Bimetallic synergy in supported Ni–Pd catalyst for selective hydrogenolysis of C–O bonds in epoxy resins

Abstract

Recycling of epoxy composites is of importance for achieving circular economy as demand for lightweight materials in the field of sustainable technologies is soaring. Although catalytic hydrogenolysis of epoxy resins provides a promising approach to recover valuable fillers and phenolic compounds from the composites, there is a lack of a reusable solid catalyst for this purpose. Here, we report a robust CeO₂-supported Ni–Pd bimetallic catalyst (Ni–Pd/CeO₂) for the hydrogenolysis of C–O bonds in epoxy resins under 1 atm of H₂. Benefiting from its heterogeneous nature, Ni–Pd/CeO₂ can be reused for several times. Furthermore, the catalyst is successfully applied to decomposition of epoxy composites to recover carbon or glass fibers and phenolic compounds, implying the potential application of our catalyst system toward recycling of epoxy composites.

Hydrogenolysis of epoxy resins is a promising method for recycling epoxy composites. Here, a CeO₂-supported Ni–Pd bimetallic catalyst is developed for facilitating such a transformation benefiting from the synergy between the two metals.

Introduction

The growing global concerns about plastic waste contamination have urged our society to develop effective ways to recycle end-of-life plastics^1–4. Epoxy resins are widely applied in various areas such as construction, electronics, aircraft, automobiles, and wind turbine blades typically in the form of fiber-reinforced epoxy composites (Fig. 1a)^5–7. Different from thermoplastics such as polyethylene terephthalate, thermosetting epoxy composites are not suitable for mechanical recycling due to the deterioration of material properties, and thus, most of the composites are landfilled, which not only wastes resources but also causes serious environmental problems^8–11. As the demand for epoxy composites is soaring with the increasing demand for lightweight materials in the field of sustainable technologies such as electric vehicles and wind power plants, the development of efficient approaches to recycle epoxy composites is getting more and more attention for achieving circular economy and carbon neutralization^5–11.

Fig. 1. Outline of this work.

a Structure of typical epoxy resin composites and traditional methods for their decomposition. b Previous works: hydrogenolysis of epoxy resins using homogeneous Ru or Ni catalysts with phosphine ligands. c This work: hydrogenolysis of epoxy resins by a reusable Ni–Pd/CeO₂ catalyst enabled by the synergy between Ni and Pd.

To date, a wide range of methods have been developed for the decomposition of epoxy composites (Fig. 1a)^8–11. For example, pyrolysis under energy-intensive high-temperature conditions is an approach to recover fillers from the composites, which often causes damage to the fillers due to the harsh conditions¹². Other methods, such as oxidative decomposition^13–15 using oxidants (e.g., H₂O₂) or solvolysis¹⁶ in concentrated acids (e.g., nitric acid) or bases (e.g., NaOH) have also been developed to deconstruct epoxy resin metrics. However, they have shortcomings of using the undesirable oxidants or a large excess amount of acids or bases which would increase the environmental burden. In addition, all the above methods mainly focus on recovering fillers in the composites, and epoxy resin building blocks such as bisphenol A (BPA) are difficult to recover. In this context, chemical recycling of epoxy resins is of importance to recover valuable fillers together with phenolic compounds from the epoxy composites, but highly challenging due to their inert nature. Recently, it has been revealed that BPA can be recovered from epoxy resins by using a super-stoichiometric amount of potassium or sodium t-butoxide (4 equiv. with respect to BPA unit)^17,18 or NaOH (6 equiv. with respect to BPA unit)¹⁹. However, tedious work-up procedures using a large quantity of acids are necessary to recover BPA because the primary products are potassium or sodium phenolates.

Quite recently, catalytic reductive decomposition has emerged as a promising method to recover fillers and phenolics from epoxy composites (Fig. 1b)^20–22. For example, Ahrens and Skrydstrup have reported a homogeneous Ru/triphos (triphos = 1,1,1-tris(diphenylphosphinomethyl)ethane) catalyst for decomposition of epoxy resin composites to recover fibers and BPA using 2-propanol as the hydrogen source^20,21. Concurrently, our group has developed a homogeneous Ni/dcype (dcype = 1,2-bis(dicyclohexylphosphino)ethane) catalyst for hydrogenolysis of epoxy resins to recover BPA using H₂ as the reductant²². In spite of the efficiency of these catalyst systems for recovering BPA and fibers from epoxy composites, they suffer from difficulties in catalyst recovery and reuse due to their homogeneous nature. Thus, homogeneous catalysts are not practical for the decomposition of large-scale epoxy composites especially when noble metal catalysts and/or expensive ligands are employed. Therefore, the development of a robust and reusable heterogeneous catalyst for hydrogenolysis of epoxy resins to recycle epoxy composites is highly desirable from a practical point of view. However, as far as we know^8–22, there is a lack of a heterogeneous catalyst for decomposition of epoxy composites to recover phenolics and fillers.

