Introduction
Nowadays, hydrogenation reactions are extensively employed in both academic and industrial context for the efficient and chemoselective reduction of unsaturated substrates like olefins (1).
The foundations were laid in 1897 by Paul Sabatier, who developed the first catalytic hydrogenation using metal catalysts supported on solids, for which he was awarded the Nobel Prize in Chemistry in 1912 (2).
In the 1930s, Melvin Calvin further expanded this work by leveraging the activation of molecular hydrogen through metal complexes in homogeneous catalysis (3). This set the basis for a broader investigation of this field, advanced by the fundamental contribution of Halpern, Harrod and James (4).
A later contribution was accompanied by Wilkinson with double bond hydrogenation exploiting ruthenium catalysis (5).
Beyond the central role that non-selective hydrogenation still represents in the industrial context, the advent of enantioselective hydrogenation marked significant progress, allowing for the targeted synthesis of enantiomerically pure compounds (6).
The true breakthrough in asymmetric hydrogenation came in the late 1960s with William Knowles, who began investigating homogeneous, asymmetric hydrogenation by using chiral ligands (7).
This paved the way for significant advancements in the 1970-1980s, with Henri Kagan’s development of the Rh/DIOP system (8) and Ryōji Noyori’s development of Rh/BINAP and Ru/BINAP complexes (9).
An efficient asymmetric synthesis is of extreme importance in the production of chiral active pharmaceutical ingredients (APIs), where the ability to efficiently generate stereochemically pure compounds is critical not only for targeting the desired therapeutic effect and minimizing adverse reactions, but also to meet increasingly stringent regulatory standards regarding enantiopurity (10).
These academic milestones established the scientific groundwork necessary for translating selective hydrogenation into robust, large-scale industrial processes. However, a time lag typically exists between academic discoveries and their industrial applications. Despite this delay, academic research continues to be a driving force for future technological innovation and process development within industry.
As highlighted by Blaser et al., enantioselective hydrogenation of olefins is one of the most thoroughly studied and industrially applied asymmetric transformations (11). However, optimal catalytic performance still requires fine-tuning of the metal-ligand combination for each specific substrate (Table 1).
General workflow for screening, optimization and scale-up
As stated above, the choice of the optimal (pre-)catalyst/ligand combination is often substrate dependent and extensive screening activities are usually required at the beginning of the project. At this stage, process parameters affecting reaction productivity (e.g., dilution, catalyst loading, etc.) are usually not included in the experimental designs and several combinations of catalysts and ligands are combined under similar conditions, namely same temperature, same solvent and same hydrogen pressure. Only the key reaction outputs like conversion, purity, enantiomeric excess (ee), and diasteromeric ratio/excess (dr/de) are evaluated to finally select the catalytic system. It is worth mentioning that this first stage could be time-consuming and high-throughput experimentation platforms are often adopted to speed up the screening process by running multiple reactions in parallel.
After having selected the best (pre-)catalyst/ligand combination, reaction optimization is required to establish the best protocol to be adopted on scale. Optimization activities should highlight all parameters affecting both product synthesis and quality attributes and, ultimately, define the experimental ranges where the reaction would reliably afford the target compound according to its final specifications. This is the stage when all parameters, including those affecting reaction productivity, are considered to maximize yield and efficiency while minimizing costs. For an asymmetric hydrogenation reaction, hydrogen pressure, solvent composition, dilution, catalyst and ligand loadings and temperature are generally taken into account.
To scale up a hydrogenation reaction, additional parameters beyond those studied during reaction optimization must be considered, as in the case of reaction calorimetry and mass transfers. The dissolution of a gas (hydrogen) into a liquid phase containing the substrate is critical as it directly affects reaction kinetics. Particularly, the rate at which hydrogen transfers from the gas phase to the liquid phase must be evaluated. This phenomenon is governed by the volumetric mass transfer coefficient, kLa (12-13).
kLa quantifies how efficiently hydrogen is transferred into the liquid phase, where it can participate in the reaction. This parameter is influenced by several factors, including mixing efficiency, hydrogen pressure, reactor geometry, viscosity and surface tension of the liquid. While some of these parameters remain consistent between laboratory-scale and industrial-scale, others, like mixing efficiency and reactor geometry, are specific to the plant design.
