Circular Upcycling of Bottlebrush Thermosets

Crosslinking of polymers into networks provides them with mechanical stability and dimensional integrity. These same covalent crosslinks that enhance material properties concurrently hinder the re-processing of products at their endof-life. [1][2][3][4][5] There are two general approaches to re-processing of polymers: recycling and upcycling, and both are intended to enhance sustainability while pursuing distinct conversion pathways and valorization targets. Closed-loop recycling technologies enable recreating the original material through a circular deconstruction-reconstruction, e.g., depolymerization-repolymerization process, thus creating a sustainable substitute with a similar value. [6][7][8][9] However, closed-loop recycling of crosslinked thermosets is not possible due to irreversible, covalent crosslinks. [10][11][12] To provide end-of-life solutions for these recalcitrant materials, upcycling approaches have been developed involving polymer functionalization and/or depolymerization [13][14][15][16] to convert a postconsumer product into a new material of higher value. [16][17][18][19][20][21][22][23] An ideal thermoset material would provide pathways for both upcycling and circularity, whereby a polymer network can be both reconfigured into a new product and/or reversibly disassembled to its constituent building blocks. [24,25] The built-in circularity (by integrating closedloop recycling and upcycling in one material) is difficult to realize in conventional thermosets given the irreversible crosslinks, [12,26,27] but has been successfully implemented in dynamic networks such as vitrimers, allowing full recovery of monomer feedstocks for reprocessing. [1,17,[28][29][30][31][32][33][34] This approach can be further advanced by taking advantage of brush-like polymer architectures that constitute a distinct subset of polymer networks with unique properties, such as tissue-mimetic mechanics, [35,36] enhanced swelling, [37,38] and structural coloration. [35,39] Brush-like polymers are already widely used in industrial products such as coatings and adhesives, while tissue-mimetic elastomers and gels with architecturally tunable mechanics open opportunities for the design of personalized biomedical devices. Unlike their linear counterparts, brush networks combine several advantageous characteristics: (i) the myriad of chain-end functionalities provide synthetic flexibility for both permanent and reversible crosslinking, [40][41][42] (ii) the large size of bottlebrush strands allows for scaffold percolation with much fewer number of nodes, (iii) the lack of chain entanglements lowers melt viscosity and facilitates molding, and (iv) the extra free volume of chain ends enhances mobility of reagents during post-crosslinking chemical modification. The outlined capabilities allow constructing a re-processing triangle that incorporates both reconfiguration on demand and recycling to a reusable melt of functional bottlebrush mesoblocks (Figure 1). For example, one may start with constructing Network 1 (high crosslink density) from melt, followed by its reconfiguration to Network 2 (low crosslink density). Switching from stiff rubber to a tissue-mimetic material leads to value-added applications in biomedical devices and soft robotics. Alternatively, N1 can be completely disassembled into bottlebrush melt to be re-crosslinked to N2. It is important to note that within this triangle, each product is stable for a long time and every transformation is reversible. [41,43] To demonstrate the feasibility of circular upcycling of bottlebrush networks, we use the thermoreversible Diels-Alder reactions (DA and retro-DA) through incorporation of furan (F) moieties at side chain ends and addition of bismaleimide (bis-M) crosslinkers. [13,33,[44][45][46] DA chemistry has been successfully used for recycling of linear thermosets by providing stable covalent crosslinking at room temperature and controlled dissociation at higher temperatures. [47][48][49][50][51] This reversible chemistry in conjunction with the aforementioned traits of brush architecture creates an intrinsic fluid-like environment to enable molecular mobility for reactions in a solvent-free state. Brush strands are synthesized through copolymerization of monomethacryloxypropyl terminated polydimethylsiloxane (PDMSMA, 1000 g mol À 1 ) and hydroxyl-terminated polyethylene glycol methacrylate (PEGMA, 500 g mol À 1 ) macromonomers via atom transfer radical polymerization (ATRP) (Figure 2A, see Supporting Information for synthetic details). The PEGMA fraction was kept under 10 wt.% to ensure miscibility with PDMSMA macromonomers. For a 600 : 1 monomer/initiator ratio, the degree of polymerization at 66.6 % conversion is determined by 1 H NMR as n bb ffi417, which is consistent with 0.67 × 600 = 402 and number average n bb ffi256, Ð = 1.1 and n bb ffi303, Ð = 1.12 from multi-angle light-scattering gelpermeation chromatography (MALS-GPC) and molecular imaging with atomic force microscopy (AFM), respectively (Supporting Information, Figures S1 and S14). Monitoring conversion of the two macromonomers over time confirmed random copolymerization (Supporting Information, Figure S2-3). Prior to functionalization, bottlebrush macromolecules were purified to remove residual monomers (Figure S1). The hydroxyl-terminated PEGMA ends are reacted with furfuryl isocyanate (FNCO) to produce bottlebrushes with a controlled fraction of F-functionalized side chains (2.65 mol % functionalization verified by 1 H NMR spectroscopy, Supporting Information, Figure S4). In parallel, a bis-maleimide PDMS crosslinker was synthesized to ensure miscibility with the PDMS side chains (Supporting Information, Figure S5).

