From Batch to Continuous Operation: Hydrogenation of Bicarbonate to Formate at Multiphase Boundaries in a Continuous Stirred-Tank Reactor

Abstract

The utilization of carbon dioxide (CO₂) as a C₁ building block has emerged as a promising strategy for sustainable chemical production. Among various CO₂-derived products, formate and formic acid are particularly attractive due to their roles as hydrogen carriers, fuel cell feedstocks, and industrial intermediates. Recent advances in Ru-based homogeneous catalysis have enabled efficient hydrogenation of bicarbonate, which was prepared from CO₂, in biphasic and triphasic systems. In this study, we employed a continuous stirred-tank reactor (CSTR) for the triphasic hydrogenation of KHCO₃ under high-pressure conditions (50 bar of H₂), and optimized the stirring conditions using a view cell to ensure efficient mixing. A kinetic model assuming a slow-reaction regime and incorporating the reverse reaction was developed, which accurately predicted the residence time–yield relationship and enabled high formate yields through residence-time optimization. Furthermore, the addition of tris­(2,4-di-tert-butylphenyl)­phosphite as an antioxidant effectively suppressed residual oxygen contamination, which is a known challenge in flow systems. Catalyst recycling and phase separation were successfully integrated into the flow setup, demonstrating the practicality and scalability of the process. These findings provide a rational framework for designing continuous triphasic hydrogenation systems and contribute to the development of resource-efficient chemical technologies based on CO₂ utilization.

Introduction

The establishment of a sustainable society through resource recycling has become one of the most critical global challenges in recent decades. Among various approaches, the utilization of carbon dioxide (CO₂) as a C₁ building block has attracted significant attention, enabling the synthesis of valuable chemicals such as carbon monoxide, formaldehyde, methylformate, methanol, and methane. In particular, the production of formic acid (HCOOH) and formate (HCOO^–) via hydrogenation of CO₂ or bicarbonate represents one of the most sustainable carbon capture and utilization (CCU) strategies since no hydrogen is lost as water during the reaction, and the products serve as industrially valuable chemical feedstocks, hydrogen carriers, and energy sources for fuel cells. Recent studies have demonstrated the effectiveness of Ru-based homogeneous catalysts for CO₂ capture and hydrogenation (Figure ). In 2014, Yadav et al. reported the hydrogenation of captured CO₂ using 1,8-diazabicyclo[5.4.0]­undec-7-ene (DBU) in MeOH with [RuCl₂(PPh₃)₃] as a catalyst to form methylformate, which was subsequently converted to formate via further hydrogenation with the same catalyst. The Pidko group also used DBU as a base to capture CO₂ and Ru complex 1 bearing a PNP-tridentate ligand as a catalyst for the hydrogenation. Later, they established a gas–liquid–liquid triphasic system using the same Ru catalyst, enabling efficient hydrogenation of potassium bicarbonate (KHCO₃) derived from KOH and CO₂ (Figure ). In such triphasic systems, the resulting formate salts typically remain in the aqueous phase, while the organometallic catalyst stays in the organic phase, allowing for easy separation of the catalyst and products by simple decantation. Similarly, successful utilization of triphasic catalyst systems for CO₂ hydrogenation has been developed by the Prakash group, , Scott et al., and Guan et al. using other PNP-ligand Ru catalysts as shown in Figure .

1.

Representative Ru catalysts and CO₂ absorbents used in formate production via hydrogenation. DBU = 1,8-diazabicyclo[5.4.0]­undec-7-ene, TMG = tetramethylguanidine, DABCO = diazabicyclo[2.2.2]­octane, and MEA = monoethanolamine.

2.

Triphasic reaction system consisting of an aqueous phase containing KHCO₃ and an organic phase containing Ru complex 1, a quaternary ammonium salt (NR₄ ⁺) as a phase-transfer catalyst (PTC), and the antioxidant tris­(2,4-di-tert-butylphenyl)­phosphite. A schematic of kinetic regimes and assumed concentration profiles at the gas–liquid interface and in the bulk phase, illustrating H₂ and KHCO₃.

