The pyrolysis of isoxazole revisited: A new primary product and the pivotal role of the vinylnitrene. A low-temperature matrix isolation and computational study

Article abstract of DOI:10.1021/ja207717k

This paper describes the pyrolysis of parent isoxazole and of its 5-methyl and 3,5-dimethyl derivatives by the high-pressure pulsed pyrolysis method, where activation of the precursor molecules occurs predominantly by collisions with the host gas (Ar in our case), rather than with the walls of the pyrolysis tube, where catalyzed processes may occur. The products were trapped at 15 K in Ar matrices and were characterized by vibrational spectroscopy. Thereby, hitherto unobserved primary products of pyrolysis of isoxazole and of its 5-methyl derivative, 3-hydroxypropenenitrile or 3-hydroxybutenenitrile, respectively, were observed. E-Z photoisomerization could be induced in the above hydroxynitriles. On pyrolysis of isoxazole, ketenimine and CO were observed as decomposition products, but this process did not occur when the 5-methyl derivative was pyrolyzed. Instead, the corresponding ketonitrile was formed. In the case of 3,5-dimethylisoxazole, 2-acetyl-3-methyl-2H-azirine was detected at moderate pyrolysis temperatures, whereas at higher temperatures, 2,5-dimethyloxazole was the only observed rearrangement product (next to products of dissociation). These findings are rationalized on the basis of quantum chemical calculations. Thereby it becomes evident that carbonyl-vinylnitrenes play a pivotal role in the observed rearrangements, a role that had not been recognized in previous theoretical studies because it had been assumed that vinylnitrenes are closed-shell singlet species, whereas they are in fact open-shell singlet biradicaloids. Thus, the primary processes had to be modeled by the multiconfigurational CASSCF method, followed by single-point MR-CISD calculations. The picture that emerges from these calculations is in excellent accord with the experimental findings; that is, they explain why some possible products are observed while others are not.

Full text of DOI:10.1021/ja207717k

ARTICLE
pubs.acs.org/JACS
The Pyrolysis of Isoxazole Revisited: A New Primary Product and the
Pivotal Role of the Vinylnitrene. A Low-Temperature Matrix Isolation
and Computational Study
†
,†
†
†
Clꢀaudio M. Nunes, Igor Reva,* Teresa M. V. D. Pinho e Melo, Rui Fausto, Tomꢀaꢁs Solomek,§,^ and
ꢁ
Thomas Bally*,§
†Department of Chemistry, University of Coimbra, P-3004-535 Coimbra, Portugal
§Department of Chemistry, University of Fribourg, CH-1700 Fribourg, Switzerland
S Supporting Information
b
ABSTRACT: This paper describes the pyrolysis of parent isoxazole and of
its 5-methyl and 3,5-dimethyl derivatives by the high-pressure pulsed
pyrolysis method, where activation of the precursor molecules occurs
predominantly by collisions with the host gas (Ar in our case), rather than
with the walls of the pyrolysis tube, where catalyzed processes may occur.
The products were trapped at 15 K in Ar matrices and were characterized
by vibrational spectroscopy. Thereby, hitherto unobserved primary pro-
ducts of pyrolysis of isoxazole and of its 5-methyl derivative, 3-hydro-
xypropenenitrile or 3-hydroxybutenenitrile, respectively, were observed.
EꢀZ photoisomerization could be induced in the above hydroxynitriles.
On pyrolysis of isoxazole, ketenimine and CO were observed as decom-
position products, but this process did not occur when the 5-methyl
derivative was pyrolyzed. Instead, the corresponding ketonitrile was
formed. In the case of 3,5-dimethylisoxazole, 2-acetyl-3-methyl-2H-azirine was detected at moderate pyrolysis temperatures,
whereas at higher temperatures, 2,5-dimethyloxazole was the only observed rearrangement product (next to products of
dissociation). These findings are rationalized on the basis of quantum chemical calculations. Thereby it becomes evident that
carbonyl-vinylnitrenes play a pivotal role in the observed rearrangements, a role that had not been recognized in previous theoretical
studies because it had been assumed that vinylnitrenes are closed-shell singlet species, whereas they are in fact open-shell singlet
biradicaloids. Thus, the primary processes had to be modeled by the multiconfigurational CASSCF method, followed by single-point
MR-CISD calculations. The picture that emerges from these calculations is in excellent accord with the experimental findings; that is,
they explain why some possible products are observed while others are not.
