Injection moldable, self-healable, and recyclable rubber-bonded NdFeB magnets with the magnetic particulates content up to 90 wt%

To apply the facile injection molding technique commonly used in thermoplastics processing to rubber-bonded magnets, typical highly filled thermosetting composites, an approach based on Diels–Alder (DA) chemistry is proposed by the authors. In the proof-of-concept experiments, the oligomers of hydroxyl-terminated polybutadiene rubber are successively functionalized by furyl and trifluoromethyl, and the magnetic NdFeB particles are modified by maleimide. By taking advantage of the reversible DA reaction, cross-linkages are established among the treated polybutadiene oligomers and NdFeB particles. The composite exhibits very low viscosity due to de-cross-linking when heated to retro-DA reaction temperature, which provides the highly filled system with necessary thermal processibility. In addition, the fluorine groups of the rubber phase are aggregated surrounding the particulate fillers allowing for well separation of the latter in the melt state. During the subsequent cooling, the DA bonds are re-built up in the rubber and at the particles/rubber interface, respectively, offering the robustness of traditional thermosets. As a result, the polybutadiene-bonded magnets with NdFeB content up to 90 wt% can be injection molded and possess rather high magnetic properties, mechanical properties, self-healability, and recyclability. The present work combines the merits of thermoplastics and thermosets and satisfies the contradictory requirements of high filling, efficient manufacturing, and balanced performance of functional polymer composites.


Introduction
Polymer-bonded magnets are composites consisting of a large number of magnetic particles mixed with a small amount of polymer binder [1][2][3]. Compared to casted and sintered magnets, their advantages include low weight, corrosion resistance, easy machining and handling, and insensitiveness to crack or break [4,5]. Bonded magnets provide various combinations of mechanical, physical, chemical, thermal, and magnetic properties due to the incorporation of different kinds of polymeric matrices and thus can be used in a broad range of applications like automotive, electronics, and medical devices, especially where both flexibility and high energy products are required [6].
It is known that the magnetic properties of bonded magnets mainly depend on the species and content of the included magnetic powders [7]. For a given kind of magnetic particle, the higher magnetic properties of bonded magnets require higher content of the magnetic fillers [8], which is used to greatly reduce the flowability of the compounds during manufacturing and affect the distribution of the additives as well. Consequently, the resultant bonded magnets often have poor coercivity field strength and mechanical performance [9]. Clearly, the contradiction between a high filling and viable processability of bonded magnets becomes a challenging issue.
Here in this work, we plan to tackle the problem by using rubber-bonded magnets as the proof-of-concept material. Traditionally, rubber-bonded magnets are produced by calendering or compression molding [3,4]. The commonly utilized injection molding of thermoplastics in the plastics industry is not suitable because of the cross-linked macromolecular structure of rubbers, which disables the high-efficient formation of rubber-bonded magnet parts. In this context, our research goal lies in the development of a new approach that allows for imparting injectability to the highly filled rubber-bonded magnets while promoting dispersion of the magnetic particles.
To this end, the core measure is the introduction of reversible covalent chemistry and fluorination to the system. Reversible covalent bonds can repeatedly break and reform under certain conditions [10][11][12]. When polymers are cross-linked by reversible covalent bonds, the materials can be transformed between thermosets and thermoplastics accordingly [11], which would not only help to raise the flowability of the polymeric networks but also bring about smart functionalities like self-healing and recycling abilities [13][14][15][16][17]. On the other hand, fluorination treatment of inorganic fillers proved to weaken the inter-particulates interaction and prevent their aggregation [18], which reduces the probability of acting as stress concentration sites and favors the enhancement of the composites. The coupling of the two techniques may realize the above design.
Specifically, low-molecular-weight hydroxyl-terminated polybutadiene (HTPB) serves as the starting material to construct the rubber phase and the reversible DA bond as the cross-linker. Polybutadiene is a conventional synthetic rubber. The unsaturated double bonds and polyhydroxy functional groups of HTPB can be easily functionalized [19,20]. Meantime, the thermally triggered retro-DA reaction of DA bonds-based cross-linkages would lead to decross-linking of the rubber networks at about 120 °C, forming low molecular weight species with low viscosity, while the forward DA reaction at ~ 80 °C would enable re-cross-linking. In addition, the isotropic NdFeB particles will act as magnetic fillers, which are widely used in bonded magnets owing to their exceedingly good magnetic properties [5,21]. Their dispersion is going to be promoted by the fluorinated polybutadiene, as described below.
