Reduction of Thermal Expansion of Ferritic/martensitic Heat Resistant Steels -Alloying Effects on Thermal Expansion of α-Fe Phase-

Abstract

Alloying effects on the thermal expansion of the α-Fe phase in connection with the magnetic states were investigated with a final goal to design ferritic/martensitic heat resistant steels with a reduced thermal expansion coefficient at high temperatures. The thermal expansion coefficient decreases with increasing the content of the alloying elements M (M = Co, Cr and V) regardless of the types of the elements in the ferromagnetic state well below the curie temperature in each alloy. The thermal expansion coefficient increases with temperature more significantly in the paramagnetic states compared to the ferromagnetic states. The temperature dependence of the thermal expansion coefficient in the paramagnetic states is not much influenced by the type of alloying element. As a result, the lower the Tc is, the higher the thermal expansion coefficient tends to become in the paramagnetic states at high temperatures. Based on the results obtained, a way is proposed to design heat resistant ferritic steels with reduced thermal expansion coefficients; to add both Cr and Co as major alloying elements such that the Tc and the A3 temperatures are placed above their service temperature.

1. Introduction

Fossil fired steam turbine power generations have been used as major power source for more than a century since the first commercial power station was established in 1880s. Because the carbon dioxide emissions from industries are suspected to be a main cause for the global warming, however, a current worldwide trend is claiming the need to replace fossil fired power with renewable energy power such as solar-power and wind-power as main power source in near future. The renewable energy power sources highly depend on weather, climate, location (country) and time (day or night) and so on, and the establishment of back-up power generation systems will thus be important to maintain the power supply robust.

Steam power generation, including combined cycle power generation, will then be a reasonable candidate for the back-up power generation to accommodate the fluctuation of power supply from the renewable energy sources. The high temperature components of the power plants will hence be exposed to high thermal stress/strain due to high fluctuation load/temperature conditions in case of back-up operation. Thermal expansion is the essential factor to generate thermal stress/strain in materials and it is, therefore, important to develop heat resistant steels/alloys with reduced thermal expansion to minimize the thermal stress and thereby suppressing degradation of the materials due to thermal fatigue conditions.^1,2,3,4,5,6)

Ferritic heat resistant steels are widely used for large components in conventional fossil fired steam turbine power plants such as main steam pipes and headers etc. because of their relatively low cost, high strength/toughness and low thermal expansion coefficient. These steels have been developed by modification of their chemical composition and microstructure to improve the heat resistance against long term creep deformation, fracture and oxidation. The ferritic heat resistant steels with the highest creep resistance include 9–12 percent Cr with some minor alloying elements to possess tempered martensitic microstructure strengthened with carbide, carbonitride and intermetallic phases.^7,8) It would be, therefore, desirable to develop ferritic (martensitic) heat resistant steels with reduced thermal expansion coefficient as well as high creep resistance for future high fluctuation load operation of steam turbine power generations.

Spontaneous magneto volume effect is known to affect the thermal expansion of iron based alloys in which its magnetic state changes with increasing temperature from a ferromagnetic state to a paramagnetic state at the curie temperature.^9,10) Gehrmann et al.⁹⁾ found that the thermal expansion coefficient of the α-Fe phase is reduced in the ferromagnetic state by Co addition and assumed that the reduced thermal expansion coefficient is attributed to an enhanced spontaneous magneto volume effect due to the addition of the element whose magnetic moment is higher than Fe. Most of alloying elements in steels are reported to decrease the magnetic moment of iron but spontaneous magnet volume effects by such alloying elements have not been so far fully understood.

It is generally explained that the thermal expansion of materials is determined by the anharmonic nature of atomic bonding interaction.^11,12) It is also known that the thermal expansion of metallic materials tends to be in inverse proportion to the melting temperature of the materials.¹³⁾ When taking into account that the melting temperatures of the materials are attributed to their atomic binding energies, it would be reasonable to expect that the thermal expansion of alloys is determined by a change in the binding energy due to alloying, i.e. by the interaction parameters between iron and alloying elements.

The purpose of the present paper is to investigate alloying effects on the thermal expansion of the α-Fe phase in connection with the magnetic state and the interatomic interaction with a final goal to design ferritic/martensitic heat resistant steels with a reduced thermal expansion coefficient at high temperatures.

