Auto-oxidation of redox electrodes for the selective recovery of platinum group metals

Abstract

The recovery and purification of platinum group metals (PGMs) from multicomponent solutions are essential for attaining a sustainable circular economy. Herein, we designed redox-active electrosorbents for the separation of PGM chloroanions by leveraging the auto-oxidation of redox electrodes. We synthesized a range of redox metallopolymers with tunable redox potentials and demonstrate their molecular selectivity in multicomponent PGM mixtures. Iridium and platinum chloroanions were shown to be capable of simultaneous auto-oxidation and binding to the redox polymers spontaneously. Thus, owing to the intrinsically high oxidation potential of the chloro-PGM complexes in leachate solutions, spontaneous electrochemical PGM recovery was possible without electrical or chemical input. As opposed to standard electrosorption, the energy consumption for iridium recovery is decreased by 75%. A combination of X-ray photoelectron spectroscopy (XPS) and ultraviolet-visible (UV-vis) spectroscopy was used to track the auto-oxidation process of the redox center and iridium chloroanion. The redox potential of ferrocene polymers was found to affect the selectivity towards PGM ions, with a high molecular selectivity of over 100 achieved between Pt and Rh. Using a porous-coated redox-polymer electrosorbent, over 186 mmol Pd uptake per mole of ferrocene was achieved in the recovery of palladium from a catalytic converter leach solution. This work demonstrates an energy-efficient, process-intensified electrochemical platform for the multicomponent recovery of PGMs from waste feedstocks.

---

Introduction

Establishing a circular economy for PGMs is essential for long-term materials sustainability, considering their core importance in the energy, chemical, and electrochemical industries.^1–4 PGMs have widespread use in the manufacturing of jewelry, automotive converters,⁵ catalysts,^6,7 pharmaceutical drugs,⁸ and fuel cells,^9,10 among others. The market price of PGMs has historically been on the rise, with that of Pd increasing from $875 USD per oz in 2017 to $2416 in 2021 (176% increase) and Rh from $1108 USD per oz in 2017 to $20 141 in 2021 (∼17 times increase), as shown in Fig. 1a.¹¹ Owing to the similar physical and chemical properties of different PGMs, separation is a significant challenge, with current industrial methods requiring several separation and purification processes^12,13 (i.e., precipitation¹⁴ and solvent extraction).^15,16 PGM mining requires a significant amount of chemicals and energy input, and these demanding processes are fraught with pollution and environmental issues.¹⁷ Therefore, the effective recovery of PGMs is necessary and has high economical potential.

Fig. 1 Market prices of PGMs, and the auto-oxidation adsorption and potential driven release between redox polymers for PGM recovery. (a) Market prices of PGMs from 2017 to 2021 (values reproduced from ref. 11). (b) Auto-oxidation of ferrocene and spontaneous adsorption of PGM ions that have higher oxidation potential, and ferrocene regeneration with the applied reducing potential. (c) [IrCl₆]²⁻ auto-oxidized ferrocene-based redox polymers and adsorbed on redox polymers as the reduced form [IrCl₆]³⁻ and oxidized redox polymers reduced by setting potential and released adsorbed [IrCl₆]³⁻.

In both mining and catalytic converter recycling, strong oxidizers are a necessity to economically leach out PGMs into solution. Thus, the resulting PGM chloro-complexes typically possess high oxidation states (Fig. 1b).^13,18–21 However, hydrometallurgy and electrodeposition methods will fully reduce PGM complexes to metallic PGMs, thereby wasting the potential energy of the PGM complexes.^13,22 Furthermore, a large input of hydrochloric acid is necessary for PGM leaching, which can be an issue due to the high energy consumption of chloralkali electrolysis (>2000 kW h ton⁻¹ Cl₂ generated).^23–25 As a result, it is critical to develop a more energy-efficient PGM recovery process that better utilizes the oxidative potential of these valuable energetic complexes.

Electrosorption has been an attractive platform employing a specific potential or current to directly control the target ion capture/release^26–30 and combines low waste generation and nearly no chemical input.^31,32 Recently, redox polymers have been extensively explored as a platform to enhance ion selectivity.^33,34 Synthetic modulation of the functional groups of redox polymers can not only change the charge-transfer behavior,^27,35 but also provide new binding sites for target ions.³⁶ For instance, the electron-donating groups next to ferrocene can tune the selectivity for heavy metal oxyanion electrosorption.²⁷ Recently, DFT studies have elucidated charge transfer as the core mechanism of ferrocene binding with platinum complexes.³⁷ However, the competitive electrosorption behavior between PGMs complexes is still unknown, which is critical to unlocking more efficient pathways to purifying PGMs, especially for sustainable mining.

Here, a reversible and spontaneous PGM adsorption system was accomplished with ferrocene polymers, by leveraging the auto-oxidation of redox couples for PGM recovery in waste recycling and mining contexts (Fig. 1c). We demonstrate that the chemical energy of high oxidation state PGMs complexes can be used for adsorption at the redox polymers, while significantly reducing the energy consumption through eliminating the energy cost of the forward electrosorption step. We also benchmark the tunable selectivity between PGMs through a combination of kinetic and equilibrium uptake studies for a range of ferrocene-containing redox-polymers, including poly(vinyl ferrocene) (PVF), poly(3-ferrocenylpropyl methacrylamide) (PFPMAm), poly(2-((1-ferrocenylethyl)(methyl)amino)ethyl methacrylate) (PFEMA), and poly(2-(methacryloyloxy)ethyl ferrocenecarboxylate) (PFcMA), to understand the interaction with PGMs ions. Through tailored functional groups (amide, alkyl, amine, ester), the redox potentials of these polymers were tuned, and as a result, their binding behavior and selectivity towards different PGMs. Moreover, X-ray photoelectron spectroscopy (XPS) and ultraviolet-visible (UV-vis) spectroscopy were used to characterize the change in oxidation state of the PGMs complexes and redox polymers. Finally, to prove our applicability for practical PGM recovery, we applied our system to the recovery of these critical elements from both mining and catalytic converter recycling streams.

Results and discussion

Auto-oxidation of redox-couples and electrode characterization

To understand the redox properties of ferrocene polymers and PGMs ions, cyclic voltammetry of polyvinylferrocene–carbon nanotube (PVF–CNT) composites and PGMs ions were carried out. Fig. 2a shows that Ir(IV) can oxidize ferrocene polymers since the redox potential of PVF–CNT (370 mV vs. Ag/AgCl) was lower than [IrCl₆]²⁻ (the standard potential of [IrCl₆]²⁻ + e⁻ → [IrCl₆]³⁻ is 0.65 V versus Ag/AgCl). Therefore, if reduced PVF–CNT is immersed in the [IrCl₆]²⁻ solution, ferrocene will be oxidized to ferrocenium, while [IrCl₆]²⁻ will be reduced to [IrCl₆]³⁻. This auto-oxidation behavior can be leveraged for PGM recovery through the spontaneous electrosorption and subsequent electrochemical release, thus requiring less energy input than a fully activated capture-and-release scheme (Fig. 1b).

Fig. 2 Redox potential and structure of ferrocene redox polymers and redox potential of [IrCl₆]²⁻. (a) Cyclic voltammetry comparing the redox potentials of polyvinylferrocene (PVF)–carbon nanotube composite (PVF–CNT) and iridium chloroanions (scan rate of 20 mV s⁻¹). The difference in the potential drives the spontaneous reaction between iridium and ferrocene. (b) Redox potentials and structures of PFPMAm, PVF, PFEMA, and PFcMA. Electrochemical potentials are presented vs. Ag/AgCl.

