Chiral dirhodium carboxylates are highly effective catalysts for synthetically useful transformations of donor/acceptor carbenes. 1,2 Most notably, they are capable of highly enantioselective cyclopropanations 3,4 and display exceptional control of regio-, diastereo-and enantioselectivity in C-H functionalization reactions. 2,5,6 Due to the versatility of the donor/acceptor carbenes, these chiral dirhodium catalysts have been utilized in many synthetic routes for the preparation of pharmaceutically relevant chiral intermediates, 7 and cyclopropanation chemistry has been applied to three different scale-up campaigns. [7][8][9] Although dirhodium carboxylate catalysts are broadly useful, a major concern with their use is the high and fluctuating cost of rhodium, leading to unpredictable production costs. As a result, effective recovery of the catalyst is a desirable and requisite practice for stable production processes. 10 Furthermore, catalyst recovery has the added benefit of reducing contamination by residual metals in the final product or the waste streams. 10 One approach for managing the cost of rhodium is to carry out the reactions with very low catalyst loadings. 8,11 A complementary approach is the recovery of these precious metal catalysts through simple separations enabled by immobilization and/or heterogenization. This approach enables reuse of the rhodium in multiple ways, either by direct catalyst recycling or via high efficiency metal recovery and reprocessing.
Immobilization of chiral dirhodium catalysts has been extensively examined and several different approaches have been explored. The most common strategy is to covalently attach the dirhodium catalysts onto solid supports such as porous silica, [12][13][14][15] monoliths, 16 cellulose, 17 hydroxyl resins 18,19 or poly(styrene). 20 This approach has been generally effective at recovery of the catalysts with minimal leaching, 12,15 but does have a drawback as incorporation of the covalent attachments can require significant synthetic effort and the modification can influence the performance of the chiral scaffold. 15 Another approach is to incorporate the dirhodium catalysts into porous materials such as metal organic frameworks (MOFs), [21][22][23][24] but so far, none of the produced catalysts are capable of high levels of asymmetric induction. 22 A simpler approach is a designed polymer with coordinating groups capable of binding to the axial sites of the dirhodium catalyst. A highly cross-linked polymer with pyridine linkers was found to be a suitable system, and could immobilize a wide range of chiral dirhodium catalysts without any further catalyst derivatization. [25][26][27] Even though the catalyst bound polymer could be recovered and recycled, the system did suffer from significant levels of rhodium leaching compared to the covalently attached catalysts.
Recently, the Davies group has designed a series of very large chiral dirhodium tetracarboxylate catalysts (MW 2000-5000) to achieve site-selective carbene transformations. 28 They are readily prepared by the self-assembly of four identical carboxylate ligands and can be further diversified by multifold Suzuki couplings on the preformed catalyst. 29 Two catalysts that have risen to prominence are the C4-symmetric bowl-shaped catalysts, Rh2(S-TPPTTL)4 1a (MW 2464) and Rh2(S-tetra-p-BrPhPTTL)4 1b (MW 3727) (Scheme 1A). Their molecular weights are approximately 7-10 times larger than the typical reagents (diazo compounds and alkene substrates) used in carbene reactions. We surmised that this significant difference in molecular weight would be sufficient to promote the separation of the catalyst from the reaction mixture using semipermeable membranes (Scheme 1B). Inspired by the simplicity of this separation approach and the potential to avoid the immobilization step that often requires tremendous synthetic effort, we herein demonstrate a "catalyst-in-bag" concept for recovering and recycling homogeneous dirhodium catalysts in asymmetric cyclopropanation reactions (Scheme 1C-D). In this report, we describe our optimization studies to develop suitable membranes and conditions to allow transfer of reagents through the membrane, while retaining the catalysts in the bag, resulting in systems that maintain good yields (75%) and enantiomeric excess (90-94% ee) for up to five cycles of recycling without loss of catalytic performance. Scheme 1. The schematic approach of cyclopropanation in the catalyst-in-bag system. A. 3D Xray structure of 1a Rh2(S-TPPTTL)4 and 1b Rh2(S-tetra-p-BrPhPTTL)4. B. Membrane-assisted reaction in the catalyst-in-bag system by size-dependent permeability. C. Model reaction for the catalyst-in-bag system. D. Illustration of reaction-recycle-recovery workflow in the catalyst-inbag system.