2. Experimental Procedures

The alloys studied are binary iron based alloys containing an alloying element M (M=Co, Cr and V). The alloys are designated by the content of alloying element in at.% through this paper unless otherwise specified. The types of alloying elements were selected primarily because of their high solubility in the α-Fe phase.¹⁴⁾ Co is known to enhance the magnetic moment of Fe, while Cr and V to weaken it.¹⁰⁾ The features of the alloying elements with regard to the solubility,¹⁴⁾ their effects on the magnetic moment of iron,¹⁰⁾ the interaction parameter against Fe¹⁵⁾ and their crystallographic structures are listed in Table 1. An unalloyed Fe and some ternary alloys were also studied.

Table 1. Features of the alloying elements with regard to the solubility,¹⁴⁾ the effects on the magnetic moment of iron,¹⁰⁾ the interaction parameter against Fe¹⁵⁾ and crystallographic structure.

Alloying element M	Solubility in ferrite at 700°C14)	Magnetic moment of Fe10)	Interaction parameter between Fe and M15)
Co	~40 at.%	Raise	−0.030 eV
Cr	~30 at.%	Reduce	−0.017 eV
V	~22 at.%	Reduce	−0.12 eV

The alloys are prepared by arc melting using 4N purity Fe, 3N Co, 3N Cr and 2N8 V in an argon atmosphere. These ingots were homogenized at 1100°C for 20 h and fabricated for dilatometric measurements and differential scanning calorimetry (DSC). The dilatometric measurements were performed for cylindrical samples with a size of 3 mm in diameter and 10 mm in height with a horizontal push-rod type dilatometry. A SiO₂ sample was used as reference. Thermal expansion was detected during heating in a temperature range between 50 and 1050°C at a rate of 5°C/min. An averaged data was selected from at least two measurements which were done for each alloy. The DSC measurements were performed for samples with 3 mm in diameter and 1 mm in height. The DSC signal was detected during heating at a rate of 10°C/min in a temperature range between 80 and 1250°C.

3. Results and Discussions

3.1. DSC Measurements

Figure 1 shows DSC curves obtained from some of the alloys studied. It can be seen that the endothermic peak from the curie temperature (Tc) increases with increasing the Co content and disappears by addition of 20% and above. The disappearance of the peaks in the high Co binary alloys suggests that the Tc exceeds the A3 temperature. Small endothermic peaks, observed at a temperature range between 600 and 700°C in the 30Co and 40Co samples, are assumed to be related to the order-disorder transition of the α’-FeCo phase, according to the reported phase diagram.¹⁴⁾ The Tc is seen to remains unchanged by addition of 10% Cr but decreases with further addition of Cr. The endothermic peak detected at ~590°C in the 30Cr sample may be caused by two factors. The first one is by the magnetic transition and the second one is related with the formation of the α’-Cr phase or the σ-FeCr phase based on the reported binary phase diagram¹⁴⁾ and previous works.^16,17) The origin for this peak is explained also in the next section. The Tc is detected to increase by 10% V addition but slightly to decrease by further addition. An endothermic peak at ~760°C in the 30V sample is assumed to be caused by a formation of the σ-FeV phase according to the reported phase diagram and a volume expansion as expected from the formation of the intermetallic phase with a higher molar volume expansion than the ferrite phase,¹⁸⁾ as explained in the next section. These observed alloying behaviors are in reasonably good agreement with the previously published data.

Fig. 1.

DSC curves obtained during heating at a rate of 10°C/min in the Fe–M binary alloys (M: Co, Cr and V): (a) Fe–Co alloys, (b) Fe–Cr and Fe–V alloys. Endothermic peaks at the curie temperature (Tc) are designated by triangles.

3.2. Dilatometric Measurements

Figures 2 and 3 show the thermal linear expansion and the thermal linear expansion coefficients of some of the alloys studied, respectively. The Tc temperatures, detected in the DSC measurements, are depicted by solid triangles in these figures. The thermal expansion coefficient, the slope of the curves in Fig. 2 and the vertical value in Fig. 3, tends to decrease with alloying. The decrease in the thermal expansion coefficients with alloying is also reported in literature.¹³⁾

Fig. 2.

