Applications of Nd–Fe–B magnets have rapidly expanded since the mid-1980s. However, because of the low Curie temperature of Nd₂Fe₁₄B compound (312 °C) and relatively low intrinsic coercivity, high-end Nd–Fe–B magnets can be used only around room temperature. Heavy rare earth, such as Dy and Tb, modification enhances the intrinsic coercivity to ∼2.4 MA/m, thus extending the operating temperature to ∼180 °C. Beyond that temperature, Nd–Fe–B magnets will be no longer appropriate for any dynamic applications.

Permanent magnet materials capable of reliably operating at high temperatures up to ∼450 °C are required in the advanced power systems of future aircraft, vehicles, and ships. A major objective of the advanced power systems is to increase device reliability, maintainability, and supportability. This advancement will be accomplished in part through the development of advanced power components such as magnetic bearings, integrated power units, and internal starter/generators for main propulsion engines. New high temperature magnets are enabling technologies for the development of these new power components. Power system designers frequently find that magnetic materials impose technological limitations on their designs. Compromises are generally required between the desired performance and the magnetic, mechanical, and electrical properties of available materials. If new materials can operate at , then new advanced designs will be possible. Air cooling, rather than complicated liquid cooling and its necessary logistics support, will become an operational capability. Likewise, oil-less/lube-less gas turbine engines and power systems will be possible.^[1]

Possessing high Curie temperatures (727 °C for SmCo₅ and 920 °C for Sm₂Co₁₇ compound), Sm–Co based magnets are still very important because of their high-temperature application capability, excellent thermal stability, and better corrosion resistance. The extensive research performed around the year 2000 resulted in a new class of Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets capable of operating at high temperatures up to 550 °C.

Another important trend in rare earth permanent magnets research is nanocrystalline and hard/soft nanocomposite magnet materials. For conventional rare earth magnets, high uniaxial anisotropy is only a necessary condition for high coercivity, but not necessarily a sufficient condition for it. Often, compositional modification and specific heat treatment have to be imposed to develop useful coercivity, as in the case of Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets. However, when the grain size is reduced from micrometer to nanometer range, a direct connection between magnetocrystalline anisotropy and intrinsic coercivity is established. This makes it possible to develop nanograin Sm₂(Co, Fe)₁₇ and nanocomposite Sm₂(Co, Fe)₁₇/Fe–Co magnets with significantly enhanced magnetization and Curie temperature, as a result of eliminating excessive Sm and completely getting rid of non-ferromagnetic Cu and Zr. The current status of research in this field will be briefly introduced, and the technical difficulties in making nanocomposite Sm–Co magnetic materials will be discussed.

In this paper, a detailed historical review of the development of Sm–Co permanent magnet materials is given and this development is compared with that of Nd₂Fe₁₄B-based magnets. Based on this comparison, prospects of future rare earth permanent magnet materials are given.

Prior to the development of Sm–Co permanent magnet materials the two important types of permanent magnets were so-called Alnico magnets and hard-magnetic ferrites. Alnico was invented in the early 1930s in Japan. It contained 24 wt.% Co, 14% Ni, 8% Al, 3% Cu, and 51% Fe (Alnico 5) and had a coercivity over 35 kA/m, or about double that of the best magnet steels. Since then, the properties of permanent magnets had been rapidly improved. The maximum energy product of Alnico 5 reached 40 kJ/m³ in the 1940s and then 103 kJ/m³ in the modified Alnico in the 1960s. The high coercivity in Alnico magnets was explained by shape anisotropy. The elongated strong ferromagnetic Fe–Co rich phase is embedded in a weak magnetic Al–Ni rich matrix, the shape anisotropy restricts the magnetization direction along the long axis of the Fe–Co phase, making demagnetization difficult.

During the years from 1933 to 1945, ferrites were developed into commercially useful materials. Hard-magnetic ferrites have the formula (where M= Ba or Sr). They have a hexagonal crystal structure and fairly large magnetocrystalline anisotropy, resulting in high coercivity. Hard ferrites exhibit greater coercive force (160–240 kA/m) but much lower remanence (0.25–0.35 T) and maximum energy product (12–28 kJ/m³) than Alnico.

A major turning point for the development of permanent magnets occurred in the 1960s. In the period 1946–1952, the study of rare earth metals was greatly accelerated because of advances in chemical separation techniques that were developed in association with the Manhattan Project, 1942–1945. Methods for producing pure rare earth metals in quantity were developed which, in turn, stimulated interest in the use of rare earth metals as alloying additions.

In 1959, Nesbit^[2] presented magnetic results for a series of Gd–Co alloys, which included the intermetallics GdCo₂, GdCo₃, and GdCo₅, and showed Gd and Co sublattices coupled antiparallel in each of these phases. This was followed in 1960 by the work of Hubbard^[3] who observed in GdCo₅ a large coercive force of 637 kA/m, which he ascribed to the large magnetocrystalline anisotropy. The significance of this work was neglected, perhaps because Gd was expensive and the magnetization of GdCo₅ was rather low, and apparently it was not recognized that GdCo₅ was only one of a family of intermetallic compounds with potential for permanent magnet applications.

In 1966, Hoffer and Strant^[4] reported that YCo₅ had an extremely large crystal anisotropy with a single easy axis of magnetization. The significance of this discovery was immediately recognized, and they suggested that YCo₅, and most other RCo₅ (R stands for rare earths) phases were potential candidate materials for new permanent magnets. Extensive studies followed to determine the permanent magnetic properties of the family of RCo₅ (1:5) compounds containing Y, Sm, Ce, La, Nd, Pr, and mischmetal (MM).^[5–17] Out of these studies evolved a new generation of permanent magnet materials with outstanding properties, which feature a useful combination of high remanence and high coercivity.

Preliminary magnets made from YCo₅ by Strnatʼs group in 1966 had energy products of only about 8 kJ/m^[34] and in 1967 the same group reported an energy product of 40.6 kJ/m³ for SmCo₅.^[5] Further improvements in energy product to 64.5 kJ/m³ and then to 147.2 kJ/m³ were reported in 1968 by Velge and Buschow^[10,12] at Philips. The development of liquid-phase sintering techniques by Das^[14] in 1969 and by Benz and Martin^[16] in 1970 made fully dense and stable SmCo₅ magnets possible. The energy products reached by the latter methods ranged from 127 kJ/m³ to 159 kJ/m³ and culminated in the basic manufacturing technology for the first generation of commercial rare earth permanent magnets (REPM). Today, the best (BH)_max of SmCo₅ is around 200 kJ/m³. Partial substitution of Pr for Sm resulted in slightly enhanced magnetization and energy product.

The promise which the rare earth-cobalt intermetallic phases R ₂Co₁₇ (2:17) held as potential permanent magnets was recognized in the early 1970s. Basic properties such as saturation magnetization, Curie temperature, and crystallographic parameters of the binary compounds were studied in the mid-1960s at the US Air Force Materials Laboratory by Strnatʼs group.^[18] From 1970 to 1973, Ray, Strant, Mildrum, and their co-workers^[19–25] at the University of Dayton expanded the study to the quasi-binary R ₂(Co_1−x Fe _x )₁₇ phases and systematically investigated the metallurgical and magnetic properties, including the magnetocrystalline anisotropy of these phases. The fact that 2:17 compounds, with or without Fe substitution, have substantially greater saturation magnetization values (1.2–1.6 T) led them to predict that a second generation of rare earth magnets with higher energy product was possible.^[19–24]

However, the realization of practical 2:17 magnets proved more difficult. It was found that all the magnet fabrication methods that had worked well for SmCo₅, when applied to Sm₂Co₁₇, yielded only very low coercive force, usually less than 0.2 MA/m. Over a period of several years, many different experimental approaches were tried but no exciting results were obtained.

