Selective Production of Hydrogen and Lactate from Glycerol Dehydrogenation Catalyzed by a Ruthenium PN³P Pincer Complex

Abstract

In the quest for cheap and abundant feedstocks for sustainable hydrogen production, glycerol is emerging as a cost-effective, promising liquid organic hydrogen-rich carrier (LOHC) that can be catalytically activated to produce hydrogen alongside valuable organic products. Selective catalytic acceptorless dehydrogenation of glycerol to lactate and hydrogen gas was achieved with a maximum turnover number (TON_max) of ca. 1600, using a pincer-type ruthenium(II) complex bearing a bis(aminophosphine)pyridine PN³P ligand as a homogeneous catalyst under moderate reaction conditions (24 h, 140 °C) in the presence of KOH as base. NMR experiments and DFT calculations provided insights into key steps of the catalytic process and the energetics of the proposed reaction pathway.

1. Introduction

The generation of sustainable energy vectors is a topic of great interest to our society because of dwindling fossil fuel sources and rising energy consumption [1]. The use of hydrogen (H₂) as fuel to feed PEM fuel cells for stationary and mobile applications has reached a mature stage of technological readiness, although critical bottlenecks still need to be overcome [2]. A critical issue is to find renewable, cost-effective and abundant sources for hydrogen production, among which water and non-edible biomass derivatives are the most promising. Although research in green hydrogen production from water electrolysis is prominent, a supply of renewable electricity (i.e., solar or wind power) is not always available [3,4,5]. Hydrogen generation from enzymatic biomass fermentation, gasification, supercritical water reforming and aqueous phase reforming by heterogeneous catalysts has been widely studied [6,7]; however, the practical use of these processes, in particular those based on lignocellulosic biomass, is still in its infancy and needs harsh reaction conditions. Thus, it is of interest to explore alternative methods for hydrogen production from renewables under milder conditions, such as those that can be achieved by homogeneous catalysis. Suitable simple hydrogen-rich organic molecules can be used as feedstocks to produce H₂ and added-value organic products by selective bond-breaking and bond-making reactions promoted by homogeneous catalysts [8]. Among these molecules, glycerol, produced as waste by-product from biodiesel industry, is receiving attention as a suitable candidate [9,10,11]. Compared to other alcohols, such as methanol (MeOH) and ethanol (EtOH), glycerol has the advantages of higher boiling point, non-flammability and lower toxicity [12]. Many useful added-value products such as propanediol, dihydroxyacetone and lactic acid (LA) [13,14,15] can be obtained from glycerol, which can also be exploited as the LOHC (liquid organic hydrogen carrier) [16,17,18] for H₂ generation. Lactic acid (LA), commonly obtained from fermentation of carbohydrates and alcohols, is a widely used chemical in food, pharmaceuticals, cosmetics, biodegradable polyethene, detergents and the manufacture of polylactic acid [19,20]. Traditional methods for the manufacturing of LA still require tedious purification and workup, leading to generally low yields and productivities [21].

Heterogeneous catalysis and biocatalysis [22] have been applied for hydrogen production from glycerol [23,24]. Heterogeneous catalysis is promising but generally requires temperature regimes above 350 °C at atmospheric pressure or over 220 °C at pressures > 2 MPa [25,26,27,28]. Recent efforts have been made to develop hydrogen from glycerol by homogeneous catalysis using acceptorless dehydrogenation (AD, Scheme 1), a cost-effective and oxidant-free method for converting (poly)alcohols into carbonyl compounds [29], releasing H₂ as byproduct [30].

