Advances and mechanistic insight on the catalytic Mitsunobu reaction using recyclable azo reagents

Catalytic Mitsunobu reactions have been substantially improved based on strict optimization, mechanistic studies and discovery of a new catalyst.


Introduction
Many reactions have been utilized as important tools in synthetic organic chemistry. "Name reactions", such as Wittig, Suzuki-Miyaura, and Mitsunobu, to name just a few, have an outstanding utility that has inuenced broad elds of academia and industry. 1 In view of economic and environmental concerns however, many of these synthetic methods suffer from serious limitations, diminishing their practical applicability. Therefore, substantial improvements of known synthetic protocols are currently an important subject in chemistry.
Indeed, the Mitsunobu reaction is a typical example including both a wide utility and serious drawbacks. 2 The reaction is one of the oxidation-reduction condensations reported by Mitsunobu and co-workers in 1967. 3 Since then, it has been widely used for the substitution of hydroxyl groups or inversion of the stereochemistry of secondary alcohols. Typically, diethyl azodicarboxylate (DEAD) and triphenylphosphine are employed as the oxidant and reducing agent in the Mitsunobu reaction, but production of a large amount of waste, i.e., diethyl hydrazinedicarboxylate and triphenylphosphine oxide, is unavoidable. These byproducts oen contaminate the desired product. In addition, DEAD is hazardous due to its toxicity and potential explosiveness. As a result, the use of the Mitsunobu reaction tends to be avoided in practical synthesis on plant scales. 4 Several modied methods have been developed to facilitate the removal of the waste generated by the Mitsunobu reaction. 5 However, there has been no substantial approach to reducing the problematic waste in the Mitsunobu reaction until the report on the catalytic Mitsunobu reaction by Toy in 2006. 6 Toy succeeded in reducing DEAD in the Mitsunobu reaction to a catalytic amount (10 mol%) by employing a sacricial oxidative reagent, i.e., iodobenzene diacetate. Recently, Mitsunobu-type reactions without azo reagents were reported. 7 In 2013, we reported the second example of the catalytic Mitsunobu reaction with azo reagents that are recyclable through aerobic oxidation with iron phthalocyanine (Fig. 1A). 8 Ethyl 2-(3,4-dichlorophenyl)hydrazinecarboxylate (1a) has been tentatively identied as the best catalyst. A catalytic concept of this reaction is benecial from the viewpoint of green chemistry because atmospheric oxygen is economically and environmentally ideal as a sacricial oxidant to generate a reactive azo form 2a (Fig. 1B). However, the scope of substrates and product yields were still moderate, and the reaction required heating conditions to obtain the products in acceptable yields. Thus, the applicability of the method was still inferior to that of the original Mitsunobu reaction.
The effect of substituents on the aromatic ring of the hydrazine catalysts was drastic. Clearly, electronic properties of catalysts affected both the Mitsunobu reactivity of the azo form as well as the aerobic oxidation of the hydrazine form. At rst glance, these seem incompatible because electron-withdrawing groups would promote the addition reaction of triphenylphosphine to the azo form but would suppress oxidation of the hydrazine form to the azo form. In the case of electron-donating groups there is the same dilemma, though the situation is interchanged. We presumed that the 3,4-dichlorophenyl group had an electronic property that made the two processes moderately compatible.
Quite recently, we have reported a detail of the aerobic oxidation process of 2-arylhydrazinecarboxylates with iron phthalocyanine, indicating two important observations. 9 First, the oxidation process was promoted in apolar solvents such as toluene or dichloromethane, and second, electron-withdrawing substituents at the aryl group did not suppress the hydrazine-toazo compound oxidation. Interestingly, halogen atoms at the para-position rather promoted the reaction. Thus, this study provided us important insights to improve the catalytic Mitsunobu reaction.
Providing the serious limitations indicated in Fig. 1C are avoided, the catalytic Mitsunobu reaction will gain a large potential in practical synthesis. 10 In this paper, we describe new advances in our catalytic Mitsunobu reaction including substantial improvement of the reaction and insights into the reaction mechanism.