In 1974, Senno and Tawara^[26] extended the range of Sm(Co, Fe) _z to z = 7.2 by adding Cu, so that the alloys could be magnetically precipitation hardened. They obtained two-phase sintered magnets in which the main phase has the 2:17 structure. Careful heat treatment made the coercivity in the 0.32–0.8 MA/m range. Later, Tawara and Senno^[27] were able to obtain high coercivity in sintered Sm(Co_0.85Fe_0.05Cu_0.10)₈. In 1976, Nagel^[28] achieved energy products in excess of those for SmCo₅ in sintered magnets Sm(Co, Fe, Mn, Cr)_8.5. However, the Mn- and Cr-containing magnets had the severe disadvantage that its coercive force drops very quickly with increasing temperature above 20 °C.

An important breakthrough was made in 1977 when a group at TDK in Japan announced that the Fe content could be increased and the Co content lowered if these compositional changes were accompanied by a small addition of Zr or similar transition metals.^[29,30] They obtained B _r=1.12 T, , and (BH)_max = 240.3 kJ/m³ for an alloy corresponding to the nominal composition Sm(Co_0.674Fe_0.213Cu_0.100Zr_0.013)_7.43. Further refinement of the isothermal aging and step aging heat treatment^[31] and compositional adjustment^[32] yielded a permanent magnet with B _r=1.2 T, , and (BH)_max = 262.6 kJ/m³ for the nominal composition Sm(Co_0.65Fe_0.28Cu_0.05Zr_0.02)_7.67. It is important to note that these magnets have a much better temperature coefficient of than the Mn- and Cr-containing magnets, making them much more useful for elevated-temperature applications. The 2:17 magnets, or the second generation rare earth permanent magnets, became commercially available in the early 1980s. The best magnetic properties of (BH)_max = 271 kJ/m³, B _r=1.22 T, were obtained in Sm(Co_0.613Fe_0.316Cu_0.052Zr_0.019)_7.88, a magnet containing low Sm, low Cu, and low Zr, but fairly high Fe.^[33]

Figure 1 shows the binary Sm–Co phase diagram.^[34–36] There exist quite a few intermetallic compounds in the Sm–Co system, including Sm₂Co₁₇, SmCo₅, Sm₂Co₇, SmCo₃, SmCo₂, Sm₉Co₄, and Sm₃Co. Among them, SmCo₅ and Sm₂Co₁₇ possess important technical significance. Sm₂Co₁₇ is the most Co-rich compound, and it has a polymorphic phase transformation (α Co₁₇Sm₂ and β Co₁₇Sm₂ in Fig. 1) and has different crystal structures at high and low temperatures. In addition, Sm₂Co₁₇ and SmCo₅ demonstrate finite homogeneity ranges, while others show as line compounds. The R–Co phase diagrams for other rare earths are generally similar, but with some minor systematic variations.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. Binary Co–Sm equilibrium phase diagram.[34–36]

As shown in Fig. 2,^[37] SmCo₅ has a hexagonal crystal structure (1:5 H, space group: P6/mmm; prototype: CaCu₅), while Sm₂Co₁₇ has a rhombohedral crystal structure (2:17 R, space group: prototype: Th₂Zn₁₇) at room temperature and a hexagonal crystal structure (2:17 H, space group: prototype: Th₂Ni₁₇) at high temperatures (1300–1340 °C). The common characteristic of all these three crystal structures is that their c-axes are the unique easy magnetization directions. This uniaxial anisotropy is the basis for SmCo₅ and Sm₂Co₁₇ to become high-performance permanent magnets.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. Crystal structures of (a) SmCo5 (1:5 H), (b) Sm2Co17 (2:17 R), and (c) Sm2Co17 (2:17 H).[37]

Three important prerequisites for high-performance permanent magnet materials are high saturation magnetization, high Curie temperature, and high uniaxial magnetocrystalline anisotropy. Saturation magnetization and Curie temperature values of the R–Co binary systems are illustrated in Figs. 3 and 4, respectively. As a general rule for the 4f–3d exchange interaction, the light rare earths couple parallel with Co, yielding high saturation magnetization, while the heavy rare earths couple antiparallel with Co, resulting in low saturation magnetization.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. Room-temperature saturation magnetization values of RCo5 and R 2Co17 compounds.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. Curie temperatures of RCo5 and R 2Co17 compounds.

Figure 5 shows anisotropy field H _a for binary RCo₅ compounds with R =Y, La, Ce, Pr, Nd, Sm, and MM. It is obvious that SmCo₅ compound has the highest anisotropy field. Figure 6 shows anisotropy field H _a for binary R ₂Co₁₇ compounds with R =Y, Ce, Pr, Nd, and Sm. It is obvious that among all 2:17 compounds, only Sm₂Co₁₇ has uniaxial magnetocrystalline anisotropy and possesses a moderately large crystalline anisotropy field.

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. Anisotropy field H a for binary RCo5 compounds.

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. Anisotropy field H a for binary R 2Co17 compounds.

It is well known that for a rare earth–transition metal (RE–TM) compound, Curie temperature is primarily determined by the TM sublattice, while crystalline anisotropy is primarily contributed by the RE sublattice unless at temperatures close to the Curie point. Research on RE–TM compounds indicates that among all 3d transition metals, Co provides the highest Curie temperature, while among all light rare earths, Sm usually provides the highest crystalline anisotropy. One exception is the cubic Laves 1:2 compounds for which the Fe compounds have higher Curie temperature than the Co compounds.

Figure 7 shows Curie temperature T _C versus Co content for Sm–Co binary compounds. In this figure, Curie temperature data for LaCo₁₃ and Sm₂Co₁₄B are also included. It can be seen from Fig. 7 that there exists a linear relationship between Curie temperatures and the Co content, which clearly demonstrates the importance of the Co content to Curie temperature. Similarly, there exists a linear relationship between saturation magnetization and the Co content, as shown in Fig. 8.

Fig. 7.

	Figure Option ViewDownloadNew Window
	Fig. 7. Curie temperature T C versus Co content for Sm–Co binary compounds. Data for LaCo13 and Sm2Co14B are also included.

Fig. 8.

	Figure Option ViewDownloadNew Window
	Fig. 8. Saturation magnetization versus Co content for Sm–Co binary compounds. Date for LaCo13 is also included.