As shown in Scheme 1a and Fig. S1a, HTPB is firstly functionalized by furyl on the side chains and then turns to be trifluoromethyl-telechelic. Next, the functionalized HTPB reacts with bimaleimide (BMI), forming the target crosslinked rubber networks (HTPB-F 3 -DA, Fig. S1a). On the other hand, the NdFeB particles are modified by maleimide (NdFeB-M, Fig. S2). When HTPB-F 3 -DA and NdFeB-M are blended at 120 ℃ (Scheme 1a and Fig. S1b), HTPB-F 3 -DA would be de-cross-linked along with breakage of DA bonds, leading to the decrease of viscosity and release of the trifluoromethyl functionalized HTPB oligomers. Hence, the hydrophobically treated magnetic particles NdFeB-M can be well mixed with the oligomers even at high concentrations. In the course of the subsequent cooling, DA bonds are formed among the oligomers giving birth to HTPB-F 3 -DA again and at the interface between NdFeB-M and HTPB-F 3 -DA (Scheme 1b and c). Moreover, the fluorine-containing groups attached to HTPB-F 3 -DA cannot be wrapped by the polybutadiene-rich domains but expelled to the rubber/ filler interface due to the low surface energy (Scheme 1b), which is supported by a recent report by Cao et al. [22], who demonstrated the similar effect of a home-made fluorinefunctionalized flow modifier on magnesium hydroxide reinforced linear low-density polyethylene. Eventually, the HTPB-F 3 -DA acquires the necessary low viscosity in the case of manufacturing and the cross-linked structure like other rubbers after fabrication. The NdFeB particles are covalently bonded to the rubber binder and separated from each other by the fluorine-containing groups of the rubbers at the same time. Such a design is believed to be able to offer a rational method for addressing the aforesaid problem. Moreover, the resultant bonded magnets would inherit self-healability and recyclability from the reversible DA bonds.
The feasibility of the proposed idea is verified hereinafter. It is hoped that the outcomes will provide a novel platform technique for producing highly filled rubber composites.

Preparation of maleimide functionalized NdFeB (NdFeB-M)
As shown in Fig. S2

Preparation of furan-functionalized HTPB oligomers (HTPB-F)
A solution of HTPB (100.0 g) in THF (300 mL) was added to a three-neck round bottom-flask equipped with a mechanical stirrer, dropping funnel, condenser, and N 2 purging. The solution was heated to 80 ℃ and stirred for 30 min. Then, the solution of 2-furanmethanetiol (26.0 g) and AIBN (5.0 g) dissolved in THF (50 mL) was poured into a funnel and dropped into the HTPB solution [19]. The reaction was completed after 12 h, and the system was cooled down to room temperature. The resultant solution was concentrated by rotary evaporation under reduced pressure, and the concentrated mixture was precipitated in methanol (500 mL each time) and repeated three times. Finally, the solution was dried in a vacuum oven at 45 ℃ to obtain the functionalized HTPB (i.e., HTPB-F) as a viscous oil (Fig. S1a)

Preparation of fluorine-functionalized HTPB-F oligomers (HTPB-F-F 3 )
HTPB-F (10.0 g), 4-(trifluoromethyl)phenyl isocyanate (1.04 g), and dibutyltin dilaurate (0.1 g) were dissolved in 100 mL THF with N 2 purging and stirred at 70 ℃ for 6 h. The resultant solution was concentrated by removing the solvent via rotary evaporation under reduced pressure, and the mixture was precipitated in methanol (100 mL at each time) and repeated three times. At last, the solution was dried in a vacuum oven at 45 ℃ to obtain the functionalized HTPB-F-F 3 as a viscous oil (Fig. S1a). The 1 H NMR signals of HTPB-F-F 3 are shown in Fig. S3. The emergence of the peaks in the phenyl region (~ 7.57 ppm) confirms the presence of trifluoromethyl in the HTPB-F chains.

Preparation of DA bonds cross-linked HTPB oligomers (HTPB-F 3 -DA and HTPB-DA)
HTPB-F-F 3 (50.0 g) and BMI (19.8 g) were dissolved in 350 mL THF with N 2 purging and stirred at 70 ℃ for 6 h. Then, the mixture was poured into a closed silicone mold and cured at 60 ℃ for 72 h to obtain a reddish-brown sheet of HTPB-F 3 -DA (Fig. S1a).
HTPB-DA was prepared in the same way as HTPB-F 3 -DA, and the only difference is that HTPB-F-F 3 was replaced by HTPB-F (Fig. S1c).

Fabrication of the control samples (NdFeB-M/ BR-DCP)
Butadiene rubber (BR) is an important synthetic elastomer based on high-content cis-1,4 polybutadiene, and the carbon backbone also possesses a double bond [24] like HTPB. In fact, HTPB is derived from the chain cleavage of BR9000 [25]. Therefore, BR9000 was chosen and cross-linked by DCP, a common vulcanizing agent of rubber that produces rapid cure under typical vulcanization temperatures [26,27], to act as the control rubber with irreversible cross-linkages. It was reported that the optimal curing time (t 90 ) of the rubber vulcanized by DCP is 24 min at 160 ℃, but it would be longer than 1 h at 120 ℃ [28]. According to this habit, when BR9000 is mixed with DCP at 120 ℃ (the processing temperature of HTPB-DA and HTPB-F 3 -DA, refer to the last sub-section), the rubber would not be completely cured and still have certain processibility.
In this context, the control sample of cross-linked BR9000 was produced as follows: BR9000 (45.0 g) was dried at 80 ℃ in a vacuum for 12 h. Afterward, DCP (3.15 g) and 4010NA (0.90 g) were added and evenly mixed together by means of Haake torque rheometer at 120 ℃ under 70 rpm for 15 min, and cured in the plate vulcanizing machine at 160 ℃ for 25 min, producing the DCP cross-linked polybutadiene (BR-DCP).
