The free radical chemistry of the azoxy group

The azoxy functional group, though relatively uncommon, is found in several natural products and in liquid crystalline compounds. Azoxyalkanes are generally very stable to heat and light and are not subject to attack by radicals. Intramolecular radical reactions, however, can lead to cyclic aminyl nitroxides and hydrazyls, which rearrange further or undergo fragmentation. The azoxy group greatly stabilizes an attached carbon-centered radical but the chemistry of the resulting α - azoxy radicals is not completely understood.


Thermolysis and photolysis of azoxyalkanes
Unlike deazatation (denitrogenation) of azoalkanes, 24 loss of N 2 O from aliphatic azoxy compounds is rarely observed.Thus they are neither free radical initiators nor thermally labile, contrary to published comments. 25For example, the common free radical initiator azoisobutyronitrile (AIBN) decomposes readily at 80 o C but the azoxy analog 1 is stable up to 180 o C. [26][27][28] Thermolysis of azoxycumene 2 is very slow and does not yield N 2 O, leading instead to a Cope amine oxide elimination. 27The same reaction takes place at 190 o C in azoxy-t-butane 3, which affords mostly nitrogen and isobutene, accompanied by a 20% yield of N 2 O and a little isobutane. 29

+
Even a triphenylmethyl group distal to oxygen does not accelerate azoxyalkane thermolysis, as illustrated by 4, which decomposes with ∆G ‡ (290 °C) = 43.0kcal/mol (t 1/2 =~50 min.) 27hen oxygen is placed proximal to the radical stabilizing substituent as in 5, homolysis of the benzhydryl to N bond is greatly facilitated. 27Now both radicals 6 and 7 benefit from resonance stabilization, reducing ∆G ‡ (150 °C) to 33.3 kcal/mol (t 1/2 = 3.2 h).The same effect is seen in 8, which cleaves only the C-N bond proximal to oxygen to generate the resonance stabilized 10. 30 Similarly, thermolysis of 11 affords 13 via the intermediate diazenyloxy biradical The greater thermal stability of azoxyalkanes than azoalkanes carries over to compounds that lose N 2 O or N 2 by concerted pericyclic retrocycloaddition.Thus the activation energy for N 2 O extrusion from 14 32 is at least 23 kcal/mol higher than that for deazatation of 15, which is unstable even at -78 °C. 33Likewise, 16 undergoes no detectable decomposition on heating at 190 °C for 28 h 32 in contrast to the azo analog 17, whose half life is 81 s at 25 °C. 34In the case of a four membered ring, the thermolysis half life of 18 is 14 h at 161 °C 35 while that of the related diazetine 19 at this temperature is 8.8 min, 36 even though 18 possesses radical stabilizing phenyl groups.+ + These large differences in stability are a consequence of thermodynamics.The endothermicities associated with loss of N 2 from azo-t-butane and N 2 O from 3 are 31.9and 68.5 kcal/mol, respectively.Likewise, deazatation of 15 is exothermic by 45 kcal/mol while N 2 O loss from 14 is far less exothermic (∆H =13 kcal/mol.)The more exothermic deazatation is associated with a lower activation energy (E a ≤ 14 kcal/mol for 15) while loss of N 2 O from 14 exhibits E a = 37 kcal/mol. 37rradiation of azoxy compounds in solution does not lead to production of radicals as it does with azoalkanes. 24Although gaseous azoxymethane 20 undergoes photofragmentation, 38 the solution phase photoreactions of azoxy compounds 21 [39][40][41][42] are ring closure to oxadiaziridines (22), [43][44][45][46] positional and geometric isomerization (23, 24), 45,47,48 and deoxygenation (25). 44,450][51] Irradiation of 26 afforded small amounts of isoindene 27 but the major reaction was closure to oxadiaziridine 28. 46,52,53