Table 1 lists the intrinsic properties of some R–Co compounds. It is shown that among all light rare earth-Co compounds, SmCo₅ possesses the highest magnetocrystalline anisotropy, while Sm₂Co₁₇ combines high saturation magnetization, high Curie temperature, and moderately high anisotropy. As for heavy rare earth–Co compounds, HRCo₅ and HR₂Co₁₇ (with HR =Gd, Tb, and Dy) have low saturation magnetization, while Tm₂Co₁₇, Yb₂Co₁₇, and Lu₂Co₁₇ have fairly high saturation magnetization, but with unfavorable magnetocrystalline anisotropy, except for Tm₂Co₁₇, which shows uniaxial anisotropy, but its anisotropy field is not high. A small amount of Fe substitution for Co in Tm₂Co₁₇ slightly increases its anisotropy field. For Yb₂Co₁₇ and Lu₂Co₁₇, a small amount of Fe substitution for Co changes the anisotropy from easy-basal-plane to uniaxial, though the anisotropy fields are not high in both cases.

Table 1.

Table 1.

Table 1. Intrinsic properties of some R–Co compounds at room temperature. . Compound μ 0 M s/T T C/K K 1/ H a/ Anisotropy type Ref. YCo5 1.06 987 6.5 10.3 uniaxial [37,38] LaCo5 0.91 840 6.3 13.9 uniaxial [37] CeCo5 0.77 653 6.4 16.7 uniaxial [37] PrCo5 1.20 893 8.1 13.5 uniaxial [37] NdCo5 1.23 910 0.7 0.4 easy basal plane [37,38] SmCo5 1.10 1020 17.2 ∼35 uniaxial [37–39] GdCo5 0.18 1008 4.6 ∼21 uniaxial [37] TbCo5 0.24 973 0.5 easy basal plane [37] DyCo5 0.30 966 2.0 uniaxial [37] HoCo5 0.53 992 uniaxial [37] ErCo5 0.64 978 4.5 8.0 uniaxial [37] TmCo5 0.67 1012 uniaxial [37] Y2Co17 1.25 1167 −0.34 easy basal plane [37,38] Ce2Co17 1.15 1053 1.2 easy basal plane [37] Pr2Co17 1.38 1153 −0.6 easy basal plane [37] Nd2Co17 1.39 1150 −1.1 easy basal plane [37,38] Sm2Co17 1.22 1190 3.3 5.4 uniaxial [37–39] Sm2(Co0.7Fe0.3)17 1.45 1113 3.0 4.1 uniaxial [37] Gd2Co17 0.75 1193 −0.5 easy basal plane [37] Tb2Co17 0.66 1188 −3.2 easy basal plane [37,90] Dy2Co17 0.68 1152 −2.6 easy basal plane [37,38] Ho2Co17 0.84 1155 −0.9 easy basal plane [37,90] Er2Co17 0.91 1186 0.72 2.5 uniaxial [37,38,92] Tm2Co17 1.21 1180 0.56 2.6 uniaxial [37,91] Tm2(Co0.82Fe0.18)17 1.42 1153 3.0 uniaxial [41,42,91] Yb2Co17 1.36 easy basal plane [37] Lu2Co17 1.40 1203 0.9 easy basal plane [37,91] Lu2(Co0.85Fe0.15)17 2.1 uniaxial [91]

Table 1. Intrinsic properties of some R–Co compounds at room temperature. .

Sm–Co permanent magnets are based on SmCo₅ and Sm₂Co₁₇ compounds. The composition of a SmCo₅ magnet is slightly Sm rich, as compared with its chemical stoichiometry. Its microstructure is basically a featureless single phase, while a small amount of oxide particles and sometimes a minor Sm₂Co₇ phase can also exist. It is a generally accepted idea that the coercivity of SmCo₅ magnet is determined by nucleation field, though it is believed that grain boundary pinning may also play an important role.

On the other hand, the composition and microstructure of Sm₂Co₁₇-based magnets are more complicated. As mentioned in section 2, in order to develop useful coercivity, extra Sm is required and considerable amounts of Cu and Zr have to be added into the alloy. In addition, a complex and long-term heat treatment procedure must be employed (the detailed heat treatment procedure is given in Section 8). Figure 9 shows a TEM microstructure of a typical Sm₂(Co, Fe, Cu, Zr)₁₇-type magnet.^[40] Sintered 2:17 magnets have grains of micrometers in size, with a fine, nanometer cellular structure within each grain. As shown in Fig. 9, the cellular structure consists of three coherent phases: a 2:17 cell interior phase; a 1:5 cell boundary phase rich in Sm and Cu; and a platelet phase rich in Zr. It is believed that the domain wall pinning in the 1:5 cell boundary phase is the origin of high coercivity.

Fig. 9.

	Figure Option ViewDownloadNew Window
	Fig. 9. TEM micrographs of a sintered Sm2(Co, Fe, Cu, Zr)17-type magnet. (a) Section perpendicular to the alignment direction; (b) section parallel to the alignment direction. Arrow ( ) in panel (b) is the alignment direction (c-axis).[40]

Demagnetization curves of commercial SmCo₅ and Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets at various temperatures are shown in Figs. 10 and 11 (By courtesy of Electron Energy Corporation). Both types of magnets can be used up to 300 °C.

Fig. 10.

	Figure Option ViewDownloadNew Window
	Fig. 10. Demagnetization curves of commercial SmCo5 magnets.

Fig. 11.

	Figure Option ViewDownloadNew Window
	Fig. 11. Demagnetization curves of commercial Sm2TM17 magnets.

Permanent magnets with very low temperature coefficients of magnetization over a wide temperature range are required for many applications. Examples are microwave tubes, gyros, accelerometers, and conventional moving-coil meters. The flux provided by most permanent magnets decreases on heating. This is an intrinsic property for all ferromagnetic materials, in which the magnetization will eventually drop to zero at their Curie temperatures. On the other hand, for most heavy rare earth–Co compounds, for example, GdCo₅ and Gd₂Co₁₇, the magnetization increases when temperature enhances before reaching a peak value. Based on this characteristic, partial substitution of heavy rare earth, such as Gd, for Sm can be made to form temperature-compensated (Sm,Gd)–Co magnets which may show a near zero temperature coefficient of magnetization from −50 °C to about 150 °C with a peak magnetization at around room temperature. Demagnetization curves of commercial temperature-compensated (Sm,Gd)Co₅ and (Sm,Gd)₂(Co, Fe, Cu, Zr)₁₇-type magnets at various temperatures are shown in Figs. 12 and 13 (By courtesy of Electron Energy Corporation). Both types of temperature-compensated magnets can be used up to 300 °C.

Fig. 12.

	Figure Option ViewDownloadNew Window
	Fig. 12. Demagnetization curves of (Sm,Gd)Co5 magnets.

Fig. 13.

	Figure Option ViewDownloadNew Window
	Fig. 13. Demagnetization curves of (Sm,Gd)2TM17-type magnets.

It is obvious from Fig. 3 that Tm₂Co₁₇, Yb₂Co₁₇, and Lu₂Co₁₇ have much higher magnetization than Gd₂Co₁₇. Figure 14 compares temperature dependence of magnetization for Sm₂TM₁₇ and Tm₂(Co, Fe)₁₇. If Tm, Yb, or Lu can be incorporated into temperature-compensated (Sm,Gd)–Co magnets, then improved magnetic performance could be expected. However, because R ₂Co₁₇ (R =Tm, Yb, Lu) have poor anisotropy, developing enough high coercivity would be a technical challenge.

Fig. 14.

	Figure Option ViewDownloadNew Window
	Fig. 14. Saturation magnetization versus temperature for Tm2(Co0.82Fe0.18)17, Tm2(Co0.94Fe0.06)17,[41,42] and Sm(Co, Fe, Cu, Zr)∼7.