The effect of the redox polymer structure for PGMs electrosorption was evaluated on PFPMAm, PVF, PFEMA, and PFcMA (Fig. 2b). The polymer synthesis followed published literature methods,^35,38–40 with the detailed procedures and polymer characterization reported in the ESI.†Fig. 2b presents the structure and measured redox potential of the redox polymers, and the cyclic voltammetry of the redox polymers is shown in Fig. S1 (ESI†). Fig. S2 (ESI†) shows the scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM-EDS) image of four redox polymers with carbon nanotubes coated on carbon paper, which demonstrated that all redox polymers were uniformly coated on the carbon paper. Among these four redox polymers, PFPMAm had the lowest redox potential (250 mV vs. Ag/AgCl), followed by PVF (370 mV vs. Ag/AgCl), PFEMA (490 mV vs. Ag/AgCl), and PFcMA (600 mV vs. Ag/AgCl). The potential difference between the redox polymers was attributed to the electron-donating and electron-withdrawing effects of the functional groups attached to the cyclopentadienyl.⁴¹ The alkyl group of PFPMAm and PVF is electron-donating, stabilizing the positive charge of ferrocenium (Fc⁺), thus making ferrocene easier to oxidize.^27,42 On the other hand, the amino methyl of PFEMA and the ester of PFcMA are electron-withdrawing groups, which increase the redox-potential of the ferrocene units by making them more difficult to oxidize.^38,43,44

Structure effect of redox polymers for electrosorption

Iridium. To investigate the auto-oxidation behavior with four redox polymers, kinetic studies of the uptake capacity with iridium (1 mM [IrCl₆]²⁻ with 20 mM NaClO₄ as a supporting electrolyte) at the open circuit potential (OCP) were conducted. Iridium was selected as the model PGM for many of the studies due to its complementary redox potential to the metallopolymers, as well as its value and importance in the PGM industry. Fig. 3a shows that the adsorption uptake of iridium rapidly came to equilibrium within 15 minutes for PFEMA–CNT and PFcMA–CNT, and reached equilibrium after 1 hour for all four polymers. PFPMAm–CNT had the highest 1 hour molar iridium adsorption uptake (388 mmol mol⁻¹), followed by PFEMA–CNT (284 mmol mol⁻¹), PVF–CNT (190 mmol mol⁻¹), and PFcMA–CNT (58 mmol mol⁻¹). The concentration profile of iridium in solution during the adsorption process is shown in Fig. S3 (ESI†). The concentration of iridium dropped from 132 to 107 mg L⁻¹ for PFPMAm–CNT, from 134 to 115 mg L⁻¹ for PVF–CNT, from 123 to 107 mg L⁻¹ for PFEMA–CNT, and from 133 to 127 mg L⁻¹ for PFcMA–CNT, while it remained at 137 mg L⁻¹ for CNT only. As for mass uptake per polymer loading, PFPMAm–CNT displayed the highest uptake (219 mg g⁻¹), followed by PVF–CNT (165 mg g⁻¹), PFEMA–CNT (155 mg g⁻¹), and PFcMA–CNT (43 mg g⁻¹). Uptake results revealed that PFPMAm–CNT has 38.8% utilization, where 38.8% ferrocene of PFPMAm–CNT was spontaneous bound to iridium. Adsorption with only CNT-coated (0.2 mg) electrodes was carried out to prove that iridium uptake was driven primarily by the selective interactions of the redox polymer (Fig. S4, ESI†). Compared with PVF–CNT (∼80 mg g⁻¹-coating, 0.2 mg PVF and 0.2 mg CNT loading), the highest iridium uptake with CNT was only 20 mg g⁻¹-coating, which was less than 25% of the iridium uptake with PVF–CNT. Therefore, the iridium uptake was mostly from the redox polymers. The profile of the potential of redox polymers (Fig. 3b) shows that the potential of redox polymers increased during iridium adsorption, while the potential of the CNT control electrode remained at ∼0.8 V vs. Ag/AgCl. Furthermore, Fig. S5 (ESI†) shows that without H₂IrCl₆ (only supporting electrolyte NaClO₄), the potential of PVF–CNT can only increase to 340 mV vs. Ag/AgCl. Meanwhile, with H₂IrCl₆ present, it increased to 652 mV vs. Ag/AgCl. These results indicate that redox polymers were oxidized over time by [IrCl₆]²⁻. In the meantime, the potential of the counter electrode was decreasing (Fig. S6, ESI†), suggesting that iridium in the solution was reducing (the counter electrode was bare carbon paper, which will not react with the solution, so the potential of the counter electrode can be seen as the potential of the solution) when exposed to the redox polymer electrode. Therefore, the iridium was adsorbed via the auto-oxidation of the redox polymers and concurrent reduction of [IrCl₆]²⁻ (Fig. 3c). Furthermore, Fig. 3b corresponds to the redox potential of redox polymers in standalone cyclic voltammetry with PFPMAm–CNT. It had the lowest open circuit potential at the beginning of iridium uptake (149 mV), followed by PVF–CNT (231 mV), PFEMA–CNT (251 mV), and PFcMA–CNT (411 mV). All potentials here are versus Ag/AgCl.

Fig. 3 Electrosorption of 1 mM H₂IrCl₆ with ferrocene redox polymers–CNTs. (a) Iridium uptake profile of 1 mM H₂IrCl₆ (20 mM NaClO₄ as a supporting electrolyte) with 0.4 mg redox polymer–CNTs (0.2 mg redox polymer with 0.2 mg CNTs) at OCP. (b) Potential profile of redox polymer–CNTs and CNT control in 1 mM H₂IrCl₆ at OCP. (c) Relationship between the measured open circuit potential and iridium uptake in 1 mM H₂IrCl₆ with redox polymer–CNTs at OCP. (d) Relationship between the 1 hour iridium uptake of 1 mM H₂IrCl₆ with redox polymer–CNTs, and the redox potentials of redox polymer–CNTs; the regression line do not include PFEMA in its calculation.

Platinum. The platinum ([PtCl₆]²⁻) kinetic uptake at OCP with the four redox polymers were also carried out and was similar to iridium, as shown in Fig. S7 (ESI†). PFPMAm–CNT had the highest uptake (491 mmol mol⁻¹), followed by PFEMA–CNT (348 mmol mol⁻¹), PVF–CNT (316 mmol mol⁻¹), and PFcMA–CNT (67 mmol mol⁻¹). The potential profile (Fig. S7, ESI†) suggested that [PtCl₆]²⁻ can auto-oxidize the reduced ferrocene redox polymers as well (the standard potential of [PtCl₆]²⁻ + 2e⁻ → [PtCl₄]²⁻ + 2Cl⁻ is 0.46 V versus Ag/AgCl). Notably, the reaction [PtCl₆]²⁻ + 2e⁻ → [PtCl₄]²⁻ + 2Cl⁻ was typically the intermediate reaction for platinum deposition ([PtCl₄]²⁻ + 2 e⁻ → Pt + 4Cl⁻, E° = 0.53 V vs. Ag/AgCl). However, XPS results (Fig. S8, ESI†) showed that the adsorbed platinum was only Pt(II) and Pt(IV), meaning there was no irreversible deposition of platinum on our electrosorption platform. Hence, the regeneration step using our electrosorption platform is simpler than electrodeposition, since Pt(II) was easier to desorb compared with deposited Pt.

In general, the adsorption uptake of PGMs ions was observed to increase with decreasing redox potential of the polymers (Fig. 3d and S7†). PFPMAm–CNT had the highest uptake of iridium or platinum and lowest redox potential, while PFcMA–CNT had the lowest uptake of iridium or platinum and highest redox potential. Therefore, PFPMAm–CNT had the largest driving force for electrosorption, originating from the highest potential difference with the PGMs ions at OCP, and resulting in a more facile adsorption of PGM complexes. The difference of the redox potential of redox polymers originated from their ferrocene-neighboring functional groups.⁴¹ Specifically, the electron donating group (alkyl group) of PFPMAm and PVF stabilized the positive charge of ferrocenium, making oxidation of the ferrocene and consecutive binding of PGMs anions easier.^27,42 Nevertheless, the ester group of PFcMA is an electron-withdrawing group, making the reaction between the PGM ions and ferrocene more difficult.^38,43,44 Interestingly, PFEMA–CNT had a higher uptake of iridium (284 mmol mol⁻¹) than PVF–CNT (190 mmol mol⁻¹), even though its redox potential was higher (490 mV vs. Ag/AgCl) than PVF–CNT (370 mV vs. Ag/AgCl), as shown in Fig. 3d. The high uptake of iridium with PFEMA–CNT was caused by the methyl amino group of PFEMA–CNT functioning as an additional binding site for iridium adsorption (the nitrogen atom binds with Ir).^36,45 Furthermore, from the uptake profile in Fig. 3a, the first 2 minutes average adsorption rate of iridium with PFEMA–CNT (75.7 mmol mol⁻¹ min⁻¹) was much faster than other redox polymers (17.0 mmol mol⁻¹ min⁻¹). Moreover, from the kinetic adsorption fitting result (Fig. S9, ESI†), only PFEMA–CNT was closer to second-order adsorption, while the other three redox polymers exhibited first-order behavior.⁴⁶ The adsorption rate constant (k_ads) of PFEMA–CNT is 2.13 min⁻¹, which is 15 times higher than PFPMAm–CNT (0.074 min⁻¹), PVF–CNT (0.103 min⁻¹), and PFcMA–CNT (0.141 min⁻¹), suggesting that PFEMA had another adsorbing mechanism (methyl amino group provides additional binding sites),^36,45 compared with the other three redox polymers (charge transfer with ferrocene only).

