Iron-Catalyzed C–H Oxygenation using Perchlorate Enabled by Secondary Sphere Hydrogen-bonds

Abstract

Perchlorate (ClO₄⁻) is a ground water pollutant that is challenging to remediate. We report a strategy to use Fe(II) tris(2-pyridylmethyl)amine (TPA) complexes featuring appended aniline hydrogen bonds (H-bonds) to promote ClO₄⁻ reduction. These complexes facilitate oxygen atom transfer from ClO₄⁻ to PPh₃ and also C–H oxygenation reactions of organic substrates. Catalytic reactions using 15 mol% afforded excellent yields for oxygenation of anthracene and cyclic alkyl aromatics, and this methodology tolerates aryl halides as well as heterocycles containing either O, S, or N.

Graphical Abstract

INTRODUCTION

An unfortunate consequence of the widespread use of perchlorates (ClO₄⁻) in pyrotechnics, rocket fuels, and munitions is groundwater contamination near point sources of generation.^1–8 Although commonly considered as an inert anion, ClO₄⁻ presents a high health risk because it competes with I⁻, disrupting normal thyroid function.³ Distinct from I⁻ is the oxidizing strength of ClO₄⁻; thus, if sequestered, ClO₄⁻ represents a strong oxidant that might have repurposed use in chemical synthesis.

The main challenge for any chemical reaction employing ClO₄⁻ as a reagent is to overcome its kinetic inertness. Select synthetic and biological systems overcome this inertness by forming thermodynamically strong M–O bonds (M= Mo, Re).^9–20 One prominent homogeneous catalyst used for ClO₄⁻ reduction was reported by Abu-Omar using an oxazoline-ligated Re(V)-oxo catalyst (Figure 1a).^9–11 Another important example is a Mo(VI)O₂ precatalyst reported by Mösch-Zanetti and co-workers (Figure 1a).^12–13 In both cases, phosphine or thioether substrates, which form strong O=PPh₃ or O=SR₂ bonds, are used as terminal O-atom acceptors. Similar catalyst designs have been employed with heterogenous systems using Pd/C supports for ClO₄⁻ reduction in aqueous media using H₂ (Figure 1a).^14–18

Figure 1.

a) Transition metal-based ClO₄⁻ reduction systems. b) ClO₄⁻ reduction using an H-bond appended ligand. c) Current approach featuring secondary sphere H-bond promoted ClO₄⁻ reduction/C–H oxygenation.

An alternative strategy to drive the challenging deoxygenation reaction is to use secondary sphere interactions to stabilize the resulting M–O bond, a design principle that is exploited by Mo-dependent perchlorate reductase enzymes.^19–24 The H-bonding network helps direct the ClO₄⁻ anion to the Mo-center and plays key roles to influence the stability as well as reactivity of the oxygenated intermediates.^20–21 Adapting this biological principle to synthetic systems, the Fout group developed an iron catalyst featuring a tripodal azafulvene-amine ligand that promotes deoxygenation/reduction chemistry (Figure 1b).^25–26 They identified an H-bond stabilized Fe(III)–OH intermediate that is formed from the reduction of ClO₄⁻.

Direct O-atom transfer from ClO₄⁻ to an Fe(II) center should form a putative Fe(IV)=O species. One ligand class that is featured prominently within the Fe(IV)=O literature is tris(2-pyridylmethyl)amine (TPA), where Fe^IV(O)(TPA) complexes are often prepared using oxidants such as PhIO, H₂O₂, CH₃CO₃H, ^tBuOOH, etc.^27–35 The oxygenation of aliphatic and aromatic C–H bonds is precedented by both heme and non-heme based Fe(IV)=O complexes.^34–64 Aliphatic substrates are proposed to engage in hydrogen atom transfer (HAT)/radical rebound pathways.^{29–51, 53–64} Alternatively, aromatic C–H hydroxylation reactions are proposed to proceed via an electrophilic aromatic substitution pathway.⁵² Harnessing intramolecular secondary sphere H-bond donors on a non-heme Fe-system presents a promising strategy to afford C–H activation reactivity using ClO₄⁻ as the oxidant.