Powder metallurgy is used to manufacture commercial Sm–Co magnets. Its processing procedures include vacuum melting, ingot crushing, ball or jet milling, powder magnetic alignment, compaction, sintering, and heat treatment. Alternatively, Sm–Co alloy powders can be produced by a reduction-diffusion process using Sm₂O₃, Co powder, and Ca or CaH₂ as a reduction agent.

Sintered Sm–Co magnets are very hard and brittle, therefore machining them into the final shape and size is often troublesome, especially for tiny magnetic parts. This led to the development of bonded Sm–Co magnets,^[37] which are made by consolidating a magnet powder with a polymer matrix. Thermosetting binders, such as epoxy resin, are employed in compression-molded magnets, while thermoplastic binders, like nylon, are used for injection-molded magnets, and elastomers, such as rubber, for extruded magnets.^[43] Table 2 lists the magnetic properties of some Sm–Co magnets.

Table 2.

Table 2.

Table 2. Magnetic properties of some commercial Sm–Co magnets (TM stands for Co, Fe, Cu, and Zr). . Magnets μ 0 M r/T SmCo5, (Sm,Pr)Co5 0.8–0.96 635–740 135–190 Sm2TM17 1.0–1.2 710–840 180–255 Temp. compensated (Sm,Gd)Co5 0.5–0.75 400–600 64–120 Temp. compensated (Sm,Gd)2TM17 0.8–0.95 480–720 80–175 Bonded SmCo5 0.4–0.5 600–1600 240–520 32–65 Bonded Sm2TM17 0.6–0.8 400–1600 310–520 64–130

Table 2. Magnetic properties of some commercial Sm–Co magnets (TM stands for Co, Fe, Cu, and Zr). .

Possessing the highest Curie temperature and moderately high magnetization and energy product among high-performance rare earth permanent magnets, Sm₂TM₁₇ magnets are the best conventional high-temperature permanent magnets.^[44,45] A conventional Sm₂TM₁₇ magnet can operate at up to 300 °C. The problem associated with higher temperature ( ) operation is that the intrinsic coercivity ( ) of these magnets drops sharply with increasing temperature. Upon heating, of the 2:17 magnets drops sharply from their room temperature values of 1.5–2.5 MA/m (or higher) to only 0.2–0.5 MA/m at 400 °C and 0.1–0.2 MA/m at 500 °C. Low intrinsic coercivity at high temperatures results in a nonlinear 2^nd quadrant induction demagnetization curve (B curve) above ∼300 °C. A linear 2^nd quadrant B curve is critical for all dynamic applications, such as for generators, motors, and actuators.

In a dynamic application, the operating point of a magnet keeps cycling on the B curve. If the intrinsic coercivity is low, then the B curve can be nonlinear and a knee appears. Under this circumstance, the operating point of the magnet can be reduced to below the knee in the B curve and the induction can be significantly reduced irreversibly. If the intrinsic coercivity of a magnet is sufficiently high, then the B curve will be linear and the induction will be reversible around the operating point even at a quite low permeance value as shown in Fig. 15. The maximum operating temperature of a magnet (T _max) can be defined as the temperature limit at which the B curve of the magnet still maintains linearity. Therefore, to increase the operating temperature of permanent magnet materials, the key is to increase their intrinsic coercivity at high temperatures, so that their induction demagnetization curves remain linear at the operating temperature.

Fig. 15.

	Figure Option ViewDownloadNew Window
	Fig. 15. Intrinsic coercivity and linearity of an induction demagnetization curve.

Around the year 2000, extensive research was carried out to substantially improve the high-temperature performance of Sm₂TM₁₇-type permanent magnets. As a result of that effort, the maximum operating temperature of permanent magnets was increased from around 300 °C to as high as 550 °C. This advance was made on systematic studies of the effects of compositions on high-temperature intrinsic coercivity of Sm₂TM₁₇-type of permanent magnets.

6.2. New high-temperature Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets

Based on systematic studies of the effects of compositions on high-temperature properties of Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets, a new series of sintered permanent magnets with significantly improved high-temperature performance were accomplished by significantly reducing the Fe content, increasing the Sm content, and adjusting the Cu and Zr contents in magnet alloys. The maximum operating temperature of these magnets was increased from previous 300 °C for conventional high-temperature magnets to as high as 550 °C. The of these new magnets reached 1 MA/m at 400 °C (two to three times higher than those of conventional magnets) and 0.72 MA/m at 500 °C (four to nine times higher than those of conventional magnets). The B curves of these new magnets remained linear up to 550 °C (250 °C to 350 °C higher than those of conventional magnets). The temperature coefficients of for the new magnets can range from a small negative value (−0.03%/°C) to near zero, or they may even be positive (up to +0.3%/°C). As a comparison, the temperature coefficients of for conventional SmCo₅, Sm₂TM₁₇, and Nd₂Fe₁₄B-based magnets around room temperature are −0.3 %/°C, −0.3 %/°C, and −0.9 %/°C, respectively.

Figure 16 shows demagnetization curves of a Sm(Co_0.79Fe_0.09Cu_0.09Zr_0.037.69 at 400 °C, 450 °C, and 500 °C, respectively.^[47] This magnet demonstrates much higher and better squareness of demagnetization curves at high temperatures than the conventional 2:17 magnets. Figure 17 shows demagnetization curves of a commercial new high-temperature magnet with its maximum operating temperature T _max = 500 °C.^[48] It can be seen from the figure that the room temperature (BH)_max of the magnet with T _max=500 °C is around 166 kJ/m³. As a result of excessive amount of Sm, Cu, and Zr, the magnetization values of these new magnets are relatively low.

Fig. 16.

	Figure Option ViewDownloadNew Window
	Fig. 16. Demagnetization curves of Sm(Co0.79Fe0.09Cu0.09Zr0.03)7.69 at 400 °C, 450 °C, and 500 °C.[47]

Fig. 17.

	Figure Option ViewDownloadNew Window
	Fig. 17. Demagnetization curves of a new high-temperature magnet with T max = 500 °C.[48]

Figure 18 is a TEM micrograph of a new high-temperature magnet with T _max = 500 °C. Comparing with that of the conventional 2:17 magnet (Fig. 9), the cellular structure has smaller cells and thicker cell boundaries.

Fig. 18.

	Figure Option ViewDownloadNew Window
	Fig. 18. TEM micrograph of a new high-temperature magnet with T max = 500 °C. 1: cell; 2: cell boundary; 3: platelet.[40]

6.3. Long-term thermal stability of new high-temperature permanent magnets

Figure 19 gives flux density loss versus time for a new magnet with T _max = 500 °C and conventional 2:17 magnets with (BH)_max=223 kJ/m³ and 234 kJ/m³ (28 MGOe and 30 MGOe) during long-term aging at 500 °C in air. It can be seen from the figure that after aging for 2000 hours, the flux density loss was around 18%, 49%, and 69%, respectively.^[49] The improvement of the new magnet is obvious.

Fig. 19.

