Liquid palladium for high-turnover carbon-carbon bond formation

Abstract

Carbon-carbon (C─C) bond formation is a key step in diverse chemical processes and requires high-performance catalysts to enable energy-efficient technologies. Here, we present liquid Pd catalysts, formed by dissolving Pd in liquid Ga, for high-turnover C─C coupling reactions. The liquid Pd catalyst achieved a turnover frequency of 2.5 × 10⁸ hour⁻¹ for a model coupling reaction at 70°C, surpassing all reported Pd catalysts by 1000-fold. Our results show that Pd atoms in the Ga matrix are liquid-like, exhibiting unique electronic and interfacial properties that substantially lower the energy barrier and enhance reaction kinetics. The system retained full activity over five cycles and showed no Pd leaching, highlighting the transformative potential of liquid-phase metals to advance high-throughput and sustainable C─C bond-forming strategies.

Liquid Pd redefines C─C coupling, achieving unprecedented turnover frequency.

INTRODUCTION

Carbon-carbon (C─C) bond formation via coupling reactions is a cornerstone process in synthetic organic chemistry and is vital for fabricating complex molecular structures prevalent in pharmaceuticals, agrochemicals, and materials science (1, 2). The efficiency of these reactions hinges predominantly on the performance of the catalysts used, with Pd being a frontrunner owing to its ability to facilitate the formation of C─C bonds (3). As homogeneous catalysts, soluble Pd complexes have long dominated the landscape of coupling reactions despite their known challenges with product purification, reusability, and sustainability (4). Heterogeneous Pd catalysts, such as Pd complexes or Pd nanoparticles anchored on a support matrix, can address these challenges but have drawbacks like complex processing steps, instability, and Pd leaching (5). Furthermore, the mechanism of the heterogeneous reaction remains unclear, with ongoing debate about whether it occurs on the surface of the catalysts (heterogeneous phase) or via Pd leaching from the catalysts into solution (homogeneous phase) (6). In terms of activity, state-of-the-art Pd catalysts are reported to offer turnover frequencies (TOFs) of ~75,000 hour⁻¹ for homogeneous Pd catalysts and 500,000 hour⁻¹ for heterogeneous Pd catalysts (7, 8).

Recently, there has been a growing interest in the concept of “liquid metal catalysts” that operate at moderate to low temperatures (9, 10). To achieve this, the dissolution of catalytically active metals in low-melting metals (e.g., Ga, Sn, and Bi) has been explored in different contexts, including the work of Upham et al. (11). Unlike molten metal alloys used for chemical conversion at high temperatures (12–14), these systems are likely to accommodate a wide range of chemical reactions that are feasible only at low temperatures. For instance, using the natural solubility of Pt in liquid Ga, which has a melting point of ~29.8°C, we previously demonstrated a Ga─Pt system that exhibited exceptional catalytic performance for different chemical reactions, including electrochemical, redox, and enzymatic reactions, in the temperature range of 50° to 70°C (9). This study further revealed that Pt dissolved in liquid Ga remains atomically dispersed without forming Pt─Pt bonds, resulting in a unique liquid-like state for Pt. Notably, the activity of the Ga─Pt system was found to be ~1000 to 10,000 times greater than that of solid Pt catalysts, including single-atom Pt catalysts. This enhancement, driven by mobile active atoms with distinct electronic properties, is unattainable in the solid state. In a system with two active metals, Sn and Ni in liquid Ga, we demonstrated the dynamic and autonomous positioning of Sn, Ni, and Ga, which enabled specific alignments with the reactants, thereby influencing the reaction pathways and selectivity (10). Notably, we found that the key to catalytic activity was the activation of adjacent Ga atoms by Pt in the Ga─Pt system and the dynamic atomic configuration induced by Ni in the Ga─Sn─Ni system. Besides, liquid metal–based catalysts have been applied for CH₄ and CO₂ conversion to form solid carbon products (15, 16). Despite this progress, such liquid catalysts remain unexplored for mainstream organic reactions such as cross-coupling reactions.