Despite the accessibility of Fe(IV)=O intermediates, Fe–TPA complexes are reported to be unreactive to ClO₄⁻– structural characterization can even include ClO₄⁻ counterion(s).^65–68 However, the Fout group’s work represents an attractive approach demonstrating that secondary sphere groups can be used to facilitate oxyanion reduction.^{25–26, 69–70} In addition to ClO₄⁻ reactivity, secondary sphere H-bonds donors have been shown to influence the stability of high-valent iron oxos.^71–75 Our group is investigating the extent to which a metal’s secondary coordination sphere can be used for targeted reaction engineering by tuning the electronics, geometry, and cooperative metal-substrate-acid binding modes.^76–78 We reasoned that an H-bond appended TPA system would provide thermodynamic stabilization of an F̵–O bond, ultimately driving the ClO₄⁻ reduction sequence. A related strategy was previously used by our group to capture O₂ and stabilize otherwise unstable peroxo adducts with Zn and Cu.^77–78 Based on the C–H oxygenation reactivity from related Fe-TPA complexes, we propose that O-atom transfer reactivity from ClO₄⁻ to Fe might be coupled with an oxygenation step of organic substrates (Figure 1c).^79–80

RESULTS AND DISCUSSION

We targeted an FeOTf₂ complex coordinated to the tetradentate tripodal ligand tris(6-phenylamino-2-pyridylmethyl)amine (L^H, Figure 2). Dropwise addition of 1.01 equiv L^H to Fe(OTf)₂ in MeCN followed by stirring for 48 h at room temperature afforded a pale-yellow powder in 87–95% yield. The ¹H NMR spectrum indicated a mixture of products (I and II) (two sets of paramagnetic resonances). Modifying the procedure (faster rate of addition of L^H) provided a single species (I; high spin, S = 2, μ_eff = 5.02 ± 0.15) (one of the two sets of resonances above) containing C₃-symmetric paramagnetic resonances in the ¹H NMR spectrum. Crystals obtained from this reaction (THF/pentane) were subjected to a single-crystal X-ray diffraction (SCXRD) experiment. The solid-state structure revealed L^H coordinated to Fe with OTf⁻ (axial) and THF (equatorial) molecules completing the octahedral coordination environment (Figure 2). The structural metrics of from the SCXRD data revealed an average N_aniline–O_axial bond distance of 3.022 Å, which is consistent with moderate-strength H-bonding interactions.⁸¹

Figure 2.

Synthesis of Fe(OTf)₂L^H complex. Molecular structures (50% probability) of I (THF adduct shown from crystals grown from THF/pentane), II, and III are shown (non-participating hydrogen atoms, counterions, and co-crystallized solvent molecules excluded for clarity).

Although a single compound could not be obtained from the mixture of products above, crystals of complex II were obtained by vapor diffusion of pentane into a THF solution and subjected to an SCXRD experiment. The solid-state structure revealed a distinct composition: an amidine ligand, likely formed by nucleophilic addition of the pendent aniline to a coordinated MeCN ligand (Figure 2). Reactions of amines with coordinated nitrile ligands are precedented,^82–85 and formation of II is likely facilitated by: a) close proximity of the aniline to coordinated MeCN, and b) formation of a stable 6-membered chelate ring upon addition.

Despite the distinct solid-state structures of I and II, we found that I can convert to II in solution, and that both undergo ligand substitution reactions (see SI, figure S32 and S33). Heating an MeCN solution of a mixture of compound I/II to 80 °C increased resonances attributed to II. In a separate set of reactions, heating a solution containing 2.0 equiv NBu₄Cl with a mixture of compound I/II to 80 °C cleanly afforded a single product. We assign the terminal product as FeCl₂L^H (III), based on spectroscopic comparisons to an authentic sample prepared by the reaction of FeCl₂ and L^H (Figure 2). These results indicate that, although mixtures of I/II are obtained during metalation in MeCN solvent, they both are competent precursors for ligand substitution reactions, and for clarity we denote their formulation throughout the manuscript as I*.⁸⁶