Preparation and feasibility of catalyst-in-bag system Scheme 1 describes the overall catalyst-in-bag system for cyclopropanation. This system utilizes a semipermeable membrane shaped like a teabag 30 that contains the rhodium catalyst in place of tea leaves. This catalyst-in-bag concept was initially introduced for recycling copper catalysts immobilized on dendrimers 30 and extended to iridium 31 and palladium 32 catalysts. However, as the reaction progressed, metal leaching was found to increase due to dendrimer instability. In comparison, our catalysts have a stable structure, mitigating metal leaching issues. The catalystin-bag design allows free passage of solvents and small substrates through the bag while effectively trapping the larger rhodium catalyst inside the bag (Scheme 1B). This selective permeability is influenced by the polarity of the solvent and the affinity of the membrane for that solvent. 33 Membrane encapsulation coupled with solubility differences in biphasic systems has been exploited in other transition metal catalyzed reactions, we well. 34 Given this prior work, we initially performed a feasibility evaluation to identify an appropriate combination of membrane and solvent conditions.
We selected two types of commercially available cellulose-based membranes, regenerated cellulose (RC) and benzoylated cellulose (Bz), both sharing the same cellulose backbone structure but having different molecular weight cut-offs (MWCO) and polarity. The catalyst-in-bag fabrication initially involves the preparation of dry membranes to eliminate the risk of water contamination, which can impact the catalysis. The commercial Bz-membrane, in contrast to the dry RCmembrane, was pre-wetted in water containing 0.002% 1-hydroxypyridine-2-thione as a preservative. We performed a solvent exchange with methanol and hexane to remove water and remaining polar species, effectively drying the membrane while minimizing shrinkage and preserving its structural integrity. 35 The stability of the benzoylate groups was probed via FT-IR spectra after the water and hexane treatments (Figure S1). Then, one end of the RC-or Bz-membrane tube was sealed using chemically stable Teflon tape, and catalyst powder was introduced into the bag through a glass pipette. Subsequently, the inside of the bag was purged with Ar gas to minimize air and moisture that may cause side reactions, and the opposite end of the tube was sealed with Teflon tape (Figure S2). Following preparation, we sequentially evaluated the feasibility of the catalyst-in-bag system by examining the permeability of membranes with different solvents (MW 85-88), substrates (MW 104-372), and dirhodium catalysts (MW 778-3727) in ascending order of molecular weight.
Solvent permeability was analyzed by immersing catalyst-in-bags containing Rh2(S-TPPTTL)4 (1a) in ethyl acetate (EtOAc, MW 88) or dichloromethane (CH2Cl2, MW 85), respectively (Figure S3a), as these solvents are representative reaction solvents in cyclopropanation. 8 As shown in Figure S3b, the EtOAc and CH2Cl2 solvents diffused into both RC-and Bz-membrane bags and wetted the Rh2(S-TPPTLL)4 catalyst powder after 24 h. The RC-membrane bags, even with higher MWCO (3kDa and 8kDa MWCO), did not inflate, whereas the Bz membrane bags (2kDa MWCO) swelled significantly.
Next, the permeability of 2,2,2-trichloroethyl 2-(4-bromophenyl)-2-diazoacetate (3a, MW 372) was analyzed by observing changes in the outer solvents in which membranes containing diazo 3a were soaked for 1 day (Figure S3a). In EtOAc, 0.3 % and 11 % of diazoacetate 3a diffused from through the 3 kDa and 8 kDa MWCO RC membranes, respectively, whereas no diffusion was shown in the CH2Cl2 solvent. In contrast, the Bz-membrane bag allowed 66 % and 63 % diffusion of diazo 3a into both EtOAc and CH2Cl2 outer solvent, demonstrating its higher compatibility with cyclopropanation conditions (Table S1).