For making the control bonded magnets, BR9000 and NdFeB-M were dried at 80 ℃ in a vacuum for 12 h. Then, the composite (Table 1) was compounded and injection molded following the same procedures as those applied for fabricating the rubber-bonded magnets containing DA bonds described in the last sub-section, except that the packing time was changed to 25 min and the mold temperature became 160 ℃.
It is worth noting that BR-DCP, HTPB-F 3 -DA, and HTPB-DA had almost the same molecular weights between crosslinks, M c (Tables S1 and S2), for the convenience of comparison. This was achieved by adjusting the amount of the initiator DCP to 7 wt% relatives to the amount of BR9000.

Basics of the modified polybutadiene
As mentioned in the Introduction, HTPB is functionalized by furyl on the side chains and then turns to be trifluoromethyl-telechelic (HTPB-F-F 3 ). The target cross-linked rubber networks (HTPB-F 3 -DA) are synthesized from the reaction between HTPB-F-F 3 and BMI. Therefore, we need to first confirm whether the furyl and trifluoromethyl are successfully grafted on HTPB, and to investigate the influence of functionalization of HTPB as well. Figure S3 reveals that the characteristic peak (i.e., -CH 2 -) of furylmethanethiol at 3.73 ppm and those of furan's = CH-at 6.14 ppm, 6.29 ppm, and 7.33 ppm appear  on the 1 H NMR spectrum of furyl functionalized HTPB (HTPB-F). The result indicates that the furan ring has been attached to HTPB as planned, and the degree of furan substitution is calculated to be 10% by using Eq. (1). As for the HTPB-F that is further functionalized by fluorine (i.e., HTPB-F-F 3 ), a new peak is perceived at phenyl region (~ 7.57 ppm) of its 1 H NMR spectrum (Fig. S3), which confirms the presence of trifluoromethyl in HTPB-F-F 3 . The effects of furan grafting and fluorine functionalization on molecular weights of the polybutadienes are inspected by GPC. As shown in Fig. S4 and Table S3, the number-average molecular weight, M n , of HTPB-F increases from 5639 to 7054 due to the introduction of 2-mercaptoethanol, which reacts with the C = C double bonds of HTPB. Other side reactions, such as cross-linking and degradation, would hardly occur for the thiol-ene click reaction [19]. Furthermore, the M n values keep unchanged before and after of trifluoromethyl-functionalization of HTPB-F, revealing that the modification does not cause cross-linking.
On the basis of the above discussion, 1 H NMR spectroscopy is used to verify the activity of the furyl on the side chains of HTPB (i.e., HTPB-F) for participating in DA reaction and check the effect of trifluoromethyl groups on DA reaction of HTPB-F-F 3 . For this purpose, a model reaction between HTPB-F-F 3 /HTPB-F and N-methylmaleimide (BM) is carried out under the inspection of a 1 H NMR spectroscope. Here 4,4′-methyenebis(N-phenylmaleimide) (BMI) is replaced by BM to avoid forming gelation of HTPB-F-F 3 / HTPB-F. By adding a 1/1 mol equivalent ratio of maleimide/ furan in o-xylene-d 10 to the NMR tube, the mixture is heated to 70 ℃ to monitor the formation of DA adduct by using 1 H NMR. Figure 1 exhibits that with a rise in time, a few new peaks assigned to DA adduct start to appear at 2.94 ppm (5′ and 6′), 5.14 ppm (1′), and 6.50 ppm (2′ and 3′), and then gradually become more and more obvious. Afterward, the samples are kept at 100 ℃ to reveal the effect of the retro-DA reaction. The above-mentioned three groups of peaks are weakened accordingly, but they are still detectable even though the reaction lasts 20 min. It means that the reversibility of the DA bonds in the modified polybutadiene is not 100% effective.
The effect of fluorine can be evaluated by taking advantage of the in situ 1 H NMR spectroscopy studies in Fig. 1a, b as follows: It is known that the time-dependent reaction conversions of DA and retro-DA reactions, x, is expressed by Eq. (2): Meantime, DA reaction obeys the second-order kinetic model Eq. (3) [29]: where k is the rate constant and t is the reaction time. Figure S5 indicates that HTPB-F-F 3 and HTPB-F have the same k values, meaning that the trifluoromethyl of HTPB-F-F 3 does not affect the rates of DA reaction and retro-DA reaction of furan groups on the side chains of HTPB. The orthogonality between DA reaction and trifluoromethyl would satisfy the requirements of both increasing fillers/ rubber interaction and decreasing inter-fillers interactions. The thermal reversibility of the DA bonds in HTPB-DA and HTPB-F 3 -DA is important for injection molding of the highly filled composites. We need to examine the reversibility of the included DA bonds at the very beginning. The cyclic DSC measurement results in Fig. 1c, d show that there are broad endothermic peaks between 100 and 150 ℃ on the spectra of HTPB-F 3 -DA and HTPB-DA due to the occurrence of retro-DA reaction. Such endothermic peaks repeatedly appear after cooling the samples to − 80 ℃, despite the fact that the second, third and fourth heating curves exhibit weaker endothermic peaks as a result of the slow formation of DA bonds when the temperature of the DSC cell drops from 160 ℃ at a relatively fast rate of 10 ºC min −1 . Comparatively, the DSC heating curves of the control BR-DCP have completely different looks because of lacking DA bonds (Fig. S6). The endothermic peaks at − 25 ℃ originate from the melting of the crystal regions of BR9000 [25]. It is thus known that the modified polybutadienes HTPB-F 3 -DA and HTPB-DA have acquired thermal reversibility as expected.