Strict optimization of the reaction conditions
We previously found that the combination of ethyl 2-(3,4dichlorophenyl)hydrazinecarboxylate (1a) and iron phthalocyanine [Fe(Pc)] formed an optimum catalytic system (both 10 mol%), and that addition of activated molecular sieves was required to induce the reaction (vide infra). 8 We tentatively improved the yields of the products by using 3,5-dinitrobenzoic acid as a nucleophile when secondary alcohols were used as substrates. 8 We employed a model reaction between (S)-ethyl lactate (3, 99 : 1 er) and 4-nitrobenzoic acid (4) using this catalytic system to strictly optimize the conditions. The reaction between 3 and 4 in heating THF (65 C) gave ester product 5 in 50% yield and in 97% inversion (Table 1, entry 1).
In the previous study, the effect of solvents was investigated at a very preliminary stage using unoptimized catalysts. 11 We could not nd a large effect of the solvents at that time, and  thereby, the effects of solvents and temperature were reinvestigated using the optimum catalytic system (Table 1). 12 Ether solvents such as 1,4-dioxane, cyclopentyl methyl ether (CPME) 13 and tert-butyl methyl ether (MTBE), except for dimethoxyethane (DME), provided product 5 in a high inversion ratio (entries 2-5), whereas acetonitrile gave a contrasting result (entry 6). 14 Reactions in hydrocarbon solvents such as n-hexane and toluene at 65 C afforded good results (entries 7 and 8). However, chlorinated solvents gave product 5 in a low inversion ratio, though the total product yield was good (entries [12][13][14]. This drastic change in the results was attributed to the presence of chlorine atoms in the solvent, and is based on the fact that the reaction in a,a,a-triuorotoluene 15 provided similar results to those in toluene (entry 15). The enantiomeric ratio was sensitive to temperature in the reaction in toluene (entries 8-10). To our delight, the reaction in toluene at room temperature provided product 5 in an excellent yield (88%) and in a perfect inversion ratio. CPME also gave a relatively good result for the reaction at room temperature. The reactions were basically clean. In the case of low yields of the product, the starting materials remained unconsumed. The effect of molecular sieves was drastic, and no reaction was induced in their absence ( Table 2, entry 1). 16 This is likely due to the high moisture sensitivity of the intermediate generated from the azo form of catalyst 1a and triphenylphosphine. Molecular sieves would serve for removing residual moisture as well as water generated by the iron-catalyzed aerobic oxidation of the hydrazine catalyst. The use of at least 500 mg MS 5Å (1.0 mmol scale), activated by heating with a heat gun (ca. 450 C) under reduced pressure (ca. 0.1 mmHg), was desirable to obtain product 5 in a good yield (entries 2-5). MS 4Å and MS 3Å were ineffective in the present reaction (entries 6 and 7).
Various "traditional methods" for the activation of molecular sieves are used in many laboratories. Representative activation methods were tested to assure a reliable experimental procedure. The use of MS 5Å without activation gave the product in a very poor yield (entry 8). MS 5Å heated for 24 h at 140 C in an oven were also ineffective (entry 9). Although heating using a microwave is sometimes used for activation of molecular sieves, this method did not afford a good result in the present reaction (entry 10). When the reaction was tested with MS 5Å activated through heating at 200 C with an oil bath under reduced pressure (ca. 0.1 mmHg), the product yield was still insufficient (entry 11). Heating using a ame under reduced pressure would be a strict method for activation of molecular sieves, and this method provided product 5 in an excellent 94% yield (entry 12). As a result, and from the viewpoints of safety and convenience, we consider the activation with a heat gun as the method of choice. Incidentally, sulfate salts did not work as a desiccant in the reaction (entries [13][14][15]. The concentration of the reactants is likely to affect the product yield (Table 3,  However, the use of a lower amount (1.1 equiv.) of triphenylphosphine diminished the yield of product 5 (entries 6 and 9). High concentration conditions would be benecial to a practical synthesis because the solvent can be saved. The good result was reproducible in a scale-up experiment (10 mmol), though the reaction time was somewhat prolonged (entry 7, results in parentheses). Triphenylphosphine is sometimes replaced with trialkylphosphines because they oen provide good results due to their high nucleophilicity. 