Binary selectivity of PGMs ions during redox polymer electrosorption

To compare the performance of the redox polymers in multicomponent separations, selectivity maps of PGM uptakes in binary mixtures were created (Fig. 4a–d). Binary PGM selectivity tests were carried out at OCP with a 1 : 1 molar ratio of mixed 1 mM PGMs solution (0.5 mM for each PGMs ion) with 20 mM NaCl as the supporting electrolyte.

Fig. 4 Separation factor heat maps of 1 hour H₂IrCl₆, H₂PtCl₆, Na₃RhCl₆, K₂PdCl₄, and K₂RuCl₅(NO) binary mixtures (1 mM total PGMs (0.5 mM for each PGM ion) with 20 mM NaCl as a supporting electrolyte) electrosorption at OCP with 0.4 mg (a) PFPMAm–CNT, (b) PVF–CNT, (c) PFEMA–CNT, (d) PFcMA–CNT, respectively (0.2 mg redox polymer with 0.2 mg CNT). The white color indicates that the separation factor is 1, INF indicates that there was no adsorption for the PGM ion indicated in the column of the table. No uptake indicates that there was no adsorption for either the PGM ion in the column or the row of the table. (e) Kinetic separation factor of [IrCl₆]²⁻ over [RuCl₅(NO)]²⁻ with redox polymer–CNTs at open circuit potential adsorption. (f) Separation factor of [PtCl₆]²⁻ over [RhCl₆]³⁻ in the H₂PtCl₆ and Na₃RhCl₆ binary mixture with 20 mM NaClO₄ supporting electrolyte for 1 hour electrosorption with redox polymer–CNTs at an open circuit potential or 0.4 V, 0.8 V versus Ag/AgCl.

The binary ion-selectivity heatmaps for the separation factors were obtained by using PVF–CNT, PFPMAm–CNT, PFEMA–CNT, and PFcMA–CNT at OCP for 1 hour adsorption, with the PGMs ions at each row labeled as species A and the PGMs ions at each column labeled as species B in separation factor α_{A, B}. Fig. 4a and b shows that in the binary PGMs solution, the binding preferences of PFPMAm–CNT and PVF–CNT were Pt > Pd > Ru > Ir ≫ Rh. Meanwhile, the binding preference of PFEMA–CNT was Pd ≈ Pt > Ir > Ru ≫ Rh (Fig. 4c) and the binding preference of PFcMA–CNT was Ir > Pd > Ru > Pt > Rh (Fig. 4d). The different selectivity between the PGMs ions originates from the redox potential difference of the redox polymers. All four redox polymers had higher selectivity toward iridium in the [IrCl₆]²⁻/[RhCl₆]³⁻ mixture. Meanwhile, in the [PtCl₆]²⁻/[RhCl₆]³⁻ mixtures, the redox polymers had higher selectivity toward platinum except PFcMA–CNT, which had no uptake for both platinum and rhodium due to the high redox potential of PFcMA–CNT. This ion selectivity can be illustrated by the spontaneous reaction between the redox polymers and [PtCl₆]²⁻ or [IrCl₆]²⁻, the auto-oxidation redox couple. [PtCl₆]²⁻ or [IrCl₆]²⁻ can auto-oxidize the ferrocene moiety of the redox polymers and bind with the oxidized ferrocene moiety of the redox polymers after the charge transfer. However, there was no reaction between [RhCl₆]³⁻ and the redox polymers at OCP, making [PtCl₆]²⁻ and [IrCl₆]²⁻ be adsorbed onto the electrode more easily compared with [RhCl₆]³⁻, resulting in the difference in the adsorption uptake.

As for the [PtCl₆]²⁻/[IrCl₆]²⁻ mixture, PFPMAm–CNT, PVF–CNT, and PFEMA–CNT showed higher selectivity toward platinum (α_Pt,Ir ranged from 1.17 to 1.54), while PFcMA–CNT showed higher selectivity toward iridium (α_Pt,Ir is 0.76). The difference of the selectivity originated from the different redox potential of the redox polymers. In detail, PFcMA–CNT possessed the highest redox potential (0.6 V vs. Ag/AgCl), and the standard reduction potential of [PtCl₆]²⁻ (0.46 V vs. Ag/AgCl) was smaller than the redox potential of PFcMA–CNT, while the standard reduction potential of [IrCl₆]²⁻ was 0.65 V vs. Ag/AgCl, which was higher than the redox potential of PFcMA–CNT. Therefore, the interaction between [IrCl₆]²⁻ and PFcMA–CNT will be more favorable compared with [PtCl₆]²⁻. These results provided a guideline for the selective redox polymer design. We can design different electron-donating/withdrawing groups next to the redox active sites of redox polymers to have a specific redox potential for selectively capturing PGMs anions. The high redox potential of PFcMA–CNT also illustrated why there was no uptake for platinum and rhodium in the [PtCl₆]²⁻/[RhCl₆]³⁻ mixtures. The redox potential of PFcMA–CNT was higher compared to [PtCl₆]²⁻.

For the [IrCl₆]²⁻/[RuCl₅(NO)]²⁻ mixture, PFPMAm–CNT (α_Ir,Ru = 0.25) and PVF–CNT (α_Ir,Ru = 0.33) showed higher selectivity toward ruthenium, while PFEMA–CNT (α_Ir,Ru = 1.46) and PFcMA–CNT (α_Ir,Ru = 1.73) showed the opposite trend. This can be attributed to PFEMA and PFcMA having higher redox potentials compared to PFPMAm and PVF. Therefore, PFEMA and PFcMA had higher selectivity of [IrCl₆]²⁻ rather than [RuCl₅(NO)]²⁻. After tracking the kinetics of adsorption (Fig. 4e), α_Ir,Ru decreased over time (α_Ir,Ru of PFEMA–CNT decreased from 1.77 at 30 minutes to 1.25 at 120 minutes, and PFcMA–CNT decreased from 1.94 at 15 minutes to 1.56 at 120 minutes), which meant that the ferrocene redox polymers favored ruthenium rather than iridium as ferrocene was being oxidized (for PFcMA–CNT, the open circuit potential increased from 341 mV to 645 mV vs. Ag/AgCl in 16 minutes, and gradually decreased to 630 mV vs. Ag/AgCl after 1 hour adsorption). The decrease of α_Ir,Ru over time was because [IrCl₆]²⁻ worked as an oxidizer to oxidize the ferrocene during the OCP adsorption, while the interaction between the redox polymers and [RuCl₅(NO)]²⁻ will be stronger once ferrocene was being oxidized (in Fig. 5a, the uptake of ruthenium was five times at 0.8 V vs. Ag/AgCl than at OCP). Furthermore, in Fig. S10 (ESI†), the uptake of iridium was decreasing after an hour of OCP adsorption, while the uptake of ruthenium increased. This points to the ion-exchange between [RuCl₅(NO)]²⁻ and [IrCl₆]³⁻ over time as ferrocene was being oxidized, resulting in a dynamic α_Ir,Ru decreasing over time.

Fig. 5 (a) 1 hour PGM uptake (1 mM PGM ions with 20 mM NaClO₄ as a supporting electrolyte) with 0.4 mg PVF–CNT (0.2 mg PVF and 0.2 mg CNT) at OCP or 0.8 V versus Ag/AgCl, and regeneration efficiency of PVF–CNTs at 0.2 V versus Ag/AgCl desorption for an hour in 20 mM NaClO₄ (Ir: H₂IrCl₆, Pt: H₂PtCl₆, Rh: Na₃RhCl₆, Ru: K₂RuCl₅(NO); *palladium adsorption was tested in 1 mM K₂PdCl₄ with 20 mM NaCl supporting electrolyte). (b) 1 hour iridium uptake and regeneration efficiency of different valence states of iridium with PVF–CNTs at OCP or 0.8 V versus Ag/AgCl (1 mM iridium chloroanions with 20 mM NaClO₄ as a supporting electrolyte) with 0.4 mg PVF–CNT. (c–e) SEM images and EDS mapping of PVF–CNTs before and after 1 hour H₂IrCl₆ adsorption at OCP and after 1 hour 0.2 V versus Ag/AgCl desorption. (f–h), Fe 2p XPS of PVF–CNTs before and after 1 hour H₂IrCl₆ adsorption at OCP and after 1 hour 0.2 V versus Ag/AgCl desorption. (i and j) Ir 4f XPS of dried 1 mM K₃IrCl₆ or H₂IrCl₆ aqueous solution on carbon paper. (k–m) Ir 4f XPS of PVF–CNTs before and after 1 hour H₂IrCl₆ adsorption at OCP and after 1 hour 0.2 V versus Ag/AgCl desorption.

