Sintered Nd-Fe-B is widely used in VCM motors, wind power generators, and driving motors (Takezawa et al., 2015; Lu et al., 2017; Liu et al., 2019; Zhou et al., 2023). The increasing demand in various applications has led to significant consumption of rare earth elements (REs). Furthermore, the scarce and expensive heavy rare earth elements Dy and Tb are being consumed in increasing amounts to satisfy the growing intrinsic-coercivity (Hcj) demands of magnets (Zeng et al., 2019; Wang et al., 2020; Jiang et al., 2022; Li et al., 2022). In contrast, the rare earth element Ce is overstocked due to its lower intrinsic magnetic properties (Shi et al., 2019; Zhang et al., 2019; Chen et al., 2020). This limits the use of Ce to low-end applications, which typically requires only small amounts of Ce. In order to increase Ce utilization, the grain boundary diffusion process (GBDP) has been employed to produce Ce-containing Nd-Fe-B magnets (Sepehri-Amin et al., 2013; Fan et al., 2018; Zeng et al., 2020). Chen et al. (2020) applied GBDP to a Ce-based permanent magnet with Nd₇₀Cu₃₀ and Dy₇₀Cu₃₀ eutectic alloys. The coercivity increments remained almost constant, given whatever diffusion source was used. Moreover, the thermal stability of the magnets treated with Nd₇₀Cu₃₀ alloy was much higher than that of the magnet treated with Dy₇₀Cu₃₀ alloy. A continuous grain boundary phase was observed in the magnet treated with Nd₇₀Cu₃₀ alloy; however, a less continuous grain boundary phase was observed in the magnet treated with Dy₇₀Cu₃₀ alloy. This indicates that Nd can optimize the GB distribution during GBDP. In Ce-containing magnets, a CeFe₂ phase is formed when the Ce proportion is above 5% in Nd-Fe-B magnets (Cui et al., 2014; Liang et al., 2014; Zhu et al., 2014; Zhu et al., 2022). This is considered to be the main factor hindering diffusion due to the high melting point of the CeFe₂ phase. Moreover, Du et al. (2022) used commercial N45 sintered Nd-Ce-Fe-B magnets processed by GBD with TbH_x and Tb₇₀Fe₃₀ alloys. The coercivity was significantly improved from 1,011.7 to 1758.4 and 1764.7 kA/m. According to Di et al. (2018), grain boundary diffusion (GBD) was applied to a sintered Nd-Fe-B magnet with Al-aided TbH₂ and TbH₂. The coercivity was enhanced from 1,092.9 to 1853.8 kA/m for Al-aided TbH₂, higher than 1771.9 kA/m of the TbH₂ GBD magnet. Microstructure analysis of the Al-aided TbH₂ processed sample showed that continuous grain boundary phases were formed and completely enveloped the grains with thin Tb-rich shells, which can be attributed to the promoted diffusion aided by Al.

Hence, this paper uses an Nd_xDy_90-xAl₁₀ alloy as a diffusion source to increase the coercivity of Ce-containing Nd-Fe-B magnets. Simultaneously, a traditional DyH₃ diffusion source was employed as a comparative sample. The influences of Nd addition on the microstructure and magnetic properties of the Ce-containing magnets were comprehensively investigated.

A commercial sintered Nd-Fe-B magnet composed of Ce₅Pr_6.5Nd_19.5M_1.6B_0.92Fe_bal (wt%, M = Cu,Co,Ga,Ti,Al) was used as the original magnet. The size of an original magnet was 10 × 10 × 5 mm, and the direction of 5 mm is the orientation direction and the diffusion direction. The original magnets were polished with an abrasive paper and cleaned with alcohol. Diffusion flakes were prepared by melt-spinning with a wheel speed of 1.3 m/s. The flakes were then nitrogen-crushed into powder with an average particle size of <17 μm. The diffusion powders were added to and mixed with alcohol, which was then used to cover the magnets on two easy planes (c-planes) homogeneously. In this work, the samples treated with Nd_xDy_90-xAl₁₀ (x = 0, 10, and 20) were marked N0, N10, and N20, respectively. In comparison, DyH₃ as the diffusion source was coated on the surface of the magnets and treated simultaneously; in this paper, it was marked as H3. In this study, 1.0 wt% by weight of the original magnet weight was coated onto the surface of the magnets. The magnets were then heat treated at 1,213.2 K for various time durations from 0 to 5 h, followed by annealing at 923.2 K for 5 h. The magnetic properties of the magnets were tested using an NIM62000C magnetometer. The microstructures of the magnets were observed using a scanning electronic microscope (SEM) (CLARA GMH), and the phase compositions were determined using an energy dispersive spectrometer (EDS). The element distribution was analyzed using an electron probe microanalyzer (EPMA-JXA-ISP100).