general mechanism shown in Scheme 1 was proposed.14,15 First,
1. INTRODUCTION
isoxazoles 1 isomerize to 2H-azirines 3 through NꢀO bond
Isoxazoles (compounds 1 in Scheme 1) are heterocyclic
compounds that have a number of applications in areas as diverse
as pharmaceuticals, agrochemistry, molecular electronics, corro-
cleavage (involving the putative vinylnitrene intermediate 2 or
via a concerted [1,3]-sigmatropic shift). The 2H-azirines 3 could
sion inhibition, and liquid crystals.1,2 They are found in natural
sources; well-known examples are ibotenic acid and muscimol,
powerful psychoactive and neurotoxic compounds isolated from
fungi.2 Much of the chemistry of isoxazoles is initiated by the
cleavage of its weak NꢀO bond, which has been shown, for
example, to occur during the biotransformation of several
isoxazole-containing drugs.3ꢀ5 Isoxazoles can also be function-
alized, which make them important building blocks in organic
synthesis.1,2,6
subsequently undergo CꢀC bond scission to yield nitrile ylides 4,
followed by cyclization to the corresponding oxazoles 5. If R3 =
H, the CꢀC scission may be accompanied by a hydrogen shift to
yield isocyanides 7. [1,2]-H-shift in 1 or 2 may lead to keteni-
mines 6. Both 6 and 7 can in turn yield nitriles 8.
The final products of the gas-phase thermolysis of isoxazoles 1
depend on the nature of the substituents R3 and R5. The reported
results on the FVP of 5-methylisoxazole,9 5-amino-4-methylisoxazole,9
and 4-acetyl-5-methylisoxazole11 indicate that when R3 = H the
preferential pathway is H migration and the formation of nitriles
8 whereas if R3 = Me or R3 = NH2 oxazoles are formed instead.14
The thermolysis of isoxazoles has been the subject of several
studies.1,2,6 Under conditions of flash vacuum pyrolysis (FVP),
most isoxazoles 1 were found to isomerize into the correspond-
ing 2H-azirines 3, oxazoles 5, or 3-oxopropanenitriles 8.7ꢀ14
Based on these results and on kinetic and theoretical studies, the
Received: August 15, 2011
Published: September 28, 2011
r
2011 American Chemical Society
18911
dx.doi.org/10.1021/ja207717k J. Am. Chem. Soc. 2011, 133, 18911–18923
|

Journal of the American Chemical Society
ARTICLE
Scheme 1. Proposed Mechanism of Isoxazole Pyrolysis14,15
Scheme 2. Pulsed Pyrolysis of Isoxazoles 1a and 1b Gener-
ates (Z)-3-Hydroxypropenenitriles 9a and 9b That Photo-
isomerize to Their E Counterparts
Concurrently, we engaged in quantum chemical calculations
to simulate and assign the observed IR spectra and to model the
potential surfaces on which the observed rearrangements take
place. We paid particular attention to the possible involvement
of the elusive yet pivotal vinylnitrene 2, which is known to have
a triplet ground state and whose lowest singlet state is of open-shell
nature.20ꢀ22 These features, which have not been considered in pre-
vious theoretical studies of isoxazole rearrangements, require
that multiconfigurational methods be used in modeling the
potential surfaces. We highlight the pivotal role of the vinylni-
trene intermediate 2, which resides on a very flat part of the
potential surface but by which the reactions leading to many of
the identified products, including the newly found hydroxy-
propenenitriles 9, proceed. The reason this has been ignored in
previous theoretical studies is that the vinylnitrene was invari-
ably assumed to be in its (excited) closed-shell singlet state,
instead of allowing it to relax to its lower-lying open-shell
singlet state.
On the other hand, only decomposition products were
formed in shock tube thermal reactor experiments of parent
isoxazole 1a and 5-methylisoxazole 1b.16,17 Acetonitrile and
carbon dioxide, followed by hydrogen cyanide and ketene,
were found in the former case, while propionitrile and carbon
monoxide, followed by ethane, methane, acetonitrile, and
hydrogen cyanide, were identified in the latter. Based on these
findings, it was proposed that the decay mechanism of
isoxazole 1a and 5-methylisoxazole 1b involved only uni-
molecular decomposition processes.16ꢀ18 However, subsequent
theoretical studies suggested that 1a and 1b should isomerize
before decomposition,15,19 in line with the above-mentioned FVP
results.