	Figure Option ViewDownloadNew Window
	Fig. 19. Long-term thermal stability of a new-high temperature magnet with T max = 500 °C.[49]

The flux density loss is caused by three different mechanisms: (i) non-linear induction demagnetization curve (B curve); (ii) oxidation starting from the surface and gradually penetrating to the magnet interior; (iii) microstructure change, such as grain growth and phase transformation. The flux density loss caused by microstructure change is very limited, especially when the operating temperature is lower than 400 °C. When a magnet is operating at a temperature higher than its T _M, its non-linear B curve will result in a large irreversible flux density loss. This is the case for conventional 2:17 magnets shown in Fig. 19. On the other hand, the loss for the new magnet is caused primarily by the oxidation. Surface coating using Cr, W, or sulfamate Ni can significantly reduce the flux density loss caused by oxidation.^[50] In addition, it was determined that Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets demonstrated much better neutron radiation resistance as compared to Nd–Fe–B type magnets. It was noted that the radiation resistance and thermal stability are somewhat related and the irradiation damage is most likely caused by a radiation-induced thermal effect. The excellent radiation resistance of 2:17 magnets was proved to be an advantage in space applications.^[51]

6.4. Abnormal temperature dependence of intrinsic coercivity

Novel temperature dependence of was observed during the research on high-temperature permanent magnets in some newly-developed magnets. A positive temperature coefficient of intrinsic coercivity in SmTM _z with z = 7 was reported in 1998.^[52] In 1999, a complex temperature coefficient in Sm(Co_0.843Fe_0.04Cu_0.09Zr_0.027)_7.26 that had a low Fe content and a high Cu content was observed.^[47] When heating this magnet, first gradually decreases and reaches a minimum at about 150 °C as shown in Fig. 20. With continued heating, rapidly increases and forms a maximum at 500 °C. The value of this magnet at 500 °C is more than 30% higher than its room temperature value. Another magnet of Sm(Co_0.825Fe_0.1Cu_0.05Zr_0.025)_7.38 that has low Cu content displays a maximum at 550 °C, which is nearly four times higher than its room temperature coercivity value as shown in Fig. 21. The abnormal temperature dependence of coercivity in 2:17 magnets was observed by Russian researchers as early as in 1990.^[53]

Fig. 20.

	Figure Option ViewDownloadNew Window
	Fig. 20. Temperature dependence of magnetic properties of Sm(Co0.843Fe0.04Cu0.09Zr0.027)7.26.[47]

Fig. 21.

	Figure Option ViewDownloadNew Window
	Fig. 21. Abnormal temperature dependence of coercivity for Sm(Co0.825Fe0.1Cu0.05Zr0.025)7.38.[53]

High uniaxial magnetocrystalline anisotropy is a key prerequisite and a necessary condition for high coercivity in rare earth magnets; however, it is not the sufficient condition for high coercivity in conventional rare earth magnets with micrometer grain structure. A convincing example to illustrate this concept is the Sm₂Co₁₇ compound. Though Sm₂Co₁₇ has moderately high uniaxial magnetocrystalline anisotropy of 3.3 MJ/m³, as mentioned previously, the coercivity of stoichiometric Sm₂Co₁₇ is very low (usually less than 200 kA/m) if its grain size is in the micrometer range. In order to develop useful coercivity, extra Sm and considerable amounts (∼10 at.%) of Cu and Zr must be added and a complex and time-consuming heat treatment procedure must be applied.^[26–32] The procedure consists of high-temperature sintering at for 1–3 h, a solid solution heat treatment at ∼1180 °C for 2–4 h, and a long-term isothermal aging at ∼800 °C for 20–50 h. Even after this long-term isothermal aging, the coercivity is still very low. As demonstrated in Fig. 31, the high coercivity is developed after a very slow cooling from 800 °C to 400 °C at 1–2 °C/min followed by another isothermal aging at 400 °C for 10–20 h. The whole procedure takes about three days (up to 80 hours) to complete, as shown in Fig. 32. These compositional modification and long-term heat-treatment are required to form the specific fine-scale cellular microstructure in which the cell boundary phase serves as pinning sites for domain wall motion.

Fig. 31.

	Figure Option ViewDownloadNew Window
	Fig. 31. Process and coercivity development comparison of conventional micrograin 2:17 (upper portion) and nanograin 2:17 magnet material[57] (lower portion).

Fig. 32.

	Figure Option ViewDownloadNew Window
	Fig. 32. TEM micrograph of a nanograin Sm2Co17 magnet sample annealed at 750 °C for 1 min with .[57]

However, if the grain size of Sm₂Co₁₇ is reduced from micrometer range to nanometer range, high intrinsic coercivity can be easily developed in the stoichiometric Sm₂Co₁₇ (without adding extra Sm and Cu, Zr and without long-term isothermal aging and slow cooling ). In 1991, Wecker^[55] obtained 0.5 MA/m after annealing a mechanically alloyed stoichiometric Sm₂Co₁₇ alloy powder at 700 °C for 30 min. A few years later, Chen^[56] obtained 0.3 MA/m after annealing a mechanically alloyed SmCo₁₀ alloy powder at 750 °C for 20 min. In 2003, a high coercivity of 1.24 MA/m was achieved after annealing a high-energy ball milled stoichiometric Sm₂Co₁₇ specimen at 750 °C for only 1 min as shown in Fig. 32.^[57] The TEM observation revealed nanograins of approximately 30 nm in average and no cellular structure was found as shown in Fig. 32. It should be noted that in order to achieve the similar level of coercivity, its micrograin counterpart must go through a sintering, a solid solution heat treatment, and a long-term isothermal aging followed by a very slow cooling, totaling 80 hours, in addition to the Cu and Zr and extra Sm addition. Therefore, magnetization reversal in nanograin Sm₂Co₁₇ must be carried out by a mechanism other than domain wall pinning.

Apparently, a fundamental change in coercivity mechanism takes place when the grain size of a rare earth magnet is reduced from the micrometer range to nanometer range. Based on novel phenomena observed in magnetic materials having nanograin structure, a new model of coercivity mechanism in magnetic materials with nanograins was proposed.^[58,59] The principal points of this model are given as follows. (i) In magnetic materials with nanograins, the formation of multiple magnetic domains in a grain is no longer energetically favorable. (ii) The magnetization reversal in nanocrystalline and nanocomposite magnetic materials is not carried out by nucleation of reversed magnetic domains or domain wall motion, but by rotation of magnetization. (iii) Therefore, in magnetic materials with nanograin structure, there is no longer a need to create a specific microstructure to prevent the formation of reversed domains or to restrict domain wall motion. (iv) High uniaxial magnetocrystalline anisotropy is not only a necessary condition for high coercivity, as it is in magnetic materials with micrograins, it is also the sufficient condition for high coercivity in magnetic materials with nanograins. (v) Thus, a direct connection between coercivity and magnetocrystalline anisotropy is established in magnetic materials with nanograin structure. (vi) Consequently, high coercivity should be readily obtained for any magnetic materials that possess high uniaxial anisotropy, provided that the materials have nanograin structure. This concept can be schematically illustrated in Fig. 33.

Fig. 33.

	Figure Option ViewDownloadNew Window
	Fig. 33. Coercivity mechanisms in rare earth permanent magnets with micrograins (a), showing indirect connection between anisotropy and coercivity; and with nanograins (b), showing direct connection between anisotropy and coercivity.