Considering most epoxy resins utilized nowadays are prepared by curing the BPA-based epoxide prepolymer with amines or acid anhydrides (Fig. 1a)^5–7,23,24, the key to recover BPA is the selective hydrogenolysis of C(sp³)–O bonds in the alkyl phenyl ether moiety without over reduction of aromatic rings^25–36. To achieve this purpose, we considered exploring bimetallic catalysts with the expectation that synergy between two metals could substantially enhance the catalytic performance for the selective hydrogenolysis of C(sp³)–O bonds in epoxy resin backbone compared to the corresponding single metallic catalysts^37–48. Here, for the first time, we successfully developed a reusable CeO₂-supported Ni–Pd bimetallic catalyst (Ni–Pd/CeO₂) for the hydrogenolysis of epoxy resins under 1 atm of H₂ (Fig. 1c). Mechanistic studies suggested that Pd induces the formation of Ni(0) species to facilitate dehydrogenation of alcohol moieties in epoxy resin backbone and that the selective cleavage of C(sp³)–O over C(sp²)–O bond is mainly promoted by Pd species. By using the present catalyst system, we demonstrate the mild decomposition of epoxy composites, including carbon fiber-reinforced plastics (CFRPs) and circuit boards, to recover carbon or glass fibers and phenolic compounds, implying its potential application to recycle epoxy composites.

Results and discussion

Catalyst development

Initially, the Ni–Pd/CeO₂ catalyst with the Ni to Pd ratio of 1/1 (Ni₁Pd₁/CeO₂) was prepared by deposition-precipitation method followed by treatment with 1 atm of H₂ at 150 °C for 0.5 h (the supported metal catalysts are designated as M_xM’_y/support, where the ratio of M to M’ is x/y; see Supplementary Information for the details on the catalyst preparation). Then, Ni₁Pd₁/CeO₂ was applied to the hydrogenolysis of epoxy resin model 1 in N-methylpyrrolidone (NMP) at 180 °C and under 1 atm of H₂. In this case, BPA (2), 4-isopropylphenol (3) and phenol (4) were obtained in 76%, 16% and 14% yields, respectively (Table 1, entry 1). When the reaction time was extended to 48 h, further cleavage of C–C bond proceeded and the yields of 3 and 4 increased to 81% and 59%, respectively (Table 1, entry 2). Other bimetallic catalysts such as Cu₁Pd₁/CeO₂, Co₁Pd₁/CeO₂, Fe₁Pd₁/CeO₂, Mn₁Pd₁/CeO₂, Ni₁Pt₁/CeO₂, and Fe₁Pt₁/CeO₂ showed much lower activity than Ni₁Pd₁/CeO₂ (Table 1, entries 3–8). Interestingly, the aromatic rings of phenolic products were not hydrogenated. The conversion of 1 and the yields of phenolic products gradually increased with increasing the amount of the supported Pd species (Table 1, entries 1, 9, and 10). However, a further increase resulted in a decrease in the catalytic activity (Table 1, entry 11). In contrast, there is no substantial change in the conversion or yields when the amount of Ni increased or decreased (Table 1, entries 1, 12–14). The single metallic Ni/CeO₂ or Pd/CeO₂ was much less effective than the Ni–Pd bimetallic catalysts for the hydrogenolysis and gave significantly lower yields of the corresponding phenolic products (Table 1, entries 15 and 16). Ni–Pd supported on other supports, such as TiO₂, Al₂O₃, or ZrO₂, resulted in much lower conversion and yields than Ni₁Pd₁/CeO₂ (Table 1, entries 17–19). Furthermore, a physical mixture of Ni/CeO₂ and Pd/CeO₂ resulted in higher activity than Ni/CeO₂ or Pd/CeO₂ alone (Table 1, entry 20 vs 15 or 16), but still gave much lower conversion and yields than Ni₁Pd₁/CeO₂ (Table 1, entry 20 vs 1). Therefore, only if Ni together with Pd were directly supported on CeO₂, the activity for the hydrogenolysis increased dramatically, and the synergy between Ni and Pd is crucial for the high activity of Ni₁Pd₁/CeO₂. As a control experiment, CeO₂ did not promote the hydrogenolysis of 1 (Table 1, entry 21). Solvent effects on the hydrogenolysis of 1 were also investigated. Among various solvents examined such as NMP, cyrene, γ-valerolactone, 1,3-dimethyl-2-imidazolidinone, and triglyme, NMP resulted in the highest conversion and yields (Supplementary Table 2).

Table 1.

Effect of catalysts on the hydrogenolysis of epoxy resin model 1

Entrya	Catalyst	Conv. of 1 (%)	Yield (%)
			2	3	4
1b	Ni1Pd1/CeO2	95	76	16	14
2c	Ni1Pd1/CeO2	99	16	81	59
3	Cu1Pd1/CeO2	25	19	1	n.d.
4	Co1Pd1/CeO2	41	25	12	11
5	Fe1Pd1/CeO2	45	21	n.d.	n.d.
6	Mn1Pd1/CeO2	49	38	3	3
7	Ni1Pt1/CeO2	17	14	n.d.	n.d.
8	Fe1Pt1/CeO2	22	6	n.d.	n.d.
9	Ni1Pd0.1/CeO2	28	12	n.d.	n.d.
10	Ni1Pd0.5/CeO2	51	36	2	n.d.
11	Ni1Pd4/CeO2	15	10	3	2
12b	Ni0.5Pd1/CeO2	95	73	14	12
13b	Ni2Pd1/CeO2	96	76	13	8
14b	Ni4Pd1/CeO2	97	66	16	12
15d	Ni/CeO2	11	1	n.d.	n.d.
16e	Pd/CeO2	18	9	2	1
17	Ni1Pd1/TiO2	25	6	n.d.	n.d.
18	Ni1Pd1/Al2O3	11	4	n.d.	n.d.
19	Ni1Pd1/ZrO2	5	1	n.d.	n.d.
20d,e	Ni/CeO2 + Pd/CeO2	49	29	4	4
21f	CeO2	3	n.d.	n.d.	n.d.