Recent reports demonstrated that, by subtle combinations of transition metals and functional ligands, efficient catalytic systems for glycerol AD to LA and H₂ can be obtained [8,31,32]. Some of the most active ruthenium complexes for this reaction are shown in Figure 1 and Table S1 (Supporting Information). For example, Beller and coworkers applied the well-known MACHO-type pincer complex [RuH(Cl)(PNP)(CO)] [PNP = bis(2-(diphenylphosphino)ethyl)amine] for the selective conversion of glycerol to LA in up to 67% yield [33]. Kumar and co-workers synthesized Ru(NNN) complexes stabilized by bis(imino)pyridine and 2,6-bis(benzimidazol-2-yl)pyridine ligands, showing that trans-[RuCl(Bim₂NNN)(PPh₃)₂)]Cl [Bim₂NNN = 2,6-di(1H-benzo[d]imidazol-2-yl)pyridine] gave remarkable efficiency, reaching 90% yield of LA with 98% selectivity [34]. More recently, Li, Xu and coworkers utilized glycerol as an in situ hydrogen source for the transfer hydrogenation of CO₂ to formate with co-generation of lactate, in the presence of [RuH(Cl)(CNN)(CO)], bearing a CNN pincer ligand based on N-heterocyclic carbene (NHC) moieties, reaching turnover numbers (TONs) up to 300,000 and 387,000 for formate and lactate, respectively [35]. Daw and coworkers reported the use of cis-[RuCl₂(NNN)(PPh₃)] catalyst [NNN = 6,6′-(methylazanediyl)bis(methylene)dipyridin-2-ol], achieving 73% yield of LA (96% selectivity) along with H₂ [36]. Good conversions (99%) and TON = 409 were obtained in water as solvent by Li and coworkers using the Ru-arene complex [(p-cymene)Ru(2-OH-6-BzlmHpy)Cl]Cl bearing a bifunctional imidazolyl-pyridine ligand (2-OH-6-BzlmHpy = 6-(1H-benzo[d]imidazol-2-yl)pyridin-2-ol) [37].

Iridium complexes were also shown to be highly efficient for glycerol AD [32]. Among recent examples, Macchioni and coworkers showed a highly efficient catalytic protocol (TON = 2498) based on [Cp*Ir(pica-L_x)Cl] complexes, bearing glucose-tethered picolylamido ligands [pica = 2-picolinamidate, L₁ = (methyl-3,4,6-tri-O-acetyl-β-D-glucopyranosid-2-yl); L₂ = (methyl-β-D-glucopyranosid-2-yl)] [38]. Weller, Kumar and coworkers disclosed the use of [Ir(ⁱPr-PN^HP)(COD)]Cl [ⁱPr-PN^HP = (ⁱPr₂PCH₂CH₂)₂NH], achieving high conversions (96% yield) with excellent selectivity (99%) at short reaction times (4 h) at moderate temperatures (140 °C) [39]. In recent years, complexes of 3d metals such as Fe and Mn bearing PNP pincer ligands have also been applied for glycerol AD [40,41,42,43]. In the last decade, we and others have been studying the use of an interesting class of pincer-type PN³P ligands based on bis(aminophosphine)pyridine cores, for successful hydrogenation and dehydrogenation reactions [44,45,46,47,48,49]. Some of their Ru, Fe and Mn complexes showed that low-energy reaction pathways can be triggered by metal-ligand cooperativity (MLC) exploiting (reversible) ligand deprotonation on the NH moiety and pyridine core dearomatization. Although formic acid dehydrogenation was previously demonstrated in the presence of Ru [50], Fe [51] and Mn [52] PN³P complexes, to the best of our knowledge their use in catalytic glycerol AD has not yet been reported. Thus, we studied the properties of two Ru(II) and one Mn(I) PN³P pincer complexes as catalysts for this reaction. The main catalytic results are hereby presented, along with mechanistic studies obtained through a combination of experimental techniques (Nuclear Magnetic Resonance, NMR) and Density Functional Theory (DFT) calculations.

2. Results and Discussion

2.1. Catalytic Tests

The PN³P pincer complexes [RuH(Cl)(CO)(PN³P-^tBu)] (1), cis-[RuCl₂(PN³P-ⁱPr)(PPh₃)] (2) and cis-[MnH(CO)₂(PN³P-ⁱPr)] (3) shown in Scheme 2 were synthesized according to literature methods [29,53] by reacting the PN³P ligands with 1 equiv. of [RuH(Cl)(CO)(PPh₃)₃] for 1, [RuCl₂(PPh₃)₃] for 2 and [Mn(CO)₅Br] for 3, respectively. The ¹H, ³¹P{¹H} NMR and Fourier Transform Infrared spectroscopy (FTIR) characterization of the products, obtained as a yellow or orange powders after workup, were consistent with the reported data.

