Optimization of the individual steps in batch
In order to be able to design a continuous setup for the multistep synthesis of valsartan precursor 13, we first examined the three individual steps in batch. Since we wanted to keep the consumption of our intermediates 5, 11 and 12 to a minimum and the optimization in flow would have required a higher amount of reagents (due to the equilibration times of our targeted continuous setups), we decided to identify a first set of suitable reaction parameters in batch. However, in view of the future improvement of the process, optimization will be performed in a continuous fashion in order to take the intrinsic dynamics and all other advantages of continuous systems into account. As the pivotal step of the reaction cascade is a Suzuki-Miyaura cross-coupling, we focused on step 2 in the beginning. For this transformation we had a number of heterogeneous Pd-catalysts in hand , which can readily be implemented in a fixed-bed reactor for continuous flow applications . Based on earlier studies, we chose to test three of them (Ce0.99-xSnxPd0.01O2-δ, x = 0.20, 0.79, 0.99) as they gave the best results in the Suzuki coupling of 2-bromobenzonitrile 7b with phenylboronic acid (standard reaction conditions: 1 mol eq. aryl halide, 1.5 mol eq. phenylboronic acid, 1.5 mol eq. K2CO3, EtOH:H2O = 7:3, 75 °C) . Our initial approaches for the synthesis of valsartan 1 involved the Suzuki coupling of boronic acid derivatives carrying a (trityl-protected) tetrazole or a nitrile group in ortho-position. These attempts failed due to deactivation of the employed Pd-catalyst and substrate degradation by rapid protolytic deboronation. However, we observed that boronic acid ester 12 couples smoothly with 2-halobenzonitrile 7b-c in EtOH:H2O = 7:3 at 75 °C (Scheme 3, step 2). In the employed aqueous reaction environment 12 is hydrolyzed to the free boronic acid, which is more reactive in Suzuki-Miyaura cross-couplings than its ester analogue . Therefore, we chose to pursue a synthetic route involving the Suzuki coupling of 12 with 7b-c leading to formation of valsartan precursor 13 as final product (Scheme 3). After having identified suitable cross-coupling partners, we wanted to determine the most active Pd-catalyst for our targeted transformation. Regarding the Suzuki coupling of 2-bromobenzonitrile 7b with boronic acid ester 12 using 1 mol% of catalyst, the activity follows the order Ce0.20Sn0.79Pd0.01O2-δ > Sn0.99Pd0.01O2-δ > Ce0.79Sn0.20Pd0.01O2-δ (Table 1, Entries 1–3). Hence, we decided to use Ce0.20Sn0.79Pd0.01O2-δ for all further experiments. Moreover, it is commonly known that the reactivity of aryl halides in Suzuki coupling increases along the halogen group (Cl < <Br < I). As expected, a higher conversion was obtained when using 2-iodobenzonitrile 7c instead of 7b (Table 1, Entry 4). Consequently, 7c was the coupling partner of choice and the catalyst loading was decreased from 1 mol% to 0.25 mol% for further studies.
In view of the performance of a multistep reaction cascade and the realization of N-acylation prior to the Suzuki coupling, we next needed to identify a suitable solvent system for achieving both steps in a sequential fashion. As reported in literature  and experienced in our lab, water is crucial for the success of the Suzuki coupling using our class of catalysts. However, step 1 needs to be performed in an aprotic organic solvent to avoid hydrolysis of the valeryl chloride reagent. Since our chosen catalyst showed no or very low activity in two-phasic solvent systems (water + dicholoromethane/toluene/ethyl acetate), we needed a single liquid phase for the Suzuki coupling step. Highly polar water-miscible organic solvents (acetonitrile; N,N-dimethyl sulfoxide; N,N-dimethylformamide) did not prove to be compatible with our catalyst. However, considerable catalyst activity was detected in aqueous ethereal solvents (tetrahydrofuran (THF); 1,4-dioxane). As THF and water are not miscible in a 1:1 ratio at elevated temperatures , in this case the addition of an alcoholic solvent was further necessary to obtain one liquid phase. The choice of organic solvent also has a substantial influence on the leaching behavior of palladium, tin and cerium into the reaction solution. Our investigations in this respect showed that palladium leaching increases along with the polarity of the reaction solvent (Table 2). Highest levels of palladium were determined in ethanol-water (3.12 mg kg−1), whereas they were found to be lowest in dioxane-water (0.37 mg kg−1). Cerium leaching shows the exact opposite trend (0.89 mg kg−1 in ethanol-water, 2.53 mg kg−1 in dioxane-water, Table 2). In contrast, leaching of tin was found to be insignificant in all tested solvents (0.06–0.08 mg kg−1, Table 2). Due to the fact that minimal leaching of the active catalytic species palladium was targeted, we chose dioxane-water as the solvent for the Suzuki coupling step.