Separation factors under varying potentials (vs. Ag/AgCl) were also evaluated. Fig. 4f summarizes the separation factor of [PtCl₆]²⁻ over [RhCl₆]³⁻ at OCP, 0.4 V and 0.8 V vs. Ag/AgCl with all four redox polymers. For PVF–CNT and PFPMAm–CNT, α_Pt,Rh reached the maximum at 0.4 V (infinite (no Rh uptake) for PVF–CNT, 395 for PFPMAm–CNT), while α_Pt,Rh of PFEMA–CNT and PFcMA–CNT increased as the applied potential increased. α_Pt,Rh of PFEMA–CNT increased from 13.8 (OCP) to 16.9 (0.4 V) and 54.9 (0.8 V), and α_Pt,Rh of PFcMA–CNT increased from 2.15 (OCP) to 7.03 (0.4 V) and 16.7 (0.8 V). The change of α_Pt,Rh at different applied potentials can be explained by the redox potential of the redox polymers. The redox potential of PFPMAm–CNT (250 mV vs. Ag/AgCl) and PVF–CNT (370 mV vs. Ag/AgCl) was lower than 0.4 V, while the redox potential of PFEMA–CNT (490 mV vs. Ag/AgCl) and PFcMA–CNT (600 mV vs. Ag/AgCl) was higher than 0.4 V, resulting in PFPMAm–CNT and PVF–CNT being oxidized at 0.4 V vs. Ag/AgCl, while PFEMA–CNT and PFcMA–CNT were reduced. Furthermore, Fig. 5a shows that there was no uptake of rhodium at OCP, indicating that the oxidation of ferrocene was necessary for rhodium electrosorption. Therefore, for PFPMAm–CNT and PVF–CNT, although 0.4 V vs. Ag/AgCl oxidized the ferrocene, it was limited to 0.4 V, which was lower than the potential at 60 minutes OCP adsorption (0.54 V vs. Ag/AgCl for PFPMAm–CNT, 0.51 V vs. Ag/AgCl for PVF–CNT). This limited the driving force for rhodium adsorption, while it had no significant effect for platinum. Hence, α_Pt,Rh reached maximum, when 0.4 V vs. Ag/AgCl was applied. However, at 0.8 V vs. Ag/AgCl, the uptake of platinum and rhodium was enhanced (uptake of platinum increased by 3 times compared to 0.4 V vs. Ag/AgCl applied, while rhodium changed from no uptake to 47 mmol mol⁻¹ with PVF–CNT), making the adsorption between [PtCl₆]²⁻ and [RhCl₆]³⁻ comparable. As a result, α_Pt,Rh decreased at 0.8 V vs. Ag/AgCl compared to 0.4 V vs. Ag/AgCl. As for PFEMA–CNT and PFcMA–CNT, applying a 0.4 V vs. Ag/AgCl reduced both polymers, thus limiting the adsorption of [PtCl₆]²⁻. Moreover, as the applied potential increased, the uptake of [PtCl₆]²⁻ and [RhCl₆]³⁻ will be enhanced, especially for [PtCl₆]²⁻ (uptake of platinum increased by 4 times compared to 0.4 V vs. Ag/AgCl applied, while the uptake of rhodium increased by 1.8 times). Hence, α_Pt,Rh increased as the applied potential increased. Fig. 4f indicates the close relationship of the operating potential and redox potential with adsorption selectivity.

Separation performance of redox electrodes towards PGMs

Single-PGM electrosorption performance. First, the adsorption of PVF–CNT towards each individual PGM was investigated in model solutions containing 1 mM of one PGMs ion salt (H₂IrCl₆, H₂PtCl₆, Na₃RhCl₆, K₂PdCl₄, K₂RuCl₅NO) and 20 mM sodium perchlorate (NaClO₄) or sodium chloride (NaCl)-supporting electrolytes in DI water. For the desorption and regeneration, adsorbed species were released in 20 mM sodium perchlorate at 0.2 V vs. Ag/AgCl applied potential. All PGMs solutions here were acidic, and H₂PtCl₆ had the lowest pH value (2.3), followed by H₂IrCl₆ (2.9), K₂PdCl₄ (3.0), K₂RuCl₅NO (3.3), and Na₃RhCl₆ (4.8). The uptake and regeneration efficiency of PGMs ions with PVF–CNT (Fig. 5a) showed that PVF–CNT had an uptake (with unit in mmol PGMs per mol ferrocene moiety) of 163 for iridium, 197 for platinum, 6.7 for rhodium, 180 for palladium, and 41 for ruthenium at OCP for an hour and had high regeneration efficiency with iridium (79%) and ruthenium (81%) after 1 hour desorption. Fig. 5c–e shows the SEM-EDS images of iridium on the PVF–CNT-coated electrode, confirming the adsorption and release of iridium. Notably, the regeneration efficiency of palladium was around zero, suggesting that palladium is deposited on redox polymers after adsorption, resulting in some Pd nanoparticles on the surface of the electrode (Fig. S11, ESI†).

To enhance the electrosorption performance, a potential was applied to oxidize PVF–CNT. In Fig. 5a, after applying 0.8 V vs. Ag/AgCl, the uptake was increased by 38% for iridium, 18% for platinum, 1125% for rhodium, and 149% for ruthenium. Platinum showed the highest uptake (252 mmol mol⁻¹), followed by iridium, ruthenium, and rhodium (249, 216, 170 mmol mol⁻¹, respectively). Furthermore, the regeneration efficiency did not decrease, and was 92% for iridium, 83% for ruthenium, 58% for rhodium, and 54% for platinum.

Tracking the oxidation state of the redox electrode. The Fe 2p and Ir 4f XPS results (reference to C–C C 1s at 284.8 eV) illustrate the charge transfer binding mechanism. Before adsorption, as shown in Fig. 5f, Fe 2p displayed strong spin-orbiting of 2p_3/2 and 2p_1/2 at binding energies of 707.7 eV and 720.5 eV for unoxidized ferrocene, respectively, meaning that the prepared PVF–CNT electrode only had the reduced form (Fc).⁴⁷ Nevertheless, after adsorption with iridium, the signal of the Fe 2p_3/2 and 2p_1/2 transition at 707.7 eV and 720.5 eV decreased and new peaks were generated at 709.9 eV and 723.1 eV, indicating that ferrocene was oxidized (Fc → Fc⁺ + e⁻), as shown in Fig. 5g. These results agreed with the previously reported Fc⁺ binding energies.⁴⁸ Specifically, 82.8% of ferrocene in the PVF–CNT got oxidized after 1 hour [IrCl₆]²⁻ adsorption at OCP. The peaks of Ir 4f_7/2 and Ir 4f_5/2 appeared at binding energies of 61.9 eV and 64.9 eV, respectively.⁴⁹ The quantitative deconvolution of Ir 4f_5/2 and Ir 4f_7/2 was carried out by using H₂Ir(IV)Cl₆ and K₃Ir(III)Cl₆ standards (Fig. 5i and j). Binding energies at 61.8 eV and 64.7 eV were assigned to Ir(III), while binding energies at 62.8 eV and 65.8 eV were specified as Ir(IV). Fig. 5l shows the Ir 4f XPS spectra of PVF–CNT after 1 hour iridium adsorption at OCP, and the deconvolution results show that there was no Ir(IV) but only Ir(III) on PVF–CNT. This indicated that during H₂Ir(IV)Cl₆ adsorption, Ir(IV) in the solution got reduced by ferrocene and became Ir(III) to bind with PVF–CNT ([IrCl₆]²⁻ + Fc → [IrCl₆]³⁻Fc⁺). When a reducing potential (0.2 V vs. Ag/AgCl) was applied to regenerate, ferrocenium was reduced back to ferrocene and released Ir(III) at the same time, where no iridium was deposited (Fig. 1c and 5m).