17 We tested a representative reaction with tri-n-butylphosphine, but the result was very poor (entry 8, results in parentheses). TLC analysis of the reaction mixture implied decomposition of the iron phthalocyanine presumably through strong coordination with the tri-n-butylphosphine. When most of the triphenylphosphine was consumed in the reaction, the Mitsunobu catalyst was detected as the azo form using TLC. The latter was easily recovered in 80-90% yield using silica gel chromatography due to its low  polarity. The hydrazine form of the catalyst, if it remained in the reaction mixture, usually did not cause problems in the purication of the product. Finally, iron phthalocyanine could be easily removed using ltration of the reaction mixture through a pad of Celite® or lter paper. The impact of decreasing the amount of hydrazine catalyst 1a seemed to be larger than that of decreasing the amount of iron phthalocyanine (entries 10-15). It is noteworthy that good results were maintained with as low as 1 mol% of iron phthalocyanine (entries 11 and 12) indicating that its amount can be exibly changed depending on the substrates or situations of the reactions. No reaction was induced in the absence of the iron catalyst. 8,9 Kinetic properties of the ethyl 2-arylazocarboxylates The catalytic cycle between the hydrazines and azo compounds would affect the efficiency of the formation of an alkoxyphosphonium intermediate to provide the nal product. We conducted kinetic experiments to investigate the substituent effect of azo compounds 2b-j in the reaction with triphenylphosphine (Fig. 2). The analysis of a mixture of 2b-j and triphenylphosphine (10 equiv.) in CDCl 3 using 1 H NMR spectroscopy revealed the presence of some starting azo compounds aer 10 hours. In contrast, in an independent experiment, 1 H NMR analysis showed that DEAD immediately disappeared under the same reaction conditions, indicating an irreversible process in this case. 18 Obviously, the addition of triphenylphosphine to ethyl 2-arylazocarboxylates is reversible, and the formation of adducts is less favorable as compared to DEAD. Therefore, reaction rates were estimated from the model reaction of azo compounds 2b-j (50 mM) with excessive   amounts (10 equiv.) of triphenylphosphine and water in THF at 25 C. The reactions were monitored by measuring the absorbance of the azo compounds 2b-j at l ¼ 419-450 nm. Rate constants were calculated from plots of a pseudo-rst-order dependence. The Hammett plot for these reactions shows a linear t with a relatively large positive slope value of r ¼ +2.71 (Fig. 2). The value is close to that of the alkaline hydrolysis of benzoate esters (r ¼ +2.51). 19 The result reects a dependence of the electronic density at the aromatic ring of azo compounds in the rate of the addition reaction of triphenylphosphine. Ethyl 2-(3,4-dichlorophenyl)azocarboxylate (2a) was also applied to the kinetic experiment, and its reaction rate (k obs ¼ 8.5 Â 10 À2 min À1 ) was approximately 13.7 times faster than that of ethyl 2-phenylazocarboxylate (2d, k obs ¼ 6.2 Â 10 À3 min À1 ). In addition, it is still 2.3 times faster compared to that of ethyl 2-(3-chlorophenyl) azocarboxylate (2g, k obs ¼ 3.75 Â 10 À2 min À1 ). This supports the high reactivity of 2a in the catalytic Mitsunobu reaction.
When benzoic acid or 4-nitrobenzoic acid (each 10 equiv.) were added to the reaction system with 2d, only a minor impact to the reaction rate was noted (2d with benzoic acid: k obs ¼ 7.1 Â 10 À3 min À1 ; 2d with 4-nitrobenzoic acid: k obs ¼ 6.8 Â 10 À3 min À1 ). This observation supports that the model reaction reects the reactivity of azo compounds toward triphenylphosphine and indicates that acids do not kinetically affect the reaction.