Complexes 1–3 were initially tested for glycerol dehydrogenation under the conditions reported by Kumar and coworkers [34], i.e., in the presence of 1.5 equiv. of KOH, ethanol (2 mL), 140 °C, 24 h (

Table 1. Metal and ligand effects screening in glycerol dehydrogenation ^a.

Entry	Cat.	Conv. b (%)	KLA b (%)	EG + FA b (%)	H2 c (ml)	TON d (KLA)	TON e (H2)
1	1	2	2	0	<1	10	10
2	2	77	77	0	18	385	385
3	3	1	1	0	<1	5	5
4	PN3P-iPr	0	0	0	0	0	0
5	[RuCl2(PPh3)3]	2	2	0	<1	10	10
6	None	0	0	0	0	0	0

^a Reaction conditions: glycerol (1 mmol), catalyst (0.2 mol%), KOH (1.5 eq. vs. glycerol), EtOH (2 mL), 140 °C, 24 h. ^b Conversions of GLY (%) and yields (%) of organic products were determined by ¹H NMR analysis using sodium acetate as the internal standard. ^c Volumes (mL) were measured by gas burette with the removal of blank volumes. ^d Turnover number (TON_KLA) = mmol KLA/mmol catalyst. ^e Turnover number (TON_H2) = mmol H₂/mmol catalyst (mmol H₂ calculated from volumes of produced H₂ using the Ideal Gas Law).

). With complex 1, potassium lactate (KLA) was obtained in low yield (2%, entry 1), likely due to poor thermal stability of the precatalyst under these reaction conditions. To our delight, complex 2, having ⁱPr substituents on the phosphorus donor atoms and different ancillary ligands (Cl, PPh₃) compared to 1, catalyzed glycerol dehydrogenation giving KLA in 77% yield and >99% selectivity (entry 2). On the other hand, complex 3 was almost inactive (entry 3). Control experiments with either free PN³P-ⁱPr ligand or the Ru precursor of 2, i.e., [RuCl₂(PPh₃)₃], gave no conversion to KLA (entries 4 and 5). Similarly, this reaction did not occur in the absence of a catalyst (entry 6).

Next, the effects of reaction parameters such as the type of solvent and base were screened in detail to further improve the efficiency of the process. The results are summarized in

Table 2. Solvent screening for glycerol dehydrogenation catalyzed by 2 ^a.

Entry	Solvent	Conv. b (%)	KLA (%)	EG + FA (%)	H2 c (ml)	TON d (KLA)	TON e (H2)
1	NMP	76	73	3	18	365	365
2	NMP/H2O	20	9	11	2	45	45
3	Toluene	23	20	3	5	100	100
4	Dioxane	27	13	14	3	65	65
5	Neat	23	20	3	5	100	115
6	THF	26	12	14	3	60	100
7	EtOH	77	77	0	18	385	374
8	iPrOH	17	17	0	4	85	85
9	tBuOH	60	57	3	14	285	285

^a Reaction conditions: glycerol (1 mmol), 2 (0.2 mol%), KOH (1.5 eq. vs. glycerol), Solvent (2 mL), 140 °C, 24 h. ^b Conversions of GLY (%) and yields (%) of organic products were determined by ¹H NMR analysis using sodium acetate as the internal standard. ^c Volumes (mL) were measured by gas burette with the removal of blank volumes. ^d Turnover number (TON_KLA) = mmol KLA/mmol catalyst. ^e Turnover number (TON_H2) = mmol H₂/mmol catalyst (mmol H₂ calculated from volumes of produced H₂ using the Ideal Gas Law).

and

Table 3. Base screening for glycerol dehydrogenation catalyzed by 2 ^a.

Entry	Base	Time (h)	Temp. (°C)	Conv. b (%)	KLA (%)	EG + FA (%)	H2 c (ml)	TON d (KLA)	TON e (H2)
1	NaOMe	24	140	37	31	6	7	155	155
2	NaOH	24	140	21	17	4	4	85	85
3	NaOtBu	24	140	18	18	0	4	90	90
4	KOtBu	24	140	85	80	5	19	400	400
5	K2CO3	24	140	27	27	0	7	135	135
6	DBU	24	140	0	0	0	0	0	0
7	Et3N	24	140	0	0	0	0	0	0
8	KOH	48	140	82	80	2	18	400	374

^a Reaction conditions: glycerol (1 mmol), 2 (0.2 mol%), base (1.5 eq. vs. glycerol), EtOH (2 mL), 140 °C, 24–48 h. ^b Conversions of GLY (%) and yields (%) of organic products were determined by ¹H NMR analysis using sodium acetate as the internal standard. ^c Volumes (mL) were measured by gas burette with the removal of blank volumes. ^d Turnover number (TON_KLA) = mmol KLA/mmol catalyst. ^e Turnover number (TON_H2) = mmol H₂/mmol catalyst (mmol H₂ calculated from volumes of produced H₂ using the Ideal Gas Law).