Continuing this line of research, here, we demonstrate liquid Pd catalysts, i.e., Pd dissolved in liquid Ga (Ga─Pd), for Suzuki-Miyaura cross-coupling (SMCC) reactions with a TOF in the range of 10⁸ hour⁻¹ without any Pd leaching, thus offering a sustainable and scalable alternative to traditional solid-state catalysts. Our findings provide a detailed account of catalyst synthesis, mechanistic insights, and transformative potential for high-throughput industrial applications. By leveraging the fluidic nature of Pd in a Ga matrix, we reveal a range of reactivity and processing advantages that are difficult to achieve with solid-state catalysts.

RESULTS AND DISCUSSION

Liquid Pd system

A schematic illustration of the preparation of the Ga─Pd system is shown in Fig. 1. We used the natural solubility of Pd in a liquid Ga at a specific temperature, where the Pd atoms are atomically dispersed and remain dynamic in the Ga matrix. With this system, we performed SMCC reactions at 70°C as model C─C coupling reactions. Our preliminary investigations into the solubility parameters and phase diagram (Fig. 1B and fig. S1) of Pd in liquid Ga suggested a Pd solubility of ~0.005 atomic % (at %) at 50°C and ~0.008 at % at 70°C (17, 18). On the basis of this information, we prepared the Ga─Pd system by dissolving a known amount of Pd in liquid Ga at ~400°C, followed by cooling to 70°C to test its catalytic activity for SMCC reactions. After reaching equilibrium at 70°C, a small portion of the Ga─Pd stock was withdrawn and analyzed by inductively coupled plasma mass spectrometry (ICP-MS), which confirmed a solubility of ~0.01 at % for Pd dissolved in liquid Ga. The slight discrepancy between the Pd solubility values obtained from ICP-MS analyses and those estimated via solubility parameters may stem from the fact that the constants for the solubility parameters were derived for a temperature range of 250° to 700°C (17, 18).

Fig. 1. Experimental and computational analysis of the Ga─Pd catalysts.

(A) Schematic illustration of the preparation and characteristics of the Ga─Pd system. (B) Phase diagram of Ga─Pd depicting the solubility of Pd in liquid Ga across a range of temperatures and compositions. (C) TEM-EDX image of Ga─Pd nanoparticles. (D and E) Atomic configurations of Pd dispersed in Ga at an interface obtained from AIMD simulations, with a panel showing the side and top views, respectively. (F) Pairwise probability diagram showing the distribution and interaction probabilities between Pd and Ga atoms at the interface. (G) Atomic density distribution for all atoms and Pd only as a function of the z position for the interfacial system at 100 ps.

Although the SMCC reactions were performed with bulk Ga─Pd droplets (a volume of 100 μl), we performed transmission electron microscopy coupled with energy-dispersive x-ray spectroscopy (TEM-EDX) analyses on Ga─Pd nanoparticles obtained from the bulk droplets to investigate the dispersion state of Pd in liquid Ga. The Pd atoms were well dispersed in Ga without any clustering (Fig. 1C). In addition, drops of Ga─Pd were flash frozen from the stock at 70°C in liquid nitrogen, and x-ray photoelectron spectroscopy was performed (fig. S2). The binding energy difference observed between the bulk Pd and Pd atoms in the Ga matrix indicates the atomic dispersion of Pd in liquid Ga (see note S1 along with Bader charge analyses). X-ray diffraction analysis (note S2 and fig. S3) further confirmed that the Pd atoms remained atomically dispersed and liquid in the Ga matrix.