Dilatation curves obtained during heating at a rate of 5°C/min in the Fe–M binary alloys (M: Co, Cr and V): (a) Fe–Co alloys, (b) Fe–Cr and Fe–V alloys. The Tc in the alloys, determined by DSC, is designated by triangles.

Fig. 3.

Change in linear thermal expansion coefficients with temperature in the alloys studied: (a) Fe–Co alloys, (b) Fe–Cr alloys, (c) Fe–V alloys. The Tc in the alloys is designated by triangles.

The detailed analysis in the present study reveals clear alloying effects in reference to the magnetic states in the ferrite phase. Figure 4 shows the thermal expansion coefficients replotted as a function of the temperature nominalized by Tc for the alloys where no significant second phase formation was identified within the ferrite phase. The following overall features can be found from this work;

Fig. 4.

The linear thermal expansion coefficients of the alloys replotted as a function of the temperature nominalized by Tc. The alloys in this figure show single-phase below the A3 temperature except 20V alloy where a small step, observed at T/Tc of ~0.7, is assumed to reflect the formation of a small amount of the σ-FeV phase.

(1) Negative peaks are visible at the Tc in the unalloyed Fe and in the alloys with alloying contents at and below 10 at.%.

(2) The depth of the negative peak is the largest in the unalloyed Fe and decreases with increasing the content of the alloying elements. The depths of the negative peaks are quantified and plotted as a function of alloying content in Fig. 5.

Fig. 5.

The depth of negative peakes observed at Tc as a function of the alloying content. The quantified method is illustrated in the figure. The extrapolated line was drwan in a linear way to pass the data between 200°C and 500°C. The difference among the alloys is not identified as significant.

(3) The thermal expansion coefficient decreases with alloying regardless of the types of elements in the ferromagnetic state well below the Tc.

(4) The thermal expansion coefficient increases with temperature more significantly in the paramagnetic states compared to the ferromagnetic states. The temperature dependence of the thermal expansion coefficient in the paramagnetic states is not much influenced by the type of alloying element.

(5) In the paramagnetic states, it is found that the lower the Tc is, the higher the thermal expansion coefficient tends to become (See Fig. 3).

The relatively large negative peak detected in the 30Cr alloy is assumed to be caused not only by the magnetic transformation but also by the formation of the second phase firstly because of the clear tendency of the alloying effect to reduce the size of the negative peak due to the magnetic transformation. The negative peak in this case would be rationalized by the formation of the σ-FeCr phase of which molar volume is reported to be smaller than that of the ferrite phase with 30 at.% Cr content and by a reduction in the lattice parameter of the ferrite due to the formation of the σ phase with the higher Cr content.^14,19) The large positive peak detected at a temperature range between 500 and 800°C in the 30V alloy is probably due to the formation of the σ-FeV phase of which molar volume is higher than that of the α-Fe.^14,18)

3.3. Alloying Effects

The obtained results are discussed in terms of the magnetic states and interatomic interaction below. Pepperhoff et al.^9,20) reported that the thermal expansion coefficients of non-magnetic metals are in inverse proportion to the melting temperatures of the metals and the following empirical equation holds:   

$${\alpha }_{(T/Tm)}\cdot T_m=\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{t}$$

where T_m and α_(T/Tm) are the melting temperature of the metal and the linear thermal expansion coefficient at a temperature normalized by T_m, respectively. By applying this empirical rule, the authors showed that the thermal expansion coefficient of the α-Fe phase deviates downward from the non-magnetic metals and this deviation increases with Co addition.⁹⁾ They state that the increased deviation by Co addition is due to an increased spontaneous magneto volume effect caused by the addition of the element that increases the magnetic moment of iron.

Normalizing the results obtained in the present study using the method adopted by Pepperhoff et al. rarely changes the overall features of the alloying effects because the melting temperatures of the binary alloys are not different much from that of unalloyed Fe, which suggests that the change in the thermal expansion coefficient values cannot be explained only in terms of the change in the magnetic moment due to alloying since the addition of Cr or V is known to decrease the moment while that of Co is reported to increase it.¹⁰⁾