Finally, to assess the permeability of the Rh2(S-TPPTTL)4 catalyst with the Bz-membrane, we examined the rhodium leaching in EtOAc solvent. We comparatively evaluated three different sizes of dirhodium catalysts, Rh2(S-TPPTTL)4 1a, Rh2(S-tetra-p-BrPhPTTL)4 1b and Rh2(Oct) 4 2, with molecular weights of 2465 Da, 3727 Da, and 779 Da, respectively (Table 1). The Bz-membrane was not able to retain the smallest catalyst, Rh2(Oct)4, as 8940 ppm of rhodium was detected in the green outer solvent by ICP-MS (inductively coupled plasma mass spectrometry) (Table 1).
Although the observed permeation of Rh2(Oct)4 through the 2kDa membrane was only ~1% of the rhodium added to the bag, presumably due to the low solubility of this catalyst in EtOAc, this rhodium would still be too significant to justify use of the membrane. By comparison, the permeation of Rh2(S-TPPTTL)4 was limited to only 1.3 ppm in EtOAc and 3 ppm in CH2Cl2. Rh2(Stetra-p-BrPhPTTL)4 showed no rhodium permeation within experimental error. These results demonstrate that the Bz-membrane with 2 kDa MWCO is effective in retaining the larger sized dirhodium catalysts Rh2(S-TPPTTL)4 and Rh2(S-tetra-p-BrPhPTTL)4.
Table 1. Loss of dirhodium catalysts with different sizes from the catalyst-in-bag.
a Permeation levels were quantified using ICP-MS and compared to the original quantity of dirhodium catalysts initially introduced into the bag. b The value is below the detection limit of the equipment. This value represents the measurement obtained directly from the ICP-MS, rather than indicating a Rh loss.
Cyclopropanation with catalyst-in-bag system
To evaluate the efficiency of the catalyst-in-bag system for cyclopropanation, model reactions were conducted with 2,2,2-trichloroethyl 2-(4-bromophenyl)-2-diazoacetate 3a and styrene 4a using Rh2(S-TPPTTL)4 1a at room temperature (25 °C). Before the catalytic reaction, the dry catalyst-in-bag was soaked for 1 day to wet the Bz-membrane and dissolve the catalyst. The soaked bags were then transferred to reaction vials containing a solution of diazoacetate 3a and styrene 4a, which were subsequently purged with argon and stirred.
Table 2. Performance comparison of catalyst-in-bag and homogeneous conditions A comparison between catalyst-in-bag conditions and homogeneous conditions showed that the Bz-membrane bag maintained its catalytic performance in both EtOAc and CH2Cl2 without significant changes in yield and enantioselectivity (Table 2). Enantioselectivities of the catalystin-bag systems reached 94% ee in EtOAc and 83% ee in CH2Cl2 (entries 2 and 4). However, yields were slightly reduced because some products remained inside the bag even after washing twice for product recovery. 1 H NMR studies also showed that cyclopropanation using the catalyst-in-bag design occurred cleanly without side reactions such as OH-insertion or dimerization. The gradual diffusion of diazo into the bag resembles the conventional slow incremental addition of a semibatch reactor, which helps minimize the risk of dimerization. This high performance indicates that the Bz-membrane was compatible with the reagents and did not affect the catalytic reaction.
In situ FTIR spectrometery (ReactIR TM , Mettler Toledo) studies were conducted to monitor the reaction using the catalyst-in-bag system. The reaction progress was analyzed by tracking the disappearance of the distinctive diazo (-N=N-) signal at 2096 cm -1 from the diazoacetate 3a (Fig- ure S5). 8 As shown in Figure 1, styrene 4a and diazoacetate 3a were sequentially added to the EtOAc solution at 30 min intervals. As diazoacetate 3a was added, the intensity of the IR peak increased rapidly and remained unchanged during stirring for 1 h. After the introduction of the catalyst-in-bag into the reaction vessel, the IR signal of diazoacetate began to gradually decrease (Figure 1). This demonstrates that the catalytic reaction was effectively facilitated by diffusion within the catalyst-in-bag system, as intended by our design, as reagents permeated through the membrane to react with the catalyst.