Tracking the oxidation state of PGMs ions in solution. The UV-vis spectra (Fig. S12, ESI†) of the releasing solution show no Ir(IV), indicating that all the released iridium was Ir(III),⁵⁰ which corresponded to the XPS results. Therefore, we propose that the binding mechanism between [IrCl₆]²⁻ or [PtCl₆]²⁻ and PVF–CNT is from the reduction–oxidation reaction of ferrocene and PGM ions (Fig. 1c). The kinetic release of iridium with PVF–CNT at 0.2 V versus Ag/AgCl (Fig. S13, ESI†) showed that the release of iridium reached equilibrium after 10 minutes. The multiple adsorption and release cycles of H₂IrCl₆ with PVF–CNT (Fig. S14, ESI†) showed that the uptake of iridium was stable at the first three cycles (around 140 mmol mol⁻¹), with some of the ferrocene units of PVF–CNT staying in the oxidized form after 0.2 V versus Ag/AgCl was applied, as shown in Fig. 5h. Furthermore, SEM-EDS and XPS results indicated some iridium residue on the electrodes after desorption (Fig. 5e and m), limiting the uptake for the next adsorption. Therefore, the uptake slightly dropped over cycles. However, the regeneration efficiency of each cycle was over 90%, indicating that there was no significant loss in capacity, and that the redox-metallopolymers can be reusable adsorbents for PGMs ions.

To justify our hypothesis for the binding mechanism, 1 mM K₃Ir(III)Cl₆ electrosorption tests with PVF–CNT were conducted (Fig. 5b). Compared with [Ir(IV)Cl₆]²⁻, the uptake of [Ir(III)Cl₆]³⁻ was significantly lower at OCP (43 mmol mol⁻¹, Fig. 5b), and the potential profile of [IrCl₆]³⁻ electrosorption only increased from 250 mV to 350 mV versus Ag/AgCl after an hour at OCP (Fig. S15b, ESI†). This meant that ferrocene was only slightly oxidized. However, the uptake of [IrCl₆]³⁻ when 0.8 V versus Ag/AgCl was applied was 3.5 times than that at OCP and comparable to the uptake of oxidized [IrCl₆]²⁻ at OCP, indicating the oxidation of ferrocene is essential for the electrosorption of iridium. Therefore, we proposed that the oxidation state of PGMs ions affected the electrosorption behavior of the ferrocene redox polymers in a way that PGMs ions, which can auto-oxidize ferrocene, had a higher uptake.

To further understand the spontaneous reaction between the PGMs ions and redox polymers, electrochemical quartz crystal microbalance (EQCM) measurements were carried out. EQCM can determine the mass change of the electrode by measuring the frequency changes under potential control.^51–53 In Fig. S16 and S17 (ESI†), the mass of the working electrode with the PVF coating increased as the potential of PVF increased during OCP electrosorption, which meant that [PtCl₆]²⁻ and [IrCl₆]²⁻ spontaneously oxidized PVF and were adsorbed. We obtained similar results from the kinetic electrosorption tests (Fig. 3a, b and S7†). Fig. S17† shows three uptake and release cycles (uptake: OCP for 1 hour, release: 0.2 V versus Ag/AgCl for 10 minutes) of [PtCl₆]²⁻ with PVF, indicating that PVF can release adsorbed [PtCl₆]²⁻ back into solution when 0.2 V versus Ag/AgCl potential was applied. Furthermore, the plot of the mass change of the PVF coating electrode versus potential for [IrCl₆]²⁻ adsorption (Fig. S16, ESI†) demonstrates classic Nernstian behavior, indicating that [IrCl₆]²⁻ was adsorbed onto PVF right after the charge transfer,⁵⁴ which corresponded to the XPS results.

Mass loading effect of electrodes. The mass loading of electrode materials can be critical to their practical applications.⁵⁵ To optimize the redox polymer loading for electrosorption, different mass loadings (0.2, 0.4, 1, 2, 4, unit in mg-PVF per cm⁻²) of PVF–CNT (1 : 1 in mass ratio) were tested for iridium adsorption at OCP and 0.8 V vs. Ag/AgCl applied potential, as shown in Fig. S18 (ESI†). Fig. S18a† shows that at 0.8 V vs. Ag/AgCl applied potential, the total uptake of iridium can be enhanced up to 90.4 μg when using 1.0 mg cm⁻² PVF loading, which was 97% more than the total uptake of 0.2 mg cm⁻² PVF loading. However, the ferrocene stoichiometric utilization decreased by 60% at these higher loadings, indicating that the iridium uptake did not increase linearly with the PVF loading. The adsorption by auto-oxidation is limited mostly at the surface of the polymers, with mass-transfer limitations at higher loadings, which limited the spontaneous charging. This can be seen by the lower open circuit potential at higher mass loadings of ferrocene (Fig. S18b†). Compared with 0.2 mg cm⁻² PVF loading (652 mV vs. Ag/AgCl), the OCP of 0.4 mg cm⁻² PVF was only 527 mV vs. Ag/AgCl after 1 hour iridium adsorption, followed by 1.0 mg cm⁻² PVF (409 mV vs. Ag/AgCl), 2.0 mg cm⁻² PVF (375 mV vs. Ag/AgCl), and 4.0 mg cm⁻² PVF (362 mV vs. Ag/AgCl). At the same time, the iridium uptake increased by 3 times by applying 0.8 V vs. Ag/AgCl compared to OCP for iridium adsorption. This increase is due to the higher fraction of ferrocene that was oxidized. As such, at higher redox polymer loadings, applying a potential to oxidize the electrodes or ensuring better dispersion of redox polymers are critical factors to enhance the redox-mediated adsorption of the PGMs chloroanions.

Recovery of PGMs from simulated mining streams and recycling of a catalytic converter

Hydrochloric acid is commonly used in the PGMs mining industry to leach PGMs from the ores,¹³ and chloride (Cl⁻) has been reported to degrade ferrocenium,⁵⁶ as shown in Fig. S19 (ESI†). A solution of 20 mM sodium chloride or hydrochloric acid was used as the supporting electrolyte to test the effect of chloride for the PGMs ions electrosorption with PVF–CNT. Results showed that OCP operation with PVF–CNT was applicable in a solution containing chloride (Fig. S20, ESI†). In fact, as shown in Fig. S20a (ESI†), iridium can be adsorbed by PVF–CNT in 20 mM sodium chloride, where the uptake of iridium was 466 mmol mol⁻¹ ferrocene, even higher than the uptake of iridium in 20 mM sodium perchlorate (312 mmol mol⁻¹ ferrocene). Furthermore, the regeneration efficiency of PVF–CNT after iridium electrosorption in 20 mM sodium chloride remained the same as that for electrosorption in 20 mM sodium perchlorate (>80%), shown in Fig. S20b (ESI†). As for hydrochloric acid, the uptake of iridium in 20 mM hydrochloric acid was 395 mmol mol⁻¹ ferrocene, demonstrating that PVF–CNT can be used for iridium electrosorption at OCP in an acidic chloride environment. However, because of the high acidic environment from hydrochloric acid, the regeneration efficiency of PVF–CNT after iridium adsorption in 20 mM hydrochloric acid was only 70% (Fig. S20b, ESI†). We observed that PVF–CNT was not leached by Cl⁻ during the iridium adsorption at OCP, [IrCl₆]²⁻ simultaneously oxidized ferrocene and became [IrCl₆]³⁻ to bind to ferrocenium right after the charge transfer – thus resulting in high selectivity towards iridium chloroanions over Cl⁻. These results demonstrated that redox polymers have the potential to be applied in the PGMs mining industry using the auto-oxidation redox couple in chloride media.

To demonstrate real-world applicability, our PVF–CNT electrosorption system was used to selectively capture PGMs from an automotive catalytic converter (Toyota 2014 Scion Tc). The catalytic converter was crushed into small pieces and leached with HCl and HClO, forming an aqueous leach solution containing 51.44 ppm Pd, 7.47 ppm Rh, 1.98 ppm Ru, and 0.23 ppm Ir (Fig. 6b and Table S1†). Adsorption with PVF–CNT electrodes was carried out directly in the leach solution at OCP, and with an applied 0.8 V versus Ag/AgCl. Pd yielded the highest molar uptake at 186 mmol mol⁻¹. Despite the low concentration of iridium in the leach solution (0.23 ppm), PVF–CNT electrodes were able to adsorb 30.0% of iridium with an applied potential of 0.8 V, and 39.1% of iridium was adsorbed from the leach solution with no applied potential (OCP), as shown in Fig. 6c.