A possible reason for the alloying effects could be different binding energy levels in each alloying element, as can be imagined from their melting temperatures. Figure 6 plots averaged thermal expansion coefficient (α) values of the alloys and pure alloying elements, obtained from the present study and from literatures,^13,21) as a function of the content of each alloying element. The α values were taken between 200 and 500°C in the ferromagnetic fields of the alloys studied. The α values obtained from the present study are in good agreement with those reported in literatures. It can be clearly seen that the values of the alloys are lower than the levels of the straight lines which ties Fe and each alloying element, which suggests that the alloying behavior should be governed by other factors than binding energy in each alloying element. It is needed to consider the validity to draw a straight line between pure Fe and pure Co since the hexagonal closed pack structure below ~420°C in the pure state of Co and other structures are expected to form in the Co rich alloys.¹⁴⁾ As can be seen in Fig. 6(a), however, the α values show a valley with a minimum at Fe-50Co and no appreciable jumps in value are observed at expected phase boundaries. Thus, it is reasonable to interpret that the reduction of the α by Co addition is also larger than expected by a difference in binding energy between Fe and Co.

Fig. 6.

Averaged thermal expansion coefficient of the alloys plotted as a function of the alloying content: (a) Fe–Co, (b) Fe–Cr, (c) Fe–V. The averaged values are taken from the temperature range between 200 and 500°C in the ferromagnetic states. All the alloys show lower thermal expantion coefficients than each linear line between Fe and each alloying element.

Another possible factor for the alloying effects on the thermal expansion coefficient can be the interatomic interaction between iron and each alloying element. As listed in Table 1, it is reported that the interaction parameters are all negative in the alloying elements selected when they are diluted.¹⁵⁾ One can imagine from these negative interaction parameters that alloying with the elements may enhance the binding energy between atoms within the α-Fe phase, and the potential curve thus becomes sharper and thereby reducing the thermal expansion coefficient values. A quantitative correlation is, however, not obtained between the interaction parameter values and the degree of the downward deviation from the linearity. The fact that the alloying effect is negligible in the paramagnetic states cannot also be explained by the interaction parameters only.

A possible scenario to interpret the observed results is described below. It is believed in theory that the atomic distance and volume are expanded by repulsive interaction between electrons with the parallel orientation of the spins in a ferromagnetic state.¹⁰⁾ The spontaneous magneto volume effect is explained by the occurrence of contraction from the expanded ferromagnetic state by disappearance of the magnitude or the parallelity of the spins in the paramagnetic state. The local negative peak observed at Tc in unalloyed Fe (Figs. 3 and 4) can be interpreted by this effect. Based on this theory, the fact that the thermal expansion coefficient and the size of the negative peak decrease with alloying in the ferromagnetic states can be explained such that the negative interaction between Fe and each alloying element would enhance the binding energy and thereby resisting the repulsive force introduced by spontaneous magnetic field. The negligible alloying effect in the paramagnetic states is not fully understood but it might imply the importance of the existence of the magnetism for the interatomic interaction to happen in this material. A theoretical calculation would be helpful for the binding energy levels of the iron based alloys in the different magnetic states to understand the observed alloying effects in connection with the magnetism.

3.4. Design of Low Thermal Expansion Ferritic/martensitic Heat Resistant Steels

The latest heat resistant ferritic heat resistant steels include 9–12 percent Cr with some minor alloying elements. Typical chemical compositions of the steels are listed in Table 2. These steels are used after forming a tempered martensitic microstructure strengthened with carbide, carbonitride and intermetallic phases.^7,8,15) Figure 7 compares the α values between 100°C and 700°C of some of commercial steels and alloys. The value of conventional ferritic heat resistant steel (Gr. 92) is comparable with that of a frequently used ferritic stainless steel (430) and lower than that of austenitic stainless steel (SUS304) and Ni based superalloy (718) by ~43% an ~13%, respectively. A way for further reduction of α in ferritic/martensitic heat resistant steels is discussed below.

Table 2. Typical chemical composition of principal alloying elements in commercial heat resistant ferritic/martensitic steels.

Steel designation		Typical concentration of main alloying elements
		Cr	Mo	W	V	Nb	Co	C
Gr.91	wt.%	9	1	–	0.2	0.1	–	0.1
	at.%	10	0.5	–	0.2	0.05	–	0.5
Gr.92	wt.%	9	0.5	1.8	0.2	0.1	–	0.1
	at.%	10	0.3	0.6	0.2	0.05	–	0.5
Gr.93	wt.%	9	–	3	0.2	0.05	3	0.1
	at.%	10	–	1	0.2	0.03	3	0.5

Fig. 7.

