Bro̷nsted Acid-Catalyzed Reduction of Furans

Abstract

Bioderived furans play a pivotal role in advancing defossilized chemical pathways. The complete reduction of furans currently relies on impractical metal-catalyzed hydrogenations at high pressures and temperatures. In addition, the Birch reduction of unbiased furans to 2,5-dihydrofurans remains an unsolved synthetic challenge. Herein, we report a mild Bro̷nsted acid-catalyzed reduction of furans to 2,5-dihydro- and/or tetrahydrofuran derivatives using silanes as reducing agents. In particular, the first formal Birch reduction of furan itself is achieved. Mechanistic investigations reveal an intricate behavior of HFIP as the crucial solvent, preventing the intrinsic polymerization behavior of furans under acidic conditions and introducing additional driving force by specific product binding.

1. Introduction

Furans are viewed as critical for transforming the fossil-based chemical industry to biobased and sustainable production pathways of fine chemicals.^1,2 For example, furfural and 5-hydroxymethylfurfural are accessed from lignocellulosic carbohydrates,² from which more elaborate secondary furanoids, such as 3-bromofuran, can be obtained. This portfolio of biobased platform chemicals challenges chemists to innovate routes toward low-volume, high-value fine chemicals (Figure 1A).³ Despite its importance, furan-based chemistry remains significantly underdeveloped relative to its potential future impact. For example, general and site-selective control over the desaturation degree of furans toward dihydrofurans could enable novel, highly sought-after valorization strategies within biorefinery channels, such as for the fragrance industry (Figure 1B).^4,5

Figure 1.

(A) Furan as a biobased source toward fine chemicals. (B) Unmet challenges of furan chemistry. (C) This work: Bro̷nsted acid-catalyzed reduction of furans.

Although homogeneous and heterogeneous metal-catalyzed reductions from furans to tetrahydrofurans employing elevated H₂ pressures and specialized Rh-, Ir- or Ru-catalysts are well-established,⁶⁻⁸ partial dearomatization strategies of the furan core are scarce. Most prominently, Birch reductions of furans to dihydrofurans are limited to a handful of electron-poor, carboxylate-substituted examples.^6,9−13 All other furans, including furan itself, suffer from uncontrolled dimerization and ring-openings, presumably due to their low resonance energy, high reduction potentials, and (vinylogous) enolether character.¹⁴

We hypothesized that Brønsted acid-catalysis might overcome these challenges by making furans susceptible to hydride attack after protonation, inverting the order of a classical Birch reduction.¹⁵ Here, we report the mild Bro̷nsted acid-catalyzed reduction of a panoply of furans to the formal Birch-reduced products 2,5-dihydrofurans and, furthermore, full reduction to tetrahydrofuran derivatives using silanes as reducing agents.

2. Results and Discussion

Our design is accompanied by two main challenges: controlling the access of isomeric intermediates can be difficult; moreover, furans are well-known to readily polymerize under acidic conditions (Figure 1B).¹⁶ Since acid-catalyzed reductions of furans are not known,¹⁷⁻²⁰ we investigated if a catalytic Bro̷nsted acid (HA) would mitigate polymerization side reactions. The model substrates 2-pentyl furan I or 3-aryl furan II with triethylsilane (Et₃SiH) were subjected to catalytic trifluoroacetic acid (TFA) or triflic acid (TfOH) along with a proton source (1.0 equiv), such as water (H₂O) or phenol, across different solvents and temperatures (Figure 2A).

Figure 2.

(A) HFIP reduces the polymerization reactivity in commonly employed solvents. (B) Screening of acids identified two sets of conditions to achieve partial reduction to dihydrofuran and complete reduction to tetrahydrofuran. Yields determined by ¹H NMR spectroscopy relative to mesitylene as internal standard.

Generally, either no reactivity or dimeric products III accompanied by polymeric material (at 50–100 °C) were obtained (Figure 2A). To circumvent this, we hypothesized that the unique non-nucleophilic properties of perfluorinated alcohols might enable sufficient stabilization of cationic intermediates without participating in solvolysis.²¹⁻²⁴ Indeed, upon employing hexafluoro-2-propanol (HFIP), productive reduction of I and II toward the 2,5-dihydrofuran and/or tetrahydrofuran at room temperature was observed with minimal polymerization degrees (0–10%). The addition of H₂O (1.0 equiv) was crucial to achieve complete conversion, although the use of H₂O as a solvent led to no conversion.