2. RESULTS AND DISCUSSION
2.1. IR Spectra of Matrix-Isolated Isoxazole and 5-Methyl-
isoxazole. Figure 1 shows the experimental and the calculated IR
spectra obtained after room-temperature pulsed deposition of 1a
and 1b. Comparison of the IR spectrum of 1a in an argon matrix
with that obtained in gas phase23ꢀ25 shows excellent agreement
(the maximum deviation in the region between 500 and
1700 cmꢀ1 is 3 cmꢀ1; see Table S1, Supporting Information,
which shows also the assignment of all observed bands to
calculated normal modes). This indicates the absence of sig-
nificant interaction between 1a and the argon host gas and
suggests that 1a is isolated as a monomer in the matrix. The
similar high quality of the spectra of 1b and 1c suggests that the
same applies for these derivatives.
The most recent ab initio calculations15 predict that the
thermolysis of 1 should involve the initial formation of the
corresponding 2H-azirines 3 in a concerted process, bypassing
the vinylnitrenes 2, which were found not to be potential energy
minima. These calculations predicted also that the rearrange-
ment of azirines 3 to isonitriles 7, which isomerize easily to
nitriles 8, has a lower activation energy than their isomerization
to oxazoles 5, either by a concerted mechanism or by a stepwise
process involving the nitrile ylides 4 as metastable intermediates.
Despite the above-described attempts to elucidate the me-
chanism of the thermolysis of isoxazoles by experimental and
theoretical means, it is still far from being well understood. In the
present work, we report the results of a study of the thermolysis
of isoxazole 1a, 5-methylisoxazole 1b, and 3,5-dimethylisoxazole
1c in an argon atmosphere using the technique of pulsed
pyrolysis and isolation of the products in a matrix at 10ꢀ20 K,
conditions under which bimolecular reactions are largely inhib-
ited, and where the products can be characterized by well-
resolved IR spectra. Thus, we were able to show that 1a and
1b engage in a hitherto unknown rearrangement leading to (Z)-
3-hydroxypropenenitriles 9, which could subsequently be photo-
isomerized by UV-irradiation to their E isomers (Scheme 2). In
contrast, 1c formed the previously identified products, azirine 3c
and oxazole 5c.8
2.2. Pulsed Pyrolysis Experiments. 2.2.1. Isoxazole 1a. On
pulsed pyrolysis of a gas phase mixture of 1a and Ar (∼1:1000),
new IR bands began to appear when the SiC tube was heated to
ca. 600 ꢀC (for details see section 3.1). As the temperature
increased to 800 ꢀC, the bands of 1a diminish in intensity, while
those of products increase. We were able to identify the presence
of carbon monoxide due to its characteristic band at 2138 cmꢀ1 26
;
two other dominant products, labeled temporarily as A and B,
were apparently formed (Figure 2).
The most intense of the product bands appears at 2037/
2039 cmꢀ1. Because it is known that ketenimines R2CdCdNR
generally absorb strongly in the 2000ꢀ2100 cmꢀ1 region,27ꢀ29
we attribute this band, which is split into two components by a
site effect, to an NdCdC antisymmetric stretching mode.30 The
formylketenimine 6a (see Scheme 1) was apparently not formed,
because no band due to the CdO stretching mode of its
aldehyde moiety was observed.31 However, all bands labeled A
in Figure 2 are in perfect accord with the reported infrared
18912
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Figure 2. Experimental IR spectra of products of pulse pyrolysis of
isoxazole 1a trapped in argon matrices at 15 K. The temperature of the
SiC tube was at (a) 600, (b) 700, and (c) 800 ꢀC. The new major bands
in spectrum (c) are assigned to two species labeled A and B. The band
marked with an asterisk is due to an impurity from the pyrolyzer system.
Scheme 3. Conformers of 3-Hydroxypropenenitrile 9aa
Figure 1. (a) Experimental FTIR spectra of isoxazoles 1a (a) and 1b (b)
isolated in argon matrices at 15 K and the respective IR spectra
calculated by the B3LYP method (details, see section 3.2).
a Values in parentheses represent the relative B3LYP/6-311++G(d,p)
calculated energies in kcal/mol (including ZPE corrections).
spectrum32ꢀ34 of the parent ketenimine (H2CdCdNH, 10;
Table 1). The presence of CO as a byproduct in the formation of
10 in the pyrolysis of 1a is expected and is readily established
(vide supra).
The IR spectrum of the other main product, B, in Figure 2
differs from those measured or calculated for any of the expected
and previously observed products of the thermolysis of 1a, that is,
2H-azirine-2-carbaldehyde 3a, oxazole 5a, 2-isocyanoacetalde-
hyde 7a, or 3-oxopropanenitrile 8a (See Scheme 1). There are
two very informative bands in the IR spectrum of B: one at
2222 cmꢀ1, in the region of CtN stretching vibrations of
nitriles,35 and another at 3544/3539 cmꢀ1, that is, in a region
of OH stretching vibrations.