Maleimide functionalized NdFeB
To enhance the interfacial affinity of the NdFeB particles toward polybutadiene via DA bonds, maleimide and furan are respectively attached to the fillers and rubber in advance (Figs. S1 and S2). It is hoped that the reaction between the maleimide groups of the modified NdFeB and the furan groups of HTPB-F 3 -DA and HTPB-DA during the thermal compounding would form the six-membered rings based on DA chemistry (Scheme 1). Since the modification of polybutadiene has been analyzed in the last sub-section, maleimide functionalization of NdFeB particles by means of N-((3triethoxysilyl) propyl) maleimide (MI-silane, Fig. S2) is the main theme of the discussion here.
Firstly, the FTIR spectrum of NdFeB-M (Fig. 2a) indicates that several new peaks are perceived in comparison with that of the untreated NdFeB. The prominent band at 1100 cm −1 is ascribed to the Si-O-Si stretching vibration, while that at 1369 cm −1 is assigned to the symmetric stretching of C-N-C. Furthermore, a distinguishing absorption at 679 cm −1 owing to the maleimide stretching appears, which suggests the presence of MI-silane on the magnetic particles [30]. That is, the maleimide-functionalized NdFeB, NdFeB-M, has been successfully obtained. Secondly, elemental analysis is utilized to examine the exact amount of the maleimide attached to the surface of NdFeB in NdFeB-M. The calculation based on Eq. (S1) reveals that NdFeB-M contains 0.54 ± 0.1 mmol of maleimide per gram of NdFeB. Thirdly, NdFeB-M is brought to react with furan to check whether the attached maleimide groups are still active enough to take part in the DA reaction. To this end, the suspensions consisting of NdFeB-M and furan in toluene are stirred for 1 day at 60 ℃. The FTIR spectrum of the reaction product, NdFeB-M-F (Fig. S2), demonstrates that the typical absorption of C-O-C is present at 1504 cm −1 , meaning that the DA reaction can indeed proceed between NdFeB-M and furan.
In addition, the furan that has reacted with the maleimide of On the other hand, the results of Fig. S7 and Table S4 indicate that the maleimide functionalization does not change the size and morphology of the NdFeB particles, and the slight decay of the magnetic properties results from nonmagnetic MI-Silane [31].

Rheological study
Because the bonded magnets are planned to be fabricated by melt mixing and injection molding (Scheme 1), the linear viscoelasticities of the composites should be measured as a function of temperature in advance to examine the effect of retro-DA reaction on melt fluidity of the systems. The higher complex viscosity often corresponds to poorer processability [32].
Besides the protagonist bonded magnets NdFeB-M/ HTPB-F 3 -DA, three reference composites, NdFeB/HTPB-F 3 -DA, NdFeB-M/HTPB-DA, and NdFeB-M/BR-DCP, are prepared in the same way as NdFeB-M/HTPB-F 3 -DA (Table 1) for the rheological study. The comparison between the viscoelasticity of NdFeB-M/HTPB-F 3 -DA and those of the controls would clarify the contributions of DA bonds and fluorine functionalization to the manufacturability of the highly filled composites.
As shown in Fig. 3a, the complex viscosities, η*, of HTPB-F 3 -DA and HTPB-DA start to drastically decrease as of 100 ℃, because the retro-DA reaction takes place, which leads to de-cross-linking of the networks and formation of the low viscosity oligomers [33]. Similarly, the composites with HTPB-F 3 -DA and HTPB-DA as binders show the same decline trend, which provides the rheological basis for melt compounding and injection molding of the highly filled composites. In contrast, the η* values of the control BR-DCP and its composites slightly increase at elevated temperatures due to the further cross-linking of free radicals.
By contrast to the complex viscosity determined by the strain rheometer, the torque measured by the torque rheometer is more directly correlated to the thermal processability of polymeric materials [34]. Figure 3b shows the balanced torques of the composites at 120 ℃, taken from the time dependences of torque (Fig. S8). It is obvious that NdFeB-M/HTPB-F 3 -DA has the lowest balanced torque with the same filling amount. The balance torque of NdFeB-M/BR-DCP is about two or three times higher than those of the composites with DA bonds cross-linked polybutadiene as the binders. The retro-DA reaction of the DA bonds plays a much bigger role in this case than the surface treatment of the magnetic particles. The maleimide of NdFeB-M could hardly react with the furan of HTPB-F or HTPB-F-F 3 oligomers released from HTPB-DA or HTPB-F 3 -DA at the retro-DA reaction temperature. As for NdFeB-M/BR-DCP, because of the presence of DCP, BR9000 is only partly cross-linked during mixing with NdFeB-M particles (refer to the Sect. 2), so its balanced torque is significantly increased. Figure 3c illustrates the repeated DSC heating curves of NdFeB-M/HTPB-F 3 -DA (85 wt%) for exploring the thermal reversibility of the built-in DA bonds and the composite, which is correlated to self-healability and reprocessability. A broad endotherm attributed to retro-DA reaction recurrently emerges between 100 and 150 ℃. In addition, NdFeB/ HTPB-F 3 -DA (85 wt%) and NdFeB-M/HTPB-DA (85 wt%) exhibit similar endothermic peaks within the identical temperature range (Fig. S9a, b). The phenomenon is consistent with those observed in HTPB-F 3 -DA and HTPB-DA (Fig. 1c, d) and the downward trend of the complex viscosity with increasing temperature as well (Fig. 3a), manifesting the composites have inherited the necessary thermal reversibility of DA reaction. For the NdFeB-M/BR-DCP (85 wt%) excluding DA bonds, their DSC heating curves also overlap each other, but there are no endotherms between 100 and 150 ℃ except for the melting peaks of the crystalline phase at ~ − 25 ℃ (Fig. S9c). Therefore, the reprocessing and recycling of the control composite NdFeB-M/BR-DCP are anticipated to be unable to carry on like the composites with DA bonds.