^aReaction conditions: 1 (200 mg, 0.52 mmol BPA unit), M_xM’_y/CeO₂ (100 mg; M ≈ 4.2x mol%; M’ ≈ 4.8y mol%, the exact amount of the metal used in the hydrogenolysis is varied slightly depending on the kind of supported metal species), NMP (2.0 mL), 180 °C, H₂ (1 atm, balloon), 12 h. All the supported metal catalysts were pre-treated with 1 atm of H₂ at 150 °C for 0.5 h. Conversion of 1 and yield of 4 were determined by ¹H NMR analysis. Yields of 2 and 3 were determined by GC analysis. ^bAn average value of two parallel experiments is shown here, and standard deviations are <4% for all data points, see Supplementary Table 1 for details. n.d. = not detected.

^c48 h.

^dNi/CeO₂ (100 mg, Ni = 4.0 mol%).

^ePd/CeO₂ (100 mg, Pd = 5.0 mol%).

^fCeO₂ (100 mg).

Characterization of the catalyst

The Ni₁Pd₁/CeO₂ catalyst was characterized by powder X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), X-ray absorption fine structure (XAFS) spectroscopy, and high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) analysis. XRD pattern of the catalyst revealed that the structure of CeO₂ was maintained well upon immobilization of Ni and Pd species and the subsequent treatment with H₂ gas (Fig. 2a). Furthermore, no apparent diffraction peaks attributable to Pd, Ni, or NiO were observed, indicating Ni and Pd species are highly dispersed on CeO₂ surface (Fig. 2a). From the X-ray photoelectron spectroscopy (XPS) analysis of Ni₁Pd₁/CeO₂, the supported Ni–Pd alloy nanoparticles contain Pd²⁺ (336.8 and 342.2 eV), Pd⁰ (335.2 and 340.5 eV), Ni²⁺ (855.4 eV), and Ni⁰ (852.4 eV) species, and the ratios of Pd²⁺ to Pd⁰ and Ni²⁺ to Ni⁰ were 17/83 and 79/21, respectively (Fig. 2b, c). Pd K-edge X-ray absorption near edge structure (XANES) spectrum of Ni₁Pd₁/CeO₂ showed that the supported Pd is mainly composed of Pd⁰ species (Fig. 2d). From Ni K-edge XANES spectrum, it is revealed that the catalyst mainly contains Ni²⁺ species (Fig. 2e). Therefore, the results obtained from XANES spectra are in good agreement with those from XPS spectra. HAADF-STEM and EDS analysis of the catalyst showed that Ni and Pd species are located on the same position of the catalyst surface, and the average particle size was approximately 2.6 nm (Fig. 2f, g).

Fig. 2. Characterization of Ni₁Pd₁/CeO₂.

a XRD patterns of (i) Ni₁Pd₁/CeO₂, (ii) CeO₂, (iii) the powder diffraction file of Pd metal (file No. 8796), (iv) the powder diffraction file of Ni metal (file No. 8783), (v) the powder diffraction file of NiO (file No. 5898), and (vi) the powder diffraction file of CeO₂ (file No. 11). b XPS spectrum of Ni₁Pd₁/CeO₂ in the region of 332–344 eV (Pd 3d). c XPS spectrum of Ni₁Pd₁/CeO₂ in the region of 849–867 eV (Ni 2p). The black line indicates the original spectrum, the blue, red, and orange broken lines indicate the deconvoluted signals and the green broken line indicates the sum of the deconvoluted signals. d Pd K-edge XANES spectrum. e Ni K-edge XANES spectrum. f STEM image and particle size distribution of Ni₁Pd₁/CeO₂ (average = 2.6 nm; standard deviation, σ = 0.8 nm, n = 256). g HAADF-STEM image and elemental mapping of Ni₁Pd₁/CeO₂ by EDS analysis. For the above analyses, Ni₁Pd₁/CeO₂ was pre-treated with 1 atm of H₂ at 150 °C for 0.5 h.

Hydrogenolysis of other epoxy resin model compounds by Ni₁Pd₁/CeO₂

As shown in Fig. 3, Ni₁Pd₁/CeO₂ can efficiently promote the hydrogenolysis of various epoxy model compounds. For model compounds 5–7 having hydroxy group adjacent to the ether bonds, the hydrogenolysis proceeded smoothly under 1 atm of H₂, giving the corresponding hydrogenolysis product 9 in high yields (Fig. 3). When the hydrogenolysis of 5 was carried out in the presence of tributylamine, substantial amounts of the C–C bond-cleaved products such as 3 (45%), 4 (13%), and anisole (29%) were obtained (Supplementary Fig. 1). Therefore, the amine moiety in model 1 is likely the reason for the formation of 3 and 4 for the hydrogenolysis of 1 (Table 1, entry 1).