To verify this concept experimentally, an YCo₅ alloy was chosen for a further test. YCo₅ was the first rare earth-transition metal compound that was discovered to have very high uniaxial magnetocrystalline anisotropy;^[4] however, useful coercivity could not be obtained in a conventional material with micron grains. A high-energy ball milled YCo₅ powder was annealed at 750 °C for 2 min, and high coercivity near 1 MA/m was readily obtained in the first experiment.^[93] Then, a moderately high coercivity of 0.6 MA/m was obtained after annealing a high-energy ball milled YCo₅/α-Fe (5 wt.%) at 750 °C for 2 min.^[93] Figure 34 shows a TEM image and selected area electron diffraction pattern of the nanocomposite YCo₅/α-Fe specimen. The electron diffraction pattern demonstrates a mixture of a 1:5 structure and an α-Fe structure, while the TEM image is characterized with small α-Fe grains and twinned YCo₅ grains. In addition, the new coercivity concept has also been supported by experimental results obtained in Sm₂Co₁₇/Co, (Sm,Gd)₂Co₁₇/Co, and Nd₂Fe₁₄B/α-Fe systems.

Fig. 34.

	Figure Option ViewDownloadNew Window
	Fig. 34. TEM image and selected area electron diffraction pattern of a mechanically alloyed YCo5/α-Fe specimen after annealing at 750 °C for 2 min.[93]

In a nanograin rare earth magnet material, the rare earth content can be reduced to lower than its chemical stoichiometric composition, resulting in a hard/soft nanocomposite magnet material. In nanocomposites, because of the hard/soft interface exchange coupling, the direction of magnetization in the soft phase is restricted by that in the hard phase and tends to be aligned in the same direction as that in the hard phase. The exchange interaction of magnetic moments at the hard/soft interface is, in a way, like a spring, leading to the term exchange spring.

Getting rid of excessive Sm content and eliminating non-ferro-magnetic elements Cu and Zr from the conventional Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets would significantly enhance magnetization and Curie temperature. Stoichiometric Sm₂Co₁₇ possesses high saturation magnetization of 1.22 T and high Curie temperature of 917 °C. Partial substitute Fe for Co further increases the saturation magnetization of Sm₂(Co_0.7Fe_0.3)₁₇ to 1.45 T. If nanocomposite Sm₂(Co_0.7Fe_0.3)₁₇/Fe–Co could be made, its magnetization would reach the same level as Nd₂Fe₁₄B (1.6 T) and it might be a new type of high-temperature and high-performance magnet, if sufficiently high coercivity could be developed.

However, there are multiple difficulties to accomplish this task. First, nanograin structures are created using rapid solidification, for example, melt spinning, and high-energy ball milling followed by crystallization. The products of these processes are ribbons or powders. Making these ribbons and powders fully dense materials without altering their nanostructure is a challenge. Second, near perfect grain alignment is necessary for any high-performance magnet materials. Aligning tiny nanograins, thus, forming anisotropic magnets is another challenge.

Melt spinning or high-energy ball milling followed by rapid hot compaction and hot deformation were successfully employed to make nanocomposite Nd–Fe–B/Fe or Nd–Fe–B/Fe–Co magnets by realizing fully dense bulk magnets and near perfect grain alignment. The hot compaction is not only a process for consolidation of powders or ribbons, but also a process for crystallization of amorphous materials. While in the followed hot deformation, the hot compacted bulk body is further made to near full density and the easy magnetization directions of all nanograins are aligned along the pressing direction.^[58,59]

However, this approach has proved not very successful for making Sm₂(Co, Fe)₁₇/Fe–Co. Experiments demonstrated that only partial grain alignment could be established in Sm₂(Co, Fe)₁₇ and Sm₂(Co, Fe)₁₇/Fe–Co after hot compaction and hot deformation.^[60,61] This may relate to the fact that there is no grain boundary Sm-rich phase in the Sm₂Co₁₇-based alloy systems, since it is well known that the grain boundary low-melting-point Nd-rich phase plays a critical role in grain alignment for Nd–Fe–B alloys during hot deformation.

Better results were obtained for hot deformed SmCo₅. Bulk, anisotropic, nanograin SmCo₅ magnets with coercivity of 795–3980 kA/m and (BH)_max of 88–135 kJ/m³ were synthesized by hot compacting the high-energy ball milled SmCo₅ powder at 700 °C, followed by hot deformation at 800–900 °C with a height reduction of 70%–90%.^[62] Figure 35 shows a TEM micrograph of a hot-deformed bulk, anisotropic, nanocrystalline SmCo₅ specimen.^[62]

Fig. 35.

	Figure Option ViewDownloadNew Window
	Fig. 35. TEM micrograph of hot deformed SmCo5 with 90% height reduction.[62]

Alternatively, surfactant-assisted high-energy ball milling was used to produce anisotropic SmCo₅ nanoflakes, and the subsequent magnetic alignment and compaction yielded bulk, anisotropic, nanocrystalline SmCo₅ magnets.^[63–65] Figure 36 shows a SEM micrograph of anisotropic SmCo₅ nanoflakes.^[64] In addition, many other processes, such as powder blending, powder particle coating, and magnetic field-assisted ball milling, were tested in recent years.

Fig. 36.

	Figure Option ViewDownloadNew Window
	Fig. 36. SEM micrograph of anisotropic SmCo5 flakes prepared by surfactant-assisted high-energy ball milling.[64]

It has been nearly 30 years since Buschowʼs group first reported magnetic properties in nanocomposite rare earth magnet materials.^[66,67] However, these new type of materials, including both Sm–Co and Nd–Fe–B systems, remain in laboratory research stage, and their magnetic performance is still far poorer than that of their conventional counterparts. Technical difficulties in developing practical nanocomposite magnets include not only how to make bulk, fully dense, anisotropic magnets using adequate processes, but also how to develop sufficiently high intrinsic coercivity to ensure linear induction demagnetization curves.

Maintaining a linear induction demagnetization curve in a hard/soft nanocomposite magnet is a very difficult task. Introducing a soft magnetic phase, such as α-Fe or Fe–Co, will certainly enhance the magnetization, however, it will definitely result in reduced intrinsic coercivity, which will most likely lead to a non-linear B curve, as illustrated for curve 2 in Fig. 37. On the other hand, to remain a linear B curve, a magnet with higher magnetization needs to have higher intrinsic coercivity. This would lead to a contradictory dilemma: making a hard/soft two-phased nanocomposite magnet that possesses the intrinsic coercivity higher than a magnet without any soft phase, as shown for curve 3 in Fig. 37. If we consider applications at an elevated temperature, this task would be even more difficult to accomplish.

Fig. 37.

	Figure Option ViewDownloadNew Window
	Fig. 37. Schematic illustration showing difficulty in developing hard/soft nanocomposite with a linear B curve.

It seems a conclusion can be made that only when sufficient high intrinsic coercivity is successfully developed (in addition to full density and perfect grain alignment), then nanocomposite magnets would be in a position to compete with conventional Sm–Co and Nd–Fe–B magnets.