Furthermore, ab initio molecular dynamics (AIMD) simulations were performed to obtain atomic insights into the configuration of Pd in the Ga matrix. To investigate whether Pd atoms aggregated or not, systems with 398 Ga and 2 Pd atoms were assembled, and AIMD simulations were run in the bulk and at the interface in the z-direction. After equilibrating for 100 ps (Fig. 1D), a configuration was achieved where Pd─Pd distances showed a minimum distribution of ~4.6 Å (typically 2.75 Å in pure Pd), with no direct contacts observed in the bulk or at the interface [similar to Pt dissolved in Ga (9)]. Previously, we reported that Pt or Ni dissolved in Ga tends to be located one atomic layer below the interface (9, 10). Here, we observed a similar behavior for Pd in Ga, with Pd found near the bottom of the interfacial layer of Ga and never directly at the interface. In contrast, Ga─Ga (2.9 Å) and Ga─Pd (2.7 Å) exhibited similar close contacts in both cases (bulk and interface). The pairwise probability distribution functions and atomic density profiles (presented in Fig. 1, F and G, respectively) further validate the above results.

Catalytic performance

The catalytic behavior of the Ga─Pd system was evaluated using a range of model SMCC reactions, where the cross-coupling reaction between iodobenzene and phenylboronic acid (batch-type reaction) was used as a standard to optimize the experimental parameters (Fig. 2, A and B). On the basis of the primary screening of different bases and solvents (tables S1 and S2), K₂CO₃ was identified as the most efficient base and ethanol as the optimal solvent for the SMCC reaction. The products were analyzed using gas chromatography–mass spectrometry (GC-MS) and nuclear magnetic resonance (¹H and ¹³C{¹H} nuclear magnetic resonance) spectroscopy and compared with previous results (19, 20). The obtained results are compiled in figs. S4 to S15. The catalytic potential of the Ga─Pd system was further expanded to encompass a diverse array of aryl halides (table S3). All these reactions afforded the corresponding biphenyl compounds in excellent yields under optimized conditions, regardless of whether the substituents were electron donating (─CH₃, ─OCH₃, ─OH, and ─NH₂) or electron withdrawing (─NO₂, ─CN, ─COOEt, and ─COMe). Notably, while iodobenzene and bromobenzene exhibited robust reactivity, chlorobenzene displayed markedly lower activity, yielding only trace amounts of products. This lower activity can be attributed to the stronger C─Cl bond (~399 kJ mol⁻¹) in chlorobenzene, which required harsher conditions (higher temperatures and stronger bases) for dissociation compared to C─I and C─Br bonds under mild conditions used in this study (21).

Fig. 2. Detailed analysis of the Ga─Pd–catalyzed SMCC reactions.

(A) Schematic illustration showing the experimental setup for the SMCC reaction catalyzed by Ga─Pd. (B) SMCC reaction showing the coupling of aryl halides with phenylboronic acid. (C) Stability (metal ion leaching) tests of Ga and Pd at different time intervals during the reaction. h, hours. (D) Kinetic analysis of the Ga─Pd catalyst for the SMCC reaction over time and the corresponding Arrhenius plot (E) showing the activation energy (E_a) of the catalyst. (F) Reusability of the catalyst over five cycles.

The kinetics and catalytic efficiency of the Ga─Pd system were also assessed for the standard SMCC reaction. Initial rates were measured at various temperatures and then used to generate Arrhenius plots, from which the activation energy (E_a) was calculated. For the Ga─Pd system, the E_a was found to be 19.1 kJ mol⁻¹ (Fig. 2E), which is much lower compared to those obtained with previously reported catalysts (table S4). For example, a Pd-bearing intermetallic catalyst (19), Y₃Pd₂, exhibited an E_a of 48.4 kJ mol⁻¹. Therefore, such a low E_a for the Ga─Pd system suggests a high reaction rate and enhanced process efficiency.

Control experiments using only Ga and solid Pd (ingots) were used to identify their roles in the SMCC reaction. For the Ga─Pd catalyst, the maximum yield (99%) of products was obtained after 12 hours. For the same duration, control experiments were performed separately using only Ga and Pd. The use of Ga alone and solid Pd (ingot) alone resulted in yields of only 13 and 3%, respectively. These results highlight the enhanced catalytic efficiency of the catalyst, which benefits from the synergistic interaction between Ga and Pd (see later discussion).