These findings led us to investigate the possible formation of
3-hydroxypropenenitrile 9a, a molecule that has hitherto never
been considered or observed in thermolyses of 1a. Compound 9a
could conceivably be formed by enolization of 8a, by a [1,4]-H-
shift from 2a or 3a, or by a [1,5]-H-shift from 6a. Compound 9a
exists in the four conformations shown in Scheme 3, which were
all found to represent potential energy minima on the B3LYP/
6-311++G(d,p) potential surface.
We found that the calculated infrared spectrum of the most
stable syn-(Z) conformer of 9a is the only one that compares well
with that of product B. Figure 3 shows the excellent agreement
between experiment and calculation for a 1:2 mixture of syn-(Z)-
9a and ketenimine 10 + CO.
When the matrix obtained on 800 ꢀC pyrolysis of isoxazole 1a
was subjected to 240 nm laser irradiation (until a photostationary
state was reached), new IR peaks, shown in Figure 4a, appeared
with concomitant intensity decrease of those of syn-(Z)-9a (note
that the peaks of 1a, for example, at 1430 cmꢀ1, did not change
during that photolysis). The calculated spectra of the four con-
formers of 9a revealed that syn-(Z)-9a was photoisomerized into
a mixture of syn-(E)-9a and anti-(E)-9a, as evidenced in Figure 4b.
Detailed results from the calculations and IR assignments are
presented in Tables S3 and S4 in the Supporting Information. The
observed shifts of ν(OH) to higher frequencies in the E-conformers
of 9a reflect the disruption of the intramolecular interaction
involving the (O)ꢀH and the in-plane CtN π-bond, which is
present only in the most stable syn-(Z) conformer (the OH bond
length in that conformer is also the longest of all four). The
absence of this hydrogen bonding interaction may also be the
reason for the lower stability of the other conformers.
Our results indicate that syn-(Z)-9a is, within the detection
limits, the only conformer obtained in pulsed pyrolysis of 1a,
suggesting that 9a is formed in a stereospecific fashion. Examples
of thermal processes that yield a specific conformer have been
reported for other systems.36ꢀ40 It is difficult to assess the
temperature of samples after they exit the pulsed pyrolysis tube,
except that it must be lower than the temperature at which the
18913
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Table 1. Comparison of Our Experimental with Previously Reported IR Spectra of Ketenimine 10
experimental spectruma this work
experimental spectrab previously reported
sym
assignmentc
3298 (w)
2039/2037 (vs)
1124 (w)
999 (s)
A0
A0
A0
A0
A00
A0
ν(NH)
ν(NdCdC)as
2040 (vs)
2033.9 (100)
1124 (w)
1000 (s)
872 (m)
690 (m)
ν(NdCdC)s + δCH2(scis)
δ(CNH)
992.7 (30)
882.8 (24)
689.3 (13)
871 (m)
τ(NH)
689 (m)
γ(H2CC)
a Argon matrix at 15 K. Band positions in cmꢀ1; experimental intensities in parentheses are presented in qualitative terms: vs = very strong, s = strong,
m = medium, w = weak. b Refs 32ꢀ34. c Symbol: ν, bond stretching; δ, bending; γ, out-of-plane bending; τ, torsion; as, antisymmetric; s, symmetric; scis,
scissoring.
Figure 3. (a) IR spectrum obtained on pyrolysis of isoxazole 1a at 800 ꢀC and the products isolated in an argon matrix at 15 K, where the bands of the
remaining precursor were subtracted. (b) IR spectrum of syn-(Z)-9a (b), and 10 (0) (ratio 0.5:1), along with CO, simulated on the basis of B3LYP
calculations (details see section 3.2). The band marked with asterisk is due to an impurity from the pyrolyzer system.
Figure 4. (a) Changes in the experimental IR spectrum of 1a subjected to pyrolysis at 800 ꢀC, with the products isolated in an argon matrix at 15 K,
which occur upon 15 min of irradiation of this matrix with UV (λ = 240 nm) light (growing bands point upward), monitoring the photoinduced
conformational isomerization of syn-(Z)-9a (negative bands) into syn-(E)-9a and anti-(E)-9a (positive bands) in matrix-isolated 9a at 15 K.