The kinetics of the catalytic aerobic oxidation of ethyl 2arylhydrazinecarboxylates (1) with iron phthalocyanine basically show zero-order dependence, but the substituent effect is of irregular tendency probably due to the participation of radical species in the mechanism. 9 The reaction rates of aerobic oxidation of ethyl 2-(4-chlorophenyl)hydrazinecarboxylate (1f) and ethyl 2-(4-bromophenyl)hydrazinecarboxylate to the corresponding azo compounds are approximately 1.5 times faster than that of ethyl 2-phenylhydrazinecarboxylate (1d). 9 In the model reaction, in dichloromethane as a solvent, the aerobic oxidation of ethyl 2-(3,4-dichlorophenyl)hydrazinecarboxylate (1a) with iron phthalocyanine is completed within 2 hours. This is clearly faster than the oxidation (4 hours) 9 of ethyl 2-phenylhydrazinecarboxylate (1d), though the kinetics of the reaction of 1a do not show a clear zero-order dependence (Fig. S14 in the ESI †). Thus, the 4-chlorine atom on the aromatic ring of 1a promotes oxidation to the corresponding azo form 2a by stabilization of the intermediary radical species, whereas the 3chlorine atom of azo compound 2a contributes to an increased electrophilicity by its inductive effect. This is the reason why azo compound 2a operates as a good catalyst in the catalytic Mitsunobu reaction. In short, two processes involving Mitsunobu activity and hydrazine re-oxidation are compatible through the 3,4-dichlorophenyl group (Fig. 3). The catalytic activity of ethyl 2-(4-chlorophenyl)hydrazinecarboxylate (1f) was insufficient under the optimal conditions compared to that of 1a (Fig. 4).
Given the above considerations, ethyl 2-arylhydrazinecarboxylates with strong electron-withdrawing groups on the aromatic ring should be more effective catalysts as these groups should promote the Mitsunobu reaction without signicantly suppressing the aerobic oxidation process. For instance, as monitored using NMR spectroscopy, the aerobic oxidation of ethyl 2-(4-cyanophenyl)hydrazinecarboxylate (1j) was completed within 5 hours, which was roughly the same reaction time as that of ethyl 2-phenylhydrazinecarboxylate (1d) (ca. 4 hours). 9 On the other hand, higher electrophilicity of 2-(4cyanophenyl)azocarboxylate (2j) over 3,4-dichlorophenyl derivative 2a is consistent with the higher (3.8 times) reaction rates of 2j (k obs ¼ 3.2 Â 10 À1 min À1 ) over 2a (Fig. 3). This suggested that ethyl 2-(4-cyanophenyl)hydrazinecarboxylate (1j) might work as a good catalyst in the catalytic Mitsunobu reaction.
When ethyl 2-(4-cyanophenyl)hydrazinecarboxylate (1j) was used in the reaction between (S)-ethyl lactate (3) and 4-nitrobenzoic acid (4) under optimal conditions, product 5 was obtained in an excellent yield, although with a slightly decreased inversion ratio (Fig. 4). On the other hand, when phenol (7) or phthalimide (8) was used as the reaction partner of 3-phenylpropanol (6), both reactions using 1j provided better results (87% and 84% yields) than the reactions with 1a (51% and 66% yields). Although 2-(4-nitrophenyl)hydrazinecarboxylate (1k) should generate a strongly electrophilic azo compound, 20 the results with this catalyst were disappointing. Gradual decomposition of 1k or its azo form was observed in the reaction with triphenylphosphine using 1 H NMR analysis, which appears to be the main reason for the poor results. 21 The reaction rate of ethyl 2-[4-(ethoxycarbonyl)phenyl]azocarboxylate (2h, k obs ¼ 6.4 Â 10 À2 min À1 ) and ethyl 2-[4-(tri-uoromethyl)phenyl]azocarboxylate (2i, k obs ¼ 1.2 Â 10 À1 min À1 ) with triphenylphosphine was roughly close to that of 2a. Good yields of ester 5 were obtained in the reaction between (S)-ethyl lactate (3) and 4-nitrobenzoic acid (4) using the hydrazine forms 1h and 1i as a catalyst, but reaction times were prolonged (Fig. 4). When phenol (7) or phthalimide (8) were used as a nucleophile in the reaction with 3-phenylpropanol (6), catalysts 1h and 1i did not provide better results than catalyst 1j, though catalyst 1i showed somewhat improved results compared with catalyst 1a. Thus, catalyst 1h showed reactivity similar to that of 1a, and the position of reactivity for catalyst 1i is likely to lie between 1a and 1j. These trends are consistent with the results of the Hammett study. Incidentally, when model experiments of iron-catalyzed aerobic oxidation of 1h and 1i were conducted in dichloromethane, the reactions were completed at 4 h and 6 h, respectively (see the ESI †). The trend of the oxidation process is similar to that of other hydrazide derivatives. 9 The above results imply that there is no perfect catalyst for the catalytic Mitsunobu reaction. Instead two catalysts can complement each other. In short, ethyl 2-(3,4-dichlorophenyl) hydrazinecarboxylate (1a) would be suitable for the reactions of carboxylic acids whereas 2-(4-cyanophenyl)hydrazinecarboxylate (1j) could serve for the reactions of other nucleophiles except for carboxylic acids.