Microstructures and interfacial interactions of the bonded magnets
Since the modified polybutadiene and NdFeB particles have been available and the retro-DA reaction has been shown to work, the highly filled rubber-bonded magnets are successively fabricated via melt compounding and injection molding. Upon cooling, DA bonds would be established not only among the cross-linked polybutadiene but also at the interface between the NdFeB particles and the neighboring polybutadiene. The fluorinated groups attached to the modified polybutadiene may gather at the fillers/rubber interface, facilitating the separation of the filler particles. The idea (Scheme 1) will be confirmed below.
For purposes of quantifying the DA cross-linkages, swelling tests are carried out (Fig. S10), and the average molecular weights between cross-links, M c , are computed with Eqs. (S4)-(S6). Table S2 indicates that both NdFeB-M/HTPB- Well dispersion of particulate fillers in the polymer is one of the key factors for bringing them into full play, which is particularly true in the case of highly filled composite. To reveal the status of the magnetic particles in the rubber-bonded magnets produced in this work, EDS mappings of the fractured surfaces of the model composites are recorded first, in which the magnetic particulates contents are intentionally reduced to 5 wt% to get rid of the interference of the overcrowded particles in the authentic highly filled systems. The mappings of carbon, iron, and fluorine elements in Figs. 4a, b and S11a, b demonstrate that no severe agglomerates of the particles are found in the composites based on fluorinated polybutadiene (i.e., NdFeB-M/HTPB-F 3 -DA and NdFeB/HTPB-F 3 -DA) and fluorine elements are enriched around the particles. Meantime, there is no gap at the fillers/rubber interface of NdFeB-M/HTPB-F 3 -DA and NdFeB-M/HTPB-DA composites (Figs. 4a and S11a), despite that larger aggregated particles are observed for the latter. It suggests that the interfacial compatibility of the two composites is rather good due to the DA reaction between the maleimide of the modified particles and the furan of the modified polybutadiene. On the whole, the above results indicate that the DA bonds at the magnetic particles/rubber interface and the accumulation of trifluoromethyl of the chain ends of the modified polybutadiene at the same interface improve the interfacial interaction and mitigate the agglomeration of the magnetic particles in NdFeB-M/HTPB-F 3 -DA, respectively, as illustrated in Scheme 1. Such effects would enhance the performance of the composite. In fact, the photos of NdFeB-M/BR-DCP (Fig. S11b) provide a counterexample to this analysis. Because of the absence of the interfacial DA bonds and fluorine in this composite, poor fillers/rubber adhesion, and larger fillers agglomeration have to be perceived. The SEM micrographs and EDS analyses of the fracture surfaces of the rubber-bonded magnets containing 85 wt% NdFeB particles (Figs. 4c, d and S11c, d) agree with the above observations of the model composites with much fewer fillers (Figs. 4a, b and S11a, b). Figure 4c, d shows that the fluorine elements are concentrated on the exposed particles of NdFeB-M/HTPB-F 3 -DA and NdFeB/HTPB-F 3 -DA. It reflects the fact that NdFeB particles can be well wrapped by HTPB-F 3 -DA networks even under low content of the rubber phase, no matter whether the particles are functionalized or not. At the same time, carbon elements are also enriched on the magnetic particles in NdFeB-M/HTPB-F 3 -DA (refer to the red circles), manifesting the appearance of cohesively failed rubber debris enabled by the interfacial DA bonds. A similar phenomenon is also found in NdFeB-M/HTPB-DA (Fig. S11c). In contrast, there are no enriched carbon elements on the exposed NdFeB particles of the fracture surface of NdFeB/HTPB-F 3 -DA, which is understandable as there is no covalent connection between NdFeB and HTPB-F 3 -DA. Figure S11d exhibits that there are quite a few pores on the fractured surfaces of NdFeB-M/BR-DCP. NdFeB particles are present without apparent traces of adhesion to the nearby rubber. There is no aggregation of carbon elements on the exposed particles of the fracture surface. It is proved that the maleimide on NdFeB cannot react with BR-DCP, leading to weak interfacial interaction, so the particles are difficult to be well dispersed in BR-DCP. The conclusion is evidenced by the average pore diameters and porosities listed in Table S5, which are the highest for NdFeB-M/BR-DCP (41.14 nm and 10.35%). The modified polybutadiene-based composites possess obviously smaller pore sizes, and their porosities are only half of that of NdFeB-M/BR-DCP.