To evaluate the extent to which appended H-bond donor groups influence ClO₄⁻ binding affinity for Fe(II), we examined the UV-Vis spectra of I* with 1.0 equiv [NBu₄][ClO₄] in MeCN at room temperature (Figure 3a). We observed an immediate change upon addition of ClO₄⁻: a band at 645 nm (after 5 min) that transformed into a new band at 605 nm over 18 h.⁸⁷ These observations differed when performing analogous control reactions using Fe-TPA variants devoid of secondary sphere H-bonding groups: Fe(OTf)₂(6-Me₃TPA) (IV, high spin) and Fe(OTf)₂TPA (V, low spin). Neither complex displayed any changes to the UV-Vis spectra upon addition of 1.0 equiv of ClO₄⁻ (Figure 3b and SI, figures S36 and S38). Importantly, even introduction of a large excess (100 equiv) of [NBu₄][ClO₄] did not change the UV-Vis spectrum of IV (see SI, figure S37).⁸⁸ These data demonstrate a clear difference between ligand variants and suggest that the appended HN-Ph groups facilitate ClO₄⁻ reactivity in a manner that is not solely due to increased binding affinity.

Figure 3.

UV-Vis studies of: Top, a) I* with [NBu₄][ClO₄] and (Ph₃C)₂. Bottom, b) IV and V with [NBu₄][ClO₄] independently.

The atypical UV-Vis result prompted us to examine the feasibility of ClO₄⁻ reduction in analogy to prior reports.^{9–13, 25–26} The UV-Vis experiment above was repeated with three variations. First, introduction of 4.0 equiv of an O-atom acceptor (PPh₃) caused only slight changes to the UV-Vis spectrum when added 18 h after [NBu₄][ClO₄] and did not bleach the absorbance when added after 5 min. However, its addition prior to [NBu₄][ClO₄] afforded a distinct UV-Vis spectrum (see SI, figures S43–S45) and O=PPh₃ in 38% yield.⁸⁹ Second, introduction of a weak H-atom donor, 9,10-dihydroanthracene (DHA, 1a; BDE= 78 kcal/mol)⁴⁶ (see SI, figures S41–S42) did not change the UV-Vis spectrum when added after 5 min or 18 h. Based on precedented reactivity of related high valent iron-oxos with DHA and PPh₃,^{46, 90} we propose that neither of the compounds observed by UV-Vis spectroscopy correspond to a stable iron-oxo.

However, addition of Gomberg’s dimer ((Ph₃C)₂, a source of •CPh₃) resulted in complete consumption of the 605 or 645 nm peak (Figure 3a and S40). When (Ph₃C)₂ was added prior to [NBu₄][ClO₄], no band at 645 nm was observed (see SI, figure S39). These observations demonstrate that the stable product formed upon introducing ClO₄⁻ reacts with •OH acceptors such as (Ph₃C)₂, and is thus consistent with an Fe(III)OH. By analogy to prior reports with O₂ derived non-heme Fe(III)OH complexes,^{75, 91} we propose that the Fe(III)OH forms from a putative Fe(IV)=O.⁹²

Following HAT to an Fe(IV)=O, the resulting Fe(III)–OH species can undergo •OH rebound to a carbon-based radical.^93–94 We assessed this reactivity mode using (Ph₃C)₂ in presence of 1a, To target a catalytic system, we introduced AgOTf to sequester the formed Cl⁻ (byproduct of exhaustive deoxygenation) and facilitate turnover. Addition of 1.0 equiv 1a (as a HAT source) and 1.0 equiv Ph₃C• (via 0.5 equiv (Ph₃C)₂) to a mixture of 1.0 equiv [NBu₄][ClO₄], 1.0 equiv AgOTf, and 25 mol% I* afforded Ph₃COH in 57% yield after heating to 80 °C for 24 h (Figure 4a). In addition to Ph₃COH, we also observed the formation of two other oxidized products: anthracene (2a, 16%) and anthraquinone (2b, 37%).⁹⁵ We found that both products were also formed in the absence of (Ph₃C)₂, albeit in lower yields (Figure 4b). Although these results may implicate •OH rebound reactivity with the product of HAT from 1a, we found that 2b was also generated when 2a was used directly as a substrate. Under identical reaction conditions, 2b formed from 2a in 40% yield (Figure 4c). These results prompted us to examine C–H oxygenation reactivity of anthracene.