Fig. 3. Hydrogenolysis of various epoxy resin model compounds by Ni₁Pd₁/CeO₂.

Reaction conditions: model 5 or 8 (0.25 mmol), Ni₁Pd₁/CeO₂ (100 mg), NMP (2.0 mL), 180 °C, H₂ (1 atm, balloon), 9 h. For model 6 or 7, 0.50 mmol of the substrate was used. The catalyst was pre-treated with 1 atm of H₂ at 150 °C for 0.5 h. The yields of 9 are shown here for the hydrogenolysis of each model compound under H₂ or N₂. The yields were determined by GC analysis.

The hydrogenolysis of models 5–7 also proceeded under N₂ with lower yields of 9 produced, indicating the transfer hydrogenolysis proceeded for which the hydroxy groups would serve as the hydrogen source. In addition, the hydrogenolysis of the ether without the hydroxy group did not proceed at all, suggesting that the hydroxy group is necessary for the efficient hydrogenolysis of the ether bonds (Fig. 3, model 8). Nevertheless, we found that the hydrogenolysis of model 8 proceeded to some extent by adding a base such as K₃PO₄, though the conversion was much lower than other models having the hydroxy group (Supplementary Fig. 2). The effect of the hydroxy group on the hydrogenolysis is discussed in the following section.

Mechanistic studies

The reaction pathway for the present hydrogenolysis was investigated. As mentioned above, the hydrogenolysis of models 5–7 proceeded even under N₂ and model 8 did not react at all (Fig. 3). Therefore, it is likely that the hydrogenolysis proceeds through dehydrogenation of the alcohol moiety to form the corresponding ketone intermediate (Fig. 4a, step 1) followed by hydrogenolysis of the C–O bond adjacent to the carbonyl group (Fig. 4a, step 2). The proposed reaction pathway was further supported by the hydrogenolysis of models 10 and 11 (Fig. 4b). When the hydrogenolysis of 10 was carried out at 180 °C and under 1 atm of H₂ for 6 h, full conversion of 10 gave 9, 2-tetradecanone (12), and 2-tetradecanol (13) in 91%, 80%, and 14% yield, respectively (Fig. 4b (i), entry 1). Notably, when Ni/CeO₂ was used as the catalyst, the dehydrogenated ketone intermediate 11 was detected (Fig. 4b (i), entry 3). In addition, the hydrogenolysis of 11 using Ni₁Pd₁/CeO₂ proceeded smoothly (Fig. 4b (ii), entry 1). Therefore, the reaction proceeds through the ketone intermediate as shown in Fig. 4a. Also, the formation of 10 during the hydrogenolysis of 11 suggests that an equilibrium between 10 and 11 exists via dehydrogenation and hydrogenation (Fig. 4b (ii), entries 1 and 3).

Fig. 4. Mechanistic studies.

a Proposed reaction pathway for the hydrogenolysis. b Hydrogenolysis of models 10 and 11 using Ni₁Pd₁/CeO₂, Pd/CeO₂, or Ni/CeO₂. c The hydrogenolysis of model 5 using Ni₁Pd₁/CeO₂ pre-treated at different temperatures. (i) reaction profiles (each data point represents an average value of two parallel experiments, and standard deviations were added as error bars), (ii) XPS spectra of the catalysts in Pd 3d region, and (iii) XPS spectra of the catalysts in Ni 2p region. The black line indicates the original spectrum, the blue, red, and orange broken lines indicate the deconvoluted signals and the green broken line indicates the sum of the deconvoluted signals.

Next, the role of the supported Ni and Pd species for the hydrogenolysis was investigated. For the hydrogenolysis of 11 using catalysts pre-treated with 1 atm H₂ at 300 °C, Pd/CeO₂ resulted in almost the same conversion and yields as Ni₁Pd₁/CeO₂ (Fig. 4b (ii), entry 1 vs entry 2), but is much more active than Ni/CeO₂ (Fig. 4b (ii), entry 2 vs entry 3). Even for Ni/CeO₂ pre-treated at 400 °C, which has almost the same ratio of Ni⁰ species (53%, Supplementary Fig. 3) as Ni₁Pd₁/CeO₂ pre-treated at 300 °C (51%, Fig. 4c (iii)), the activity was still much lower than Pd/CeO₂ (Fig. 4b (ii), entry 2 vs entry 4). These results indicate that the supported Pd species are highly active and mainly responsible for the hydrogenolysis of the ketone intermediate (Fig. 4a, step 2). On the other hand, Pd/CeO₂ showed much lower activity than Ni₁Pd₁/CeO₂ for the hydrogenolysis of 10 (Fig. 4b (i), entry 1 vs entry 2), which suggests the Pd species alone have lower efficiency for the dehydrogenation step, and Ni together with Pd substantially promote the dehydrogenation of 10 to 11 (Fig. 4a, step 1).