Figure 1A shows the room temperature demagnetization curves of the original magnet and the diffused magnets treated at 1,213.2 K/5 h, followed by 923.2 K/5 h. The remanence (Br), intrinsic coercivity (Hcj), coercivity increment, and Dy utilization are summarized in Table 1. After diffusion, the coercivity increases from 1,124.7 kA/m to 1,656.4, 1,673.9, 1,584.8, and 1,490.9 kA/m for the N0, N10, N20, and H3 diffused magnets, respectively. The increment rates for coercivity are 47.2%, 48.8%, 40.9%, and 32.5%, respectively. The deposited Dy in N10 diffusion is 0.8% wt of the original magnet, which is lower than N0 (0.9%wt) and H3 (1%wt), but the coercivity increment rate is higher than that of the two others. Although the coercivity increment rate for N20 diffusion is 40.9%, which is lower than N10 (48.8%), Dy is inadequate in N20 diffusion. At the same time, the remanence (Br) of diffused magnets decreases from 1.354 T to 1.316, 1.319, 1.329, and 1.303T, and N20 possesses the highest Br. The decrease in Br is most likely caused by the substitution of Dy for Pr and Nd, as well as the increased volume fraction of non-magnetic phases in the grain boundary (Zeng et al., 2019).

Figure 1B shows the temperature coefficients of coercivity β at 280–420 K. β is usually used to measure the temperature stability of a magnet and is given by the following equation: β = H c j T − H c j T 0 H c j T 0 T − T 0 × 100 % (1)

The lower the absolute value of β, the higher the temperature stability of the magnet. Figure 1B shows that the value of β improved from −0.65%/K in the original magnet to −0.55%/K, −0.48%/K, −0.43%/K, and −0.56%/K in N0, N10, N20, and H3, respectively. The enhancement in coercivity is responsible for the increased temperature stability. Moreover, the temperature stability of N20 and N10 is better than that of N0 and H3. The addition of Nd optimizes the wettability between grains and RE-rich layers, which leads to higher coercivity enhancement in N10 and N20. Furthermore, the substitution of Pr with Nd increases the temperature stability of N10 and N20 due to the forming of a high Curie temperature Nd₂Fe₁₄B (585K) phase than Pr₂Fe₁₄B (565K) (Zhang et al., 2019).

Figure 2 presents the back-scattered electron images of original and diffused magnets positioned at varying distances from the surface. There are two main types of areas, black and white areas, referring to the 2:14:1 phase and the RE-rich GB phase, respectively. After diffusion, the grain size is clearly smaller. Ramesh et al. (1988) derived the relationship between coercivity and grain size as 1/lnDⁿ, where D is the grain size. During GBDP, the melting RE-rich phase dissolves the bumps and corners of the main phases, thus reducing the size of the main grains. The finer the grains, the higher the Hcj of a magnet. The boundary in the original magnet is blurred, which means that the matrix grains are not well-isolated. After diffusion, the grain boundaries are optimized and become clearer. Moreover, a Dy-rich core shell was observed in all diffused samples. At a distance of 50 μm from the magnet surface, the Dy-rich core shell is notable in all diffused samples. However, as the diffusion distance increases, the shells gradually become thinner. At a depth of 100 μm, the shells can hardly be observed in the N0 and N20 diffused samples, whereas they remain visible in the N10 diffused sample. This means that Dy is diffused deeper into the N10 diffused sample and Dy distribution is more uniform. On the contrary, for the N0 and H3 diffused samples, Dy seems consumed in the surface, and less Dy diffused deeper.

Figure 3 demonstrates the EDS mapping of diffused samples at a depth of 50 μm. Nd is enriched in the triple junctions (TJs). In contrast, Dy enriches around the matrix main phases and evenly enters the main phases. Dy will replace Pr or Nd with (Pr/Nd)₂Fe₁₄B to form the Dy₂Fe₁₄B phase (Liu et al., 2021). This explains the remanence decreases after the GBD process because the saturation magnetization (Js) for Dy₂Fe₁₄B is lower than (Pr/Nd)₂Fe₁₄B. Furthermore, Ce is enriched in the TJs among all diffused magnets. According to Yan et al. (2014), Ce tends to enrich in the grain boundaries; when the Ce content is higher than 6.0 wt%, a CeFe₂ phase is formed at the intergranular boundary of the sintered Nd-Ce-Fe-B magnet. Too much CeFe₂ phase is harmful for the coercivity of the magnet because it deteriorates the wettability of grain boundary phases.