Regarding the optimal dioxane-to-water ratio for the Suzuki coupling step, solubility issues leave a narrow margin. If the reaction mixture is rich in water (≥ 55 v%), employed organic compounds are not entirely soluble. If, on the other hand, dioxane is present in ≥70 v%, the inorganic base potassium carbonate (K2CO3) does not dissolve completely. An equivolume mixture of dioxane and water provided the best results for our reaction (Table 1, Entries 5–7). Apart from that, as expected the use of a larger excess of boronic acid ester 12 leads to an increased reaction rate (Table 1, Entries 8–9). Nevertheless, we decided to use 1.10 mol equivalents of 12 for the continuous reaction setup considering atom economy as well as by-product formation. Particularly, homocoupling and oxidation of the organoboron reagent were reported to occur upon depletion of the aryl halide coupling partner , causing a more complex reaction mixture.
After having identified optimum conditions in batch for the Suzuki coupling step, we concentrated our attention on step 1 of the reaction cascade (Scheme 3, step 1). N-acylation of 11, which was synthesized using a literature procedure , with valeryl chloride was performed in dioxane in presence of an organic base. We aimed for quantitative conversion of boronic acid ester 11 as previous experiments showed literature-known  deactivation of our Pd-catalyst, which is employed in the subsequent Suzuki coupling step, by N-H functional groups. We investigated the effect of temperature as well as molar equivalents of reagents (valeryl chloride, organic base) on the conversion (Table 3). N,N-diisopropylethylamine (DIPEA) was chosen as organic base because the formed hydrochloride salt proved to be soluble in dioxane at elevated temperatures. The rate of N-acylation of 11 increases with higher temperature (Table 3, Entries 1–3) as well as a larger excess of valeryl chloride (Table 3, Entries 4–5). Using equimolar amounts of valeryl chloride and DIPEA further accelerates the reaction (Table 3, Entry 6), achieving almost full conversion after 5 min (98.8%).
The final step of our targeted reaction cascade is the saponification of 5 to yield valsartan precursor 13 (Scheme 3, step 3). As a dioxane-water mixture was found to be optimal for the preceding Suzuki coupling step, ester hydrolysis of 5 was performed in the same solvents. At a temperature of 80 °C, the rate of methyl ester hydrolysis was found to increase with a larger molar excess of sodium hydroxide (Table 4, Entries 1–4). Substitution of the base by potassium hydroxide did not have a considerable impact on the conversion (Table 4, Entry 5). Raising the water content in the reaction solution to 55 v% slightly decelerated the conversion (Table 4, Entry 6) and a further increase to 60 v% caused solubility issues.
Optimization of sequential steps 1 and 2 in batch
For Pd-catalyzed Suzuki-Miyaura cross-couplings, the presence of a base is essential for the reaction to occur . Hence, the pH of the reaction environment affects the rate of C-C bond formation . In view of the performance of the Suzuki coupling subsequently to step 1, different influencing factors on the pH of the reaction mixture have to be considered. These include the formation of hydrogen chloride during the N-acylation step and hydrolysis of the excess acid chloride to the corresponding carboxylic acid upon addition of water. Consequently, the pH of the reaction mixture after step 1 is acidic and part of potassium carbonate, which is added for the Suzuki coupling, gets neutralized. In order to study the effect of the pH on the Suzuki coupling step and to identify the necessary amount of potassium carbonate to be added, we performed steps 1 and 2 in a sequential fashion in one-pot (Scheme 4).
For the sequential performance of steps 1 and 2, N-acylation of 11 was accomplished using 2 mol eq. of both valeryl chloride and DIPEA in dioxane at 80 °C. Upon completion of the reaction, the reaction mixture was quenched with an equal volume of water. Then, potassium carbonate, 2-iodobenzonitrile 7c and Pd-catalyst Ce0.20Sn0.79Pd0.01O2-δ were added to the reaction flask in order to achieve formation of 5. We compared the Suzuki coupling step performed in one-pot with the optimized separate Suzuki coupling, exhibiting a pH of 11.3 in the presence of 1.5 mol eq. of potassium carbonate (Table 5, Entry 1). When the same amount of base was used in the one-pot experiment, a pH of only 9.7 was obtained and the conversion in the Suzuki coupling step was considerably lower (Table 5, Entry 2). Therefore, another 0.17 mmol of K2CO3 were added, corresponding to the employed amount of valeryl chloride reagent in step 1. In doing so, the pH approaches the reference value and conversion was enhanced significantly (Table 5, Entry 3). Further increasing the amount of base in the reaction mixture gave comparable conversion (Table 5, Entry 4). Upon addition of 0.39 mmol of base, formation of a second liquid phase was observed (Table 5, Entry 5), which was confirmed by following the reaction with the Crystalline (Technobis) . The device is designed for studying crystallization processes by turbidity measurements but also allows real time monitoring of reactions via a particle view camera. Regarding obtained results and in view of the sequential performance of steps 1 and 2 in the multistep continuous setup for the synthesis of 13, the use of 3.7 mol eq. potassium carbonate in the Suzuki coupling step was considered as most appropriate.