Figure 1. Kinetic investigation to evaluate the suitability of the catalyst-in-bag system using in situ FTIR spectrometer, ReactIR.
While the catalyst-in-bag design was successfully applied to the cyclopropanation of diazo 3a with styrene 4a, the reaction rate was significantly slower than that of homogeneous conditions, with the complete conversion taking up to 2 days instead of minutes, 8 presumably due to diffusion limitations of the Bz-membrane. An in situ IR spectrometer was utilized to gain a better understanding of the kinetic behavior of the catalyst-in-bag system. To establish optimal reaction conditions for a kinetic perspective, we investigated the effects of temperature, catalyst concentration, and substrate concentration on the reaction rate. The effect of temperature on the cyclopropanation within the catalyst-in-bag system was analyzed at 25 °C, 40 °C, and 60 °C. The reaction showed a 1 st order relationship with diazoacetate
0 6 12 18 24 30 36 42 48 0.0 0.2 0.4 0.6 0.8 1.0 Relative Concentration of Diazo compound 3a (a.u.) Time (h) 25 °C 40 °C 60 °C
3a. As the temperature increased, both the reaction time and half-life (t1/2) decreased from 47 h (t1/2 of 12.5 h) at 25 °C to 16 h (t1/2 of 3 h) at 40 °C, and further to 9 h (t1/2 of 1.5 h) at 60 °C.
Consequently, the reaction rate coefficients (k) increased with temperature, as shown in Figure S6.
Rhodium leaching was observed to remain constant at lower temperatures, measuring 1.9 ppm at 25 °C and 2.2 ppm at 40 °C, but increased slightly at 60 °C to 19 ppm. The yields of cyclopropane 5a slightly increased with increasing temperatures to 77%, 82%, and 81%, respectively, while the enantioselectivity (ee%) decreased to 94%, 92%, and 87%, respectively. These results indicate that temperature not only improves membrane diffusion but also impacts catalyst performance. Considering that the catalyst-in-bag system showed reasonable reaction rates at 40 °C for ReactIR studies, we pursued further kinetic studies at this temperature.
Control experiments were conducted to evaluate the reliability of the kinetic analysis in the catalyst-in-bag system. Based on the detection of 2.2 ppm of rhodium (1.1x10 -5 mg of Rh2(S-TPPTTL)4, 0.002x10 -3 mol % catalyst loading) in the external solvent after reaction at 40 °C (Figure 2), we evaluated whether the observed catalytic activity could be due solely to rhodium that permeated through the membrane. In comparative tests shown in Figure 3a, a homogeneous catalytic reaction using 0.001 mol % catalyst loading (0.5x10 -2 mg of Rh2(S-TPPTTL)4) required more than 42 h to complete the reaction, significantly longer than the catalyst-in-bag system, suggesting that the minimal rhodium permeation could not be the sole source of the observed reactivity with the catalyst-in-bag system and has an overall negligible impact on the observed rate. Additionally, an on/off experiment confirmed the direct influence of the catalyst-in-bag on the reaction progress; removal of the bag halted the reaction, which resumed upon reintroduction (Figure 3b).
This demonstrates that the kinetic evaluation via in situ IR of the catalyst-in-bag system is not only reliable but also the integrity of the system. The catalyst loading was varied at 0.2, 0.5, 1, 2, and 5 mol % with constant concentrations of styrene and diazoacetate (Figure 3c). The reaction rate remained consistent at catalyst loadings between 1 and 5 mol %, but gradually decreased as the catalyst loading was reduced to 0.5 and 0.2 mol %. Although the rate-determining step is diffusion of substrate through the membrane, these results suggest that the activity of the catalyst can influence the diffusion process due to its effect on the concentration gradient. In a previous study, the variable time normalization analysis (VTNA) method demonstrated that the Rh2(S-TPPTTL)4 catalyst has first-order kinetics. 8 The rate law for the cyclopropanation using rhodium (II) is known as follows:
The reaction profiles were plotted against a normalized time scale Time[cat]T n (([cat]T = Total catalyst concentration, n = reaction order of the catalyst to be determined) with n=1, and it showed a similar kinetic profile from 0.2 to 1 mol % catalyst loading, but gradually slowed down at 2 mol % and 5 mol % (Figure S7b). This indicates that diffusion limitations hinder the full utilization of the catalyst at higher concentrations, such as 2 mol % and 5 mol %. The optimal catalyst concentration that balances both diffusion and reaction rates at 0.02 M substrate concentration is 1 mol %.