To have an in-depth understanding of the interfacial interactions, surface free energies of the relevant materials are estimated from the contact angle measurements. Table S6 shows that the surface free energy of the treated NdFeB particles is lower than that of the untreated versions, owing to the coverage of MI-Silane on the surface of the former, which reduces aggregation of the particles. In addition, HTPB-F 3 -DA has lower surface free energy than HTPB-DA, which is attributed to the introduction of trifluoromethyl to the former.
Accordingly, adhesive work, W a , interfacial tension, γ mf , and spreading coefficient, S, are further calculated using Eqs. (S7)-(S11) to quantify the fillers/rubber interfacial interaction (Table 2). Clearly, the W a of NdFeB-M/HTPB-DA is the highest, and its γ mf is the lowest. In general, W a reflects the interfacial combination ability. The larger the W a , the greater the adhesion between NdFeB-M and HTPB-DA. Besides, γ mf expresses the infiltration and dispersion abilities among phases. The smaller the γ mf , the more easily NdFeB-M is dispersed in HTPB-DA. In this context, the ratio of W a /γ mf can be used to comprehensively characterize the interfacial interaction. The larger W a /γ mf implies the stronger the interaction between the magnetic particles and rubber. Table 2 demonstrates that NdFeB-M/HTPB-DA has the largest W a /γ mf as a result of the DA bonds between NdFeB-M and HTPB-DA built up during manufacturing.
Along with the incorporation of fluorine to HTPB-DA, the interfacial interaction of the composite is somewhat decreased, so that the W a /γ mf value of NdFeB-M/HTPB-F 3 -DA is decreased from 7.11 to 4.69 [35]. Meantime, the spreading coefficient of the system is improved from 21.86 to 24.21 mJ m −2 . This parameter describes the wettability of liquid/solid contact. The positive spreading coefficient means that the spontaneous spreading of liquid on solid occurs, and the negative value predicts that liquid has to form a droplet on the solid [36]. Therefore, the higher spreading coefficient of NdFeB-M/HTPB-F 3 -DA than that of NdFeB-M/HTPB-DA at the expense of a decrease of W a /γ mf means that the former gains a better compounding effect by slightly reducing interfacial bonding. Such a trade-off is necessary for balancing the processability and properties of the composites. With respect to NdFeB/HTPB-F 3 -DA, its spreading coefficient (27.07 mJ m −2 ) is greater than that of NdFeB-M/HTPB-F 3 -DA (24.21 mJ m −2 ), as the interfacial DA bonds of the latter result in decreased wettability of the rubber phase to the particulates.
With regard to the control NdFeB-M/BR-DCP, its spreading coefficient is relatively larger because the hydrophobicity of NdFeB-M is raised after surface treatment and somewhat matches the hydrophobic BR-DCP, while the W a value is the lowest due to the insufficient interfacial interaction, which coincides with the results of Figs. 4 and S11.

Magnetic properties
The room temperature magnetization data of the injectionmolded rubber-bonded magnets are shown in Figs. 5, S12, and Table S4. Herein, remanence, B r , refers to the magnetization remains in a ferromagnetic material when the external magnetic field is dropped to zero, and coercivity, H c , describes the field strength at which the magnetic flux density of a material was previously magnetized to saturation falls to zero. Intrinsic coercivity, H ci , is the strength of the magnetic field required to reduce the magnetic polarization to zero. Maximum energy product, (BH) max , is the product of remanence and coercivity, which represents the magnetostatic energy a permanent magnet material can store and is therefore a direct indicator of magnetic strength [37,38]. The higher values of B r , H c , H ci , and (BH) max suggest the stronger the magnet. It is found from Fig. 5a-c that the rubber amount remarkably influences the B r , H c , and (BH) max of the HTPB-bonded magnets. The parameters increase with decreasing the content of the rubber, which fits the general imagination. Moreover, NdFeB-M/HTPB-F 3 -DA and NdFeB-M/HTPB-DA have similar magnetic properties in the case of identical content of the magnetic particles, implying that the fluorine attached to the modified HTPB has no influence in this regard. Meantime, the values of B r , H c , and (BH) max of NdFeB/HTPB-F 3 -DA are higher than those of NdFeB-M/HTPB-F 3 -DA and NdFeB-M/HTPB-DA when the feeding amount of the particles is the same, because the NdFeB particles are not modified by MI-Silane in advance and the actual content of NdFeB particles in the former composite is higher than those in the latter (Table S7). After all, the magnetic properties of composites depend on the amount of the included magnetic powders [7,21].
Interestingly, the H ci values of NdFeB-M/HTPB-F 3 -DA, NdFeB/HTPB-F 3 -DA, and NdFeB-M/HTPB-DA nearly do not change with the loading of the magnetic particles (Fig. 5d) and keep up with that of the original NdFeB (520 kA m −1 ), which agrees with the observation of other researchers [5]. However, the H ci value of NdFeB-M/BR-DCP decreases to 502 (80 wt%) and 477 kA m −1 (85 wt%). It should be attributed to the damage of NdFeB-M [9]. Owing to the poor interaction between NdFeB-M and BR-DCP, BR-DCP fails to effectively fill up the spaces among the particles, and the particles have to directly grind each other during the compounding of the highly filled system. The more particles are involved, the more apparent the drop in H ci . The phenomenon proves the importance of improved processing flowability of the composites, which allows for not only injection molding of the rubber-bonded magnets but also for avoiding damaging the magnetic particles.