Continuous setups for the individual steps
After optimization of the individual steps in batch, we first targeted the realization of N-acylation as well as saponification in continuous flow. As both transformations exhibit a monophasic reaction solution, we decided to utilize a stainless steel coil (L x O.D. x I.D. 3.0 m × 1/16 in. × 0.030 in., V = 1.368 mL) as reactor in combination with a split-and-recombine unit (V = 0.565 mL) as static mixer (Scheme 5). The mixing efficiency of the latter and its applicability for rapid chemical transformations has been demonstrated lately for the aerobic oxidation of Grignard reagents in continuous flow . The reactor comprises a precooling section, after which the two reagent streams are merged and introduced into a split-and-recombine section for enhanced mixing.
Regarding the continuous performance of steps 1 and 3, a dual syringe infusion pump (LA-120, Landgraf) was utilized for steady reagent delivery into the reactor setup (V = 1.933 mL), which was assembled using standard HPLC fittings. The performance of the continuous setup at ambient pressure was determined at three different total flow rates based on the conversion of 11 and 5, respectively (for details see Supporting Information). In both cases, substrate conversion of >99% was obtained at the lowest flow rate of 0.10 mL min−1, which corresponds to a calculated residence time of τ~19 min (Table 6, Entry 1). Whereas the N-acylation step proved to be similarly efficient at higher flow rates of 0.15 and 0.20 mL min−1, for methyl ester hydrolysis conversion decreased noticeably (Table 6, Entries 2–3). However, the observed general trend of higher conversion at lower flow rates suggests that the mixing obtained with the split-and-recombine reactor is sufficient for both transformations.
Next, we focused on the continuous setup for the second step of our cascade towards 13, Suzuki-Miyaura cross-coupling. The applicability of our heterogeneous Pd-catalyst for the continuous Suzuki coupling of various substituted bromoarenes with phenylboronic utilizing the Plug & Play reactor has already been reported by our group [29, 30]. Therefore, we decided to adopt this approach for the C-C cross-coupling of boronic acid ester 12 with aryl halide 7c. In view of the performance of our three-step reaction cascade, we based the experiment design on the preceding N-acylation step. For step 1 quantitative conversion of 11 is targeted, which is achieved in dioxane using a flow rate of 0.10 mL min−1. As we identified an equivolume dioxane-water mixture as optimal for subsequent Suzuki coupling, we intended to merge the reagent stream of step 1 with an aqueous stream of the same flow rate. Consequently, for testing step 2 separately in a continuous fashion, we chose a flow rate of 0.20 mL min−1. In the experiment, a pre-mixed reagent solution containing both coupling partners as well as potassium carbonate was pumped through the Plug & Play reactor , which was equipped with an HPLC column (L x I. D. 120 × 8 mm) filled with catalyst particles (4.7 g) (Scheme 6, for details see Supporting Information). Due to the backpressure of the catalyst bed (2–3 bar) and incompatibility of the solvent mixture with the available HPLC pump, a high-pressure syringe pump (VIT-FIT HP, LAMBDA Instruments) in combination with stainless steel syringes was utilized. At the chosen flow rate of 0.2 mL min−1, the mean residence time inside the HPLC column containing the catalytic species was determined by measurement of a tracer-determined residence time distribution curve to be 22.3 min. Utilizing the abovementioned setup, step 2 of the targeted reaction cascade was successfully performed in a continuous fashion. After an initial equilibration phase of 50 min (consuming 10 mL of stock solution), the target compound 5 was obtained with an average 95% yield (Fig. 1).