, respectively. The choice of solvent had a significant effect on the conversion of glycerol and on the selectivity toward lactate (Table 2). Using N-methylpyrrolidone (NMP) as solvent and a 2/KOH ratio of 1:750, KLA was obtained in 73% yield with a corresponding TON_H2 of 365 (entry 1). Interestingly, ¹H NMR analysis of the post-catalytic reaction mixture revealed that under the described conditions NMP partially underwent ring opening to give sodium-4-N-methylaminobutanoate, suggesting a potential side reaction involving nucleophilic attack on NMP by hydroxide under basic reaction conditions [41]. When an NMP/H₂O (1:1) mixture was used as the solvent, only 9% of lactate was produced (entry 2). The adverse effect of added water is unclear, maybe due to catalyst decomposition, as evidenced by the white-colored solution observed at the end of the reaction. Changing the solvent to toluene resulted in only 20% KLA yield at the end of the reaction (entry 3). Using dioxane produced only 13% lactate (entry 4), probably because in this solvent complex 2 is only partially soluble. Additionally, we attempted to perform the reaction under solvent-free conditions. Only a 20% yield of KLA was obtained (86% selectivity), likely due to the high viscosity of the reaction mixture, which hindered effective mixing of catalyst and substrate (entry 5). Similar results were observed when dry THF was used as solvent (entry 6). After several trials with aprotic solvents, alcohols were examined as protic solvents, and EtOH gave the best yield in KLA (77%, entry 7). In case of ⁱPrOH, only 17% conversion was observed (entry 8), whereas in ^tBuOH a 57% yield of KLA was observed (entry 9). Based on the results of this screening, EtOH was chosen as the solvent for the rest of the study.

Next, different bases were tested, revealing that potassium hydroxide (KOH) produced the highest lactate yields and selectivities (Table 3). When replacing KOH with sodium methoxide (NaOMe), 31% KLA yield was obtained, with TON_H2 = 155 (entry 1). Using NaOH and NaO^tBu led to even lower yields of 17–18%, respectively (entries 2 and 3). Notably, employing KO^tBu as base produced a higher yield of 80% lactate and a TON_H2 of 400 (entry 4), indicating its superior performance among the tested bases. Conversely, utilizing K₂CO₃, a weaker base, resulted as expected in a lower yield and poor selectivity (entry 5). Organic bases such as DBU and NEt₃ were ineffective, producing negligible amounts of lactate and volumes of hydrogen (entries 6 and 7). The effect of the reaction time and temperature were also examined to optimize the reaction protocol (entry 8). The results showed that, for reactions carried out at 140 °C, runs of 24–48 h were necessary to obtain high glycerol conversion. Lowering the temperature to 120 °C led to reduced yields but still maintained high selectivity, suggesting that higher temperatures accelerate the reaction without compromising selectivity.

Finally, the effect of catalyst amount was studied, and the results are shown in

Table 4. Screening of catalyst loading for glycerol dehydrogenation catalyzed by 2 ^a.

Entry	2 (mol%)	Base	Conv. b (%)	KLA (%)	EG + FA (%)	H2 c (ml)	TON d (KLA)	TON e (H2)
1	0.2	KOH	77	77	0	18	385	385
2	0.05	KOH	86	80	6	19	1600	1600
3	0.2	KOtBu	85	80	5	19	400	400
4	0.02	KOH	18	18	0	4	900	900
5	0.05	KOtBu	9	9	0	3	180	180

^a Reaction conditions: glycerol (1 mmol), 2 (0.02–0.2 mol%), KOH (1.5 eq. vs. glycerol), EtOH (2 mL), 140 °C, 24 h. ^b Conversions of GLY (%) and yields (%) of organic products were determined by ¹H NMR analysis using sodium acetate as the internal standard. ^c Volumes (mL) were measured by gas burette with the removal of blank volumes. ^d Turnover number (TON_KLA) = mmol KLA/mmol catalyst. ^e Turnover number (TON_H2) = mmol H₂/mmol catalyst (mmol H₂ calculated from volumes of produced H₂ using the Ideal Gas Law).