Reactions of radicals with azoxyalkanes
The fact that nitrones (e.g.29) are commonly used "spin traps" 54,55 might suggest similar behavior for azoxy compounds 21.However, we have been unable to find any published examples of this reaction.Gas phase photolysis of 20 gave no products above m/e = 80, ruling out addition of one CH 3 • to nitrogen of 21 to give 30 (R = R' = R'' = CH 3 ) followed by recombination of a second CH 3 • with 30. 38We irradiated azomethane to completion with neat azoxy-t-butane (3) at ambient temperature to test the possibility of methyl radical attack on 3; however, both NMR and GC analysis showed no disappearance of 3 and no new products. 56

R
Theoretical calculations at the UB3LYP/6-311+G(d) level 57 are consistent with these results.While addition of methyl radicals to the basic nitrone structure is exothermic by 51.4 kcal/mol (eq.1), reaction at nitrogen of an azoxy group is exothermic by only 33.8 kcal/mol (eq.2).The much greater driving force for the nitrone case leads to a lower activation energy for radical trapping (E a (1) = 0.3 kcal/mol, E a (2) = 2.9 kcal/mol), allowing the use of nitrones as spin traps.
Attack of methyl radical on oxygen of the azoxy structure (eq. 3) is also plausible but is considerably less exothermic and has a higher activation energy than aminyl nitroxide formation (eq. 2).

β-Azoxy radicals
In contrast to intermolecular radical attack on the azoxy group, the intramolecular case is well established.Thermolysis of 31 or 32 at 120 °C affords a mixture of azoxyalkanes 3 and 36,