The reduction stage of II was found to be controllable by acid strength (Figure 2B). The use of phosphoric acid or TFA allowed clean conversion toward 2,5-dihydrofurans without overreduction. Notably, this selectivity was only achieved under catalytic acid regime; increasing the acid to 1.0 equiv led to complete reduction to tetrahydrofurans. Stronger acids, such as p-toluenesulfonic acid (PTSA) or aqueous hydrochloric acid (HCl), lead to parallel overreduction to a mixture of dihydro- and tetrahydrofurans (THF). For triflic acid (TfOH), only THF products were observed. In general, the degree of reduction was not efficiently controlled by the stoichiometry of the silane.

Based on these screening results, two sets of reaction conditions were selected to probe their general applicabilities to furan substitution and electronic character: (A) TFA (5 mol %), Et₃SiH (1.5 equiv), H₂O (1.0 equiv), and (B) TfOH (2 mol %), Et₃SiH (2.5 equiv), both in HFIP (0.2 M) at room temperature.

Upon employing Condition (A), a wide range of 3-aryl substituted furans 1a–1g can be selectively converted to 2,5-dihydrofurans 2a–2g in good to excellent yields (68–95%, Figure 3). Variations of the electronic nature of the 3-aryl substituent were well tolerated, spanning the whole Hammett σ_p-scale,²⁵ albeit the stronger acidic Condition (B) was required to obtain the electron-poor nitro-derivate 2g successfully. Similarly, 3-alkyl substituted furans 1h–1l were converted in near-quantitative yields to 2,5-dihydrofurans 2h–2l by employing Condition (B). Surprisingly, chloromethylene (2m, 59%) and terminal olefins (2n, 82%) were tolerated, as being commonly reduced in ionic silane reductions.²⁶ As a result, the partial reduction of Perillene 1o, a natural monoterpene in the essential oil of perilla frutescens, to 2,5-dihydroperillene 2o in 57% yield was possible, leaving the homoprenyl-side chain untouched. Furthermore, conversion of the flavoring agent menthofuran²⁷1p was feasible in 75%. The 2,4-substituted furan 1q was cleanly reduced to 2q. Further, 1r can be coverted to 2r in 40% yield, tolerating (thio-)esters.

Figure 3.

Substrate scope toward 2,5-dihydrofurans with superscripts A and B indicate Conditions (A) and (B). Reactions were performed on a 0.2–30.0 mmol scale. Yields are reported as isolated yields after column chromatography, except † indicates ¹H NMR yields. *Incomplete conversion achieved: addition of further portions of Et₃SiH. **5 mol% TfOH. See the Supporting Information for detailed reaction conditions.

Limitations of the methodology were shown by meat-flavoring supplement²⁸1s yielding only 21% 2,5-dihydrofuran 2s, presumably being too electron-rich and prone toward reductive disulfide cleavage. For electron-poor furans, only ester 1t was successfully converted to the 2,5-dihydrofuran. Substrates lacking the stabilizing methyl group, such as 2-carboxyl furans or 3-carboxyl furans, remain unreacted, rendering the methodology complementary to the existing dissolving-metal reduction of electron-deficient furanoic acids.⁶

To our delight, furan itself (1u) proved to be a suitable substrate and was cleanly converted to 2,5-dihydrofuran (2u) in quantitative yield on a 1.0 g scale. This constitutes the first reduction of plain furan to a “Birch”-type product.

Condition (B) was employed to fully reduce the 3-aryl furans to their respective THF derivatives 4a–4e in good to very good yields (72–83%, Figure 4). This example showcases how selective control over the reduction stage via acid strength is possible under Conditions (A) and (B). Electron-deficient substrates 3f and 3g could not be reduced further than their 2,5-dihydro analogs. In contrast, 2-substituted furans, such as 3h–3m, could not be selectively reduced to a dihydrofuran and instead consistently delivered their tetrahydrofuran product in high trans-diastereoselectivities (4j, 4k). In particular, 2-aryl substituted furans were sensitive to the conditions employed. Electron-deficient 3n–3p could readily be reduced using Condition (B), whereas 2-phenyl furan 3q required Condition (A) to avoid side reactivity. The proton-affinity of the respective dihydrofuran intermediate is likely to be responsible for the susceptibility to different acid strengths (vide infra, Figure 6).