. Using 0.05 mol% of 2 yielded an 82% yield of KLA with 95% selectivity and a TON_H2 of 1600 (entry 2). When the catalyst loading was reduced to 0.02 mol%, 18% yield was obtained with 99% selectivity after 24 h, with TON_H2 = 900.

Catalyst recycling experiments were also carried out under the catalytic conditions reported for entry 1, Table 4, by adding new aliquots of substrate to the vessel after a set reaction time (Figure 1). In the first run, complex 2 gave an 82% conversion of glycerol with 99% selectivity for KLA, while in the second and third runs yields of 80% and 75% were obtained with 98% and 97% selectivities, respectively. In the fourth run, a marked decrease in conversion was observed, with a low 19% yield of KLA (Figure 2). No significant improvement was observed by extending the reaction time to 30 h. These data indicate that catalyst deactivation and/or decomposition may occur after three consecutive runs, hampering direct application for industrial use.

2.2. Mechanistic Studies

The mechanism of glycerol catalytic dehydrogenation in the presence of 2 was initially studied using NMR spectroscopy to understand the precatalyst activation step. In the first series of experiments (Figure S24a), complex 2 (5 mg) was dissolved in CD₂Cl₂ under nitrogen at room temperature and ¹H NMR was collected. A sample of EtOH was analyzed by ¹H NMR showing the presence of a signal at 4.39 ppm due to the OH group (Figure S24b). After the addition of EtOH (100 equivalents) to 2, the ¹H NMR spectrum displayed a shift in the signal due to OH from 4.39 ppm to 2.52 ppm, indicative of coordination of EtOH to the Ru center, presumably by dissociation of one chloride ligand, now resulting as a counter ion in the putative [RuCl(PN³P-ⁱPr)(PPh₃)(EtOH)]Cl complex (Figure S24c). By addition of KOH or other strong bases to the sample, this complex then underwent deprotonation at one or more acidic sites, for example the metal-coordinated EtOH ligand, giving [RuCl(PN³P-ⁱPr)(PPh₃)(EtO)], or by PNP ligand N-H bond deprotonation to give [RuCl(⁻PN³P-ⁱPr)(PPh₃)(EtOH)] (⁻PN³P-ⁱPr = monodeprotonated PNP ligand). As ¹H NMR spectroscopy was not sufficient to clarify this point, we turned to Density Functional Theory (DFT) calculations to pinpoint details on precatalyst activation and to find out the lowest energy reaction pathway for the entire process.

The DFT study was carried out at the M06-2X level of theory using the Gaussian 16 software package (for the full description see Supporting Information) [54,55,56,57,58,59,60]. Solvent effects were modeled using the CPCM solvent model for ethanol, the same employed in the experimental tests. All calculations were performed at 140 °C, the reaction temperature used in optimized experimental runs. First of all, the free energy change for the two possible overall reactions shown in Scheme 3 were evaluated. The neat reaction of glycerol (GLY) with KOH to give KLA, H₂O and H₂ was estimated to be strongly exergonic (−44.5 kcal·mol⁻¹). In contrast, the conversion of glycerol to glyceraldehyde (GLYA) and H₂ in the absence of base (bottom of Scheme 3) was calculated to be endergonic by +10.1 kcal·mol⁻¹. This suggests that the driving force of the overall reaction resides in a substantial free energy gain in the (uncatalyzed) conversion of the glyceraldehyde, formed upon glycerol AD, to the final products.

After structural optimization of complex 2 (Figure 3, left), the most favored precatalyst activation mechanism was assessed (full details in Supporting Information). The initial step involves the dissociation of a chloride ligand from the metal center. Calculations showed that the five-coordinate cationic complex 4⁺ (Figure 3, right) resulting from the dissociation of Cl1 trans to P1 with ΔG = −1.3 kcal·mol⁻¹ is preferred over 5⁺ (removal of Cl2 cis to P1, Figure S26), a process with ΔG = +10.3 kcal·mol⁻¹.