Fig. 6 PGM recycling from catalytic converters using redox-mediated electrosorption. (a) Catalytic converter (2014 Scion Tc was purchased new from Toyota) disassembled and digested by HCl and HClO. (b) Element composition of the catalytic converter digested by HCl and HClO. (c) Electrochemical recovery platform and removal percentage of PGMs and other elements at open circuit potential or 0.8 V applied potential versus Ag/AgCl with PVF–CNTs from the leaching of the catalytic converter.

The uptake performance was similar between OCP and the applied potential experiments for all elements tested, shown in Fig. 6c. However, compared with 0.8 V vs. Ag/AgCl applied potential, the removal of Pd was enhanced at OCP by 14% and Ir was enhanced at OCP by 30%. Similar to previous results, the measured potential response of the open circuit adsorption experiment suggested that the initially reduced ferrocene sites on the electrode surface were progressively oxidized, going from an initial potential of 0.28 V to 0.38 V (Fig. S21, ESI†). Furthermore, the leach solution was simultaneously reduced, shown as a decrease in potential from 0.86 V to 0.5 V, indicating that the PVF–CNT electrode was oxidized by the leach solution spontaneously without any energy input. Comparatively, when a constant 0.8 V vs. Ag/AgCl was applied to the PVF–CNT electrode, 150 J g⁻¹-PVF of energy was consumed to oxidize the PVF–CNT electrode in the catalytic converter leach solution (Fig. S22, ESI†). Our results showed that the PVF–CNT electrodes resulted in the selective and energy efficient capture of PGMs from the catalytic converter leach solution.

Energy consumption of PGMs adsorption with PVF–CNT

In conventional hydrometallurgical methods, the chemical energy from PGM ions is not utilized and lost since it uses chemicals to fully reduce PGMs.¹³ However, the auto-oxidation redox couple utilized the wasted chemical energy of PGMs (oxidizing potential) for selective capture, while simultaneously reconstituting the oxidizing leachates during PGMs release, thus cutting down the energy cost since no extra energy was needed in the adsorption step. Specifically, only the electrochemical desorption step required energy input, as the adsorption step leveraged the spontaneous redox-reactions to adsorb the metal chloroanions. Fig. 7 shows the energy consumption of the processes when comparing the OCP or 0.8 V vs. Ag/AgCl adsorption, coupled 0.2 V vs. Ag/AgCl desorption with PVF-CNT. Results showed that although the energy consumption of desorption is slightly higher after OCP adsorption compared to 0.8 V adsorption, OCP adsorption did not need energy input in the adsorption step. Thus, in total, the OCP adsorption can save around 75% for Ir recovery, 68% for Rh recovery, and 20% for Pt recovery compared to the 0.8 V adsorption. These results indicated that OCP adsorption of PGMs complexes with ferrocene polymers was a significantly more energy-effective method compared with applying a potential.

Fig. 7 Energy consumption of PGM complexes with PVF–CNT at OCP or 0.8 V vs. Ag/AgCl adsorption and 0.2 V vs. Ag/AgCl desorption.

Energy saving with PVF–CNT in chloralkali electrolysis

To prove that adsorption with redox polymers is more energy-efficient compared to traditional hydrometallurgy processes, a pre-oxidized PVF–CNT electrode in chloralkali electrolysis was carried out. Results showed that using pre-oxidized PVF–CNT coated carbon paper as the cathode can save 50% of the energy consumption compared to using carbon paper without redox polymers (Fig. S23, ESI†). Specifically, in Fig. S24a (ESI†), the traditional chloralkali electrolysis generates hydrogen at the cathode (−0.8 V vs. Ag/AgCl), which needed more applied potential compared to the reduction of PVF–CNT (0.4 V vs. Ag/AgCl), while the same potential was used at the anode for generating chlorine (1.36 V vs. Ag/AgCl). Therefore, the overall cell potential of chloralkali electrolysis by using carbon paper as the cathode (generating hydrogen) will be two times that of using PVF–CNT as the cathode (reducing ferrocene), as shown in Fig. S24b (ESI†). Combining the desorption step and chloralkali electrolysis, it can not only save 49% of the energy used for generating chloride for PGMs leaching, but also reduce PVF–CNT to release the adsorbates. Hence, a more energy-efficient chloralkali electrolysis process was invented by the recyclable adsorbent PVF–CNT.

Conclusion

Our study demonstrated the energy-efficient recovery of PGMs through the synthetic tuning of redox-metallopolymers, to leverage the auto-oxidation between the redox-active electrode and the redox-active PGM chloroanions for spontaneous capture of the valuable metals. Ferrocene-based redox polymers were shown to auto-oxidize in the presence of [IrCl₆]²⁻ or [PtCl₆]²⁻, thus promoting the spontaneous and selective PGM chloroanion adsorption. The platform was applied to the recycling of valuable elements from waste catalytic converters, through the leaching and electrochemical recovery of the PGMs. The energy consumption during the spontaneous capture using the auto-oxidation redox couple (5.3 kJ g⁻¹) decreased by 75% compared to electrosorption at 0.8 V vs. Ag/AgCl (21.5 kJ g⁻¹) for Ir recovery. Tuning the polymer structure was shown to be able to modulate the selectivity and uptake of the PGMs chloroanions. Remarkably, PFcMA–CNT showed selectivity for [IrCl₆]²⁻ over [RuCl₅(NO)]²⁻ and [IrCl₆]²⁻ over [PtCl₆]²⁻, while PFPMAm–CNT and PVF–CNT showed reverse selectivity in the binary competitive adsorption at the open circuit potential. The separation factor of [PtCl₆]²⁻ over [RhCl₆]³⁻ was the highest, which can be up to 40 at the open circuit potential and increased to over 100 at 0.4 V vs. Ag/AgCl. PFPMAm–CNT had the highest Ir uptake (388 mmol mol⁻¹) because of the lowest redox potential of the polymer. The amino methyl group of PFEMA can be an additional adsorption site, resulting in a high Ir uptake (284 mmol mol⁻¹) even though the redox potential of PFEMA–CNT is high (490 mV vs. Ag/AgCl). For real world application, the ferrocene-based redox polymer can be used for recovering PGMs from automotive converter leaching streams (186 mmol Pd uptake per mole of ferrocene moiety). The auto-oxidation based binding mechanism was tracked through a combination of electrosorption and spectroscopy. Fe 2p XPS showed that the ferrocene of the redox polymer was oxidized after adsorption in [IrCl₆]²⁻ solution, and Ir 4f XPS showed that [IrCl₆]²⁻ can oxidize ferrocene and become [IrCl₆]³⁻ to bind with PVF–CNT. In summary, our work presents a promising electrochemical PGMs recovery system with less energy input that can facilitate the adsorption of PGMs ions by the potential difference. Furthermore, the selectivity between PGMs ions can be controlled by modifying the applied potential or the polymer structure. We envision that the system can be applied in industry with a combination of process optimization and further improvements of the separation factor and regeneration efficiency of the redox-polymers through continued materials design .

Experimental

Materials and synthesis

All chemicals were obtained from Sigma Aldrich, Fisher Scientific or Polysciences Inc., and used as received. Poly(3-ferrocenylpropyl.methacrylamide) (PFPMAm), poly(2-((1-ferrocenylethyl)(methyl)amino)ethyl methacrylate) (PFEMA), and poly(2-(methacryloyloxy)ethyl ferrocenecarboxylate) (PFcMA) were synthesized as reported previously (ESI†).^35,38–40 All nanostructured redox polymer/carbon nanotubes electrodes (PVF–CNT, PFPMAm–CNT, PFEMA–CNT, and PFcMA–CNT) were made by the drop-casting of a polymer ink solution (redox polymer mixed with carbon nanotube in 1 : 1 mass ratio in chloroform solution).³⁷ The addition of carbon nanotubes (CNT) is to increase the conductivity for electrochemical experiments, as well as surface area.

Chemical characterization

NMR spectra were recorded on a 500 MHz spectrometer with UI500NB. NMR sample was prepared with 10–15 mg of polymer in 700 μL solvent. GPC analysis for PFPMAm and PFEMA used the column of Tosoh (7.8 mm × 30 cm × 5 μm), and LiBr DMF was used as the eluent. Poly(methyl methacrylate) standards were used for the calibration with the flow rate of 0.6 mL min⁻¹. GPC analysis for PFcMA used the column of TSKgel GMHhr-H (7.8 mm × 30 cm × 5 μm), and THF was used as the eluent. Poly(styrene) standards were used for the calibration with the flow rate of 0.3 mL min⁻¹.