A significant mismatch could lead to slower reaction rates at industrial scale, with detrimental effects on plant scheduling.
Horizontal autoclaves can ensure optimal mixing and, consequently, more effective hydrogen dissolution into the liquid phase compared to vertical reactors. Procos, with extensive experience using this equipment, has been selecting this configuration since the 1950s for industrial-scale manufacturing (Figure 2).
In certain industrial applications, especially when very low catalyst loadings are used or when stringent purity requirements apply, polishing the autoclave to a “mirror finish” is a commonly recommended practice. This treatment minimizes surface roughness, reducing the risk of catalyst residues and facilitating cleaning (Figure 3).
Case study
Procos is actively engaged in the development of enantioselective hydrogenation processes, consistently integrating academic insights into scalable and industrially robust applications. Building on the key aspects discussed in the previous sections, the following case study presents catalyst screening and process optimization efforts conducted in preparation for the scale-up of an asymmetric hydrogenation of a tetrasubstituted double bond (Scheme 1).
Catalysts reported as effective for the hydrogenation of tetrasubstituted olefins (Table 1) were initially shortlisted as promising candidates for the hydrogenation reaction reported in Scheme 1.
However, the screening campaign was broadened, as the primary objective was to identify a catalyst–ligand system capable of delivering both high conversion and enantioselectivity, while operating at low catalyst loadings. For industrial viability, several criteria had to be met: the system needed to be stereo-selective, commercially accessible, and cost-effective, ideally exhibiting high turnover number (TON) and turnover frequency (TOF).
A systematic screening strategy was therefore implemented, evaluating a range of commercially available transition metal catalysts and chiral ligands sourced from established suppliers. Catalysts based on rhodium (Rh) and ruthenium (Ru), primarily from Takasago and Johnson Matthey, were tested in combination with various chiral ligands. Among the systems evaluated, Ru–BINAP complexes (Table 2, Entries 3 and 7) emerged as the most promising, offering a favorable balance of enantioselectivity and diastereomeric excess.
Process optimization
The subsequent optimization activities were aimed to increase both the ee and the de of the reduced product while minimizing catalyst loading, thus decreasing manufacturing costs and increasing productivity. Theoretically, these activities should be performed with a robust and well-reproducible protocol with a limited experimental variance in the shortest time possible.
To eliminate potential sources of experimental variability (such as operator influence or manual additions) and to maximize data collection by conducting numerous reactions in parallel, Procos adopted ChemSpeed automated platform for highly-efficient automated reaction screening (Figure 4).
With the two preferred catalysts in hands, an in-depth investigation on the performance of CAT-1 (Table 2, Entry 3) and CAT-2 (Table 2, Entry 7) was conducted employing this automated platform.
Specifically, the ee of the product with increased catalyst turnover numbers was established running the enantioselective hydrogenation with decreasing catalyst loadings. Notably, a slight increase in ee was observed when pressure was increased at high turnover numbers (Figure 5).
Data clearly demonstrated the superior performance of CAT-1 over CAT-2. Accordingly, the former was selected for the following optimization activities aimed to a general increase in reaction productivity.
As usually performed at this stage of the process development, all reaction variables were considered as potentially impactful on yield, ee and all other reaction metrics (e.g., space-time yield and process mass intensity). In case of the asymmetric hydrogenation of interest, the following variables were considered: catalyst loading, dilution (solvent volumes), hydrogen pressure and temperature. To assess both the impact of the single variables and their interaction on the most critical process output, namely the ee, a multivariate approach was conducted. Specifically, a fractional factorial design of experiment (DoE) was performed.
By mean of a 24-1 fractional factorial, 16 experiments (in place of 32 employing a full factorial design) were sufficient to establish the experimental behaviour of the system inside a cubic space (Figure 6).
The analysis of the response surfaces clearly showed that higher catalyst loadings, higher dilutions and lower temperatures, combined, resulted in increased ee values (target >98 %).
Accordingly, a second set of experiments was performed around an improved condition emerged from this first DoE study. A deeper investigation of the experimental space was performed by mean of a D-Optimal design. This design is associated with a more accurate estimation of the model coefficients, and it was preferred in the fine-tuning stage of the process optimization and in the definition of the Proven Acceptable Ranges (PARs).