Continuous synthesis plays a vital role in industrial chemical production, offering significant advantages in process efficiency, scalability, and cost reduction compared with conventional batch operations. Based on the hydrogenation of bicarbonate using Ru complex 1 in a batch reactor, , we previously examined the reaction under continuous-flow conditions using a tubular reactor to evaluate its performance and the recyclability of the catalyst. The maximum formate yield (67%) was obtained with a 0.5 mol L^–1 KHCO₃ aqueous phase; however, increasing the bicarbonate concentration to 1.25 mol L^–1 and 1.5 mol L^–1 resulted in lower formate yields of 33% and 31%, respectively. This decrease is likely due to mass transfer limitations among the three phases. To address similar challenges in triphasic hydrogenation systems, other groups have developed well-mixed continuous reactors, notably stirred-tank and jet-loop designs. − Continuous stirred-tank reactors (CSTRs) mitigate mass transfer limitations through uniform mixing and high interfacial area while offering a design similar to that of successful batch reactors, which simplifies scale-up and reduces development risk. Moreover, for reactor design, kinetic analysis is essential since it not only elucidates the underlying reaction mechanisms and rate-determining steps but also enables quantitative prediction of reaction rates under various operating conditions, which is essential for scale-up. By establishing reliable rate expressions, it becomes possible to simulate and optimize reactor performance based on parameters such as temperature, pressure, and concentration. For instance, in such a triphasic system for the hydroformylation of 1-octene, it is known that the overall reaction rate is controlled by mechanical agitation, gas sparging, and residence time, which influence the environment at the liquid–liquid interface, gas–liquid interface, and within the bulk liquid phase.

In continuous synthesis using flow reactors, maintaining catalyst activity over extended periods is essential for achieving a high production efficiency. However, flow systems are particularly susceptible to oxygen ingress through connection points, which can lead to catalyst deactivation. Accordingly, extensive research efforts have been devoted in both academia and industry to understanding and preventing catalyst deactivation, as it remains a critical challenge in long-term catalytic processes. − In our previous study, we reported that the antioxidant, tris­(2,4-di-tert-butylphenyl)­phosphite, effectively suppressed catalyst deactivation during KHCO₃ hydrogenation by preventing oxidation of the ligand caused by trace oxygen contamination. This strategy enabled the highest total turnover number (TON) reported to date for bicarbonate hydrogenation, reaching 9.43 million, and underscored the importance of oxidative stability in continuous reaction systems.

Building on this background and motivation, we herein established a kinetic model for the triphasic hydrogenation of KHCO₃ under high-pressure conditions, enabling the prediction of residence time vs yield relationships in a continuous stirred-tank reactor (CSTR). For gas–liquid reactions, diffusion of gaseous and liquid components and their subsequent reaction typically occur simultaneously within the liquid boundary layer near the gas–liquid interface. The relative magnitudes of diffusion and reaction rates determine the concentration profiles of reactants, thereby influencing the overall rate expression. In this study, the H₂ pressure was set at 50 bar to minimize the influence of interfacial mass transfer limitations (i.e., achieving sufficient supply of H₂ in the liquid phase, especially toward the gas–liquid interface), and the kinetics were formulated under the assumption that the overall reaction rate is governed by the catalytic reaction in the bulk liquid phase. Under these conditions, as illustrated in Figure , the concentration of dissolved H₂ decreases linearly across the boundary layer, while the KHCO₃ concentration remains constant. Here, $C_{{\mathrm{H}}_2}$ and $C_{{\mathrm{KHCO}}_3}$ are the concentrations of KHCO₃ and H₂, respectively, in the aqueous phase [mol m^–3], and $p_{{\mathrm{H}}_2}$ is the partial pressure of H₂ [Pa]. This situation corresponds to the “slow-reaction regime” in which residence time becomes the key design parameter, because the overall reaction rate is largely defined by the reaction rate in the bulk liquid and thus the reaction time can be effectively approximated by the residence time, i.e., how long reactants remain in the continuous reactor. , The developed kinetic model exhibited excellent agreement with the experimental data, demonstrating that high formate yields can be achieved in a CSTR through residence-time optimization. Furthermore, phase separation and catalyst recycling were successfully demonstrated within the same flow system, establishing a rational and scalable foundation for the industrial implementation of triphasic hydrogenation.