(b) Simulated difference IR spectrum obtained as [20% anti-(E)-9a (4) + 80% syn-(E)-9a (2)] minus [syn-(Z)-9a (b), based on B3LYP calculations
(details, see section 3.2).
tube is held (because thermal equilibrium is not established in a
pulsed flow system). According to thermochemical calculations,41
the other three conformers should be within 2ꢀ3 kcal/mol of
syn-(Z)-9a on a free energy scale and should therefore be present
in proportions of 6ꢀ9% at 300 ꢀC up to 14ꢀ18% at 700 ꢀC,
which should be detectable. If enough energy is available to
cleave the NꢀO bond in 1a, this should also suffice to induce
rotation around the double bond in 9a,42 let alone the CꢀO
bond, so the other conformers should be kinetically accessible.
Nevertheless, we found other reaction channels with lower
energy than that required to rotate the double bond in syn-(Z)-
9a (see the section on mechanistic molecular modeling).
18914
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Figure 5. (a) Experimental IR spectra of products of pulse pyrolysis (700 ꢀC) of 5-methylisoxazole 1b trapped in an argon matrix at 15 K; (b) simulated
IR spectrum of syn-(Z)-9b (b) and 1b (0) (assuming the ratio of 1:9) on the basis of B3LYP calculations (details see section 3.2).
Figure 6. (a) Experimental IR spectrum of the products of pulse pyrolysis of 5-methylisoxazole 1b (when the temperature of the SiC tube is kept at 800 ꢀC),
trapped in an argon matrix at15 K. The absorptions dueto the reagent 1b and syn-(Z)-9b were eliminatedbysubtraction of thescaledspectraof nonpyrolyzed
1b and syn-(Z)-9b (see Figure S2 in the Supporting Information). Bands labeled with (2) correspond to syn-(E)-9b; those labeled ACN correspond to
acetonitrile. The peaks labeled with a question mark are tentatively assigned to the CtN stretching vibrations of syn- and anti-8b. (b) Simulated IR spectrum
of syn-8b ([) and anti-8b (]) (ratio 65%/35%, adjusted to experiment), based on B3LYP calculations (details see section 3.2).
Scheme 4. Summary of Observations from the Pulsed
Pyrolysis of Isoxazole 1a
As an example, Figure 5a shows the spectrum obtained on pulsed
pyrolysis of 1b at 700 ꢀC. Most of the bands in that spectrum
belong to nonreacted 1b or can be readily assigned to syn-(Z)-9b,
as shown in Figure 5b. The same photoisomerization as found for
9a can also be observed for the methyl derivative 9b (see Figure
S1 and Table S4, Supporting Information) further confirming
the assignment.
The spectrum in which both the peaks of the precursor 1b and
of 9b are subtracted is depicted in Figure 6a (details of the
procedure are documented in the Supporting Information). The
most conspicuous signal in this resultant spectrum is a double-
humped band at 1750 cmꢀ1, which hints at the presence of two
conformers of a ketone. The ketenimine 6b, the isonitrile 7b, or
the nitrile 8b, which might all be formed upon 1b pyrolysis, all
possess a keto group (cf. Scheme 1).
Ketenimine 6b and isonitrile 7b should have a strong band at
around 2070 and 2170 cmꢀ1, respectively.47 However, these
spectral regions are almost empty in Figure 5a. In contrast,
B3LYP calculations predict that the CtN stretching vibration in
In addition, “conformational cooling” occurs on supersonic
expansion of the jet exiting the 1 mm ID pyrolysis tube, which
might also lead to the depopulation of less stable conformers.43ꢀ46
2.2.2. 5-Methylisoxazole 1b. Similar experiments as those
described for 1a were carried out with 5-methylisoxazole 1b.
18915
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Figure 8. The two singly occupied MOs of formylvinylnitrene 2a in its
lowest singlet state from a CAS(8,7)SCF calculation (light blue is the N,
red the O-atom).
weak NꢀO bond, which leads to the syn-(Z)-form of carbonyl-
vinylnitrenes (structure 2 in Scheme 1). Conceptually these
vinylnitrenes can decay to azirines 3 or, by different H-shifts, to
carbonyl-ketenimines 6, nitriles 8, their enol forms 9, or imino-
ketenes 11, as shown in Scheme 5 for the parent compound. We
engaged in quantum chemical calculations to show why some of
these products are apparently formed, while others are not, and
how the secondary products that were observed in our study
(ketenimine 10 + CO or oxazoles 5) may arise.
In Scheme 5, it is implied that all stable products (3, 6, 8, 9, and
11) originate from different conformations of the vinylnitrene 2.