Figure 4.

Initial C–H oxygenation experiments: a) with (Ph₃C)₂ in presence of 1a, b) with only 1a, c) with 2a.

Catalytic oxidation of anthracene to anthraquinone was optimized by varying loading, solvent, additives, temperature, and time (Table 1 and SI, table S2). Increasing the amount of [NBu₄][ClO₄] relative to I* and keeping other parameters the same increased the yield from 40% to 51% over 18 h at 12.5 mol% catalyst loading (Table 1, entry 1). The TON was calculated based on the mmol of [NBu₄][ClO₄] (TON_ClO4) and O-atoms (TON_O) consumed (by 2a to form 2b) over mmol of catalyst used.⁹⁶ Changing the solvent to propionitrile (EtCN) enabled higher reaction temperatures (120 °C) which further improved the yield (74%; TON_ClO4 = 1.48; entry 2).⁹⁷ Activity was maintained at lower catalyst loadings, with 2.5 mol% I* affording a TON_ClO4 of 2.47 (entry 3).

Table 1.

Optimization of C–H oxidation of anthracene^a

entry	catalyst (mol%)	solvent	temp (°C)	yield (%)b	TONClO4 (TONO)
1	I* (12.5)	MeCN	80	51±1	1.01±0.02 (4.1±0.1)
2	I* (12.5)	EtCN	120	74±3	1.48±0.02 (5.9±0.2)
3	I* (2.5)	EtCN	120	25±2	2.47±0.21 (9.9±0.8)
4	VI (2.5)	EtCN	120	42±2	4.17±0.15 (16.7±0.6)
5	VII (2.5)	EtCN	120	11±2	1.10±0.17 (4.4±0.7)
6	IV (2.5)	EtCN	120	5±1	0.50±0.10 (2.0±0.4)
7	V (2.5)	EtCN	120	17±2	1.67±0.15 (6.7±0.6)
8	VI (7.5)	EtCN	120	82±2	2.72±0.05 (10.9±0.2)
9c	I* (12.5)	MeCN	80	16±1	0.31±0.02 (1.3±0.1)
10c,e	-	MeCN	80	nd	-
11c,d,e	I* (12.5)	MeCN	80	nd	-

^aReaction conditions: [NBu₄][ClO₄] (0.08 mmol, 1.0 equiv), 2a (0.5 equiv), catalyst (mol%), AgOTf (0.5 equiv), solvent (1 mL).

^b1H NMR yield using 1,3,5-trimethoxybenzene as internal standard.

^cwithout AgOTf.

^dwithout [NBu₄][ClO₄].

^ereaction was run for 24 h. TON was calculated by dividing the mmol of [NBu₄][ClO₄] consumed (TON_ClO4) and O-atoms (TON_O) transferred (to form 2b) by mmol of I*used. Averages with associated standard deviations for yields and TONs are based on triplicate measurements.

We note that anthraquinone formation is precedented with Mn(IV)-oxo or hydroxo complexes to form H₂O in addition to the ketone products.^98–99 Addition of 5.0 equiv Me₃SiCl after 18 h in conditions analogous to entry 1 indicated formation of ~3.0 equiv of (Me₃Si)₂O with respect to 2b (see SI, figure S72). This result implicates the generation of ~3.0 equiv of H₂O during the reaction. As an alternative technique, we directly quantified ClO₄⁻ by ion-selective electrode (ISE) conductivity measurements (see SI, tables S4–S7 and figure S73). This analysis was performed after repeating the experiment in entry 1, followed by solvent removal in vacuo, then reconstitution in H₂O. This procedure facilitated the measurement of dissolved ClO₄⁻ and afforded a value of ~0.025 mmol of [ClO₄⁻] consumed (0.08 mmol added). This value corresponds to an overall loss of ~2.5 equiv with respect to I* (~10 TON with respect to O-atoms transferred). The TON exceeds that observed from the ¹H NMR calculations (4.1). This result is consistent with the formation of ~3.0 equiv of H₂O in addition to 2b. To simplify the TON analysis (quantification of [ClO₄⁻] and O-atoms consumed with respect to I*), we report the TON by only taking account of the O-atoms transferred to anthraquinone from [NBu₄][ClO₄], measured using NMR spectroscopy.⁹⁶