Liquid-phase analytics

The concentration of PGMs in the solutions was measured by Inductively Coupled Plasma Optical Emission Spectroscopy (5110 ICP-OES, Agilent Technologies) using an eluent of 5 wt% hydrochloric acid. Samples were diluted with 5 wt% hydrochloric acid and were analyzed with the ICP-OES in 8 replicates. The wavelengths for determining concentration of PGMs and other elements were Pt 214.424 nm, Ir 224.268 nm, Ru 267.876 nm, Pd 340.458 nm, Rh 343.488 nm, Fe 238.204 nm, La 333.749 nm, Sm 359.259 nm, Sc 361.383 nm, Y 371.029 nm, Pr 417.939 nm, and Ce 418.659 nm.

Electrochemical characterization and separation

Cyclic voltammetry (CV) was performed at 20 mV s⁻¹ scan rate to evaluate the reversibility of the redox-copolymer and the redox potential. A three-electrode system, with redox polymer–CNT coated carbon paper as the working electrode and carbon paper as the counter electrode and Ag/AgCl (3 M KCl) as a reference, was used. The measurements were performed in 2 mL of 20 mM NaClO₄ in aqueous solution.

Electrosorption/release and separation factor tests were conducted with 3D-printed electrochemical cells (Fig. S25, ESI†) as previously reported.³⁷ The 3-D printed electrochemical cells were constructed with polypropylene with a layer thickness of 0.1 mm and parallel spacing of 1 cm² for working and counter electrode with a hole for kinetic sampling. All electrosorption/release and separation factor tests were performed with a 3 × 1 cm redox polymer–CNT working electrode (1 × 1 cm coated), a 3 × 1 cm plain carbon paper counter electrode, an Ag/AgCl reference electrode, and a magnetic stir bar. For electrosorption/release tests, to remove any surface impurities and make sure all redox electrodes were fully reduced, five cycles of cyclic voltammetry from −0.2 to 0.8 V at 20 mV s⁻¹ were run in 20 mM NaClO₄ solution and stopped at zero current at the end of the fifth cycle with a BioLogic SP-200 single-channel potentiostat before the electrosorption tests. After that, redox electrodes were transferred to another 3-D printed cell containing 2 mL of 1 mM PGMs anions (H₂IrCl₆, K₃IrCl₆, H₂PtCl₆, Na₃RhCl₆, K₂PdCl₄, K₂Ru(NO)Cl₅, Sigma-Aldrich) and 20 mM NaClO₄ or NaCl (for K₂PdCl₄ only) in an analytical solution, unless otherwise specified. A supporting electrolyte was used to simulate competing anions from waste streams and leachates, and to maintain a more stable conductivity regime for both the adsorption and desorption measurements. An open circuit potential or 0.8 V versus Ag/AgCl was applied onto the polymer–CNT electrode for 1 hour for electrosorption, unless otherwise specified. Regeneration of the polymer–CNT redox electrode was carried out by applying 0.2 V versus Ag/AgCl for 1 hour in a clean 20 mM NaClO₄ solution. For kinetic adsorption/release tests, 50 μL aliquots of the solution were retrieved for analysis at different time points.

Separation factors (α) were calculated by the equation below:

α_A,B = N_A,ads/C_A,sol × C_B,sol/N_B,ads

where N_A,ads is the uptake of species A in molar quantity, N_B,ads is the uptake of species B in molar quantity, C_A,sol is the concentration of species A in the remaining solution, and C_B,sol is the concentration of species B in the remaining solution. If the separation factor is higher than 1, it means that a redox polymer has higher selectivity toward species A rather than B (indicated by red colored square); if the separation factor is lower than 1, it means that a redox polymer has higher selectivity toward species B rather than A (indicated by blue colored square). The calculation of the uptake, regeneration efficiency, and energy consumption can be found in ‘ESI I: Experimental procedures, ESI’.†

Chloralkali electrolysis used the same setup as the electrosorption tests. For the control chloralkali electrolysis, carbon paper was set as both working and counter electrodes, with an applied constant current of −100 μA for 15 minutes in 25 wt% NaCl aqueous solution. For chloroalkali electrolysis with PVF–CNT, PVF–CNT was pre-oxidized at 0.8 V versus Ag/AgCl in NaClO₄ for 10 minutes, and then transferred to 25 wt% NaCl aqueous solution with an applied −100 μA constant current for 15 minutes.

In situ electrosorption was carried out using an Electrochemical quartz crystal microbalance (EQCM, BioLogic BluQCM QSD (QSD-TCU)) to measure the frequency change with Au-coating 5 MHz quartz crystal, with a piezo electroactive area of 0.2 cm² (diameter: 14 mm, polished finish, AW-R5CUP, BioLogic) working electrode, platinum wire counter, and Ag/AgCl (in 3 M NaCl) reference electrode. The working electrode was spun-coated (2000 rpm for 1 min with an acceleration of 1000 rpm) with 50 μL PVF solution (7.5 mg mL⁻¹ in chloroform), and 1 mM PGMs analyte solution was added to the electrochemical cell before analyzing. The mass change was determined by Sauerbrey equation. The calculation detail can be found in ‘ESI I: Experimental procedures, ESI’.†

Catalytic converter recycling

A new catalytic converter from a 2014 Scion Tc was purchased from Toyota, and the internal PGM-coated catalyst material was removed from the stainless-steel tubing with a grinding wheel. 663 g of catalyst material was recovered from the catalytic converter, which was then finely ground with mortar and pestle. In a typical digestion, 1 g of crushed catalyst material was added to 25 mL of 38% hydrochloric acid. Chlorine gas was generated in situ by adding 5 mL of a 9% sodium hypochlorite solution. The vessel was sealed shut with Teflon tape and left to stir for 24 hours. The solution was filtered and evaporated at 40 °C until only 1 mL of solution remained to remove excess chlorine gas and HCl. Finally, 24 mL of DI water was added to the 1 mL concentrate for a final catalyst digestion solution. The solution was analyzed with ICP-OES, and its composition is shown in Table S1 (ESI†). Adsorption experiments were carried out in a similar manner as previous tests: a PVF–CNT electrode with 0.2 mg of PVF and a 1 × 1 cm area was placed in 1.5 mL of digested catalyst solution, along with the Ag/AgCl reference and carbon paper counter electrodes. The cell was operated either at an open circuit potential or at a constant 0.8 V vs. Ag/AgCl for 1 hour.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was primarily supported by the NSF DMREF grant #2323988. This material is based upon work partially supported by the U.S. Department of Energy, Office of Basic Energy Sciences under Award Number DOE DE-SC0021409. We also thank the University of Illinois Urbana-Champaign and the School of Chemical Sciences for their support through startup funds. SEM was carried out in the Frederick Seitz Materials Research Laboratory Central Research Facilities, University of Illinois. The authors thank YiCheng Rong for his help with GPC characterization.