In order to reveal the coercivity enhancement and diffusion mechanism, the magnets were GBDP-treated by diffusing Dy₉₀Al₁₀ and Nd₁₀Dy₈₀Al₁₀ alloys. The diffusion heat-treatment is 1,213.2K/x h, followed with 923.2K/5 h (x = 1,2,3,4). EDS was conducted to measure the Dy and Nd concentrations at a depth of 100 μm from the surface of a diffused sample. Since the original magnets contain no Dy, all Dy in the magnets are diffused. In addition, the Nd concentration is deducted from the Nd content in the original magnet. The Nd and Dy concentrations as functions of heat-treatment time were examined, and the results are shown in Figure 4A. As diffusion time increases, Nd and Dy contents increase in all diffused samples. The Dy content of the Nd₁₀Dy₈₀Al₁₀ diffused sample is larger than that of the Dy₉₀Al₁₀ diffused sample. Figure 4B shows the concentration ratio Dy/Nd at GBs in the Nd₁₀Dy₈₀Al₁₀ diffused sample. At the first-step diffusion time of 1 h, the value is 6.21, which is lower than the atomic number ratio in Nd₁₀Dy₈₀Al₁₀ (8:1). For times of 2, 3, and 4 h, the Dy/Nd ratio decreases to 4.15, 3.42, and 2.92, respectively. This shows that the diffusion of Nd is faster than that of Dy, providing diffusion channels for subsequent Dy diffusion. This explains Dy is more uniform and diffused deeper in the N10 diffused sample. According to Liu et al. (2018), an HRE-rich shell considerably increases coercivity due to the decoupling of matrix grains. The thinner and deeper the distribution of the HRE-rich shell, the higher coercivity and more homogeneously magnets can be obtained.

The original magnet contains 5% Ce, and SEM images show that there are two kinds of grain boundary phases in Nd₁₀Dy₈₀Al₁₀ diffused magnets (Figure 5), a gray grain boundary phase and a white grain boundary phase (Figure 5A1). In order to ascertain the components of the grain boundary phases in the Nd₁₀Dy₈₀Al₁₀ diffused sample, EDS line scanning was used (Figures 5A2–B5). Figures 5B3, B4 show that the ratio of Ce to Fe is 1:2 in the gray grain boundary. According to the results obtained by Zhang et al. (2017) and Xiong et al. (2018), there is a CeFe₂ phase in the original magnet. Figures 5A5, B5 show that, after diffusion, the Dy content in the gray grain boundary phase is higher than that in the white grain boundary phase, indicating that Dy is mostly consumed in the gray grain boundary phase. This observation suggests that Ce and Fe impede the diffusion of Dy, causing it to accumulate in the gray area. It is worth noting that the melting temperature of the metallic RE-rich phase is approximately 919.2 K (Zeng et al., 2019), which is lower than that of the CeFe₂ phase (1,198.2 K). The high melting point of CeFe₂ contributes to poor wettability in the gray region, hindering the diffusion of Dy and resulting in its accumulation within this region. This reduces the coercivity enhancement effect of Ce-containing magnets.

In order to explore the effect of Nd addition on the diffusion of Ce-containing Nd-Fe-B magnets, the EPMA measurements of the N10, N20, and H3 diffused samples were conducted (Figure 6). The Dy content in N10 and N20 is higher than H3, and the diffusion depth of Dy in N10 is significantly deeper than H3. This indicates that the addition of Nd is conducive to the diffusion of Dy in Ce-containing Nd-Fe-B. According to Tang et al. (2017), Nd can optimize the wettability of grain boundary phases. A contiguous grain boundary is beneficial for the diffusion of Dy. Therefore, the diffusion depth of Dy is deeper in N10 than that in H3. This is consistent with the previous discussion that the coercivity of the N10 diffused sample is higher than that of N0 and H3.

Figure 7 demonstrates the Dy distribution at the depth of 100 and 150 μm in the N10, N20, and H3 diffused samples. Dy is enriched around the matrix main phases in all diffused samples. At the same diffusion depth, the Dy content in N10 is higher than that in N20 and H3. In addition, the Dy content in N20 is higher than that in H3. This indicates that more Dy is diffused into the magnet in Nd-containing diffusion. The Nd-containing diffusion source promotes the diffusion of Dy, which is beneficial to improve the coercivity of the magnet. The aforementioned work provides a method of the synergetic effects between Nd and Dy to develop the low-cost Nd-Ce-Fe-B magnets.

Coercivity enhancement of 5% wt Ce-containing Nd-Fe-B sintered magnets was achieved by diffusing Nd_xDy_90-xAl₁₀ alloys. The coercivity increased significantly from 1,124.7 to 1,656.4, 1,673.9, and 1,584.8 kA/m, for Dy₉₀Al₁₀, Nd₁₀Dy₈₀Al₁₀, and Nd₂₀Dy₇₀Al₁₀ samples, respectively. Although the Dy deposited in the Nd₁₀Dy₈₀Al₁₀ diffused sample was 0.8%wt of the original magnet, which is lower than that of the Dy₉₀Al₁₀ diffused sample (0.9%wt), its coercivity increment rate is significantly higher. The microstructure of the N10 diffused sample showed that the addition of Nd improved the distribution of grain boundary phases, which provides channels for subsequent Dy diffusion. In addition, a high melting point CeFe₂ phase was present in the Ce-containing Nd-Fe-B magnets, and significant amounts of Dy accumulated in this phase, resulting in Dy wastage. This situation was optimized using Nd-containing alloy as a diffusion source. Here, Dy could diffuse further due to the increased fluidity of the grain boundary phase by the addition of Nd.