Given that many Fe^IV(O)(TPA) complexes form bis-μ-oxo dimeric structures that are less active for O-atom transfer reactivity than the monomers,¹⁰⁰ we hypothesized that catalytic activity could be improved by modifying the ligand architecture with sterically encumbering groups. We prepared electronically distinct, yet sterically analogous compounds VI and VII featuring two new ligand variants containing 3,5-substituents at the aniline unit: L^tBu (electron-rich) and L^CF3 (electron-deficient) (Figure 5). Both complexes are high spin (S = 2); (μ_eff = 5.27 ± 0.13) (VI) and (μ_eff= 5.20 ± 0.20) (VII) and show similar spectroscopic data to I. To evaluate the redox properties of each TPA variant, we performed electrochemical analysis experiments in 0.1 M NBu₄Cl in dimethylformamide, conditions that form the corresponding FeCl₂L^R complexes in-situ.¹⁰¹ Complexes I, VI, and VII all exhibit reversible Fe^II/III redox couples at a similar potential (E_1/2 = –0.560 V vs Fc). The <10 mV difference in redox potentials between the three variants demonstrates that changes in ClO₄⁻ reactivity are unlikely to be attributed to differences in their reducing ability.

Figure 5.

Synthesis of Fe(OTf)₂L^R complexes and stacked cyclic voltammograms of in-situ generated FeCl₂L^R complexes derived from I, VI, and VII. Molecular structure (50% probability) of VI shown (co-crystallized solvent molecules and non-participating hydrogen atoms excluded for clarity).

The solid-state structures of VI and VII are pseudo-octahedral and contain two coordinated OTf⁻ ligands (Figure 5 and SI, figure S103). SCXRD data show an average N_aniline–O_axial bond distance of 3.08 Å for complex VI and 3.044 Å for complex VII which is comparable to 3.022 Å (I) and similarly fall in the region of intermediate strength H-bond distances. These observations, along with the cyclic voltammetry results indicate that differences in ClO₄⁻ reactivity are most likely derived from the secondary sphere aniline groups, rather than redox tuning at Fe. Complexes VI and VII similarly exhibit ¹H NMR spectra consistent with C₃ symmetry, while the ¹⁹F NMR spectrum exhibited no free OTf⁻ signals. However, ¹H and ¹⁹F NMR recorded at −40 °C (VI) showed increased numbers of paramagnetic signals consistent with rapid OTf⁻ exchange on the NMR timescale. This process was reversible when warmed to room temperature (see SI, figures S34–S35). In addition to these aniline TPA derivatives, we also evaluated two complexes devoid of secondary H-bond donors: TPA ligands containing either –CH₃ (IV) or –H (V) in place of the –HNAr units.

Following their preparation, we directly compared catalytic reactivity across the series. Complex VI, containing ^tBu groups, provided the highest yield (43%; TON_ClO4 = 4.17, entry 4).¹⁰² In contrast, complex VII with CF₃-substituents, afforded a lower yield (11%, entry 5). Control complexes (IV and V) similarly showed low catalytic activity, providing 5% and 17% yields of 2b, respectively (entries 6, 7). We found that a catalyst loading of 7.5 mol% provided an optimal yield of 82% with a TON_ClO4 of 2.72 (entry 8). The role of AgOTf additive was apparent: in its absence, the yield reduced to 16% (entry 9).¹⁰³ Finally, control experiments indicated no detection of the desired product in the absence of either I* or [NBu₄][ClO₄] (entries 10, 11).