Returning to the values of B r , H c , and (BH) max of NdFeB-M/BR-DCP (Fig. 5a-c), they slightly decrease when the content of NdFeB-M increases from 80 to 85 wt%. Besides injury of the magnetic particles mentioned above, the more remarkable decrease in the actual content of NdFeB is another reason (Table S7). Because the viscosity of NdFeB-M/BR-DCP (80 wt%) is already rather high, an increase in the loading of the magnetic particles from 80 to 85 wt% would make them harder to be blended into BR-DCP.
The discussion so far reveals that NdFeB-M/HTPB-F 3 -DA has the highest magnetic properties in comparison with other composites under the same filling content. To explore the ceiling of the proposed methodology, the particle content is further raised to 90 wt%, but the composites cannot be injection molded (Figs. S13a, a′), obviously because too little rubber binder is involved. On this occasion, we decide to use magnetic particles of two sizes. That is, the 150 μm particles are introduced besides the 7 μm particles we have been using.
As shown in Fig. S13b, b′, even when the content of the particles is as high as 90 wt%, the proof-of-concept composite with the mixing ratio of NdFeB-M particles of the two sizes of 1:1 (Table 1), NdFeB-M(1:1)/HTPB-F 3 -DA, can indeed be injection molded. The smaller particles must have filled up the interstitial space between the larger ones, increasing the packing density of the composite and the effect contacts between the particles and rubber binder. The critical importance of the combination of the particles with different sizes is also revealed by Fig. S13c, c′, which indicate that the composite containing 90 wt% 150 μm particles alone cannot be injection molded. And the hybrid particles only take effect under the aegis of DA bonds (Fig. S13d, d′).
Accordingly, the magnetic properties of NdFeB-M(1:1)/ HTPB-F 3 -DA are superior to those of the other composites at the same particle contents (i.e. 80 and 85 wt%), as shown in Fig. 5 and Table  To demonstrate the capability of the NdFeB-M(1:1)/HTPB-F 3 -DA (90 wt%) magnet straightforwardly, two cylinder-like samples are injection molded and magnetized by an impulse magnetizer at 4 T field strength. The samples attract to each other, and the lower one can lift a weight of 270 g (about 77 times its own weight and equivalent to 8.43 kPa), as shown in Fig. 5e. Figure 5f compares remanence and maximum energy product of NdFeB-M(1:1)/HTPB-F 3 -DA (90 wt%) with those of the reported bonded magnets containing NdFeB by means of the enhancement ratio relative to the NdFeB powders [3][4][5][38][39][40][41][42][43][44]. Evidently, NdFeB-M(1:1)/HTPB-F 3 -DA (90 wt%) is superior to other composites, and its relative improvement of (BH) max reaches the record-high value of 52%. This should be attributed to the low viscosity during processing, which largely avoids the damage of NdFeB-M at high loading. Considering most of the bonded magnets cited in the figure are made by traditional compression molding, except for NdFeB/ 38 Page 12 of 18 polyphenylene sulfide that was prepared from the thermoplastic binder with the aid of lubricant, the manufacturing technique based on covalent adaptable composite network proposed in this work does have merits in producing highly filled polymer composites.

Mechanical properties
As revealed by Figs. 6, S14, and Table S8, Young's moduli and tensile strengths of all composites are higher than those of the respective rubber binders at the expense of elongation  (Table S2). Clearly, besides the interfacial interactions discussed above, cross-linking densities of the composites also play a critical role in static stress transfer in the small deformation range. The analysis is evidenced by the very low Young's moduli of NdFeB-M/BR-DCP because it possesses poorer interfacial interaction and lower cross-linking density.
When tensile strength is concerned, an obvious strengthening effect is observed for the composites containing HTPB-F 3 -DA and HTPB-DA. Figure 6b exhibits that the ordering of the composites' tensile strengths resembles that of Young's modulus (Fig. 6a). Because of the presence of interfacial DA bonds, NdFeB-M/HTPB-F 3 -DA and NdFeB-M/HTPB-DA have higher cross-linking densities. The networked structure also works to ensure stress transfer at the interface under larger deformation conditions. The uniform deformation of the composites (NdFeB-M/HTPB-F 3 -DA and NdFeB-M/HTPB-DA) under tension (Fig. S15a, c) supports the discussion.