Results and Discussions

Kinetic Modeling Using Batch Reaction Data

As the initial step of this study, we developed and validated a kinetic model for the hydrogenation of KHCO₃ to potassium formate (HCOOK) (eq ) based on general reaction kinetics in gas–liquid biphasic systems

$${\mathrm{KHCO}}_3+{\mathrm{H}}_2\leftrightarrows \mathrm{HCOOK}+{\mathrm{H}}_2\mathrm{O}$$ 1

As illustrated in Figure , under the present reaction conditions, the overall reaction rate was assumed to be governed by the catalytic reaction in the bulk liquid phase, while mass transfer across the gas and liquid films was balanced with the bulk reaction rate. Based on the molar balance, eq is obtained with the assumptions that the hydrogen absorption rate into the liquid phase is equal to the formate production rate as well as the potassium bicarbonate consumption rate

$$-r_{{\mathrm{H}}_2}=k_G·a_b·(p_{{\mathrm{H}}_2}-p_{{\mathrm{H}}_2,i})=k_L·a_b·(C_{{\mathrm{H}}_2,i}-C_{{\mathrm{H}}_2})=k_1·C_{{\mathrm{H}}_2}·C_{{\mathrm{KHCO}}_3}-k_2·(C_{{\mathrm{KHCO}}_3[0]}-C_{{\mathrm{KHCO}}_3})$$ 2

Here, the H₂ consumption rate per unit volume was expressed as $-r_{{\mathrm{H}}_2}$ [mol m^–3 s^–1]. The parameters k _G [mol m^–2 s^–1 Pa^–1] and k _L [m s^–1] represent the gas- and liquid-side mass transfer coefficients, respectively. The hydrogen partial pressure is denoted as $p_{{\mathrm{H}}_2}$ [Pa], and the gas–liquid interfacial area is represented by a _b [m² m^–3]. Moreover, the net reaction rate in the bulk liquid phase is determined by the forward and reverse reaction rates, where k ₁ [mol m^–3 s^–1] and k ₂ [s^–1] are the respective rate constants. In the reverse reaction term, the concentration of water was assumed to remain effectively constant because water was present in large excess in the aqueous phase; therefore, the reverse reaction was treated as pseudo-first-order with respect to formate concentration. This approximation may not hold if the water content changes significantly. The subscript “i” refers to the concentration or partial pressure at the gas–liquid interface (film region), while “ $C_{{\mathrm{KHCO}}_3\left[0\right]}$ ” denotes the initial concentration of KHCO₃ in the aqueous phase.

The equilibrium relationship between the partial pressure of hydrogen in the gas phase and its concentration in the liquid phase is described by Henry’s law, as shown in eq :

$$p_{{\mathrm{H}}_2,i}=H_{{\mathrm{H}}_2}·C_{{\mathrm{H}}_2,i}$$ 3

Here, $p_{{\mathrm{H}}_2,i}$ represents the partial pressure of hydrogen in the gas phase [Pa], $C_{{\mathrm{H}}_2,i}$ denotes the concentration of dissolved hydrogen in the liquid phase [mol m^–3], and $H_{{\mathrm{H}}_2}$ is the Henry’s constant for hydrogen [Pa m³ mol^–1]. This equation quantitatively describes the solubility equilibrium of H₂, indicating that its partial pressure is directly proportional to its concentration in the liquid phase.