Complex 4⁺ may then coordinate either a molecule of EtOH (solvent) or GLY (substrate), yielding 6⁺ or 7⁺ respectively (Figure S27, Supporting Information), with only a slight preference for 7⁺ (7.5 kcal·mol⁻¹ vs. 8.4 kcal·mol⁻¹). Next, the effect of KOH in the preactivation step was studied. It was established that the favorite path (exergonic by −28.0 kcal·mol⁻¹) is the deprotonation of coordinated glycerol in 7⁺, to provide neutral complex 7_O (Figure S28, Supporting Information), whereas deprotonation of either EtOH in 6⁺ or the PN³P ligand in 7⁺ is less favored. After dissociation of the second Cl⁻ ligand in 7_O, the active catalytic species 8⁺_N, featuring a glycerate ligand trans to the pyridine N atom (Figure 4, left), is preferentially formed (full details in Supporting Information). The β-elimination step then occurs, overcoming a barrier of 23.0 kcal mol⁻¹ followed by a free energy gain of −10.2 kcal mol⁻¹, giving 9⁺_N (Figure 4, center), bearing a hydrido ligand trans to PPh₃ and glycerate trans to pyridine N atom. From 9⁺_N, glyceraldehyde can then be released with a free energy gain of −6.5 kcal·mol⁻¹, yielding the five-coordinate complex 10⁺_P (Figure 4, right). By coordination of a second glycerol molecule in the vacant coordination site, the octahedral intermediate 11⁺_N (Figure S33 left, Supporting Information) is formed with a free energy cost of only 4.0 kcal·mol⁻¹.

At this point, formation of a η²-H₂ ligand occurs by proton transfer from coordinated glycerol to the hydrido ligand, overcoming a calculated energy barrier of ca. 20.2 kcal mol⁻¹, to give the most favored isomer 12⁺_P (Figure S35, Supporting Information). Finally, H₂ is released from 12⁺_P, restoring the initial 8⁺_N with a free energy gain of −12.1 kcal mol⁻¹, and the system is ready to restart the catalytic cycle. The free energy pathway for H₂ formation through glycerol activation (Figure S36, Supporting Information) has a free energy cost of 10.1 kcal mol⁻¹. Glyceraldehyde undergoes an uncatalyzed, off-cycle organic reaction under alkaline conditions to give KLA with a free energy gain as large as −54.6 kcal mol⁻¹. Based on these results, we propose the simplified catalytic mechanism shown in Scheme 4.

3. Experimental Section

General Procedure for Glycerol Catalytic Dehydrogenation

In a typical experiment, the catalytic mixture containing solvent, catalyst and base was prepared in a Schlenk tube under a nitrogen atmosphere. The tube was then placed into an oil bath preheated to the desired temperature and stirred at 500 rpm for the specified reaction time. After the run, the Schlenk tube was cooled to <10 °C using an ice bath. The pressure was gently released into a manual gas burette to measure hydrogen evolution, then the solvent was evaporated under reduced pressure. The Schlenk tube was thoroughly rinsed with H₂O, and the washings were added to the rest of the mixture. Sodium acetate (1 equiv. to glycerol) was added as an internal standard, and the lactate content was determined by integration of the corresponding ¹H NMR signal vs. NaOAc. D₂O (ca. 0.7 mL) was added as a deuterated solvent. All tests were repeated at least twice to ensure reproducibility (average error ca. 6%).

4. Conclusions

In conclusion, complex [RuCl₂(PN³P-ⁱPr)(PPh₃)] (2) showed high activity for glycerol dehydrogenation, reaching TON = 1600 in single batch runs, at a catalyst loading as low as 0.05 mol% under optimized conditions. Mechanistic studies suggest a fully inner-sphere mechanism centered on complex [Ru(GLY⁻)(PN³P-ⁱPr)(PPh₃)]⁺ (8⁺_N, GLY⁻ = glycerate) that represents the active form of the catalyst. These results demonstrate the potential of PN³P-type Ru(II) molecular complexes in glycerol dehydrogenation, that may contribute to greener chemical processes and the utilization of biomass-derived feedstocks. Further optimization of the catalysts by design of modified PN³P ligands is currently in progress in our laboratories.