Cyclopropanation reactions were then conducted using different concentrations of diazoacetate 3a styrene 4a and at 0.01 M, 0.02 M and 0.1 M, keeping the catalyst loading at 1 mol%.
As shown in Figure 3d, the reaction under the higher concentration diazoacetate 3a and styrene 4a conditions converted more to cyclopropane in the same amount of time, but the half-life (t1/2) for all conditions is the same, indicating that the catalyst-in-bag is a diffusion-limited system. To mitigate the impact of concentration differences, we normalized the cyclopropanation kinetics based on [3a]. The results showed that conversion rates remained constant across all tested concentrations (Figure S8). This consistency suggests the catalyst-in-bag system is diffusion-dominated and exhibits pseudo-first-order behavior (eq 1). However, after the first half-life (t1/2), the reaction kinetics at 0.01M and 0.02M became slower compared to the 0.1 M condition. This appears to be a result of the reduced diffusion rate due to the lower absolute concentrations of 3a and 4a in the 0.01 M and 0.02 M conditions. Thus, the catalyst-in-bag systems have reaction kinetics in which an additional diffusion element is introduced compared to the homogeneous system. To improve conversion rates to get higher product amount within the same timeframe, increasing the mass of substrates can overcome the diffusion limitations, highlighting the importance of substrate concentration in catalyst-in-bag systems.
Compatibility of catalyst-in-bag system Reaction conditions: 3 (0.2 mmol, 1.0 equiv), 4 (5 equiv), Rh2(S-TPPTTL)4 (1 mol %, in membrane), and 4Å molecular sieves (100 wt%) in ethyl acetate (10 ml, 0.02 M) at 25 °C for 2 days.
The catalyst-in-bag system utilizing benzoylated cellulose membranes works well using the standard reagents, diazoacetate 3a and styrene 4a. To further evaluate the scalability and compatibility of the catalyst-in-bag system, we conducted 9 different cyclopropanation reactions using 5 diazoacetates and 4 alkenes, each with diverse functional groups (Figure 4 and Figure S9). The addition of functional groups to the substrate did not interfere with its ability to pass through the 2 kDa Bzmembrane, and the cyclopropanations of the 9 different substrate combinations were completed successfully. The reaction of product 5a served as the standard. Moreover, the resulting products (5b-5j) were produced with good yields and enantioselectivities. These data show that the catalystin-bag system can be broadly applied to various cyclopropanation reactions.
To investigate the recyclability and stability of the catalyst-in-bag system, the bags were reused multiple times in both EtOAc and CH2Cl2 at 25 °C. The recycling experiments were conducted under identical conditions for each cycle. After completion of each reaction, the catalystin-bag was washed twice with fresh solvent before it was introduced into a new batch of reaction solvent. This process was repeated to achieve 5 cycles of reactions (Figure S10).
Recycling studies of the catalyst-in-bag system (a) using EtOAc as solvent with average reaction times of 1-2 days, and (b) using CH2Cl2 as solvent with average reaction times 1-1.5 days.
With both solvents, EtOAc (Figure 5a) or CH2Cl2 (Figure 5b), the membrane showed reproducible performance and that it could be reused for up to 5 cycles without any significant changes in yields and enantioselectivity compared the first catalyst-in-bag cycle. These values are consistent with the results obtained under homogenous conditions (Table 2). As the bag was reused, the apparent yield slightly increased (EtOAc) or fluctuated (CH2Cl2). The increasing yield is
73 77 81 87 95 94 95 93 93 1 2 3 4 5 0 20 40 60 80 100 Cycle number Yield (%) 0 20 40 60 80 100 ee (%) 78 71 75 70 83 85 84 86 84 1 2 3 4 5 0 20 40 60 80 100 Cycle Yield (%) 0 20 40 60 80 100 ee (%)
because the products were not completely recovered during the washing process in the prior cycle.