For the hydrogenolysis of model 5, an induction period was observed (Fig. 4c (i), and Supplementary Fig. 4). To investigate the origin of the induction period, we pre-treated Ni₁Pd₁/CeO₂ with 1 atm of H₂ at different temperatures, and the hydrogenolysis of 5 was carried out using these catalysts. The reaction profiles showed that the induction period was decreased with an increase in the pre-treatment temperature (Fig. 4c (i)). XPS analysis of the catalysts showed that the ratio of Pd²⁺ to Pd⁰ was approximately the same when the catalyst was treated at different temperatures (Fig. 4c (ii)). However, the ratio of Ni²⁺ to Ni⁰ decreased from 79/21 to 49/51, indicating a substantially larger amount of Ni²⁺ species was reduced to Ni⁰ at higher pre-treatment temperature (Fig. 4c (iii)). Therefore, the supported Ni species rather than Pd species are related to the induction period, and the higher amount of the surface Ni⁰ species resulted in a shorter induction period. This result was further supported by the following experiments. When Ni₁Pd₁/CeO₂ pre-treated with H₂ at 150 °C was subjected to the hydrogenolysis of 5 and recovered after 2 h, the ratio of Ni²⁺ to Ni⁰ was decreased from 79/21 to 57/43, and the ratio was kept almost unchanged until the end of the reaction (61/39), supporting that Ni²⁺ was reduced to Ni⁰ during the induction period (Supplementary Fig. 5).

In our bimetallic catalyst system, Pd induces the reduction of Ni²⁺ to Ni⁰. When Ni/CeO₂ was reduced with 1 atm of H₂ at 300 °C for 1 h, the ratio of Ni²⁺ to Ni⁰ was 73/27 (Supplementary Fig. 6). On the other hand, the ratio was 49/51 for Ni₁Pd₁/CeO₂ reduced under the same conditions, indicating a larger amount of Ni²⁺ was reduced to Ni⁰ in the presence of Pd. As the supported Pd species rather than Ni species play the major role in the hydrogenolysis of the C–O bond in the ketone intermediate 11, the reduction of Ni²⁺ to Ni⁰ is indispensable for promoting the dehydrogenation of the alcohol moiety. Overall, the Pd-induced reduction of Ni²⁺ to Ni⁰ enhances the rate of the dehydrogenation step shown in Fig. 4a. Consequently, the bimetallic Ni₁Pd₁/CeO₂ is much more active than the corresponding single metallic catalysts for the hydrogenolysis of the β-hydroxy ether moieties in epoxy resins.

Hydrogenolysis of epoxy resins

The applicability of our catalyst system was demonstrated by the hydrogenolysis of epoxy resins and the decomposition of epoxy composites. As shown in Fig. 5, Ni₁Pd₁/CeO₂ was successfully applied to the hydrogenolysis of acid anhydride- or amine-cured epoxy resins. For the hydrogenolysis of the BPA-based epoxy resin cured with 4-methylhexahydrophthalic anhydride (MHHPA), a catalytic amount of 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) was added to promote decomposition⁴⁹, and BPA was obtained in 52% yield (Fig. 5a). The hydrogenolysis of the BPA-based epoxy resin cured with 4,4’-methylenebis(cyclohexylamine) (MBCHA) also proceeded efficiently in the presence of a catalytic amount of K₃PO₄ (Fig. 5b). In this case, C–C bond cleavage occurred to give 4-isopropylphenol and phenol in 99% and 81% yield, respectively. As described above, the aliphatic tertiary amine moiety in the epoxy resin is likely responsible for the C–C bond cleavage. Considering the hydroxy groups in the resin are often involved in crosslinking, the role of K₃PO₄ in our catalyst system is likely to promote the hydrogenolysis of the ether bond even without the β-hydroxy group, which is supported by the experimental result that hydrogenolysis of model 8 indeed proceeded in the presence of K₃PO₄ (Supplementary Fig. 2). The hydrogenolysis of commercial BPA-based epoxy resins cured with amines other than MBCHA (Supplementary Table 3) and bisguaiacol F (BGF)-based epoxy resin cured with MBCHA (Supplementary Fig. 7) also proceeded to give the corresponding phenolic products. Scale-up of the hydrogenolysis was also successful. As shown in Fig. 5a, b, the corresponding phenolic products were formed in yields comparable to the small-scale hydrogenolysis, and the products were successfully isolated using silica gel column chromatography. Moreover, the present catalyst system was also applicable to the bisphenol S (BPS)-based epoxy resin, giving BPS in 39% isolated yield (Fig. 5c). Control experiments using base additives in the absence of Ni₁Pd₁/CeO₂ resulted in the formation of much lower yields of phenolic products (<15% yields, Supplementary Table 4), indicating Ni₁Pd₁/CeO₂ is indispensable for the hydrogenolysis.