31
presumably via intermediate 35. 58This intermediate nitroxyl radical was not detected by ESR in this experiment nor when generated independently at room temperature by trapping the aziridinyl radical from 37 59,60 with nitroso-t-butane. 61Instead, the observed ESR signals were consistent with structure 38, suggesting that 35 is quite labile even at 25 °C.This behavior stands in contrast to the persistence of carbon analog 39a. 62Part of the reason for the greater lability of 35 is surely the gem dimethyl group 63 but theoretical calculations at the U3BLYP/6-311G(2s,2p) level also indicate that ring opening of 39b (∆E r = 15.1 kcal/mol) is 3.7 kcal/mol more endothermic than opening of 40b. 57If the activation energies for ring opening run parallel to the endothermicities, 40a and 40b are less likely to persist at ambient temperature than are 39a, 39b.
The interconversion rate of β-azoxy radicals 33 and 34 was determined by the radical clock technique to be k f = 1.5 x 10 7 s -1 and k r = 1.5 x 10 3 s -1 at 120 °C.[67] • • • 41 Surprisingly, cyclization of 33 occurred not only at nitrogen but also at oxygen to form proposed intermediate 42, which then fragmented to hydrazonyl radical 43.A similar intermediate was generated at 25 °C by irradiation of azo-azoxy compound 44.The β-azoxy radical 45 cyclized to 46 in competition with its usual reactions with t-butyl radical; 58  In 1999, we reported the results of UQCISD-FC/6-31+G(d) calculations on simplified models A, B, and F of the 33→35 and 33→42 cyclizations, where the lettered structures are the same as in the published paper. 29Aminyl nitroxide F is the analog of 35 while A and B are analogous to 33 and 42, respectively.The computed exothermicity of the A→F process was  57 Since A→F and A→ B are merely intramolecular versions of equations ( 2) and ( 3) above, their exothermicities should resemble those of the acyclic cases, if ring strain is taken into account.Under the crude assumption that the strain energy of F is the same as that of cyclopropane (27.6 kcal/mol) and the strain energy of B equals that of cyclopentane (6.3 kcal/mol) the corrected (strainless) exothermicity of A→F is -10.3 -27.6 = -37.9kcal/mol and that of A→B is -23.6 -6.3 = -29.9kcal/mol.Formation of the aminyl nitroxide is favored by 8.0 kcal/mol, which can be compared to the ∆E difference between equations ( 2) and (3) of 16.8 kcal/mol.Despite the disagreement in these numbers, one may conclude that aminyl nitroxide is more stable than hydrazyl in the acyclic series (eq.2, 3) but ring strain reverses the order in the B, F comparison.This explanation rationalizes the initially surprising formation of 42 and 46 from β-azoxy radicals.
In contrast to the behavior of 31, isomer 47 did not lead to cyclization but instead gave N 2 O in ~80% yield. 58The intermediacy of β-azoxy radical 48 was demonstrated by trapping with 1,4-cyclohexadiene, an experiment that also provided the fragmentation rate k fr of 48 as 4.1 x 10 8 s -1 at 120 °C.Because the calculated enthalpy change for fragmentation of both 33 and 48 to t-Bu•, isobutene, and N 2 O is the same at 4.8 kcal/mol, 58 the very different chemical behavior of these two radicals is not a matter of overall thermodynamics.Instead, radical 49 is better resonance stabilized than the analogous radical 50 from 33, similar to the cases of 4, 5 and 8 presented above.9][70] The 4.8 kcal/mol endothermicity for fragmentation of 48 is far below the 68.5 kcal/mol for thermal decomposition of azoxy-t-butane (3) to 2 t-Bu• + N 2 O, where the starting material is not a radical.On this basis, we can understand why 47 falls into the small group of azoxy compounds that lose N 2 O.
Despite their lack of stabilization, aromatic analogs of 50 are viable intermediates.Treatment of aryl-N'-fluorodiimide-N-oxides or aryl-N'-tosyloxydiimide-N-oxides (51) with Grignard reagents in THF yielded solvent-containing azoxy product 55. 71The proposed mechanism begins with electron transfer to the azoxy moiety followed by loss of X -to form radical 53.This species recombines with the Grignard-derived alkyl radical or one generated by hydrogen abstraction from THF to afford the observed products 54 and 55.An alternate mechanism wherein THF• adds to the azoxy moiety to give 56 suffers from the disadvantage that the azoxy group is not readily subject to intermolecular radical attack.Radicals 53 are also likely intermediates in the reaction of N-bromodiazene oxides 57 with alkenes to form α,β- unsaturated azoxy compounds 58. 72The regiochemistry of the addition, the effect of alkene structure on reactivity, and the accelerating effect of light and added benzoyl peroxide all point to the radical mechanism shown below.Having discussed the chemistry of β-azoxy radicals, we now turn to higher homologs.Thermolysis of 59 gave typical perester products along with a 61% yield of ethylene. 29This fragmentation, whose rate constant was 2 x 10 5 s -1 at 120 °C, might proceed via β-scission of 60 or it could involve prior cyclization to 61, analogous to 33 → 42.Theoretical calculations yielded no low energy transition structure for cyclization of 60, suggesting that it fragments directly to 62, 29 a radical whose chemistry will be discussed below.Since 64 was definitely absent, 60 does not cyclize to the four membered aminyl nitroxide 63.Experimental support for direct fragmentation of 60 was found in studies of pyrazoline oxide 65, which cannot cyclize to 67 without engendering about 22 kcal/mol of ring strain.The fact that 65 gives a 34% yield of ethylene shows that it fragments directly to 66, which implies that 60 follows the same mechanism.Isomeric radical 68 gave no ethylene because β-scission would produce 69, which is far less stabilized than 66 (see Table I and its associated discussion below).t-butyl ether from 71.Although the major product was the carboxylic acid precursor of 70, the presence of 74 was proven unambiguously, suggesting that 71 cyclizes to 72, which then reopens to 73. 73 Six-membered analogs of 72 were shown to be persistent by ESR at room temperature 59 but in our case, the gem dimethyl group, five-membered ring strain, and high temperature would all contribute to rapid opening of 72.