The reverse reaction rate is defined as being proportional to the forward reaction rate constant, and the proportionality factor α is given by the following ratio (eq ):

$$\alpha =\frac{k_2}{k_1}$$ 4

Substituting eq into eq yields a function of $C_{{\mathrm{H}}_2,i}$ , eliminating $p_{{\mathrm{H}}_2,i}$ . Further mathematical transformation leads to eq , in which the consumption rates of KHCO₃ and H₂ are equal (see SI for detailed derivation):

$$-r_{{\mathrm{H}}_2}=-r_{{\mathrm{KHCO}}_3}=\frac{p_{{\mathrm{H}}_2}·C_{{\mathrm{KHCO}}_3}+\mathit{C}·(C_{{\mathrm{KHCO}}_3}-C_{{\mathrm{KHCO}}_3[0]})}{\mathit{A}·C_{{\mathrm{KHCO}}_3}+\mathit{B}}$$ 5

where the parameters are defined as

$$\mathit{A}=\frac{1}{a_b·k_G}+\frac{H_{{\mathrm{H}}_2}}{a_b·k_L}[{\mathrm{Pa\; s\; mol}}^{-1}]$$

$$\mathit{B}=\frac{H_{{\mathrm{H}}_2}}{k_1}[\mathrm{Pa\; s}]$$

$$\mathit{C}=H_{{\mathrm{H}}_2}·\alpha [\mathrm{Pa}]$$

Here, $‐r_{\mathrm{K}\mathrm{H}\mathrm{C}{\mathrm{O}}_3}$ [mol m^–3 s^–1] represents the consumption rate of KHCO₃ per unit liquid volume.

To evaluate the proposed kinetic model, batch hydrogenation of KHCO₃ was conducted under various conditions, including substrate concentration ( $C_{\mathrm{K}\mathrm{H}\mathrm{C}{\mathrm{O}}_3\left[0\right]}$ = 2.0–3.6 mol L^–1), catalyst concentration (C _{Ru catalyst} = 60–240 μmol L^–1), and H₂ pressure (30–50 bar), as summarized in Table . Formate was obtained in comparable yields (65–75%) after 2.5 h (Table ), while the reaction rates varied depending on the conditions (see Figure S1 for raw reaction rate data). The kinetic parameters A , B , and C were estimated by fitting the experimental data to eq using nonlinear least-squares optimization, which minimized the residuals between the measured and simulated KHCO₃ concentrations over time. To ensure an accurate representation of the primary reaction phase, fitting was performed using conversion data up to 10% below the final yield. No side products were detected through the experiments, confirming the high selectivity of the hydrogenation. As shown in the parity plot (Figure ), the model demonstrated excellent predictive performance under all tested conditions, with data points closely aligned along the diagonal (y = x), indicating strong agreement between experimental results and model predictions. For example, at a Ru complex 1 of 60 μmol L^–1, the fitted parameters were A = 3.47 × 10⁶ [Pa s mol^–1], B = 9.66 × 10⁹ [Pa s], and C = 6.36 × 10⁵ [Pa].

1. Summary of the Experimental Conditions Used for Kinetic Validation,,,,

entry	$C_{\mathrm{K}\mathrm{H}\mathrm{C}{\mathrm{O}}_3\left[0\right]}$ [mol L–1]	C Ru catalyst [μmol L–1]	H2 pressure [bar]	formate yield after 2.5 h [%]
1	3.6	120	40	75
2	3.6	60	40	70
3	3.6	240	40	72
4	3.6	120	30	65
5	3.6	120	50	73
6	2.0	60	50	66

^aBatch hydrogenation of KHCO₃ was performed at varying substrate concentrations, catalyst loadings, and H₂ pressures (entries 1–6).

^bReaction conditions: the organic phase contained 60–120 μmol L^–1 Ru complex 1, 54 mmol L^–1 trioctylmethylammonium chloride, and 6 mmol L^–1 tris­(2,4-di-tert-butylphenyl)­phosphite; the aqueous phase contained 2.0–3.6 mol L^–1 KHCO₃.

^cH₂ pressure was 30–50 bar.

^d50 mL toluene and 50 mL aqueous.

^eReaction time = 90–120 min.

3.

Parity plot of measured versus simulated concentrations of KHCO₃ under various hydrogenation conditions. The close alignment with the diagonal line confirms the validity of the proposed kinetic model.