Fig. 5. Ni₁Pd₁/CeO₂-catalyzed hydrogenolysis of epoxy resins.

a Hydrogenolysis of the BPA-based epoxy resin cured with MHHPA. Reaction conditions (0.52 mmol scale): resin (chunks, 333 mg, 0.52 mmol BPA unit), Ni₁Pd₁/CeO₂ (100 mg), DBU (0.125 mmol), NMP (1.0 mL), 180 °C, H₂ (1 atm, balloon), 7 d. Reaction conditions (3.1 mmol scale): resin (chunks, 2.0 g, 3.1 mmol BPA unit), Ni₁Pd₁/CeO₂ (500 mg), DBU (1.0 mmol), NMP (5.0 mL), 180 °C, H₂ (1 atm, balloon), 7 d. b Hydrogenolysis of the BPA-based epoxy resin cured with MBCHA. Reaction conditions (0.50 mmol scale): resin (powder, 223 mg, 0.50 mmol BPA unit), Ni₁Pd₁/CeO₂ (100 mg), K₃PO₄ (0.125 mmol), NMP (1.0 mL), 180 °C, H₂ (1 atm, balloon), 3 d. Reaction conditions (5.5 mmol scale): resin (chunks, 2.4 g, 5.5 mmol BPA unit), Ni₁Pd₁/CeO₂ (500 mg), K₃PO₄ (0.60 mmol), NMP (5.0 mL), 180 °C, H₂ (1 atm, balloon), 3 d. c Hydrogenolysis of the BPS-based epoxy resin cured with MBCHA. Reaction conditions: resin (chunks, 286 mg, 0.50 mmol BPS unit), Ni₁Pd₁/CeO₂ (100 mg), K₃PO₄ (0.125 mmol), NMP (1.0 mL), 180 °C, H₂ (1 atm, balloon), 3 d. For all the above experiments, Ni₁Pd₁/CeO₂ was pre-treated with 1 atm of H₂ at 300 °C for 1.0 h. Yields of BPA, 4-isopropylphenol, and phenol were determined by GC analysis, and BPS by ¹H NMR. Isolated yields are shown in parentheses. Pictures of epoxy resin substrates are also shown. Synthetic procedures for the epoxy resins are shown in the Supplementary methods.

Decomposition of epoxy composites

The present catalyst system can be applied to the decomposition of epoxy composites. For example, CFRPs composed of carbon fibers and epoxy resins cured with methylcyclohexene-1,2-dicarboxylic anhydride (CFRP_MCDA) or dicyandiamide (CFRP_DICY) were successfully decomposed, and carbon fibers and phenolics were recovered from the CFRPs (Fig. 6a, b). The recovered carbon fibers were analyzed by scanning electron microscope (SEM), which revealed that the carbon fiber surface was clean after the decomposition (Fig. 6a, b). Furthermore, decomposition of a circuit board composed of epoxy resins and glass fibers was also successful, and glass fibers and phenolic products 3 and 4 were recovered (Fig. 6c). From SEM analysis of the recovered fibers, it was revealed that the fiber surface was quite clean (Fig. 6c). For the decomposition of CRFPs or the circuit board, although decomposition proceeded using base in the absence of the catalyst, the yields of recovered phenolics were much lower (Supplementary Fig. 8). In addition, although the surface of fibers recovered from CFRP_MCDA was clean, those from CFRP_DICY and the circuit board were not. Therefore, Ni₁Pd₁/CeO₂ is necessary for recovering phenolics and/or clean fibers.

Fig. 6. Ni₁Pd₁/CeO₂-promoted decomposition of epoxy composites.

a Decomposition of CFRP_MCDA, and a SEM image of the carbon fiber recovered after the decomposition experiment. Reaction conditions: CFRP_MCDA (200 mg), Ni₁Pd₁/CeO₂ (100 mg), DBU (0.30 mmol), NMP (2.0 mL), 180 °C, H₂ (1 atm, balloon), 3 d. b Decomposition of CFRP_DICY, and a SEM image of the carbon fiber recovered after the decomposition experiment. Reaction conditions: CFRP_DICY (300 mg), Ni₁Pd₁/CeO₂ (100 mg), K₃PO₄ (0.50 mmol), NMP (2.0 mL), 180 °C, H₂ (1 atm, balloon), 3 d. c Decomposition of a circuit board (epoxy composite with glass fibers), and a SEM image of the glass fiber recovered after the decomposition experiment. Reaction conditions: circuit board piece (426 mg), Ni₁Pd₁/CeO₂ (100 mg), K₃PO₄ (0.50 mmol), NMP (2.0 mL), 180 °C, H₂ (1 atm, balloon), 3 d. For all the above experiments, Ni₁Pd₁/CeO₂ was pre-treated with 1 atm of H₂ at 300 °C for 1.0 h. Yields of the phenolic products were determined by GC analysis. Pictures of epoxy composites and recovered fibers or metal parts are shown.

Verification of heterogeneous nature and reuse experiment of Ni₁Pd₁/CeO₂

The heterogeneous nature of the present catalyst system was investigated as follows. For the hydrogenolysis of 5 under the conditions shown in Fig. 7, the reaction completely stopped when the catalyst was removed by hot filtration at the conversion of approximately 60% (Fig. 7a). In addition, when the filtrate obtained after the hydrogenolysis was analyzed by inductively coupled plasma optical emission spectroscopy (ICP-OES), Ni and Pd species were hardly detected (below the detection limit). Therefore, the Ni₁Pd₁/CeO₂ catalyst worked as a heterogeneous catalyst and the reaction proceeded on the catalyst surface⁵⁰.