Considering the number of surface Pd atoms on the Ga─Pd droplet (note S3), the TOF was estimated to be 2.5 × 10⁸ hour⁻¹ for the standard SMCC reaction (iodobenzene and phenylboronic acid). This value is substantially greater than the activity of ~8.1 × 10⁵ hour⁻¹ reported for the highly efficient silica-based Pd(II) catalyst (22) or those of other reported catalysts such as Pd@CNPC, SiliaCat Pd(0), polymer-anchored Pd(II), and Pd/Fe₃O₄@PC, which typically have TOFs in the range of ~10³ to 10⁴ hour⁻¹(table S5). Therefore, the Ga─Pd system results in a TOF nearly three orders of magnitude greater than those of the reported catalysts, implying its superior performance and potential to considerably advance catalytic processes involving C─C bond formation.

Resistance to Pd leaching

The standard SMCC reaction between iodobenzene and phenylboronic acid was conducted in an ethanolic solution where the reactants were mixed with the Ga─Pd catalyst. To assess the potential leaching of the corresponding elements from the Ga─Pd catalyst during the reaction, an aliquot of the reaction mixture was periodically sampled. The collected samples were subjected to ICP-MS analyses to determine the concentrations of Ga and Pd in the solution at various time intervals. The ICP-MS results, shown in Fig. 2C, demonstrated a negligible change in the Ga concentration and no change in the Pd concentration over time, confirming the stability of the Ga─Pd catalyst, without any notable leaching of Ga and no observed leaching of Pd. Furthermore, considering the small amount of leached Ga³⁺ ions (in the range of ~0.8 parts per million), we further investigated its role in the coupling reactions. Control experiments were performed using a solution of Ga³⁺ ions with an equivalent concentration according to the leaching analysis. However, the Ga³⁺ solution was found to be ineffective for the coupling reactions.

The problem of Pd leaching from conventional heterogeneous catalysts poses two major challenges: (i) It obscures the true catalytic mechanism, and (ii) it leads to the contamination of the final products (23). The latter is particularly concerning for the pharmaceutical industry, as toxic Pd ions compromise product purity and complicate regulatory compliance. In this context, Ga─Pd catalysts (as demonstrated above) offer a stable, nonleaching (Pd ions) alternative, providing key advantages for industrial applications requiring stringent safety and purity standards such as in pharmaceutical synthesis.

Recyclability

Recyclability is another crucial factor for probing the practical suitability of a catalyst because it ensures sustainability over extended periods while reducing the need for frequent replacements. In this context, the recyclability of Ga─Pd was evaluated using the standard SMCC reaction. The Ga─Pd system was investigated for five consecutive cycles of reactions, and no loss in activity was observed (Fig. 2F). After five cycles, the Ga─Pd system was analyzed to confirm its compositional stability. ICP-MS analyses confirmed that there was no leaching of Pd atoms from the Ga matrix. The TEM-EDX images (Fig. 1C and fig. S16) of the catalyst in the form of nanoparticles, before and after five runs, revealed that the dispersion of Pd atoms remained consistent without any clustering. These results affirm the outstanding recyclability of the Ga─Pd system, with its efficacy maintained across multiple cycles without any observable performance degradation. We also conducted a preliminary cost-benefit analysis and assessed the environmental impact of our catalyst. Compared to several high-performing catalysts reported in the literature, the Ga─Pd catalyst demonstrated superior performance across key metrics, including TOF, recyclability, and environmental sustainability, highlighting its potential for industrial applications (note S4 and table S6).