Figure 4.

Substrate scope toward tetrahydrofurans with superscripts A and B indicating Conditions (A) and (B). Reactions performed on 0.1–7.2 mmol scale. Yields are reported as isolated yields after column chromatography, except † indicates ¹H NMR yields. See the Supporting Information for detailed conditions.

Figure 6.

2D maps plotting ΔG_rel(H⁺) and ΔG_rel(H^–) for the reduction of (A) furan to dihydrofuran and (B) dihydrofuran to tetrahydrofuran at CPCM(HFIP)-DLPNO–CCSD(T)/def2-TZVPP//CPCM(HFIP)-B2PLYP-D3BJ/def2-SVP (298 K/1 M) level of theory. See the Supporting Information for details. PMP = para-methoxyphenyl.

Interestingly, the highly acidic Condition (B) enables the total reduction and deoxygenation of biobased substrates, avoiding the use of commonly employed precious metal catalysts and high H₂ pressures. For example, furfural 3r and 4-hydroxymethyl furfural (HMF) 3s were converted to methyl-THF 4r (83%) and 2,4-dimethyl-THF 4s (quantitative). Furthermore, the reductive cyclization of levulinic acid 3t to γ-valerolactone 4t was possible in 89% yield. Besides, benzofuran 3u was successfully converted to 2,3-dihydrobenzofuran 4u in 55% yield.

To understand the observed selectivities, the reaction of 1b under Condition (A) was followed via No-D ¹H NMR spectroscopy, varying the amounts of each reagent. First, the reaction kinetics was not significantly influenced by the equivalents of Et₃SiH used, indicating a zeroth-order dependence on Et₃SiH (see Supporting Information). Second, TFA-catalyst loading was investigated, and two regimes in the kinetic time course could be identified (Figure 5A). The fast consumption of furan 1b in the first few minutes shows a clear dependency on the catalyst loading, where, as expected, higher catalyst loading leads to faster consumption. The profile then changes to a relatively linear decay. Lastly, the H₂O equivalents were varied. Without H₂O, the reaction stagnates at approximately 25% conversion; upon addition of 1.0 equiv H₂O at t = 24 h, the conversion resumes, proceeding as previously observed (see Figure 5B).

Figure 5.

(A) Concentration profile of different catalyst concentrations for the reduction of 1b under Condition (B) was monitored by No-D ¹H NMR. (B) Concentration profile of varying water amounts and delayed water addition to the reaction mixture. (C) Identification of the silylated catalyst employing 1b under Condition (B) with Et₂MeSiH instead of Et₃SiH. (D) Combined representation of Hammett- and KIE-analysis: The line diagram depicts the Hammett slope of substrates 1a–1f in an intermolecular competition experiment with 1d; the bar diagram shows the obtained KIE in intermolecular competition experiments between Et₃Si–H and Et₃Si–D. (E) Binding isotherm of 2-MeTHF and 2-methylfuran and HFIP at [HFIP] = 0.1 M host concentration and derived 1:1 binding constant K. See the Supporting Information for further details. (F) Proposed mechanism in combination with computed structures at CPCM(HFIP)-B2PLYP-D3BJ/def2-SVP (298 K/1 M) level of theory. E⁽²⁾ value stems from second order perturbation theory analysis of fock matrix in NBO basis. HFIP-solvation cloud was generated with ORCA 6.0-Docker at GFN2-xTB level of theory. See the Supporting Information for further details.

Therefore, we hypothesized that catalyst 5a is silylated to 5b and is hydrolytically regenerated by H₂O, as the HFIP solvent lacks sufficient nucleophilicity to efficiently drive the solvolysis.²¹ The hydrolysis may regulate the kinetics in the second phase and attenuate the initial reaction rate. To probe this, Et₃SiH was exchanged for Et₂MeSiH at different concentrations, and indeed, the formation of the silylated catalyst Et₂MeSi-TFA 5c could be monitored by its indicative Me-singlet assigned via ¹H–²⁹Si HMBC (Figure 5C).