Fig. 7. Leaching test and reuse of Ni₁Pd₁/CeO₂.

a Effect of removal of the catalyst for the hydrogenolysis of model 5. The filled squares indicate the yield of 9 without removal of the catalyst, and the open squares after removal of the catalyst by hot filtration. b Reuse experiments. After the reaction, the catalyst was retrieved by filtration, washed with acetone, water, and ethanol, dried at room temperature, and then applied to each reuse experiment. Reaction conditions: 5 (0.25 mmol), Ni₁Pd₁/CeO₂ (100 mg), NMP (2.0 mL), 180 °C, H₂ (1 atm, balloon), 7 h. For each experiment, the catalyst was pre-treated with 1 atm of H₂ at 300 °C for 1.0 h. The conversion and yields were determined by ¹H NMR and GC analysis, respectively. Each data point represents an average value of two parallel experiments, and standard deviations were added as error bars.

Then, the reusability of Ni₁Pd₁/CeO₂ was examined for the hydrogenolysis of 5. The catalyst can be easily retrieved from the reaction mixture by simple filtration with >95% recovery after the hydrogenolysis. The recovered catalyst was washed with acetone, water, and ethanol followed by drying at room temperature and reducing with 1 atm of H₂ at 300 °C for 1 h before each reuse experiment. As shown in Fig. 7b, the catalyst can be reused at least five times without significant loss of performance. The HAADF-STEM analysis of Ni₁Pd₁/CeO₂ recovered after the 5th reuse experiment revealed that the Ni–Pd alloy nanoparticles remain highly dispersed on the CeO₂ surface without significant aggregation of particles (Supplementary Fig. 9). Furthermore, by comparing the powder XRD patterns of the fresh Ni₁Pd₁/CeO₂ and that recovered after the 5th reuse experiment, the structure of the CeO₂ support remained unchanged (Supplementary Fig. 10). These results support the robustness of Ni₁Pd₁/CeO₂ for the hydrogenolysis.

Furthermore, Ni₁Pd₁/CeO₂ can be reused for the decomposition of CFRP_MCDA several times. After the decomposition experiment of CFRP_MCDA, the catalyst can be easily separated from the carbon fiber by washing with acetone and water, benefiting from the different shape of the catalyst and the fiber (powder vs fiber). The recovered catalyst was washed with acetone, water and ethanol several times, followed by calcination at 300 °C under air for 3 h and H₂ for 1 h. Following these regeneration procedures, the catalyst was again subjected to the decomposition of CFRP_MCDA. As shown in Table 2, the catalyst can be reused at least 5 times without notable loss of its performance. After the 5th reuse experiment, 96% of the catalytic performance was still preserved compared to the decomposition experiment using the fresh catalyst. These experiments demonstrate the potential applicability of our catalyst system for the recovery of carbon fiber and phenolic compounds from CFRP_MCDA.

Table 2.

Reuse of Ni₁Pd₁/CeO₂ for the decomposition of CFRP_MCDA

Entry	Catalyst	CFRP (mg)	Catalyst recovery (%)	Recovered carbon fiber (mg)	Yield of 2 (mmol)	Reusability compared to the fresh catalyst (%)
1	Fresh	200	92	106	0.087	-
2	1st reuse	192	92	106	0.080	95
3	2nd reuse	199	94	107	0.083	95
4	3rd reuse	201	96	106	0.084	97
5	4th reuse	205	92	102	0.083	93
6	5th reuse	204	93	103	0.085	96

Reaction conditions: CFRP_MCDA (approximately 200 mg), Ni₁Pd₁/CeO₂ (100 mg), DBU (0.30 mmol), NMP (2.0 mL), 180 °C, H₂ (1 atm, balloon), 3 d. The recovered catalyst was washed with acetone, water, and ethanol, and calcined at 300 °C under air for 3 h and H₂ for 1 h. The yield of 2 was determined by GC analysis. Pictures of CFRP_MCDA, and recovered fiber and catalyst are shown.

In summary, we have successfully developed the heterogeneous Ni–Pd/CeO₂ catalyst for the hydrogenolysis of epoxy resins. Notably, the catalyst can be applied to the decomposition of epoxy composites including CFRPs and the circuit board to recover phenolics and fibers. The catalyst can be recycled and reused several times even for the decomposition of the epoxy composite. Mechanistic studies suggested that the reaction proceeds through the dehydrogenation/hydrogenolysis sequences. In addition, while the Pd-induced reduction of Ni²⁺ to Ni⁰ is the key to promoting dehydrogenation of the alcohol moiety, Pd species are mainly responsible for the C–O bond hydrogenolysis. The robustness and reusability of Ni–Pd/CeO₂ for the decomposition of CFRP imply its potential application in the recycling of epoxy composites.