This must not necessarily be the case, as indicated by the dashed
black lines which imply that, in the case of azirines 3, CꢀN bond
breaking and H-shifts may occur concomitantly, thus bypassing
the vinylnitrenes 2.
Figure 7. (a) Experimental IR spectrum of products of pulse pyrolysis
of 3,5-dimethylisoxazole 1c (when the temperature of the SiC tube is
kept at 600 ꢀC), trapped in an argon matrix at 15 K. The absorptions due
to the reagent were eliminated (by subtraction of a scaled spectrum of
the nonpyrolyzed 1c isolated in an argon matrix, see Figure S3,
Supporting Information, for the original spectra). (b,c) Experimental
IR spectra of 2-acetyl-3-methyl-2H-azirine 3c (b) and 2,5-dimethylox-
azole 5c (c) isolated in argon matrices at 15 K in separate experiments.
In fact, earlier computational studies of the thermolysis of
isoxazoles15,19 have never found any evidence for the involve-
ment of vinylnitrenes 2. In contrast, it was noted that these
species lie very high in energy and do not even represent
stationary points on the potential energy surface. However, these
studies failed to account for the fact that the lowest singlet states
of vinylnitrenes are of open-shell (biradicaloid) nature, as was
shown in the 1990s.20,21 Any calculations that force isoxazole to
remain closed shell on the way to one or the other stable product
will thus necessarily bypass the vinylnitrenes 2, because the
(excited) closed-shell singlet states of these species lie much
higher in energy.
We thus decided to reinvestigate the rearrangements shown in
Scheme 5 using the multiconfigurational CASSCF method, which
allows for a smooth passage between closed- and open-shell
singlet states, provided that the orbitals are averaged for the two
states. Because dynamic correlation effects are known to be
important in determining activation energies, we carried out
single-point MR-CISD calculations at the CASSCF geometries
(for details of these calculations see Methods).
Figure 8 shows the two MOs that are occupied by electrons of
opposite spin in the lowest singlet state of formylvinylnitrene 2a.
2.3.1. The Formation and Decay of Vinylnitrenes 2. In their
seminal paper on the ring expansion of phenylnitrene,21 Karney
and Borden pointed out, in a footnote, that on a CASSCF(4,4)/
6-31G* potential surface, the lowest singlet state of parent
vinylnitrene is a transition state for the interconversion of
enantiotopic azirines. Our first step was thus to check whether
formylvinylnitrenes 2a are also transition states on a correspond-
ing CASSCF surface. It turned out that three out of the four
planar structures are actually (shallow) minima (the anti-(Z)
geometry corresponds to a flat transition state), which decay,
however, in almost (or completely) barrierless processes to one
of the two conformations of azirine 3a. Thus, the different planar
nitrile 8b gives rise to an IR band at 2306 cmꢀ1 that is ∼20 times
less intense than that for the ketone CdO stretch. Indeed some
weak new bands appear between 2200 and 2300 cmꢀ1, in concert
with the CdO stretch at 1750 cmꢀ1, which are not incompatible
with the predictions for 8b (Figure 6). Figure 6b shows a
simulated spectrum of the two conformers of 8b (in a ratio that
is in accord with the relative intensity of the two CdO bands).
Therefore, ketonitrile 8b appears to be the only viable candidate
for the ketone that is formed on pyrolysis of 1b.
No CO and ketenimine (H2CdCdNMe, 10b) are formed in
the experiments with 5-methylisoxazole 1b. Such a process
would require a shift of the methyl group with concomitant
CꢀC bond rupture to release CO. This reaction channel seems
to be inaccessible when a substituent other than hydrogen is
present in the 5-position of isoxazole. The fact that we observe
instead ketonitrile 8b in the pyrolysis of 1b indicates that these
fragments arise from nitrile 8b.
2.2.3. 3,5-Dimethylisoxazole 1c. Figure 7 shows the results of
the pyrolysis of 3,5-dimethylisoxazole 1c at 600 ꢀC, where we can
distinguish, upon subtraction the bands due to the remaining
precursor, the IR bands of two products, namely, azirine 3c (main
bands around 1715 and 1360 cmꢀ1) and oxazole 5c (most of the
remaining bands) by comparison with their authentic matrix IR
spectra, which are shown in traces b and c. At higher temperatures,
the bands of 3c disappear, while those of oxazole 5c continue to
grow. In addition, several new products begin to appear, whose
definitive assignment will be presented in a separate publication.