Proposed reaction pathways

We constructed multiple configurations (AIMD) of aromatic rings at the interface to investigate the possible ways in which the SMCC reaction could occur at the Ga─Pd interface. For the oxidative addition step, the most stable configuration was the aromatic ring interacting with a Ga atom adjacent to the Pd atom, with the I atom interacting with two Ga atoms above an adjacent Pd atom (Fig. 3A). No stable configurations with the aromatic ring interacting directly with a Pd atom were found, as the ring would spontaneously move to an adjacent Ga atom in both MD simulations and geometry optimizations. In the next step, a deprotonated ethanol molecule (represented by CH₃CH₂OK) displaced the I atom above the Pd atom adjacent to the aromatic ring. In the subsequent transmetalation step, the aromatic organoboron ion was exchanged with the ethoxide ion, leaving two aromatic rings on the surface.

Fig. 3. Proposed pathway of the SMCC reaction and the electronic factors influencing the high activity of Ga─Pd catalysts.

(A) Energy profile diagram illustrating the key steps in the SMCC reaction, namely, oxidative addition, transmetalation, and reductive elimination, and highlighting the energy changes associated with each step catalyzed by the Ga─Pd system. (B and C) Detailed visualization of the reductive elimination step, comparing the process occurring on a single and different Pd sites, respectively. (D) Calculated DOSs of the Ga─Pd system and bulk Pd(111), reflecting a modified electronic environment for Pd in Ga─Pd compared with that of pure Pd.

In the proposed reaction pathway, two Pd atoms are present near the interface with a “bridging” Ga atom between them and an additional seven to eight Ga atoms surrounding each Pd atom with an average Pd─Ga distance of ~2.7 Å. While the bridging Ga atom has a partial charge of +0.25e⁻ and slightly shorter distances to Pd (~2.5 Å), the remaining neighboring Ga atoms maintain an average charge of +0.10e⁻, with atoms closer to the interface exhibiting larger partial charges than those closer to bulk Ga (fig. S17). This creates a cluster of adjacent Ga atoms with partial positive charges at the interface that would not occur in the absence of Pd. In the oxidative addition step, the I atom interacts with the bridging Ga atom and another Ga atom above one Pd atom, while the aromatic C atom interacts with a Ga atom above the other Pd atom. While the three Ga atoms in direct contact with I and C atoms become more positively partially charged, the partial charge on the Pd atoms does not change substantially. In the next step, the ethoxide oxygen takes the place of iodide, and as oxygen is more electronegative than iodine, the two Ga atoms become more partially positive to compensate, with the Ga atom in the bridging position moving to a position directly above Pd and another Ga atom taking its place in the bridging position between two Pd. This demonstrates the ability of the system to adjust dynamically to different steps in the reaction, and here, Pd does become slightly more partially negative in response. In the transmetalation step, the ethoxide oxygen interacting with two Ga atoms is replaced by an aromatic carbon interacting with a single Ga atom above Pd. While the partial charges on the Pd atoms are identical at this step, the partial charges on the Ga atoms interacting with the aromatic C atoms differ by 0.07e⁻ (fig. S17). This asymmetry is due to the differing arrangement of Ga atoms surrounding each Pd with varied Ga─Pd distances while retaining the average distance of 2.7 Å for each Pd. In the reductive elimination step, the Ga─Pd system returns to an electronic and structural configuration similar to the reactants, with an asymmetry around the Pd sites because of the nature of the dynamic liquid metal system.

Earlier, it was suggested that for solid-state Pd catalysts, both aromatic rings should interact directly with a single surface Pd before the final reductive elimination step (19). We therefore examined configurations with both aromatic rings on a single Pd, as well as configurations with each aromatic ring on a different Pd. In every case, during AIMD simulations, the rings spontaneously moved from Pd to Ga adjacent to Pd, converging to two distinct configurations: each ring interacting with separate Ga above a single Pd (Fig. 3B) or each ring interacting with Ga above a different Pd (Fig. 3C). These configurations are in line with a recent study showing that aluminum interfaces with buried transition metals could catalyze Suzuki coupling reactions (24). For both configurations, the reductive elimination step was highly energetically favorable: −2.3 ± 0.4 eV for the configuration involving a single Pd and −1.3 ± 0.7 eV for the configuration involving two Pd. However, the initial reactant configuration involving two Pd ions is more energetically favorable by ~1 eV.