In addition to bulk magnets, Sm–Co thin films have been used in areas including, but not limiting to, microelectromechanical systems (MEMS) and magnetic recording. MEMS are miniaturized electromechanical devices, such as motors, actuators, sensors, mini-pumps, and micro-systems with coupled electric, mechanical, radiant, thermal, magnetic, and chemical effects. Some MEMS applications require a permanent magnet film up to a few hundred nanometers in thickness, while others use a permanent magnet ‘thick layer’ of a few microns, sometimes even to a few tenths of millimeters.^[68]

Permanent magnets used for MEMS should have (i) proper coercivity, high remanence, high maximum energy product, and high Curie temperature; (ii) adequate to MEMS processing; and (iii) environmental stability, including mechanical stability, chemical stability, and thermal stability. Among all potential candidates, SmCo₅ thin film demonstrates the best magnetic performance; however, its corrosion resistance is poorer than that of some other materials, such as Pt–Co.^[68]

On the other hand, the perpendicular magnetic recording can provide the storage density three times more than the traditional longitudinal recording. For the perpendicular recording, a thin film possessing high uniaxial anisotropy, with its c-axis perpendicular to the substrate surface, is required. This can be accomplished by epitaxially growing SmCo₅ thin film on an appropriate underlayer.^[69]

Epitaxial SmCo₅ thin films with strong perpendicular magnetic anisotropy have been developed using sputtering or pulse laser deposition on various substrates, including Cu, Cu/Ti, W, Cr/Cu, Al₂O₃ (0001), heated Ru buffered Al₂O₃ (0001), Cr buffered single crystal MgO (110), and Ru/Cu/Ru sandwich, etc. and large perpendicular anisotropy and high coercivity have been achieved.^[69–72]

In addition, it was reported that a high maximum energy product of 310 kJ/m³ was achieved in a SmCo₅/Fe nanocomposite multilayer system,^[94] which is higher than the theoretical value of the anisotropic single-phase SmCo₅ magnet (241 kJ/m³).

Three generations of rare earth magnets appeared around the mid-1960s, 1970s, and 1980s, respectively, which made people believe that we might have a new generation of rare earth magnets in about every ten years. However, 33 years have passed since the discovery of Nd–Fe–B magnets, there is still no sign of any new generation on the horizon. This fact made some pessimists even think that Nd–Fe–B might be the last high-performance rare earth permanent magnets.

In order to have reasonable prospects of future rare earth permanent magnet materials, we have to realize the distinguished differences between the developments of the 1^st, 2^nd generations and the 3^rd generation rare earth magnets. It is obvious from section 2 that the developments of SmCo₅- and Sm₂Co₁₇-based magnets were the outcome of systematic studies of binary R–Co compounds. When research on RCo₅ compounds started from the late 1950s, preliminary versions of R–Co phase diagrams were available and the existence of RCo₅ and some other R–Co compounds were already known.^[73] Researcherʼs tasks, then, became to prepare R–Co intermetallic compounds and to test their basic magnetic parameters, including saturation magnetization values, Curie temperatures, and crystalline anisotropy fields, and to determine compounds that have potentials to be developed into practical permanent magnets.

The development of Nd–Fe–B magnets was very different from that of SmCo₅- and Sm₂Co₁₇-based magnets. To understand this important difference and its significance, it is necessary to review the discovery of Nd₂Fe₁₄B compound. In fact, searching candidates for permanent magnet materials was conducted simultaneously in both R–Co and R–Fe systems. Extensive investigation of R–Fe systems (R=Ce, Pr, Nd, Sm, Gd, Tb, Dy, Ho, Er, Tm, Lu, and Y) was carried out in the mid-1960s by Ray, Strnat, and their co-workers^[74–77] at the US Wright-Patterson Air Force Base and the University of Dayton. Unfortunately, R–Fe binary compounds have neither high Curie temperature, nor uniaxial crystalline anisotropy, and therefore, did not appear promising.

In the 1970s, research on amorphous materials, including soft magnetic materials, using rapid solidification became very active and stimulated the hope for finding new metastable phases in R–Fe systems. In 1973, Clark^[78] obtained an energy product of 69 kJ/m³ by heating a TbFe₂ amorphous ribbon to 500 °C. Starting from 1980, Croat^[79–82] studied melt-spun R–Fe-alloys (R=Pr, Nd, Sm, Gd, Tb, and Er) and obtained (BH)_max = 24–32 kJ/m³ in Nd_0.4Fe_0.6 and Pr_0.4Fe_0.6. Apparently, Koon^[83,84] was the first person adding B to melt-spun R–Fe alloys. The purpose of adding B was to restrain the tendency of the melt-spun alloys to crystallize. Then, Hadjipanayis^[85,86] also added Si and/or B into R–Fe systems to make it easier to obtain amorphous state during rapid quenching, as he mentioned that “the metalloid was included in the system to make the ribbons more glassy”.^[85] He also increased the Fe content to enhance the magnetization and obtained (BH)_max = 103 kJ/m³ in Pr₁₆Fe₇₆B₅Si₃ and Pr₁₆Fe₇₆B₈.

The fact that the x-ray diffraction spectrum of heat treated Pr₁₆Fe₇₆B₅Si₃ resembled that of a Fe₂₀ R ₃B tetragnal phase discovered by Stadelmaier^[87] caused Hadjipanayis to attribute the hard magnetic properties of R ₁₆Fe₇₆(B,Si)₈ to this highly anisotropic phase. Since R ₃Fe₂₀B is a stable equilibrium phase, it was realized that the new R–Fe–B magnet might be made by traditional powder metallurgy method in addition to melt-spinning. Finally, in November 1983, Sagawa^[88] in Japan reported that a (BH)_max = 279 kJ/m³ was obtained in Nd₁₅Fe₇₇B₈ using the same conventional powder metallurgy technique as that for producing SmCo₅, which symbolized the birth of the third generation of rare earth permanent magnets. Later studies revealed that the exact composition of the new compound is R ₂Fe₁₄B, not R ₃Fe₂₀B as previously realized.

Apparently, the new R ₂Fe₁₄B compound was created in early 1980s without even its creatorʼs (Koon and Hadjipanayis) recognition. The addition of B and/or Si into R–Fe was to make it easier to obtain an amorphous phase in a hope that new metastable phases could be formed in the heat treatment after rapid solidification. Therefore, the discovery of Nd₂Fe₁₄B is a fortunate incidental or accidental event, or a lucky hit, rather than the outcome of systematic studies like what happened for SmCo₅ and Sm₂Co₁₇.

From the discovery of the Nd₂Fe₁₄B compound, there are at least two lessons we can learn. The first lesson is that we must pay close attention to every incidental or accidental event in research. More often than not, the outcome of a research may be different from the original intention. This is not only true for Nd₂Fe₁₄B, but also true for the discovery of hard/soft nanocomposite rare earth magnetic materials.

When Buschowʼs group extended compositions of melt-spun R–Fe–B alloys to a more Fe-rich and B-rich range, their original intention was, again, to try to find metastable ferromagnetic materials for permanent magnets.^[66] More Fe was added for a higher magnetization and more B was added for easier glass formation. But, the R ₂Fe₂₃B₃ (R=Pr, Nd, Sm, and Gd) metastable ternary Fe-rich compounds they obtained all have cubic crystal structures and are not suitable for permanent magnet materials.