Scope of substrates using the optimized protocol
The discovery of new catalyst 1j largely expanded the scope of the catalytic Mitsunobu reaction. Fig. 5 shows the results of catalytic Mitsunobu reactions applying catalyst 1a or 1j to various substrates. Typically, the reactions were performed under the optimal conditions that provided the best result (Table 3, entry 7), but more practical conditions (e.g., Table 3, entry 11) were also applicable to several substrates. Reactions between 3-phenylpropanol and various carboxylic acids with catalyst 1a provided the corresponding esters 11-15 in excellent yields. The reaction of the alcohol with phenols gave the corresponding ethers 9 and 16 in improved yields when catalyst 1j was employed. An iodine atom was intact under the present conditions in the reaction of 4-iodophenol to give 16. Reactions with phthalimides and the nosylamide needed to be performed in 0.5 M solution due to the solubility issues. In such cases, heating the reaction mixture at 65 C improved the results in reaction time and product yield. Alcohols sensitive to oxidative conditions were tested with several nucleophiles and were transformed into the corresponding Mitsunobu products 21-24 in good yields. It is noteworthy that a trisubstituted olen, a thiophene and an indole were intact under the aerobic oxidation conditions. The catalytic Mitsunobu reaction using catalyst 1j was applicable to intramolecular reactions of alkyl sulfonamides having a hydroxyl group to give the corresponding cyclic amines 25 and 26 (ref. 23b) in reasonable yields.
Next, various combinations of secondary alcohols and nucleophiles were tested (Fig. 6). Reactions of (S)-ethyl lactate (3) with several aromatic carboxylic acids gave the corresponding esters 34-36 in good yields with almost full inversion of stereochemistry. The reaction of alcohol 3 with 3,5-dinitrobenzoic acid in toluene gave ester 35 in a moderate level of enantioenrichment (er, 83 : 17). The reaction of 3,5-dinitrobenzoic acid, under the previous conditions (in THF at 65 C) provided 35 in a higher level of enantioenrichment. 8 In the reaction of 3 with 3-phenylpropionic acid, the enantioenrichment of ester 37 was not good (er, 78 : 22), but the reaction at low temperature (0 C) gave an improved result (er, 90 : 10). Other nucleophiles such as phenol and phthalimide were applicable to reactions of chiral secondary alcohol 3 to provide the corresponding Mitsunobu products 38 and 39, though the product yields were somewhat moderate. Reactions of other representative secondary alcohols 27-32 with 4-nitrobenzoic acid (4) readily provided the corresponding inversion products 40-45 in good yields. There was a slight loss of the optical purity of ester 42, which was also observed in the typical Mitsunobu reaction with DEAD. 6a However, the case of (À)-menthol (33) was still a limitation in the catalytic Mitsunobu reaction even though a highly acidic carboxylic acid was employed. 24 For instance, the reaction of 33 with 4-nitrobenzoic acid gave inversion product 46 as a minor isomer. Fortunately, we found out that inversion product 47 was produced exclusively when the 2-methyl-6-nitrobenzoic acid was used as a nucleophile. These contrasting results could be attributed to the catalytic system. The reaction with a catalytic amount of the azo reagent maintains a low concentration of an intermediary alkoxyphosphonium salt. There would be an equilibrium process between the alkoxyphosphonium intermediate and an acyloxyphosphonium intermediate. 25 If a subsequent reaction of the alkoxyphosphonium intermediate with a carboxylic acid to give an inversion product is slow, a retention product would increase via the equilibrium process to give the acyloxyphosphonium intermediate because the concentration of a free carboxylic acid is sufficiently higher than that of the alkoxyphosphonium intermediate in the catalytic system. 2-Methyl-6-nitrobenzoic acid has a sufficient acidity but is sterically hindered. Therefore, conversion of the alkoxyphosphonium intermediate into the corresponding acyloxyphosphonium intermediate would be an unfavourable process due to a steric factor of the carboxylic acid. 26 Mechanistic studies of the reaction using NMR spectroscopic methodologies Does the reaction of the ethyl 2-arylazocarboxylates with triphenylphosphine form Morrison-Brunn-Huisgen betaine intermediates 27 like in the typical Mitsunobu reaction? Precedent mechanistic studies indicate the formation of betaine intermediates from azo reagents and phosphines. To obtain insights into the intermediates in the present reaction, we monitored the reactions of ethyl 2-arylazocarboxylates with triphenylphosphine using multinuclear ( 1 H, 13 C, 31 P, 15 N) 1D and 2D NMR spectroscopy. To assist an unambiguous structure elucidation and assignment of NMR parameters, two kinds of 15 N-labeled ethyl 2-phenylazocarboxylates 2d-15 N and 2d-15 N 0 , 15 N-labeled potent Mitsunobu reagents 2a-15 N, 2j-15 N and doubly 15 N-labeled DEAD (di-15 N-DEAD) were prepared and used in the study, along with some unlabeled analogues (Fig. 7).