We also observed that the yield without washing gradually increased with each cycle (Figure S11), eventually approaching the yield obtained after washing (Figure 5).
Kinetic studies for recycling reactions using the catalyst-in-bag system.
a Rhodium loss via permeation through the bag was determined by ICP-MS with internal standardization.
The robustness of the Bz-membrane during continuous use was evaluated by measuring the rhodium permeation and the reaction kinetics in EtOAc solvent (Figure 6). The rhodium content in the external reaction solutions in each of 5 cycles was measured using ICP-MS spectroscopy to detect the rhodium permeation from the bag to the outer solution surrounding the bag. From the first cycle to fifth cycle, similar rhodium permeation occurred, yielding rhodium concentrations of 1.9 to 3.5 ppm in the external solution. The kinetic profile measured via ReactIR also showed nearly identical reaction rates over the 5 cycles. These results demonstrate that the Bz-membrane maintains a robust catalyst-in-bag structure during the 5 cycles of reaction, allowing the catalyst to be stored and reused.
0 6 12 18 24 30 36 42 48 0.0 0.2 0.4 0.6 0.8 1.0 Relative Concentration of Diazo compound 3a (a.u.) Time (h) Cycle 1 Cycle 2 Cycle 3 Cycle 4 Cycle 5
Catalyst recovery was carried out simply by cutting the used bag with scissors (Scheme 1c) and collecting the internal catalyst solution without further purification processes. The soluble rhodium recovery from the catalyst-in-bag system was measured at 91% in EtOAc as determined by H 1 NMR experiments based on the initial catalyst loading after 5 cycles, and 9% of the rhodium loss was estimated to be retained on the ends bound to the Teflon tape or lost during sample handling.
Deformability of the Catalyst-in-Bag System
The flexible nature of the Bz-membrane allows the customization and scaling of the chemical synthesis processes to suit specific reaction conditions by adjusting the bag size or converting the catalyst-in-bag system into various structures. As shown in Figure 7a and 7b, reactions proceed efficiently in different vessels using the same type of catalyst-in-bag, demonstrating its adaptability to various reaction environments. Scaling up the reaction is straightforward by simply increasing the size and enclosed volume of the membrane (Figure 7c and Figure S13). In gram-scale experiments, the system maintained 83% yield and 92% ee in the cyclopropanation with no significant impact of scale on the performance. Additionally, after four washing steps, the rhodium catalyst was recovered in 99% yield with 92% of purity without further purification (Figure S14).
Rhodium catalyst was also successfully deployed in a cylindrical, capsule reactor with disc-shaped membrane windows, demonstrating the versatility and adaptability of the system, allowing the use of a variety of structural morphologies, including catalysts-in-vessels beyond the traditional tube configuration that was the focus here (Figure 7d). However, cyclopropanation produces N2 gas as a by-product, so for truly large-scale applications it will be necessary to engineer rapid gas permeation out of the catalyst-in-bag system to prevent membrane bursting during the initial fastest reaction rate period.
In conclusion, we have developed a simplified method for the recycling and recovery of dirhodium catalysts in cyclopropanation reactions utilizing a catalyst-in-bag system. The kinetic analysis identified that substrate/product diffusion is the critical rate-limiting step in the Bz-membrane system. Although use of non-optimized commercial membranes slowed down the reaction rates, this catalyst-in-bag system with Rh2(S-TPPTTL) catalyst demonstrated compatibility with a variety of substrates and achieved comparable yields and enantioselectivities. Moreover, the catalystin-bag system is an interesting framework for enhancing precious metal and ligand recovery, as the bag is flexible, easy to store, and scalable. Further studies may advance this catalyst-in-bag system with improved reaction efficiency by developing more optimized membranes for specific target reactions like cyclopropanation.