α-Azoxy radicals
Hydrazones can give a variety of products upon oxidation 1,[74][75][76][77][78] but in this review, we will be concerned only with their conversion to α-azoxy radicals 75a, alternately represented as • + hydrazonyloxides 75b.The chemistry of such radicals is less well understood than the topics covered above.Bergmann and coworkers reported in 1923 that oxidation of benzaldehyde Nphenylhydrazone 76 with perbenzoic acid led to the unlikely structure 77. 79After several more misassignments, [80][81][82] the correct structure of the "Bergmann oxide" was finally established as 78. 83,84The activation energy for breaking the central C-C bond of 78 was a remarkably low 14 kcal/mol, indicating that α-azoxy radicals 79 are particularly stable.These radicals were observed by ESR in a hot toluene solution of 78; in fact, the ESR spectra of many hydrazonyloxides have been analyzed in detail. 85,86The elevated temperature caused no permanent change but the signals disappeared immediately on cooling as 79 reverted to 78.Radicals 79 could also be trapped with nitrosobenzene or phenyl-t-butyl nitrone. 85terically hindered hydrazonyloxides such as 80 were much more persistent, as would be expected if their main reaction is indeed C-C dimerization to a structure like 78.MCPBA oxidation of ketone hydrazones 81 yielded the hydrazonyloxide radicals as well, but they were stable for only a few hours in solution and no radical dimers could be isolated.When both R groups of the ketone were alkyl as in 82, no radicals were seen by ESR unless at least one of the groups was especially sterically demanding. 85Irradiation of 81 and 82 with t-Bu-OO-t-Bu, t-BuOOH and air gave the corresponding hydrazonyloxides but in low concentration and accompanied by secondary radical products.
We attempted to oxidize acetone phenylhydrazone 83 with MCPBA in CH 2 Cl 2 at -65 °C but the initially bright yellow solution turned to a brown sludge on warming to 0-10 °C.TLC showed a streak and NMR showed acetone to be the dominant product.In contrast to the successful oxidation of benzaldehyde N-methylhydrazone, 84,87

X + +
In order to evaluate the stabilization of aliphatic α-azoxy radicals, we and Xavier Creary studied the rearrangement rate of p-substituted phenylmethylenecyclopropanes 86. 88Substituents X accelerate the rearrangement, depending on their ability to stabilize biradical 87.Selected values of k rel (80 °C) are shown in Table I.The much smaller rate enhancement of proximal azoxy (k rel = 2.00) compared to distal azoxy (k rel = 9.45) is in excellent accord with ethylene loss from 65 but not 68.Furthermore, theoretical calculations showed that the spin density on the substituent is much greater and that on the benzylic position is much smaller for distal azoxybenzyl radical 89 than for the proximal analog 90.Distal azoxy is among the most effective radical stabilizing groups but is not as strong as nitrone (k rel = 13.5).Obviously, both of these groups benefit from nitroxyl resonance forms such as 75b.The k rel values in Table I correlated nicely with B3LYP calculated energy changes for the isodesmic reaction of substituted benzylic radicals with toluene.
Theoretical calculations also indicate a large stabilization energy for aliphatic α-azoxy radicals.Dissociation of 91 was calculated at the QCISD/6-311++G**//B3LYP/6-31G* level to be 7.4 kcal/mol less endothermic than the analogous process in 1,5-hexadiene 92. 57n view of the stability of 93, it would be interesting to determine the contribution of each set of phenyl groups, terminal and internal, to the lability of 78, whose activation energy is at least 22 kcal/mol lower than that of 93.Compound 94 would be useful because it possesses the gemdimethyl groups of 93 and the terminal phenyl groups of 78.Attempts to oxidize 95 89 with one equivalent of MCPBA gave mainly 96 with lesser amounts of the isomeric mono + + + azoxyazoalkane plus two of the three bisazoxyalkanes, the absent one being 94.Likewise, application of Kovacic's method 28 to 2,3-diamino-2,3-dimethylbutane 97 did not yield 94. Since both methods failed, it is possible that 94 is thermally labile but that unlike 79, the hydrazonyloxide radicals undergo some irreversible reaction like disproportionation.One can estimate the stability of 94 by two approaches, which, however, give much different answers.
The first approach is based on the known, very stable bisazoxy compound 98. 90 With no added scavenger, the thermolysis half-life of 98 at 185 °C is 29 min and the complex product mixture includes aniline and biphenyl.Addition of thiophenol reduces the yield of biphenyl, increases the yield of aniline, and enhances the thermolysis rate by ~25%. 73If the reaction being measured is irreversible C-C bond homolysis, its ∆G ‡ (185 °C) is 34.3 kcal/mol.This value is far above the 14 kcal/mol for 78, implying a very large radical stabilizing role for the internal phenyl groups.Interestingly, this behavior is opposite that of hydrazonyl radicals, where the resonance energy of 99 (18.4 kcal/mol) exceeded that of 100 (15.3 kcal/mol) by only 2.9 kcal/mol. 89In order to estimate of ∆G ‡ of 94, we assume that addition of four methyl groups to 98 should lower its ∆G ‡ by about 5.5 kcal/mol, the known torsional strain energy of 95. 89 Therefore ∆G ‡ for 94 should be 28.8 kcal/mol, making it an easily handled compound.However, it is possible that C-C homolysis of 98 is highly reversible, even in the presence of thiophenol, so that 34.3 kcal/mol is a maximum for ∆G ‡ .In that case, the estimated ∆G ‡ for 94 is also a maximum.
The second estimate of ∆G ‡ for 94 is based on the ∆G ‡ 's of 101 (27.3 kcal/mol) and 95 (32.2 kcal/mol).If we ignore ∆S ‡ and add the 4.9 kcal/mol difference between these figures to the 14 kcal/mol E a of 78, we obtain for 94 E a = 18.9 kcal/mol.It is therefore possible that our failure to prepare 94 is due to thermal lability.Whether this is true or not, there are inconsistencies in the thermal stability of hydrazonyloxide dimers that call for clarification by careful kinetic studies of both the new cases and of known compounds.
While generation of aryl substituted α-azoxy radicals like 79 is not difficult, 84,85 almost nothing is known about their aliphatic counterparts.One reaction that presumably gives such a radical is β-scission of 60 but in this case we were unable to isolate any product attributable to 62.Although we have found no published ESR spectra of purely aliphatic hydrazonyl oxides, such radicals must exist because free radical bromination of azoxyalkanes (e.g.20) gives the expected substitution products. 91 It is conceivable that α-azoxy radicals cyclize and then fragment at elevated temperatures, analogous to 33→42.Although fragmentation of the alkoxy radical 102 would be rapid, 93 the cyclization shown below is calculated to be endothermic by 32 kcal/mol.Attempts to prepare azo and perester precursors that would generate 62 under mild conditions met with failure.Recently however, we synthesized azoxyketones 103a,b which are expected to give α-azoxy radicals by irradiation at room temperature. 94The outcome of these experiments will be reported in the near future.