The averaged thermal expansion coefficient α between 100 and 700°C of commercial heat resistant steels and a Ni based alloy and of a steel designed to have a low α value in the present study.

Figure 8 summarizes the alloying effects on the α value of the α-Fe phase in the temperature range between 100 and 700°C. Cr, an essential alloying element in the heat resistant steels, is effective in reducing the coefficient and an increase in the content is good to meet the purpose. On the other hand, the increase in the Cr content stabilizes the ferrite phase and the addition of an austenite-stabilizing element should then be needed to obtain martensitic microstructures. Among austenite-stabilizing elements (Co, Ni, Mn, Cu, Re), Co is found to be best element since the element moderately stabilize the austenite phase with a largest solubility in the ferrite phase and effectively reduces the thermal expansion coefficient in the ferromagnetic state of the ferrite phase. Co addition has some advantages due to the raising of the Tc. First, the thermal expansion coefficient of the ferrite phase drastically increases above the Tc, as revealed by the present study. Second, it is considered that the self-diffusion coefficient of Fe is decreased by magnetic transformation from paramagnetic to ferromagnetic below the Tc and a reduced self-diffusion coefficient may be expected by raising the Tc.^22,23) These considerations lead us to propose a following way to design heat resistant ferritic steels with reduced thermal expansion coefficients; to add both Cr and Co as major alloying elements in such a way that the A3 temperature and the Tc are placed above their service temperature.

Fig. 8.

Alloying effects on averaged thermal expantion coefficient between 100 and 700°C of the α-Fe.

The α/γ phase boundaries and Tc were calculated in the Fe–Cr–Co ternary system with commercial software and are shown in Fig. 9. It is noted that these calculations are in reasonably good agreement with our experiments.²⁴⁾ It can be seen that a chemical composition range with an equi Cr/Co atomic ratio allows us to make martensitic steels where the γ phase is thermodynamically stable at high temperatures and the α phase is stable at low temperatures below ~700°C. In the ternary composition range the α decreases with the amount of Cr+Co contents and show a minimum at the Cr/Co ratio slightly higher than 1, as shown in the Fig. 9(b). Adding a high amount of Cr and Co may, however, promote the formation of brittle phase such as the σ phase and α’-Cr phase. Taking the above points into account, the chemical composition around Fe-(15-20)Cr-(15-10)Co would be good for the base composition in the design of ferritic heat resistant steels with tempered martensitic microstructures and a reduced thermal expansion coefficient at high temperatures up to 700°C. The α of 11.0×10⁻⁶ K⁻¹ is currently achieved in a ferritic/martensitic heat resistant steel designed by the indicated way (See Fig. 8).

Fig. 9.

Calculated α/γ phase boundaries in the Fe–Cr–Co ternary system at: (a) 1000°C, (b) 700°C. Calculated Tc lines at 700°C and 750°C and averaged thermal expansion coefficient values are also shown in (b).

4. Summary

Alloying effects on the thermal expansion of the α-Fe phase were investigated in connection with the magnetic states with a final goal to design ferritic/martensitic heat resistant steels with a reduced thermal expansion coefficient at high temperatures. The main results are:

(1) The thermal expansion coefficient decreases by alloying with Co, Cr and V regardless of the types of elements in the ferromagnetic state well below the Tc.

(2) The thermal expansion coefficient increases with temperature more significantly in the paramagnetic states compared to the ferromagnetic states.

(3) The temperature dependence of the thermal expansion coefficient in the paramagnetic states is not much influenced by the type of alloying element.

(4) The lower the Tc is, the higher the thermal expansion coefficient tends to become in the paramagnetic states at high temperatures.

(5) A way to design heat resistant ferritic steels with reduced thermal expansion coefficients is to add both Cr and Co as major alloying elements such that the Tc and the A3 temperatures are placed above their service temperature.

Acknowledgement

The present research was financially supported by Autorace.