The addition of triphenylphosphine (10 equiv.) into the solution of azo compounds in CDCl 3 resulted in the appearance of low-eld resonances in the 31 P NMR spectra (2d: +33.9 ppm, 2a: +34.5 ppm, 2j: +35.4 ppm) that are supportive of the formation of betaine intermediates 48. In light of the electron density of a nitrogen atom, these chemical shis are roughly consistent with that of di-15 N-DEAD (+44.2 ppm) and DEAD (+44.8 ppm). 27 Although it is predicted that Michael-type addition of triphenylphosphine to ethyl 2-arylazocarboxylates (an attack to N2) takes place to form betaines, 28 the formation of other intermediary structures should be considered. Unlike for the symmetric DEAD, 29 the issue of the regiochemistry of the triphenylphosphine attack to ethyl 2-arylazocarboxylates is raised as a consequence of their non-symmetric nature and the potential electrophilicity of the azo benzene derivatives toward triphenylphosphine. 30, 31 15 N NMR spectroscopy was sought as a probe for the in situ investigation of the regiochemistry. The formation of adducts formed between the triphenylphosphine and azo reagents was monitored using 1 H, 13   To further support the structure of the intermediate we carried out a trapping experiment in which betaine 48a, formed in situ from 2a-15 N and triphenylphosphine in CDCl 3 , was treated in an NMR tube with iodomethane. 15 N NMR chemical shis of the starting compounds and products are shown in Scheme 1. The reaction of 2a-15 N with triphenylphosphine followed by treatment with iodomethane readily afforded a methylated product holding a phosphine, as conrmed using 1 H-31 P HMBC. A correlation between the N-CH 3 proton resonance with that of the C]O carbonyl in the 1 H-13 C HMBC spectrum, along with the absence of N-CH 3 correlations with aromatic carbons, strongly suggested the formation of 15 N-methylated phosphonium salt 49. An upeld 15 N NMR shi from 182 ppm (in 48a-15 N) to 109 ppm upon methylation additionally supports the structure of 49. By repeating the trapping experiment with 15 Nunlabeled 2a in a preparative way, the corresponding phosphonium salt decomposed during chromatographic purication on silica gel into ethyl 2-(3,4-dichlorophenyl)-1-methylhydrazine-1carboxylate (see the ESI †). Although the intermediates generated from the dialkyl azodicarboxylates and triphenylphosphine are generally presented in a form of a resonance structure with a negatively charged nitrogen atom and a C]O double bond, our NMR data suggest that the alternative with the sp 2 hybridized nitrogen atom more accurately represents the true structure of the betaine (Fig. 7). This is also in agreement with oxygen being more electronegative than nitrogen.
Overall, the NMR experimental results support the formation of P-N betaines such as 48 in the Mitsunobu reaction using our reagents and indicate that other structures such as regioisomer 51 and P-O betaine 53 are unlikely. Formation of O,N-phosphorane 52 could not be ruled out but was not detected in our NMR analysis.
This suggests that an equilibrium toward betaine 48a from 2a is unfavorable but the reactivity of 48a toward an alcohol is sufficiently high in toluene. Thus, the fate of the betaine generated from the ethyl 2-arylazocarboxylates and triphenylphosphine appears to be very similar to that from the typical Mitsunobu reaction using DEAD.