The experimental findings for the pulsed pyrolysis of 1a are
summarized in Scheme 4. In the following section we will try to
explain these findings by means of quantum chemical calculations.
2.3. Mechanistic Molecular Modeling. Undoubtedly, the
first process in the thermolysis of isoxazoles is cleavage of the
18916
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Scheme 5. Possible Decay Paths for Formylvinylnitrenes 2a
Scheme 6. State Correlation Diagram for 1a, 2a, and 9a in Cs
forms of vinylnitrenes 2a also serve as transition states for (or
fleeting intermediates in) the interconversion of pairs of enan-
tiomeric azirines 3. The different planar forms of 2a can in turn
only interconvert via the azirines, to which the vinylnitrenes
collapse readily on twisting either of the central CꢀC bonds.
In a second step, we investigated the process of formation of 2a
from isoxazole 1a. Thereby we noted that this reaction cannot
take place in an adiabatic fashion in Cs symmetry because the
ground state of isoxazole (as that of all planar closed-shell species
in Scheme 5) is 1A0 whereas the lowest singlet state of 2a has one
electron in a σ- and one in a π-MO (see Figure 8) and has
therefore 1A00 symmetry. Thus, there will be a conical intersection
in Cs symmetry, which must be circumvented by an out-of-plane
distortion on the way from 1a to 2a and on the path from 2a to
other planar products, such as 9a, shown as an example in
Scheme 6.
Symmetrya
Indeed we found that the transition state for the 1a f 2a
reaction is slightly nonplanar, and the wave function at this transi-
tion state is a mixture of closed- and open-shell configurations.
At the MRCI level, the barrier is just under 50 kcal/mol
(cf. Scheme 6), very close to the energy of syn-(Z)-2a, which is
thus barely protected from decaying back to isoxazole 1a (note,
however, that the other conformers of 2a lie slightly lower in
energy).
Next we searched the transition states for the different
hydrogen shift reactions that are indicated with differently
colored arrows in Scheme 5. The lowest-lying of these transition
states is for the stereospecific [1,3]-H-shift leading from the
syn-(E) conformer of 2a (which is the most stable of the four
a Semiquantitative, based on MR-CISD(Q) calculations.
18917
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Scheme 7. MRCI//CASSCF-Energies of Transition States Levels Relative to That of 1aa
a CS and OS stand for “closed shell” and “open shell”, respectively. For structures, see Figure S4 of the Supporting Information.
conformers) to the syn-(Z) conformer of the hydroxypropeneni-
trile 9a, which was very gratifying to note because 9a is the first
product that is seen on pulsed pyrolysis of 1a at the lowest
temperature where decomposition is observed. Interestingly, the
transition state for that reaction, which at the MRCI(Q) level lies
slightly below that for the breaking of the NꢀO bond in 1a,
turned out to be almost planar, and its wave function is solidly
closed shell, that is, the surface crossing must have occurred prior
to the transition state, in the region leading up to it.
The next decay channel that becomes accessible at the MRCI-
(Q)/CASSCF level is that leading from anti-(Z)-2a to iminoke-
tene 11a, but since this product was not observed in the pulsed
pyrolysis experiments, there must be another decay channel
associated with lower barriers leading to the other observed
products, that is, ketenimine + CO. The barriers for the [1,2]-H-
shifts from any of the conformers of 2a to syn- or anti-formyl-
nitrile 8a are slightly higher than that for the decay to 11a (they
are also closed-shell in nature, and possibly they directly connect
azirines 3a to 8a). The [1,2]-H-shifts leading from 2a or 3a to the
(also unobserved) ketenimines 6a have even higher barriers, so
they will not be considered further.
The results of the above-described calculations are summed up
in Scheme 7, which shows the energies of the transition states
corresponding to the different decay channels relative to the
energy of isoxazole 1a (exact numbers and all structures are given
in Table S5 and Figures S4 and S5 of the Supporting In-
formation). We note that the relative energies of transition states
that are of closed- and open-shell nature, respectively, are affected
to surprisingly different extents by the Davidson correction for
quadruple excitations (“Q”) to the MR-CISD results. The origin
18918
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Scheme 8. Pathways for Secondary Rearrangements and Cleavages of Primary Products of the Pyrolysis of Isoxazoles a
a All numbers are energies in kcal/mol for the parent compounds, that is, R3 = R5 = H, relative to isoxazole, 1a. The numbers next to the arrows are
energies of transition states (those in parentheses are relative to the preceding minimum). The colors of the arrows are consistent with those in Scheme 6.