Considering the same four process parameters, and three levels for each one (-1, 0, +1), the iterative algorithm selected 26 experiments (over 81 possible experiments) plus 3 repetitions of the centre point (0, 0, 0, 0) to compute an informative model (Figure 7).
A comprehensive analysis of the response surface facilitated the accurate identification of the PARs for the critical process parameters, thereby ensuring process robustness in the context of scale-up activities.
Conclusion
The development of an efficient and scalable enantioselective hydrogenation process demands a multidisciplinary and integrated approach to ensure robust and reproducible protocol to be applied to industrial scale in the pharmaceutical sector.
In the case study presented, the combination of systematic metal catalyst screening, multivariate design of experiments, and the use of automated platforms enabled rapid and resource-efficient optimization of reaction parameters. The employment of ChemSpeed also ensured high experimental reproducibility, significantly reducing variability associated with manual operations and providing reliable and consistent results during the optimization phase. This strategy not only minimizes development time and resources consumption but also represents a forward-looking model for the process development of asymmetric hydrogenation processes.
References and notes
1. P. N. Rylander, Catalytic Hydrogenation in Organic Synthesis, Academic Press, 1979.
2. P. Sabatier, Nobel Lecture, December 11, 1912.
3. M. Calvin, J. Am. Chem. Soc. 1939, 61, 8, 2230-2234.
4. J. Halpern, J. F. Harrod, B. R. James, J. Am. Chem. Soc. 1961, 83, 3, 753–754.
5. J. A. Osborn, F. H. Jardine, J. F. Young, G. Wilkinson, J. Chem. Soc. A. 1966, 1711–1732.
6. I. Ojima, Catalytic Asymmetric Synthesis, Wiley, 2010.
7. W. S. Knowles, M. J. Sabacky, Chem. Commun. (London). 1968, 1445-1446.
8. T. P. Dang, H. B. Kagan, J. Chem. Soc. D, 1971, 481.
9. H. Takaya, R. Noyori, Acc. Chem. Res. 1990, 23, 345-350.
10. V. Farina, J. T. Reeves, C. H. Senanayake, J. J. Song, Chem. Rev. 2006, 106, 7, 2734–2793.
11. H. U. Blaser, B. Pugin, F. Spindler, J. Mol. Catal. A. 2005, 231, 1, 1-20.
12. B. Fillion, B. I. Morsi, K. R. Heier, R. M. Machado, Ind. Eng. Chem. Res. 2002, 41, 12, 3052.
13. D. J. Ende, Chemical Engineering in the Pharmaceutical Industry: R&D to Manufacturing, Wiley, 2010.
14. Т.А. Ruzova, B. Haddadi, Results in Surfaces and Interfaces. 2025, 19, 6, 100441.
15. R. Leardi, C. Melzi, G. Polotti, CAT (Chemometric Agile Tool), freely downloadable from http://gruppochemiometria.it/index.php/software.
Figure 1. Main contributors to catalytic hydrogenation in the past century.
Table 1. Catalyst types for different substrates.
Figure 2. 1500L horizontal autoclave for hydrogenation reactions at Procos SpA.
Figure 3. Ra Roughness Chart to establish surface polishing (left) (14).
Internal surface of an industrial autoclave, polished to “mirror finish” (right).
Scheme 1. Scheme of synthesis of the case study.
Table 2. Enantiomeric excess (ee) and diastereomeric excess (de) for different commercially available transition metal catalysts and chiral ligands. The results were recorded in the same experimental conditions: mol% catalyst loading (100 TON), 25 bar H2 pressure, 80°C.
Figure 4. ChemSpeed equipment installed at Procos: 12 reactions under pressure with mechanical stirring can be conducted independently in parallel. All dosages and samplings are performed by the robotic arms of the platform.
Figure 5. Enantioselectivity of the asymmetric hydrogenation depending on catalyst loading. Reactions were conducted at 80 °C, at 20 bar employing catalyst loading ranging from 100 to 2000 TON (left). Enantioselectivity increase at higher hydrogen pressures at 2000 TON (right).
Figure 6. Experimental design of the fractional factorial 24-1 (15).
Figure 7. Experimental design of the D-optimal design for the four parameters at three levels (15).