An intermolecular competition experiment of substrates 1a–1f relative to 1d yielded a log(k_x/k_H) graph based on the relative ¹H NMR integration of products. A nonlinear, concave, downward-shaped Hammett plot was obtained (Figure 5D, line chart), suggesting a change in the rate-limiting step for 1a–1f under the same general mechanism.^25,29,30 More specifically, for the electron-poor furans 1e and 1f, a negative reaction rate constant ρ_I = −1.7 was found, consistent with furan protonation being involved in the rate-determining step. The electron-rich substrates 1a–1c form a slightly positive Hammett-slope of ρ_II = +0.25, associated with a negative charge build up, which would be in line with the reduction becoming the rate-limiting step.²⁹

To investigate this reduction step further, the kinetic isotope effects (KIEs) of an equimolar intermolecular competition experiment between Et₃Si–H and Et₃Si–D with limiting amount of substrates 1a–1f were determined (Figure 5D, bar diagram, see Supporting Information for further details). A positive KIE of 1.11 for 1f was found, whereas a secondary inverse KIE of up to 0.82 was obtained for the electron-rich substrates 1a–1c.^31,32 Taking the obtained concave Hammett data into consideration, this suggests that the hydridic reduction of carbocation 5e is the rate-determining step for electron-rich substrates, whereas for electron-poor substrates, it is the protonation of 5d.³³

To gain insight into the role of HFIP as a crucial solvent, an EXSY-NMR of 1a with TFA (1.0 equiv) was measured. No direct TFA furan ion-pair was detected, instead there was a protic exchange between HFIP and H^A of furan 1a (see Supporting Information for further details). This hints that HFIP functions as a reliable cationic-stabilizing solvent, inhibiting the dimerization and polymerization as observed in conventional solvents (Figure 2A) without interfering solvolytically as a nucleophile in the reaction. Implicit solvent modeling confirms the presence of multiple H-bonds between the HFIP solvent cage and protonated furan 5e (Figure 5F).³⁴

Further, titration studies (Figure 5E) show high 1:1 binding affinities of HFIP to dearomatized intermediates (e.g., 4i: K = 17.9 ± 0.72 mol L^–1) and negligible complexation constants for furans itself (e.g., 3i: K = 0.32 ± 0.045 mol L^–1). This was corroborated by computational analysis, locating the complexes between HFIP for 5f. Natural bonding orbital (NBO) analysis confirms the presence of a strong H-bond between HFIP and 5f (E⁽²⁾ = 28.6 kcal mol^–1). This HFIP-complexation of ethereal oxygen adds additional driving force to the reduction reaction.

Collectively, these results support the catalytic cycle depicted in Figure 5F for 3-substituted furans. Upon protonation of the furan 5d, the oxocarbenium cation 5e is reduced by Et₃SiH at the 2-position, resulting in a 2,5-dihydrofuran product 5f. The catalytic trifluoroacetate anion recombines with the remaining silylium cation to form the stable intermediate 5b. Subsequently, it is regenerated through a hydrodesilylation with H₂O to the active catalyst TFA. This step becomes rate-limiting after the first induction period decelerating the reaction kinetics.

To put the found conditions in a broader context of furan reactivity, a 2D plot was constructed to rationalize the observed reactivity as described above. Computed Gibbs free energies ΔG_rel(H⁺) of the protonation of furan were plotted on the x-axis, and on the y-axis were plotted the ΔG_rel(H^–) of the reduction of the protonated intermediate toward dihydrofuran relative to the chosen reference 3-phenyl furan 6e (ΔG^I_rel(H⁺) = 0; ΔG^I_rel(H^–) = 0; Figure 6A). A similar map was constructed from dihydrofuran to tetrahydrofuran with 3-phenyl-2,5-dihydrofuran 6i chosen as the reference (ΔG^II_rel(H⁺) = 0; ΔG^II_rel(H^–) = 0; Figure 6B). Both maps are based on computations at the DLPNO–CCSD(T)/def2-TZVPP//B2PLYP-D3BJ/def2-SVP (298 K/1 M) level of theory.

For each step, all possible protonation and reduction sites were computationally considered, and the energetically lowest-lying isomer was selected. The maps can be divided into segments for different selectivities based on the experimental results (Figures 3 and 4). For example, in Figure 6A, 6a marks the marginal example of segment ②, which TfOH can still productively protonate; for 6b or 6c in segment ③, a stronger acid than TfOH would have been needed for reactivity. Similarly, the plot shows the substrate set 6d to 6f, spanning a range of ΔΔG^I_rel(H⁺) = 7.8 kcal mol^–1, where substrate 6d is sufficiently protonated by TFA (segment ①), whereas for substrate 6f TfOH is needed.