Methods

Instruments and reagents

Gas chromatography (GC) analyses were performed on Shimadzu GC-2014 equipped with flame ionization detector (FID) and InertCap 5, 5MS/Sil capillary column. GC mass (GC–MS) spectra were recorded on Shimadzu GCMSQP2010 equipped with an InertCap 5MS/Sil capillary column at an ionization voltage of 70 eV. Nuclear magnetic resonance (NMR) spectra were recorded on BRUKER Ascend500 (¹H: 500 MHz, ¹³C: 126 MHz) spectrometer at ambient temperature. Chemical shift values for protons were determined using tetramethylsilane (TMS) (δ = 0 ppm) as the internal reference. Chemical shift values for carbons were determined using CDCl₃ (δ = 77.2 ppm) as the internal reference. High-resolution mass spectrum (HRMS) was taken with the electron spray ionization time-of-flight (ESI-TOF) method on JEOL JMS-T100LP AccuTOF LC-plus mass spectrometer. ICP-OES analyses were performed on a ThermoFisher iCAP PRO Duo. HADDF-STEM and EDS analysis were carried out using JEOL JEM-ARM 200F Thermal FE STEM operated at 200 kV. Powder X-ray diffraction (XRD) patterns were measured on a Rigaku Miniflex 600-C (CuKα, λ = 1.5406 Å, 40 kV, 15 mA). X-ray photoelectron spectroscopy (XPS) measurements were carried out on PHI5000 VersaProbe III using Al Kα radiation (hν = 1486.6 eV, 15 kV, 1.7 mA). The binding energies were calibrated by using the C 1s signal at 284.8 eV. Ni K-edge and Pd K-edge XANES measurements were performed at the BL01B1 beamline at SPring-8, which was operated at 8 GeV. A Si(111) and Si(311) double-crystal monochromator were used for Ni K-edge and Pd K-edge XANES measurements, respectively. The XAFS spectra were obtained at room temperature in transmittance or fluorescence mode. The data was reduced using xTunes (Science & Technology Institute Co.)⁵¹. Scanning electron microscopy (SEM) observation was performed on a JEOL JSM-7500FA. ZrO₂ (BET surface area: 279 m² g⁻¹, JRC-ZRO-6), Al₂O₃ (BET surface area: 148 m² g⁻¹, JRC-ALO-8), and TiO₂ (BET surface area: 269 m² g⁻¹, JRC-TiO-14) are provided by Catalysis Society of Japan. CeO₂ (BET surface area: 111 m² g⁻¹, Aldrich, cat no. 544841-25G) is commercially available. CFRP_MCDA and CFRP_DICY were kindly supplied by Toray Industries, Inc., for which epoxy resins cured with MCDA or DICY were used, respectively. The circuit board was commercially available (KAUMO universal board, double-sided through hole type, glass epoxy composite, 2.54 mm pitch, 2 × 8 cm; the composition of epoxy resin in the composite is unknown), which was cut into small pieces and used for the decomposition experiment. Solvents, substrates, and products were purchased from Kanto Chemical, TCI, Wako, or Aldrich, and used as received.

Preparation of Ni₁Pd₁/CeO₂

First, CeO₂ (2.0 g) was added to an aqueous solution (60 mL) of NiCl₂ (8.33 mM), PdCl₂ (8.33 mM), and KCl (2.0 equiv. with respect to PdCl₂, 16.7 mM). The resulting mixture was stirred vigorously at room temperature for 10 min, followed by adjusting the pH to 10 ± 0.5 using aqueous NaOH (1.0 M). After vigorous stirring at room temperature for 24 h, the solid was filtered off, washed with water (2.0 L), dried in vacuo, and pre-treated at 150°C and under H₂ (1 atm) for 0.5 h, affording the Ni₁Pd₁/CeO₂ catalyst (1.9 g, Pd contents: 0.25 mmol g⁻¹, 2.7 wt%, Ni contents: 0. 22 mmol g⁻¹, 1.3 wt%).

Typical procedures for the hydrogenolysis reaction (e.g., model 1)

Ni₁Pd₁/CeO₂ (100 mg) was added into a Schlenk tube (volume: ca. 20 mL) connected to a balloon filled with H₂ (1 atm). Then, the catalyst was pre-treated at 150 °C for 0.5 h. Following the pre-treatment, epoxy resin model 1 (200 mg, 0.52 mmol BPA unit), NMP (2.0 mL), and a Teflon-coated magnetic stir bar were successively added into the Schlenk tube under N₂ atmosphere. The reaction mixture was degassed twice by the freeze-pump-thaw method, and the Schenk tube was connected to a balloon filled with H₂. The reaction mixture was vigorously stirred at 180 °C for 12 h. After the reaction was completed, internal standards (1,1,2,2-tetrachloroethane and dodecane) were added to the reaction mixture. Conversion of 1 and yield of 4 were determined by ¹H NMR analysis using 1,1,2,2-tetrachloroethane as the internal standard, and yields of 2 and 3 were determined by GC analysis using dodecane as the internal standard.

Author contributions

X.J. and K.Nozaki designed the studies and conceived the main idea. Y.H. and Y.Y. executed all the experimental works except for the XAFS measurements. K.Nomoto, H.M., and T.S. measured and analyzed the XAFS spectra. All authors discussed the results and wrote the paper.

Peer review

Peer review information

Nature Communications thanks Alexander Ahrens and Yao Fu for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data supporting the findings of this study are available within this article and its Supplementary Information file or from the authors upon reasonable request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-56488-4.