Application of Kinetic Parameters to Predict CSTR Performance

Based on the reaction kinetics established through batch experiments, a kinetic model for a CSTR was developed. Under steady-state conditions and assuming perfect mixing, the material balance for KHCO₃ in the CSTR is expressed as follows:

$$F_{A0}-F_A=v(C_{\mathrm{i}\mathrm{n}}-C_{\mathrm{o}\mathrm{u}\mathrm{t}})=-r_{A}V$$ 6

Where F _{A 0} and F _A are the inlet and outlet molar flow rates [mol s^–1], C _in and C _out are the inlet and outlet concentrations [mol m^–3], v is the volumetric flow rate [m³ s^–1], V is the reactor liquid volume [m³], and τ = V/v is the residence time [s]. Substituting eq into eq and rearranging yields eq :

$$C_{\mathrm{i}\mathrm{n}}-C_{\mathrm{o}\mathrm{u}\mathrm{t}}=\tau \frac{p_{{\mathrm{H}}_2}{C}_{\mathrm{o}\mathrm{u}\mathrm{t}}+\mathit{C}(C_{\mathrm{o}\mathrm{u}\mathrm{t}}-C_{\mathrm{i}\mathrm{n}})}{\mathit{A}{C}_{\mathrm{o}\mathrm{u}\mathrm{t}}+\mathit{B}}$$ 7

Further algebra manipulation of eq leads to eq :

$$C_{\mathrm{o}\mathrm{u}\mathrm{t}}=\frac{\mathit{A}{C}_{\mathrm{i}\mathrm{n}}-\{\mathit{B}+\tau (p_{{\mathrm{H}}_2}+\mathit{C})\}+\sqrt{{\{\mathit{B}+\tau (p_{{\mathrm{H}}_2}+\mathit{C})-\mathit{A}{C}_{\mathrm{i}\mathrm{n}}\}}^2+4\mathit{A}{C}_{\mathrm{i}\mathrm{n}}(\mathit{B}+\tau \mathit{C})}}{2\mathit{A}}$$ 8

The single-pass conversion of KHCO₃ is defined as x = (C _in – C _out)/C _in, corresponding to the formate yield under the assumption of 100% selectivity, which was confirmed by our experiment within the detection limit. Subsequently, the relationship between the residence time and formate yield was predicted using the developed kinetic model, as illustrated in Figure . These insights provided the basis for applying the model to a CSTR reactor, enabling further evaluation under flow conditions.

4.

Predicted formate yield as a function of residence time in CSTR on the overall reaction rate equation with constants A = 3.47 × 10⁶ [Pa s mol^–1], B = 9.66 × 10⁹ [Pa s], and C = 6.36 × 10⁵ [Pa]. Conditions: The organic phase contained 60 μmol L^–1 Ru complex 1, 2 mol L^–1 KHCO₃ and 50 bar of H₂ pressure.

Flow Reaction

Next, we prepared a CSTR bearing a custom-built 200 mL vessel with a height of 90 mm and a diameter of 56 mm, equipped with a 20 mm four-blade stirring wing positioned at the interface between the 20 mL of organic and 20 mL of aqueous phases (Figures and S2). Prior to each experiment, the reactor was flushed with N₂, the pressure of which was adjusted with a rotameter. The organic phase was prepared in a glovebox under a nitrogen atmosphere using degassed toluene containing Ru complex 1, trioctylmethylammonium chloride, and tris­(2,4-di-tert-butylphenyl)­phosphite. The aqueous phase was prepared using standard Schlenk techniques with degassed distilled water containing 2 mol L^–1 of KHCO₃ under argon. Those prepared solutions were filled into canisters, which were connected to the stirred tank, and a total of 40 mL of the organic and aqueous phases was introduced. The mixture in the reactor was pressurized with H₂ to 50 bar and then heated to 90 °C and stirred for a designated period. Continuous operation was initiated by simultaneously feeding and withdrawing both phases at controlled flow rates. The resulted mixture was collected to a phase separator to sample the aqueous phase containing formate while recycling the organic phase, which contains the catalyst.

5.

(A) Simplified diagram of the CSTR. (B) Schematic diagram of the stirred tank reactor for the hydrogenation of KHCO₃. (C) Stirring wing.