Regarding the performance of Suzuki cross-coupling reactions in continuous flow, leaching of catalytically active Pd species from the solid support into the reaction solution is known to be a major issue [39, 40]. Therefore, the contents of palladium, cerium and tin in the outlet flow of the continuous experiment were measured by ICP-MS after certain time points (Table 7). As already observed in the batch experiments, the amount of tin in the reaction solution was found to be negligible in all samples (<0.02 mg kg−1). The concentration of cerium in the outlet flow instead was determined to decrease over the course of the continuous experiment and was quantified to be 0.70 mg kg−1 after 4 h. As already reported previously , the heterogeneous catalysts with the molecular formula Ce0.99-xSnxPd0.01O2-δ show a significant loss of palladium during the initial phase (0–120 min) of the Suzuki cross-coupling in continuous flow (literature conditions: 1.10 g of Pd-catalyst, flow rate 0.45 mL min−1). However, after equilibration of the system the palladium content drops below the limit of quantification . Likewise, considerable leaching of palladium into the reaction solution was also observed in the continuous Suzuki coupling step for the synthesis of valsartan precursor 5 over the entire runtime of the experiment. Presumably, due to the lower flow rate of 0.2 mL min−1 and the larger amount of catalyst compared to mentioned literature experiment , the applied system has not reached its steady-state within the duration of the experiment, which was limited by the availability of boronic acid ester 12. However, in view of a possible application of the reaction setup for continuous Suzuki coupling on a larger scale, the implementation of a palladium scavenging strategy is advisable. Galaffu et al. reported the successful use of different sulfur-based silica scavengers for the removal of palladium from a variety of advanced synthetic intermediates including valsartan precursor 5 . In their study, after the Suzuki coupling step (1 mol% Pd(PPh3)4, DME/EtOH/H2O, Na2CO3) the residual amount of palladium in product 5 was effectively decreased from 2100 ppm to <1 ppm by the utilization of respective functionalized silica frameworks .
Multistep continuous setup for the synthesis of 13
After confirming the feasibility of the three single steps of the reaction cascade in continuous flow, an integrated setup for their consecutive performance was developed by simple combination of the individual reaction modules (Scheme 7). The N-acylation step was performed in the coil reactor setup already described using a flow rate of 0.10 mL min−1 by means of high-pressure syringe pumps (VIT-FIT HP, Lambda Instruments). The outlet stream was then quenched with an aqueous potassium carbonate solution delivered by an HPLC pump (0.10 mL min−1, P4.1S, Knauer) using a T-mixing element. For the key step of the reaction cascade, Suzuki-Miyaura cross-coupling, the reaction solution was introduced at resulting 0.20 mL min−1 into the Plug & Play reactor  equipped with a fixed-bed of heterogeneous Pd-catalyst Ce0.20Sn0.79Pd0.01O2-δ. Finally, the process stream was merged with a sodium hydroxide solution supplied by a syringe pump (0.1 mL min−1, LA-120, Landgraf) to achieve methyl ester hydrolysis. For this step, a longer reactor coil (PTFE, L x O.D. x I.D. 10.0 m × 1/16 in. × 0.031 in.) was chosen in comparison to the individual setup in order to compensate for the higher flow rate of 0.3 mL min−1 and to achieve a similar residence time. With this setup exhibiting an estimated total residence time of roughly 1 h, the three steps of the reaction cascade were successfully performed continuously in a consecutive fashion for over 6 h. After an initial equilibration phase (260 min), quantitative conversion of 2-iodobenzonitrile 7c was achieved and the target compound 13 was obtained with up to 96% yield (90 ± 4% yield, values fluctuating around the mean) and 73% enantiomeric excess (determined by achiral and chiral HPLC, respectively; see Supporting Information). Performing the analogous three-step reaction sequence from 11 to 13 in batch including purification of intermediates 5 and 12, the obtained overall yield was only 28% (see Supporting Information for the performance of the individual steps in batch). Therefore, translation of the multistep cascade from batch to continuous flow allowed a significant increase in yield of valsartan precursor 13. Addressing the moderate enantiopurity of compound 13 synthesized in continuous flow, racemization apparently occurred during the saponification step as intermediates 5 and 12 were shown to be enantiopure (see Supporting information for determination of enantiopurity of 5 and 12). Consequently, in view of a potential application of the developed continuous approach for actual API production, further process optimization is required to achieve the synthesis of an enantiomerically pure compound.
Regarding the selectivity of the reaction steps, in Ce0.99-xSnxPd0.01O2-δ catalyzed Suzuki cross-coupling oxidation and homocoupling of the boronic acid species were reported to cause minor side product formation upon depletion of the aryl halide coupling partner . Therefore, only a slight excess of 1.10 mol eq. of boronic acid ester 12 was employed to minimize formation of the respective by-products, which were presumably present in the reaction mixture and can be removed by chromatographic techniques. Apart from that, N-acylation as well as methyl ester hydrolysis proved to be rather clean reaction steps. This is supported by the fact that HPLC analysis of the outlet flow of the thee-step continuous process revealed side products to the extent of 9 area% compared to the product peak (λ = 230 nm), which most likely originate from the 10% excess of 12.