Factors influencing the high catalytic activity

The unprecedented catalytic activity of the Ga─Pd system can arise from multiple factors, such as a favorable electronic structure, activation of surrounding Ga atoms, and dynamic Pd atoms that continuously refresh the interface. For the mobile atoms, we previously developed a simplified model considering the change of reactant concentrations with time over a dynamic surface with mobile active sites; details can be found therein (9). Briefly, we compared a solid-state Pd catalyst and liquid Ga─Pd system, as schematically presented in fig. S18. For the Ga─Pd system, we considered that the number of surface-Pd atoms remains statistically constant over time, as the vertical movement of Pd atoms to and from the surface is mutually balanced. The lateral (x, y) components of the diffusion coefficient for a single Pd atom in the Ga─Pd system, derived from AIMD simulations, were used to estimate the lateral diffusion length. From this, the number of new positions per second was determined to be 1.9 × 10⁵/s (note S5), which allows the active sites to avoid localized reactant depletion (concentration well per second) (25, 26). As a result, the mobility of Pd in liquid Ga─Pd facilitates reaction kinetics by relocating active sites to regions with higher reactant concentrations (C_r). In contrast, solid Pd or Pd in a solid matrix suffers from immobility, leading to suppressed activity because of reactant depletion.

Density of states (DOS) calculations (Fig. 3D) show that the interactions between Pd and surrounding Ga atoms reduce the Pd state density below the Fermi level compared to bulk Pd. Bader charge analyses provided insights into the electronic charge distribution within this dynamic liquid system. In these analyses, the Pd atoms present a negative Bader charge of ˗0.72e⁻ compared to 0e⁻ for pure Pd, indicating electron acquisition from neighboring Ga atoms, which has a positive Bader charge of +0.10e⁻ (against 0e⁻ for pure Ga). This is corroborated by the x-ray photoelectron spectroscopy analysis (fig. S2), which reveals a greater binding energy for Pd (3d_5/2 at 336.4 eV) in the Ga─Pd system than for the bulk Pd (3d_5/2 at 335.0 eV). This shift in the binding energy can be attributed to the local environment of Pd, which is influenced by Ga. This shift is also consistent with the higher electrophilicity of Pd relative to Ga on the Pauling scale (Ga = 1.81 and Pd = 2.20) (27), altering the electronic orbitals of Pd. These findings are consistent with previous studies (9, 13) confirming the atomic dispersion of Pd atoms within Ga and highlighting their synergistic properties, which are highly favorable for low-temperature catalysis.

MATERIALS AND METHODS

Preparation of the Ga─Pd catalyst

A Pd ingot weighing ~200 mg was carefully placed inside a glass vial within a glove box, followed by the addition of ~14 g of Ga. The mixture was subsequently heated to ~400°C for 4 hours to ensure complete dissolution of Pd in Ga. The resulting alloy was then allowed to cool and equilibrate at the desired temperature of 70°C before use. Last, a volume of 100 μl was withdrawn from this initial stock solution and used for each SMCC reaction.

General procedure for SMCC reactions

In a typical batch-type reaction, 0.82 mmol of organohalide, 0.92 mmol of aryl boronic acid, and 1.64 mmol of K₂CO₃ were dissolved in 10 ml of ethanol, to which 100 μl of the Ga─Pd (droplet) catalyst was added. The progress of the reaction was initially monitored by thin-layer chromatography. The products were analyzed and quantified by GC, GC-MS, and nuclear magnetic resonance spectroscopies. The recycling experiments were performed by isolating the catalyst after each completion of the reaction, thoroughly washing it with ethanol, and reusing it in a subsequent reaction batch under identical conditions.

Supplementary Materials

This PDF file includes:

Supplementary Materials and Methods

Supplementary Notes S1 to S5

Figs. S1 to S18

Tables S1 to S6