Thermal stability of the developed Mitsunobu reagents
When typical azo reagents such as DEAD are used, sufficient care is oen required from the viewpoint of their thermal instability. Ethyl 2-(3,4-dichlorophenyl)hydrazinecarboxylate (1a), ethyl 2-(4-cyanophenyl)hydrazinecarboxylate (1j) and their azo forms 2a and 2j are stable crystalline solids under ambient conditions, and no decomposition of these compounds was observed aer two months. Incidentally, when di(2methoxyethyl) azodicarboxylate (DMEAD), that is a crystalline solid, was exposed to ambient conditions for two months, a partial but clear decomposition was observed using 1 H NMR analysis. It is known, from differential scanning calorimetry (DSC), that DEAD, diisopropyl azodicarboxylate (DIAD) and di(2-methoxyethyl) azodicarboxylate (DMEAD) show a large exothermic peak at 210-250 C, indicating exponential decomposition of these compounds. 33 We investigated the thermal properties of ethyl 2-(3,4dichlorophenyl)azocarboxylate (2a) and ethyl 2-(4-cyanophenyl) azocarboxylate (2j) using thermogravimetry-differential thermal analysis (TG-DTA). Interestingly, it indicated the absence of exothermic peaks, whereas endothermic peaks were observed at 191.3 C (3,4-dichlorophenyl derivative 2a, mp: 52.1 C) and 225.7 C (4-cyanophenyl derivative 2j, mp: 55.4 C) with a loss of weight of the samples. These peaks likely show boiling points of the azo compounds that are accompanied by some evaporation. A possibility of endothermic decomposition is unlikely because decomposition of azo compounds is generally exothermic. To eliminate the possibility of the endothermic decomposition, we representatively tested by heating 2a in the solution-phase. A solution of 2a in benzene-d 6 was kept for 10 min at 200 C in an autoclave and then analyzed using 1 H NMR spectroscopy, which indicated no decomposition (see the ESI †). Similarly, TG-DTA of ethyl 2-(3,4-dichlorophenyl)hydrazinecarboxylate (1a, mp: 114.0 C) and ethyl 2-(4-cyanophenyl)hydrazinecarboxylate (1j, mp: 138.1 C) showed endothermic peaks with a loss of weight of the samples at 250.3 C and 267.4 C, though partial decomposition seems to occur around this temperature in the case of 1j. Thus, we did not observe clear exponential decomposition of our azo and hydrazine compounds under the ambient pressure unlike in typical Mitsunobu reagents, though we did not test the thermal stability of these compounds at higher temperatures in a pressured vessel. 34 Overall, the experimental results support that our Mitsunobu catalysts can be safely stored and used without special precautions.

Conclusions
Ethyl 2-arylazocarboxylates can operate in the Mitsunobu reaction like typical Mitsunobu reagents such as diethyl azodicarboxylate (DEAD). The former, however, are recyclable using aerobic re-oxidation of the resultant ethyl 2-arylhydrazinecarboxylate with cheap and nontoxic iron phthalocyanine. This outstanding ability enables catalytic Mitsunobu reactions by using these reagents as organocatalysts. Our systematic study reveals that Mitsunobu activity of azo forms of these catalysts is compatible with an oxidation process of hydrazine forms. Two effective catalysts have been identied. Ethyl 2-(3,4-dichlorophenyl)hydrazinecarboxylate (1a) is suitable for catalytic Mitsunobu reactions with carboxylic acids, working best for the inversion of stereochemistry of secondary alcohols. Ethyl 2-(4-cyanophenyl)hydrazinecarboxylate (1j) provides excellent results in reactions with nucleophiles other than carboxylic acids, serving for the transformation of the hydroxyl groups of alcohols to other functional groups. Thus, the catalytic Mitsunobu reaction has been complemented by two potent reagents and strict optimization of the reaction conditions. The present catalytic protocol is comparable to the original Mitsunobu reaction in both, reactivity and scope. It is also noteworthy that these reagents are stable solids, and their thermal behavior is different from the typical Mitsunobu reagents. Our study has illustrated that serious limitations of the Mitsunobu reaction are avoidable using new reagents and improved procedures. We expect that the improved method will promote the use of the Mitsunobu reaction in practical synthesis.