Notes and references

1. U.S. Department of Energy, Critical Materials Assessment 2023, accessed August 2023, https://www.energy.gov/sites/default/files/2023-07/doe-critical-material-assessment_07312023.pdf.
2. M. K. Debe, Nature, 2012, 486, 43–51.
3. M. F. Li, Z. P. Zhao, T. Cheng, A. Fortunelli, C. Y. Chen, R. Yu, Q. H. Zhang, L. Gu, B. V. Merinov, Z. Y. Lin, E. B. Zhu, T. Yu, Q. Y. Jia, J. H. Guo, L. Zhang, W. A. Goddard, Y. Huang and X. F. Duan, Science, 2016, 354, 1414–1419.
4. S. Park, Y. Y. Shao, J. Liu and Y. Wang, Energy Environ. Sci., 2012, 5, 9331–9344.
5. J. Kaspar, P. Fornasiero and N. Hickey, Catal. Today, 2003, 77, 419–449.
6. E. Antolini, ACS Catal., 2014, 4, 1426–1440.
7. R. Lang, T. B. Li, D. Matsumura, S. Miao, Y. J. Ren, Y. T. Cui, Y. Tan, B. T. Qiao, L. Li, A. Q. Wang, X. D. Wang and T. Zhang, Angew. Chem., Int. Ed., 2016, 55, 16054–16058.
8. C. Y. Zhang, C. Xu, X. Y. Gao and Q. Q. Yao, Theranostics, 2022, 12, 2115–2132.
9. E. Antolini, Energy Environ. Sci., 2009, 2, 915–931.
10. R. R. Adzic, J. Zhang, K. Sasaki, M. B. Vukmirovic, M. Shao, J. X. Wang, A. U. Nilekar, M. Mavrikakis, J. A. Valerio and F. Uribe, Top. Catal., 2007, 46, 249–262.
11. Johnson Matthey 2022 PGM market report, accessed January 2023, https://matthey.com/pgm-market-report-2022.
12. P. Sinisalo and M. Lundstrom, Metals, 2018, 8, 12.
13. F. Crundwell, M. Moats, V. Ramachandran, T. Robinson and W. G. Davenport, Extractive Metallurgy of Nickel, Cobalt and Platinum Group Metals, Elsevier, 2011.
14. B. Raju, J. R. Kumar, J. Y. Lee, H. S. Kwonc, M. L. Kantam and B. R. Reddy, J. Hazard. Mater., 2012, 227, 142–147.
15. M. Rzelewska-Piekut and M. Regel-Rosocka, Sep. Purif. Technol., 2019, 212, 791–801.
16. T. H. Nguyen, C. H. Sonu and M. S. Lee, Hydrometallurgy, 2016, 164, 71–77.
17. B. J. Glaister and G. M. Mudd, Miner. Eng., 2010, 23, 438–450.
18. S. Harjanto, Y. Cao, A. Shibayama, I. Naitoh, T. Nanami, K. Kasahara, Y. Okumura, K. Liu and T. Fujita, Mater. Trans., 2006, 47, 129–135.
19. M. Fan, S. L. Li, H. Deng, X. G. Zhang, G. T. Luo, Z. J. Huang and M. H. Chen, Sep. Purif. Technol., 2022, 289, 11.
20. S. Ilyas, R. R. Srivastava, H. Kim and H. A. Cheema, Sep. Purif. Technol., 2020, 248, 8.
21. A. M. Yousif, J. Chem., 2019, 2019, 7.
22. G. Nicol, E. Goosey, D. S. Yildiz, E. Loving, V. T. Nguyen, S. Riano, I. Yakoumis, A. M. Martinez, A. Siriwardana, A. Unzurrunzaga, J. Spooren, T. A. Atia, B. Michielsen, X. Dominguez-Benetton and O. Lanaridi, Johnson Matthey Technol. Rev., 2021, 65, 127–147.
23. L. N. Zhang, R. Li, H. Y. Zang, H. Q. Tan, Z. H. Kang, Y. H. Wang and Y. G. Li, Energy Environ. Sci., 2021, 14, 6191–6210.
24. S. Dresp, F. Dionigi, M. Klingenhof and P. Strasser, ACS Energy Lett., 2019, 4, 933–942.
25. I. A. Moreno-Hernandez, B. S. Brunschwig and N. S. Lewis, Energy Environ. Sci., 2019, 12, 1241–1248.
26. D. I. Kim, G. Gwak, P. Dorji, D. He, S. Phuntsho, S. Hong and H. Shon, ACS Sustain. Chem. Eng., 2018, 6, 1692–1701.
27. R. L. Chen, J. Y. Feng, J. Jeon, T. Sheehan, C. Ruttiger, M. Gallei, D. Shukla and X. Su, Adv. Funct. Mater., 2021, 31, 11.
28. M. E. Suss, S. Porada, X. Sun, P. M. Biesheuvel, J. Yoon and V. Presser, Energy Environ. Sci., 2015, 8, 2296–2319.
29. S. E. C. Sener, V. M. Thomas, D. E. Hogan, R. M. Maier, M. Carbajales-Dale, M. D. Barton, T. Karanfil, J. C. Crittenden and G. L. Amy, ACS Sustain. Chem. Eng., 2021, 9, 11616–11634.
30. G. Lota, K. Fic and E. Frackowiak, Energy Environ. Sci., 2011, 4, 1592–1605.
31. K. Kim, R. Candeago, G. Rim, D. Raymond, A. H. A. Park and X. Su, iScience, 2021, 24, 31.
32. X. Su, H. J. Kulik, T. F. Jamison and T. A. Hatton, Adv. Funct. Mater., 2016, 26, 3394–3404.
33. K. Kim, S. Cotty, J. Elbert, R. L. Chen, C. H. Hou and X. Su, Adv. Mater., 2020, 32, 8.
34. X. Su, K. J. Tan, J. Elbert, C. Ruttiger, M. Gallei, T. F. Jamison and T. A. Hatton, Energy Environ. Sci., 2017, 10, 1272–1283.
35. H. Vapnik, J. Elbert and X. Su, J. Mater. Chem. A, 2021, 9, 20068–20077.
36. T. H. Nguyen, C. H. Sonu and M. S. Lee, J. Ind. Eng. Chem., 2016, 36, 245–250.
37. S. Cotty, J. Jeon, J. Elbert, V. S. Jeyaraj, A. V. Mironenko and X. Su, Sci. Adv., 2022, 8, 12.
38. J. Jeon, J. Elbert, C. H. Chung, J. Chae and X. Su, Adv. Funct. Mater., 2023, 10,  DOI:10.1002/adfm.202301545.
39. B. Y. Kim, E. L. Ratcliff, N. R. Armstrong, T. Kowalewski and J. Pyun, Langmuir, 2010, 26, 2083–2092.
40. N. Kim, J. Elbert, C. Kim and X. Su, ACS Energy Lett., 2023, 8, 2097–2105.
41. M. L. Hu, M. Abbasi-Azad, B. Habibi, F. Rouhani, H. Moghanni-Bavil-Olyaei, K. G. Liu and A. Morsali, ChemPlusChem, 2020, 85, 2397–2418.
42. A. Paul, R. Borrelli, H. Bouyanfif, S. Gottis and F. Sauvage, ACS Omega, 2019, 4, 14780–14789.
43. M. W. Cooke, T. S. Cameron, K. N. Robertson, J. C. Swarts and M. A. S. Aquino, Organometallics, 2002, 21, 5962–5971.
44. P. O. Schwartz, S. Fortsch, E. Mena-Osteritz, D. Weirather-Kostner, M. Wachtler and P. Bauerle, RSC Adv., 2018, 8, 14193–14200.
45. A. M. Wilson, P. J. Bailey, P. A. Tasker, J. R. Turkington, R. A. Grant and J. B. Love, Chem. Soc. Rev., 2014, 43, 123–134.
46. J. L. Wang and X. Guo, J. Hazard. Mater., 2020, 390, 18.
47. C. M. Woodbridge, D. L. Pugmire, R. C. Johnson, N. M. Boag and M. A. Langell, J. Phys. Chem. B, 2000, 104, 3085–3093.
48. X. Y. Wu, Y. K. Xu, C. Zhang, D. P. Leonard, A. Markir, J. Lu and X. L. Ji, J. Am. Chem. Soc., 2019, 141, 6338–6344.
49. I. G. Casella, M. Contursi and R. Toniolo, J. Electroanal. Chem., 2015, 736, 147–152.
50. M. Schnippering, P. R. Unwin, J. Hult, T. Laurila, C. F. Kaminski, J. M. Langridge, R. L. Jones, M. Mazurenka and S. R. Mackenzie, Electrochem. Commun., 2008, 10, 1827–1830.
51. H. Hübner, R. Candeago, D. Schmitt, A. Schießer, B. Xiong, M. Gallei and X. Su, Polymer, 2022, 244, 124656.
52. R. Chen, H. Wang, M. Doucet, J. F. Browning and X. Su, JACS Au, 2023, 3, 3333–3344.
53. R. Candeago, H. Wang, M.-T. Nguyen, M. Doucet, V.-A. Glezakou, J. F. Browning and X. Su, JACS Au, 2024, 4(3), 919–929.
54. A. J. Bard, L. R. Faulkner and H. S. White, Electrochemical Methods: Fundamentals and Applications, John Wiley & Sons, 2022.
55. Y. Gogotsi and P. Simon, Science, 2011, 334, 917–918.
56. R. Prins, A. R. Korswagen and A. G. Kortbeek, J. Organomet. Chem., 1972, 39, 335–344.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ta01384k

‡ Ching-Hsiu Chung and Stephen Cotty contributed equally to this work.

---