After optimizing the conditions (Table 1, entry 8), we next explored the substrate scope on a 0.1 mmol scale (Figure 6). Alkylaromatic substrates with C–H bond BDEs in the range of 68–86 kcal/mol, anthracene (BDE = 111 kcal/mol), and its chloro-derivative were examined.^{46, 104–105} Anthracene converted to anthraquinone (2b) in an excellent 92% isolated yield. The methodology was compatible with a halide functionality as shown by 2-chloroanthracene, which furnished 86% of the corresponding anthraquinone derivative (2c). Benzylic systems composed of relatively weak C–H bonds were next evaluated. DHA furnished a mixture of anthracene and anthraquinone in 7% and 58% yields respectively. When anthrone was used as a substrate, a potential intermediate en route to anthraquinone, 2b formed in almost quantitative yield (97%).

Figure 6.

Scope of C–H substrates. ^a1H NMR yield using 1,3,5-trimethoxybenzene as internal standard. ^bI (25 mol%) used as catalyst. ^c0.04 mmol PhCH(OH)Ph used as a substrate.

Heterocyclic benzylic substrates were tolerated, forming products derived from xanthene (2d) (75.2 kcal/mol), thioxanthene (2e) (74.6 kcal/mol), N-substituted (methyl, 2f) (80 kcal/mol) and (phenyl, 2g) dihydroacridines in the range of 61–98% yields. The 5-membered carbocyclic substrate fluorene (~82.2 kcal/mol) gave only trace amounts of the fluorenone product (2h, 7%). This result could be attributed to the comparatively higher BDE of the fluorenyl C–H bond, compared to DHA. However, 1,4-dihydronaphthalene, a substrate with a similar BDE (83 kcal/mol) to fluorene, generated naphthalene (2i) in an excellent 95% yield but did not afford 1,4-naphthaquinone. Acyclic substrates, diphenylmethane (82 kcal/mol) formed benzophenone (2j) only in trace amounts (<2%), whereas ethylbenzene (~85.4 kcal/mol) did not yield the respective ketone product, acetophenone (2k), at all. The trace or non-formation of products in case of fluorene, diphenylmethane, and ethylbenzene provide support that these reactions are not strictly governed by BDE values.

We performed isotope labelling experiments using H₂¹⁸O as a mechanistic probe. We introduced 5.0 equiv of H₂¹⁸O prior to addition of [NBu₄][ClO₄] and I* and then subjected the sample to the reaction conditions analogous to entry 1 (Table 1). Ex-situ analysis by GC-MS showed ¹⁸O-incorporation in anthraquinone as a mixture of [M], [M+2], and [M+4] products (see SI, figures S75–S81). This observation is consistent with either: 1) ¹⁸O exchange at Fe prior to oxygenation, or 2) addition of H₂O or OH⁻ to the oxidized anthracene. Note that (1) has been previously reported with related Fe(IV)=O species,^{52, 106} and (2) has been implicated in the direct oxidation of anthracene by ClO₂.¹⁰⁷

Based on the above observations, we propose these reactions proceed analogous to prior reports with related systems (see SI, figure S82).^{37, 46, 52, 99} Alkylaromatics often undergo a HAT-initiated mechanism,^{37, 46} while anthracene may undergo either an electrophilic attack or outer-sphere oxidation by an Fe(IV)=O species.^{52, 99} Based on the distinct reaction outcomes when using fluorene, diphenylmethane, and 1,4-dihydronaphthalene, which all feature similar BDE values, we propose that the H-atom transfer step does not govern the reaction success. Instead, the oxygenation step(s) determines the net reaction. This conclusion was supported by separate reactions when diphenylmethanol was used as a substrate: benzophenone formed in 87% yield (see SI, figure S98); however, this product was only formed in trace amounts when diphenylmethane was treated as a substrate.

CONCLUSION

In summary, we have reported a secondary sphere H-bond assisted ClO₄⁻ reduction protocol coupled with C–H oxidation. This system employs appended aniline (–NH) H-bond donors that are proposed to function by stabilizing the product of O-atom transfer. This catalytic system exhibits radical rebound chemistry via sequential HAT/OAT reactions to afford C–H oxidized products. Ongoing work aims to identify the product of O-atom transfer formed from ClO₄⁻ in addition to improving the C–H oxygenation scope as well as elucidating the underlying mechanistic details.