process. 145 (2015) 108–113.
• 9)   D.R.  Mac Rae: Plasma Chem. Plasma Process. 9 (1989) 85S–118S.
• 10)   C.  Hagelüken: Platin. Met. Rev. 56 (2012) 29–35.
• 11)  S. Ivanović, V. Trujuć, M. Gorgievski, L. Mišić and D. Božić: 15th International Research/Expert Conference, (TMT, 2011) pp. 701–704.
• 12)   S.  Suzuki,  M.  Ogino and  T.  Matsumoto: J. MMIJ 123 (2007) 734–736.
• 13)   W.  Nishijima and  K.  Yamaguchi: J. Japan Inst. Met. Mater. 78 (2014) 267–273.
• 14)   S.  Nakamura and  N.  Sano: Metall. Mater. Trans. B 28 (1997) 103–108.
• 15)   S.  Nakamura,  K.  Iwasawa,  K.  Morita and  N.  Sano: Metall. Mater. Trans. B 29 (1998) 411–414.
• 16)   M.  Chen,  K.  Avarmaa,  L.  Klemettinen,  H.  O’Brien,  D.  Sukhomlinov,  J.  Shi,  P.  Taskinen and  A.  Jokilaakso: Metall. Mater. Trans. B 51 (2020) 1495–1508.
• 17)   K.  Baba and  K.  Yamaguchi: J. MMIJ 129 (2013) 208–212.
• 18)   M.  Benson,  C.  Bennett,  J.  Harry,  M.  Patel and  M.  Cross: Resour. Conserv. Recycling 31 (2000) 1–7.
• 19)   G.  Kolliopoulos,  E.  Balomenos,  I.  Giannopoulou,  I.  Yakoumis and  D.  Panias: OALib. J. 1 (2014) e736.
• 20)   T.  Murata and  K.  Yamaguchi: J. Japan Inst. Met. Mater. 84 (2020) 115–120.
• 21)  The Japan Institute of Metals (ed.): Thermodynamics of Microstructures, (Maruzen, Tokyo, 2006) pp. 46–47.
• 22)  W. Haynes (ed.): CRC Handbook of Chemistry and Physics 93rd Edition, (Taylor & Francis Group, Oxford, 2013).
• 23)  V. Eisenhüttenleute (ed.): Slag Atlas 2nd Edition, (Verlag Stahleisen Gmbh, Dusseldorf, 1995).