Comparatively, the magnetic particles in NdFeB/HTPB-F 3 -DA cannot strengthen the HTPB-F 3 -DA networks as effectively as NdFeB-M, due to the lack of interfacial DA bonds. Nevertheless, NdFeB particles are relatively well dispersed in HTPB-F 3 -DA (Fig. 4b, d) with the help of the fluorine groups and low viscosity of the oligomers released from de-cross-linking of HTPB-F 3 -DA in the course of processing. Although stress concentration appears in the stretched specimen (Fig. S15b), as a result, the tensile strength of NdFeB/HTPB-F 3 -DA can still increase with the filler content (Fig. 6a). The tensile strength of NdFeB-M/BR-DCP is the lowest, which is attributed to the poor interaction between NdFeB-M and BR-DCP. The uneven strain mapping of the specimen of NdFeB-M/BR-DCP (Fig. S15d) must originate from the serious aggregation of the particles (Fig. S11b, d). Figures 6c and S14 show that the values of elongation at the break of all composites are lower than those of the respective rubber binders (HTPB-F 3 -DA, HTPB-DA, and BR-DCP). This follows the general regulation of particulate-filled polymer composites, as the particulate fillers restrict the deformation of the surrounding rubbers. In this case, the rather low cross-linking density of the control sample NdFeB-M/BR-DCP takes effect. The retention of its elongation at break is much higher than in other composites. On the other hand, although the elongation at break of NdFeB-M/HTPB-F 3 -DA (85 wt%) composite is lower than that of NdFeB-M/BR-DCP, it maintains good flexibility. For example, it can be bent to 90°, and recover to the original state after bending (Fig. S16a).
Just to be consistent with what we do in the last section, tensile properties of NdFeB-M(1:1)/HTPB-F 3 -DA (90 wt%) composites are measured (Figs. 6 and S14). Its Young's modulus and tensile strength are higher than those of NdFeB-M/HTPB-F 3 -DA under the same particle loading, while the elongation at break is lower than the latter. It is interesting to see that when the content of the hybrid magnetic particles is raised from 85 to 90 wt%, the tensile strength of the composite is still increased from 20.8 to 23.6 MPa. It means that with the aid of the molecular design of the rubber binder, the high packing density of the composite benefits the improvement of both magnetic and strength properties of the bonded magnets with the magnetic particles up to 90 wt%. Generally, the tensile strengths of rubber composites decrease when filler content exceeds ~ 33 wt% (Fig. S17). Compared with these results in the literature, our NdFeB-M(1:1)/HTPB-F 3 -DA (90 wt%) can include much more particulate fillers and has a decent tensile strength compared with the reported values.

Self-healability and recyclability
DA bonds cross-linked polymers are known for their selfhealability and recyclability, but it is unclear whether the smart properties can be inherited by the present materials, which only contain very few DA-bonded networks.
Furthermore, a quantitative assessment of self-healability is conducted in terms of tensile tests (Figs. 7a-c, S19, and Table S9). Similar to the above results of qualitative evaluation, tensile strengths and Young's modulus of NdFeB-M/ HTPB-F 3 -DA (85 wt%) and NdFeB-M/HTPB-DA (85 wt%) can be fully restored, which forms a sharp contrast to the incapability of NdFeB-M/BR-DCP (85 wt%) that does not contain reversible bonds. Benefiting from the DA bonds all over the composite, the healing efficiency (calculated from Eq. (S12)) of NdFeB-M(1:1)/HTPB-F 3 -DA (90 wt%) is still as high as 91.1%. After all, however, there is too little rubber to achieve a 100% repair in this case. For NdFeB/ HTPB-F 3 -DA (85 wt%), the healing efficiency is relatively lower (84.8%), as DA bonds only exist in the rubber but not at the particles/rubber interface. In addition, the values of elongation at break of the healed composites with DA bonds cross-linked polymer as the binder are higher than those of the virgin ones (Fig. 7c), which is due to the longer repair time than the injection molding time and hence reformation of DA bonds is favored. Compared with the DAbonded cross-linked particles/polymer composites reported by other groups [45][46][47], the current rubber-bonded magnets contain higher loading of particles and acquire higher healing efficiency and tensile strength at the same time.
Meanwhile, the reversible DA bonds should also allow for repossessing of the composites. The following experiments  Table S4) at the weight ratio of 1:1. The magnetic particles in other composites are all 7 μm in size prove this is the case. Figure S16b depicts that NdFeB-M/ HTPB-F 3 -DA (85 wt%) can be reshaped when the specimen is winded around glass rods at 110 °C and then cooled down. The malleability originates from the de-cross-linking/recross-linking of the material as a function of retro-DA/forward DA reactions. The irreversibly cross-linked NdFeB-M/ BR-DCP (85 wt%), by contrast, still stands straight after the same treatment. In addition, we also examine the recyclability of NdFeB-M/HTPB-F 3 -DA (85 wt%), NdFeB-M(1:1)/ HTPB-F 3 -DA (90 wt%), and NdFeB-M/HTPB-DA (85 wt%) by repeated pulverization and injection molding. That is, the original injection-molded dumbbell-shaped specimens are ground into powders and remolded three times (Fig. S20). As shown in Figs. 7d, S21, and Table S10, the recycled composites maintain not only the mechanical properties (Fig. 7d), but also the magnetic properties as the original ones (Figs. 7e, f and S22).
By taking advantage of the reversible DA bonds, the recyclability of our rubber-bonded magnets can be displayed in another way [48]. As exhibited in Fig. S23, the mixture of NdFeB-M/ HTPB-F 3 -DA (85 wt%) and 1,1,2,2-tetrachloroethane is heated at 120 ℃ for 2 h for triggering retro-DA reaction and then filtered at elevated temperature. Because of the removal of the de-cross-linked HTPB-F 3 -DA, NdFeB-M particles are isolated.