The mixing behavior within the reactor was initially assessed by direct visual inspection, made possible by the incorporation of a sapphire view cell into the reactor design (Figure B). This approach is notably rare in high-pressure CSTRs, where internal observation is typically restricted due to system constraints. The ability to visually monitor the mixing process under actual operating conditions allowed for direct evaluation of phase behavior and facilitated the optimization of both the stirring blade position and rotation speed. This was particularly important in the three-phase reaction system, where efficient interfacial mixing is essential for enhancing mass transfer between phases and improving the overall reaction efficiency. At a stirring rate of 300 rpm, the organic and aqueous phases remained clearly separated, while uniform mixing of the two phases was observed at stirring rates of 600 rpm or higher (Figure ). Moreover, stirring at 800 rpm was necessary to achieve homogeneous mixing and sample the mixture with the desired ratio of the organic to aqueous phase.

6.

Liquid-phase behavior of toluene and water at 300 rpm (A) and 600 rpm (B) in the CSTR. Those figures were taken from a view port of the CSTR.

Next, to evaluate the effect of the phase volume ratio in the reactor, the ratio between the organic and aqueous phases was varied upon different catalyst loadings while maintaining a constant total volume of 40 mL and a constant residence time (Table ). At a 1:1 organic-to-aqueous phase ratio and a residence time of 60 min, the collected aqueous phase through pump 3 (Figure A) showed a 26% formate yield (entry 1). On the other hand, when the volumes of each phase were adjusted to 8 mL of organic phase and 32 mL of aqueous phase, the formic acid yield decreased to 18% (entry 2), even at a catalyst loading higher than that in entry 1 (entry 2). The vertical position of the stirring blades close to the phase interface (under the static caseorganic phase above and aqueous phase below) was found at the 1:1 ratio, presumably enhancing interfacial mass transfer and resulting in improved overall catalytic reaction efficiency (see Figure S3). Thus, the 1:1 phase ratio was selected for subsequent experiments. The effects of the phase volume ratio on reaction performance were systematically evaluated in a batch reactor, as summarized in Figures S4 and S5 and Tables S2 and S3.

2. Effect of Phase Volume Ratio in the Reactor with Different Ru Complex Loading,,,

entry	organic phase [mL]	aqueous phase [mL]	Ru complex 1 [μmol L–1 ]	formate yield [%]
1	20	20	30	26
2	8	32	50	18

^aOrganic phase: [Ru complex 1] = 60 or 250 μmol L^–1 (Overall Ru concentration in the reactor: 30 μmol L^–1 (entry 1), 50 μmol L^–1 (entry 2)), [trioctylmethylammonium chloride] = 54 mmol L^–1, [tris­(2,4-di-tert-butylphenyl)­phosphite] = 6 mmol L^–1.

^bAqueous phase: [KHCO₃] = 2 mol L^–1.

^cH₂ pressure was 50 bar.

^dResidence time: 60 min.

^eCalculated based on the total liquid volume (organic + aqueous phase) in the reactor.

To evaluate the reusability of the catalyst, recycling experiments were performed by introducing 20 mL of each of the organic and aqueous phases into the reactor (Figure ). Initially, the hydrogenation of KHCO₃ was first carried out as a batch reaction for 90 min to reach equilibrium, after which the system was switched to continuous operation with a residence time of 90 min. Following each reaction, the organic phase was collected, diluted to its original volume, and stored for 20 h. The recovered organic phase was then used in the subsequent continuous-flow experiment combined with a fresh aqueous phase.

7.

Catalyst recycling experiment on the continuous flow reaction. Reaction conditions: organic phase contained 60 μmol L^–1 Ru complex 1, 54 mmol L^–1 trioctylmethylammonium chloride, and 6 mmol L^–1 tris­(2,4-di-tert-butylphenyl)­phosphite; aqueous phase contained 2 mol L^–1 KHCO₃. H₂ pressure was 50 bar. Phase volumes in the reactor: 20 mL of organic and 20 mL of aqueous. Residence time: 90 min.