Conclusions
In stark contrast to azoalkanes, azoxy compounds rarely form radicals on heating or irradiation.Furthermore, they are unreactive to alkyl radical attack unless the reaction is intramolecular.For example, β-carbon centered radicals cyclize to azoxy nitrogen or oxygen and produce short-lived aminyl nitroxides that reopen or hydrazyl radicals that undergo fragmentation.The azoxy group is a powerful stabilizer of an adjacent radical center but the chemistry of α-azoxy radicals (hydrazonyloxides) and their dimers is not fully understood.
then, 46 fragmented to 43 plus acetone.The photolysis of 44 was clean enough to determine the cyclization rate of 45 as k c =1.7 x 10 6 s -1 at 25 o C and to estimate its activation energy as E c ~3 kcal/mol.

2 +
Proceeding to the next homolog, we find that thermolysis of 70 produces δ-azoxy radical 71, which leads to products typical of perester thermolysis, namely, the alkane, alkene, and

92 The facile dissociation of 78 ,
the large stabilizing effect of the distal azoxy group on 87, and the calculations on 91 stand in contrast to the behavior of bisazoxyalkane 93.Although thermolysis of 93 might be expected to cleave the central C-C bond, this compound is nearly as stable as azoxy-t-butane 3 and it gives the same kind of products, presumably by the Cope amine oxide elimination (product yields shown below are moles/mole).The ∆G ‡ for C-C cleavage must exceed that for the elimination reaction, ∆G ‡ (190 o C) = 36.8kcal/mol.

95
61milar treatment of 84 gave a mixture of products, none of which was 85.61