Reversible assembly-disassembly of DA-crosslinked networks was monitored by measuring the storage and loss moduli as a function of temperature (T) and time (Figure 2B). At lower T, the M and F moieties react to produce a covalent bottlebrush network. The crosslinking demonstrates typical Arrhenius behavior, where the rate of the crosslinking reaction is lower at lower temperatures (Figure 2B inset). Upon heating, the network dissociates into a melt of PDMS bottlebrushes (Supplementary Videos 1 and 2). In the corresponding T sweeps, three distinct regions are identified: (i) at T < 60 C, DA coupling reactions dominate over retro-DA, indicative of stable network where the crosslink density remains invariable, (ii) at 70 < T < 130 C, the DA and retro-DA reactions are in equilibrium, resulting in softening at G 0 > G 00 , which enables network upcycling while preserving its structural integrity, and (iii) at T > 130 C, the retro-DA reaction is much faster than DA cycloaddition, which permits chemical recycling back to the bottlebrush precursor. Once the heat is removed, the DA cycloaddition converts the network to its original state (Supporting Information, Figure S6). [46,52] By varying the M/F feeding ratio, elastomers are obtained with different crosslink densities (% n À 1

x ), where n x is a backbone DP between DA crosslinks. These individual materials demonstrate the Young's modulus, E 0 , ranging from 2 to 20 kPa in tensile tests (Figure 2C and Figure S6). Elastomers made that contain a higher concentration of furan compared to maleimide (M=F < 1) show a linear increase in modulus as bis-maleimide crosslinker is added, which corresponds to the stoichiometric increase in modulus (E 0 / n À 1

x / M=F). However, in systems with a higher concentration of bis-maleimides (M/F > 1), dangling maleimides are evident and serve as side-chain extenders to decrease modulus accordingly. In this study, we consider M/ F < 1 where E 0 can be accurately controlled through crosslinker concentration (Figure 2C inset, dashed line). The corresponding crosslink densities varying from 4.7 to 18.9 mol m À 3 were deduced from the Young's modulus using the recently developed network forensics methodology (Figure S15). The synthesized networks exhibited high gel fractions of 99.9 AE 0.1 % due to (i) high crosslink efficiency of large size macromolecules and (ii) purification of brushes prior to crosslinking. However, the reported gel fraction might be overestimated because of extremely slow diffusion of bulky bottlebrushes in a polymer network with a mesh size smaller than individual macromolecules.

For closed-loop recycling of the DA-crosslinked elastomers, we exploit the ability of anthracene to react irreversibly with maleimide moieties (Figure 3A,B). To accomplish network disassembly, the crosslinked networks are preloaded with 9-anthracenemethanol and heated to 130 °C, producing a free-flowing viscoelastic fluid (Figure 3C inset). Complete recovery of the polymer melt is achieved at a 9anthracenemethanol concentration of 0.1 wt %, which represents a twofold excess of anthracene compared to furan groups (Supporting Information, Figure S7). The disassembly process is monitored my measuring the shear storage and loss moduli as a function of time at constant T = 130 °C (Figure 3C, Figure S8). The anthracene-loaded network yields melt of lower G 00 , which suggests lower molecular weight products of the disassembly process. To verify the capturing of maleimide moieties by anthracene, the melts were cooled down from 130 to 60 °C to re-initiate DA cycloaddition. As expected, the formulation without anthracene rapidly (< 180 sec) re-crosslinks to restore the original modulus value of G 0 ffi 7kPa (Figure 3D). In contrast, the anthracene-loaded formulation does not re-crosslink (G 0 < G 00 ) over a period of 15 days. As corroborated by NMR of urethane peaks (Supporting Information, Figure S9), the furan-containing polymer melts remain active and undergo DA cycloaddition with newly introduced cross-linkers (Figure 3B) to enable reassembly of the elastomer with nearly identical dynamic and equilibrium mechanical properties compared to the original elastomer (Figure 3E and F). Within the same framework, we hypothesized that it is possible to make elastomers softer or stiffer on demand through controlled addition of crosslinker scavengers (Figure 4A) or crosslinkers (Figure 4B), respectively. For elastomer softening, we prepared a series of anthracene-loaded samples with different sub-stoichiometric ratios of 9-anthracenemethanol to furan functional groups (0, 0.19, 0.38, 0.57). These elastomers remain stable at room temperature for a period of 12 days with no change in modulus (Supporting Information, Figure S10-S11). Upon heating to 80 °C, the DA-crosslinks are rendered dynamic and a subset of maleimide groups reacts irreversibly with the anthracene functionality; thus, decreasing crosslink density without compromising the overall network integrity. The stability of the reconfigured network is verified by constant G 0 and G 00 during % 16 hrs after decrosslinking (Supporting Information, Figure S11). Tensile experiments determine that the decrease in Young's modulus is precisely controlled by the relative concentration of anthracene (Figure 4C). For example, a M/F = 0.76 network loaded with 0.19 eq of anthracene became 25 % (0.19/0.76) softer compared with the original

elastomer: E 0;0:19 ¼ 0:75E 0;0:76 ¼ 0:75 Á 18:9 ¼ 14:2 kPa (Fig- ure 4D).

For elastomer stiffening, the DA crosslinked brush elastomers are pre-loaded with different concentrations of furan protected bis-maleimides as inert, dormant crosslinkers (Figure 4B and Figure S12). Heating to 80 °C results in deprotection of maleimides through retro-DA and simultaneous evaporation of furan (b p % 31 °C), which enables the activated crosslinkers to react with the F-functionalized side chains. The reconfigured networks demonstrate a linear increase in crosslink density with the corresponding stoichiometric addition of crosslinker (Figure 4E,F and Supporting Information Figures S12-S13