The catalyst storage conditions were optimized through both batch and flow hydrogenation reaction (see Tables S5 and S6). A clear difference in catalyst stability was observed depending on the storage atmosphere. The Ru catalyst stored under a hydrogen atmosphere exhibited higher stability than that stored under nitrogen. ¹H NMR measurements revealed that pressurization of the organic phase with H₂ led to the formation of Ru dihydrido and formate complexes with suppressing the decomposition of complex 1 (Figure S6). Moreover, the stability of the isolated Ru formate complex was examined (Figure S8). A solution of the Ru formate complex in C₆D₆ was stirred with 1 mL of a 5 mol L^–1 D₂O solution of HCOOK for 30 min. The obtained organic phase was subsequently divided into two portions and stored for 4 days under either a hydrogen or nitrogen atmosphere at 6 bar. For the sample stored under nitrogen, decomposition of Ru complexes was observed (Figure S8a), whereas the sample stored under hydrogen retained complex 1 and the formate complex (Figure S8b). Furthermore, the addition of 6 mmol L^–1 tris­(2,4-di-tert-butylphenyl)­phosphite to the organic phase significantly increased the formate yield in the subsequent reaction, likely due to suppression of oxidation of the PNP ligand. Under the optimized condition with a residence time of 90 min, the formate yield reached 67–70% over four recycling tests (total TON = 114 × 10³), showing good agreement with the model predictions. Formate yields were calculated using the aqueous phase collected from the reactor after each reaction since it is the yield during the steady-state reaction. As shown in Table S4, the time-dependent formate yields indicate that the formate yield was 39% in the initial aqueous phase collected at 90 min, whereas it increased to 60% in the later aqueous phase collected at 150 min. These results indicate that a certain period is required for the system to reach steady state and that the aqueous phase collected from the reactor represents the composition at a given steady-state condition. Additional catalyst recycling data are provided in the Supporting Information (Tables S7–S10).

To verify catalyst recovery, ICP-MS analysis was conducted to quantify the Ru content in each phase. When the initial Ru concentration in the organic phase was 60 μmol L^–1, the Ru concentration in the organic phase remained unchanged after hydrogenation. However, approximately 3% of the Ru originally present in the organic phase was detected in the collected aqueous phase, which can be attributed to the partial transfer of the organic phase itself into the aqueous phase rather than intrinsic leaching of the Ru catalyst. In Figure , 47 mL of the organic phase was recovered from an initial volume of 60 mL after the first reaction, while an additional 5–8 mL of the organic phase could not be recovered in subsequent recycling runs (Table S10). The loss of the organic phase is attributed to two main factors: partial dispersion of the organic phase into the aqueous phase during the reaction and retention of a portion of the organic phase in the reactor and tubing.

In conclusion, we successfully established a CSTR system for the hydrogenation of KHCO₃ under high-pressure conditions. By visually confirming and optimizing the stirring conditions using a view cell, we ensured efficient mixing, which is essential for maintaining the reaction performance in the triphasic flow system. Furthermore, by analyzing the batch reaction data within a unified kinetic framework that assumes a slow-reaction regime and consideration of the reverse reaction, a consistent set of kinetic parameters was obtained. These parameters were applied to the CSTR model, enabling reliable prediction of the residence time–yield relationship. A kinetic model developed from batch experiments accurately predicted the residence time–yield relationship in the CSTR, with formate yields closely matching the model’s predictions. Regarding scalability, the current model is based on a well-mixed stirred tank and incorporates interfacial and transport parameters and is therefore expected to remain applicable at larger scales provided hydrodynamic and mass-transfer performance are preserved. In practice, this requires maintaining power input per unit volume and matching the overall gas–liquid mass-transfer coefficient under geometric similarity. For pilot or industrial translation, we recommend verifying the gas–liquid mass-transfer coefficient and assessing dispersion characteristics to ensure stable operation. These measures provide a practical framework for scaling continuous triphasic hydrogenation processes based on CO₂ utilization.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.6c00388.

• Experimental procedures, kinetic model derivation, supplemental results for batch and flow hydrogenation, including liquid phase-ratio effects and steady-state confirmation, catalyst storage and recycling data, and NMR spectra (PDF)

#.

M.I. and G.A. contributed equally.

The authors declare no competing financial interest.