References

• 1)   O.  Kwon,  M.  Myers,  A. D.  Karstensen and  D.  Knowles: Int. J. Press. Vessel. Pip., 83 (2006), 349. https://doi.org/10.1016/j.ijpvp.2006.02.012
• 2)   N. K.  Mukhopadhyay,  B. K.  Dutta and  H. S.  Kushwaha: Int. J. Fatigue, 23 (2001), 549. https://doi.org/10.1016/S0142-1123(01)00009-3
• 3)   B.  Fournier,  M.  Sauzay,  C.  Caës,  M.  Noblecourt,  M.  Mottot,  A.  Bougault,  V.  Rabeau and  A.  Pineau: Int. J. Fatigue, 30 (2008), 649. https://doi.org/10.1016/j.ijfatigue.2007.05.007
• 4)   T. P.  Farragher,  S.  Scully,  N. P.  O’Dowd and  S. B.  Leen: Int. J. Fatigue, 49 (2013), 50. https://doi.org/10.1016/j.ijfatigue.2012.12.007
• 5)   M.  Subanović,  J.  Pirón,  F.  Zeller,  M.  Jarrar and  A.  Schneider: Proc. ASME 2018 Symp. on Elevated Temperature Application of Materials for Fossil, Nuclear, and Petrochemical Industries, ETAM2018-6748, ASME, New York, (2018), 1.
• 6)   T.  Fischer and  B.  Kuhn: Int. J. Fatigue, 124 (2019), 288. https://doi.org/10.1016/j.ijfatigue.2019.03.005
• 7)   F.  Masuyama: ISIJ Int., 41 (2001), 612. https://doi.org/10.2355/isijinternational.41.612
• 8)   K.  Sawada,  H.  Kushima,  M.  Tabuchi and  K.  Kimura: Mater. Sci. Eng. A, 528 (2011), 5511. https://doi.org/10.1016/j.msea.2011.03.073
• 9)   B.  Gehrmann,  M.  Acet,  H. C.  Herper,  E. F.  Wassermann and  W.  Pepperhoff: Phys. Status Solidi B, 214 (1999), 175. https://doi.org/10.1002/(SICI)1521-3951(199907)214:1<175::AID-PSSB175>3.0.CO;2-J
• 10)   W.  Pepperhoff and  M.  Acet: Constitution and Magnetism of Iron and Its Alloys, Springer-Verlag, Berlin, Heidelberg, (2001), 115.
• 11)   N. F.  Mott and  H.  Jones: The Theory of the Properties of Metals and Alloys, The Oxford University Press, Oxford, UK, (1936), 1.
• 12)   K.  Hirano: Netsu Sokutei, 15 (1988), 10 (in Japanese).
• 13)   Y. S.  Touloukian,  R. K.  Kirby,  R. E.  Taylor and  P. D.  Desai: Thermophysical Properties of Matter, The TPRC Data Series, Vol. 4, IFI/Plenum, New York-Washington, (1975), 3a.
• 14)   H.  Okamoto: Phase Diagrams of Binary Iron Alloys, ASM International, Materials Park, Ohio, (1993), 102.
• 15)   T.  Tsuru,  C.  Suzuki,  Y.  Kaji and  T.  Tsukada: J. Appl. Phys., 107 (2010), 061805. https://doi.org/10.1063/1.3340521
• 16)   K.  Bungardt,  H.  Borchers and  D.  Kölsch: Arch. Eisen., 34 (1963), 465.
• 17)   S.  Kobayashi and  T.  Takasugi: Scr. Mater., 63 (2010), 1104. https://doi.org/10.1016/j.scriptamat.2010.08.015
• 18)   W.  Liu,  X.-G.  Lu,  P.  Boulet,  M.-C.  Record,  H.  Wang and  Q.-M.  Hu: Intermetallics, 98 (2018), 95. https://doi.org/10.1016/j.intermet.2018.04.008
• 19)   W.  Liu,  X.-G.  Lu,  H.  Wang,  Y.  Liu and  Q.-M.  Hu: J. Alloy. Compd., 791 (2019), 994. https://doi.org/10.1016/j.jallcom.2019.03.366
• 20)   T.  Schneider,  M.  Acet,  B.  Rellinghaus,  E. F.  Wassermann and  W.  Pepperhoff: Phys. Rev. B, 51 (1995), 8917. https://doi:10.1103/physrevb.51.8917
• 21)   M. E.  Fine and  W. C.  Ellis: Trans. AIME, 175 (1948), 742.
• 22)   C.  Zener: Trans. AIME, 203 (1955), 619.
• 23)   F. S.  Buffington,  K.  Hirano and  M.  Cohen: Acta Metall., 9 (1961), 434.