Moreover, we sought to explain why 2-substituted furans could not be halted at the partially reduced dihydrofuran. Subjecting a potential 2,5-dihydro intermediate 6g to Condition (B) yielded no product, whereas the potential 2,3-dihydro intermediate 6h did. This supports a variation of the mechanism depicted in Figure 5F for 2-substituted furans, potentially proceeding via a 2,3-dihydro instead of a 2,5-dihydro intermediate. Additionally, the enol 2,3-dihydro intermediate 6h was found to be thermodynamically more stable by 4.3 kcal mol^–1 and hence selected as the preferred intermediate for the 2D map (Figure 6B). Their high relative proton-affinities ΔG^II_rel(H⁺) explain why 2-substituted furans in segment ④ always reduce to tetrahydrofuran, as the second protonation is always preferred (|ΔG^I_rel(H⁺)| < |ΔG^II_rel(H⁺)|).

Relative to 6i, the protonation of 3-alkyl substituted furans 6j, as well as plain 2,5-dihydrofuran 6k in segment ⑥, is too disfavored to achieve further conversion toward tetrahydrofuran. All these substrates halt at the partially reduced furan. For the intermediate segment ⑤, selectivity control can be achieved by acid strength (see Figure 2).

Finally, we were keen to demonstrate the utility of the dearomatization strategies in potential novel industrial pathways to fine chemicals. For example, 2,5-dihydrofuran is conventionally produced by epoxidation of butadiene and further thermal rearrangement (Figure 7).^35,36 The direct access from furan might offer novel opportunities in biorefineries if HFIP recycling can be achieved efficiently. To avoid the use of Et₃SiH, polymethylhydrosiloxane PMHS, a waste product of the silicon industry, was found to be a viable reduction agent for furan too.

Figure 7.

Novel synthetic pathways. † indicates NMR yields.

When 2-phenyl furan is sequentially treated with TFA and TfOH in an excess of Et₃SiH, complete opening to 4-phenylbutanol was achieved in quantitative yields without chromatographic purification. So far, this chemical has been produced in multistep sequences, employing stoichiometric amounts of AlCl₃ and NaBH₄ for the use as a key intermediate of the leukotriene receptor antagonist Pranlukast (Figure 7).^37,38

3. Summary

In summary, a Bro̷nsted acid-catalyzed reduction of furans to 2,5-dihydro- and/or tetrahydrofuran derivatives using silanes as reducing agents is reported, leveraging the stabilizing effect of the HFIP solvent in preventing polymerization side reactivity. Mechanistic and computational analysis rationalize the selection of suitable acid strength, enabling control over the reduction degree for 3-aryl substituted furans. Further, a broad scope of unbiased furans was reduced in a substrate-controlled way to either their 2,5-dihydrofuran or tetrahydrofuran analogs. In particular, the first formal Birch reduction of furan itself was achieved. We anticipate that the now easy-to-access 2,5-dihydrofuran structures from a wide range of furans enable novel retrosynthetic pathways. Further investigations into the HFIP-effect and an asymmetric variant of the reported reaction are ongoing in our laboratories.

Glossary

Abbreviations

TFA

2,2,2-Trifluoroacetic acid

TfOH

Trifluoromethanesulfonic acid, triflic acid

HFIP

1,1,1,3,3,3-Hexafluoropropan-2-ol

PTSA

p-Toluenesulfonic acid

THF

Tetrahydrofuran

HMF

Hydroxymethyl furfural

KIE

Kinetic isotope effect

NMR

Nuclear magnetic resonance

HMBC

Heteronuclear multiple bond correlation

EXSY

Exchange spectroscopy

ΔG

Gibbs free energy

PMHS

Polymethylhydrosiloxane

NBO

Natural Bonding Orbital

HA

Brønsted Acid

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c18485.

• Experimental and mechanistic details and analytical data for all new compounds (PDF)

• Cartesian coordinates (ZIP)

Open access funded by Max Planck Society.

The authors declare no competing financial interest.