Later in 1989, after annealing amorphous melt spun flakes, they obtained a two-phased Nd₂Fe₁₄B (15%)/Fe₃B (85%) nanocomposite magnet material with remarkable isotropic hard magnetic properties. The remanence μ ₀ M _r is 1.2 T, intrinsic coercive fields μ ₀ H _c is almost 0.4 T, and (BH)_max=95 kJ/m.^[67] From that time on started the extensive research on nanocomposite rare earth permanent magnet materials. Therefore, the discovery of hard/soft two-phased nanocomposites is very similar to that of Nd₂Fe₁₄B compounds.

It is obvious that Nd₂Fe₁₄B is the first rare earth-transition metal compound with technical importance discovered in a ternary system. Since all binary R–Co and R–Fe systems have been thoroughly investigated, the chance of finding new promising equilibrium or metastable compounds in those systems is rare. The discovery of Nd₂F₁₄B has opened up a new avenue for future research, which is to explore more ternary and multiple systems and it expands a vast new field of opportunities. It is reasonable to believe that Nd₂Fe₁₄B-based magnets are not the last high-performance rare earth magnets, but they are the first high-performance rare earth magnets in ternary and multiple systems. However, because of the greatly increased complexity in ternary and multiple systems, it will most likely take quite a long time before the next potential compound to emerge. We should not ignore the possibility that the next new high-performance magnet material would be discovered by another fortunate incidental event, like the case of Nd₂Fe₁₄B.

The second lesson we can learn from the discovery of the Nd₂Fe₁₄B compound is that we must pay particular attention to the effect of atomic spacing on intrinsic magnetic properties. Nd₂Fe₁₇ is the only stable compound in the binary Nd–Fe system. It has a rhombohedral crystal structure with a = 0.857 nm and c = 1.246 nm, unfavorable easy-basal-plane anisotropy, and unfavorable low Curie temperature of 57 °C. By adding B, a new ternary Nd₂Fe₁₄B compound forms. It has a tetragonal crystal structure with a = 0.879 nm, c = 1.218, uniaxial anisotropy with H _a=6 MA/m, and Curie temperature of 312 °C. It is believed that these significant changes are originated from the atomic spacing variation.

In addition, when nitrogen or carbon was introduced into R ₂Fe₁₇ as interstitial atoms, the small increase of lattice constants yielded vast enhancement in magnetization, Curie temperature, and anisotropy field.^[89,95] For example, comparing Sm₂Fe₁₇N₃ with Sm₂Fe₁₇, only about 2% increase in lattice constants was demonstrated, but 93% increase in Curie temperature (from 116 °C to 476 °C), 54% in saturation magnetization (from 1.0 T to 1.54 T) were observed, and the anisotropy changed from easy basal plane to uniaxial with H _a=11 MA/m.

We simply cannot overemphasize the significance of atomic spacing, since according to the Bethe–Slater curve, the basic types of magnetization, such as paramagnetism, ferromagnetism, and antiferromagnetism, are closely related to atomic spacing. Altering atomic spacing through forming new compounds may result in a change of magnetic types. For example, Mn is antiferromagnetic below 100 K and paramagnetic at room temperature. However, Mn shows ferromagnetic in Mn–Al, Mn–Bi, Cu–Mn–Sn, and Cu–Mn–Al systems. If in the relatively distant future we are able to modify materials not only in micrometer and nanometer ranges, but also in an angstrom range by forming ternary or multiple system compounds, then it might be not impossible to change a material from paramagnetic or ferrimagnetic to ferromagnetic.

It is assumed that one of the future high-performance rare earth magnets might still take the form of R–T–M, where R is one or more rare earths that primarily contribute high crystalline anisotropy, and R =Ce, Pr, Nd, Sm, Gd, Dy, Ho, or Er; T is one or more 3d (or 4d) transition metals that primarily contribute high Curie temperature and high magnetization, and T =Co, Fe, Mn, Cr, or Mo, Nb; and M are one or more metals, or semi-metals, or non-metal elements, that are to adjust atomic spacing, and M = Al, Si, Ga, Ge, etc.

In advanced power/propulsion systems for future aircraft, vehicles, and ships, permanent magnet materials capable of reliably operating at high temperatures up to ∼450 °C are required. Those operating temperatures are far beyond the capability of Nd–Fe–B magnets.

Extensive research efforts performed around the year 2000 resulted in a new class of Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets capable of operating at high temperatures up to 550 °C and these new magnets are commercially available in the market. However, as a result of excessive amount of Sm, Cu, and Zr, the magnetization values of these new magnets are relatively low.

When the grain size is reduced from micrometer to nanometer range, a direct connection between coercivity and magnetocrystalline anisotropy is established in magnetic materials. Consequently, high coercivity should be readily obtained for any magnetic materials that possess high uniaxial anisotropy, provided that the materials have nanograin structure. Therefore, it is possible to obtain high coercivity in stoichiometric Sm₂Co₁₇ and Sm₂(Co, Fe)₁₇, and this was confirmed by experiments.

Getting rid of excessive Sm content and eliminating non-ferro-magnetic elements Cu and Zr from the conventional Sm₂(Co, Fe, Cu, Zr)₁₇-type magnets would significantly enhance magnetization and Curie temperature. Obviously it is of great significance if nanocomposite Sm₂(Co, Fe)₁₇/Fe–Co with high magnetic performance could be made.

However, this effort has encountered many technical difficulties, including how to find appropriate processes to fabricate fully dense, anisotropic, nanocomposite magnets with perfect grain alignment. Obtaining sufficiently high intrinsic coercivity for maintaining a linear induction demagnetization curve in hard/soft nanocomposite magnets has proved to be very difficult. Only when sufficient high intrinsic coercivity is successfully developed (in addition to full density and perfect grain alignment), would nanocomposite magnets be in a position to compete with conventional Sm–Co and Nd–Fe–B magnets.

This paper gives a detailed historical review of the development of Sm–Co permanent magnets and this development is compared with that of Nd₂Fe₁₄B-based magnets. The developments of SmCo₅- and Sm₂Co₁₇-based magnets are the outcome of systematic studies, while the discovery of Nd₂Fe₁₄B compound is quite different. The original purpose of adding boron into binary R–Fe (R = La, Pr, and Nd) systems was to make it easier to obtain an amorphous phase in the hope of finding promising metastable binary phase. At the time researchers did not realized that they were actually creating a totally new ternary Nd₂Fe₁₄B compound.

The discovery of Nd₂F₁₄B has opened up a new avenue for future research, which is to exploring more ternary and multiple systems and it expands a vast new field of opportunities. In future research, more attention should be paid to the effects of atomic spacing on intrinsic magnetic properties. It is reasonable to believe that one of the future high-performance magnets might still take the form of R–T–M, where R is a rare earth that primarily contributes high crystalline anisotropy, T is a 3d transition metal that primarily contributes high Curie temperature and high magnetization, and M is another metal, or a semi-metal, or a non-metal element, that is added to form a new ternary compound and to adjust atomic spacing, just like boron does in Nd₂Fe₁₄B.

The author would like to thank Mr. Y. Wang, Prof. D. T. Zhang, and Prof M. Yue of Beijing University of Technology, Prof. J. Edmondson of Michigan State University, Lucia Liu of Georgia Institute of Technology, and Prof. J. P. Liu of University of Texas at Arlington for their help in preparing this manuscript. Special gratitude goes to Yuhui Shen, my colleague at the University of Dayton Magnetics Laboratory, for her effort to search important references, correct errors in the manuscript, and useful discussions.