Note that a direct pathway leading from 3 to 5 probably does not exist (discussion, see text). Structures are depicted in Figures S4 and S5 of the
Supporting Information.
of this discrepancy will be analyzed in a separate paper, but we
will base our discussion on the MR-CISD(Q) results (actually, it
turns out that the chemical conclusions are not affected by this
choice).
ketonitrile 8. In both cases, a migration of R5 is required to occur
in concert with the CꢀC cleavage ([1,2] in the case of 6, [1,4] in
the case of 8), which explains why this cleavage is suppressed
when R5 = CH3 (see Scheme S3 in the Supporting Information).
Because instead 8 is observed on pyrolysis of 1b, we focused on
the cleavage of 8 in the case where R5 = H. Indeed we found a
rather low-lying transition state involving H-migration from the
carbonyl group to the N-atom and concurrent CꢀC cleavage. If
the temperature is high enough to induce rearrangement of 1a to
form 8, then there is also enough energy available for the cleavage
of the latter compound to 10 + CO, even if partial thermalization
takes place in the molecular beam.
If 10 is formed from nitrile 8, then the question arises whether
8 could be formed by a lower energy pathway than that which
involves the vinylnitrenes 2. There are two possibilities: (a) by
ketonization of the Z-hydroxypropenenitrile 9 or (b) by rearran-
gement of azirine 3 to the isonitrile 7 (by way of CꢀC cleavage +
migration of R3), which may then isomerize to nitrile 8. Which
pathway is followed will depend on how rapidly the primary
products, 3 and 9, can dissipate their excess energy under the
conditions of an experiment: If thermalization is complete, then
the complicated rearrangement of azirine 3 to isonitrile 7 involves
2.3.3. Rearrangements of the Primary and Formation of
Secondary Products. Some of the products observed in one or
the other type of thermolysis experiment are formed by pathways
that do not involve the carbonylvinylnitrenes, and in this section,
we will discuss these. Because multideterminant methods are not
needed to calculate the structures and transition states here, we
will base our discussion on results from two high-level compo-
site procedures, the CBS-QB3 method of Petersson et al.48 and
the G4 method of Curtiss et al.49 (for details, see Methods).
Scheme 8 sums up the corresponding energies including ZPV
contributions relative to isoxazole or, for transition states, also
relative to the reactants, respectively (in parentheses).
The first question that we will address concerns the formation
of ketenimine 10 + CO, which was observed in the pulsed
pyrolysis of parent isoxazole 1a (but not of its 5-methyl derivative,
1b). These products must arise by cleavage of a compound that
has a carbonyl group and a CꢀCꢀN moiety of which only two
may be formed directly from isoxazole, namely, ketenimine 6 and
18919
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
indeeda much lower barrier (ca. 39 kcal/mol) than ketonization of
9 (ca. 62 kcal/mol). Conversely, if there is no thermalization then
the ketonization of 9 is spontaneous (because the barrier for this
process is lower than that for the formation of 9), whereas the
rearrangement of 3 to 7 involves a barrier that is on the same order
as that for direct formation of 8 from 2.
and ring-opening of the azirines to yield nitrile ylides 4 and
eventually oxazoles 5 takes over.
Finally, we can try to account, on the basis of our calculations,
for the decomposition products that were observed in the shock-
tube experiments,16 where apparently much more energy was
available to drive cleavage reactions. In the case of 1a, the major
products were ketene + HCN, which were hypothesized to arise
by decomposition of the carbene that is formed from 1a by a
[1,2]-H shift. This shift involves, however, a fairly high barrier
(ca. 69 kcal/mol), but on the other hand, the carbene that is
formed decomposes in a practically activationless process. Alter-
natively, H2CdCO + HCN may be formed by way of iminoke-
tene 11a, which undergoes a [1,3] (NꢀC)-H-shift, concomitant
with cleavage of the CꢀC bond. Although, if one starts from
thermalized precursors, the barrier for this process is higher than
that for the formation of the carbene, the second process
becomes competitive if thermalization is incomplete.
The formation of CH3CN + CO may occur by [1,2]-H-shift plus
CꢀC bond cleavage from carbonylnitrile 8. Again the barrier for
this process is high, but if excess energy from the formation of 8 is
available to drive the process, it may be competitive with some of
the rearrangements in Scheme 8 (an alternative path, starting
from ketenimine 6 and involving two successive H-shifts, was
found to have a much higher activation energy). Other products
that were observed in the shock-tube studies, such as CO2,
methane, and ethane, cannot be accounted for in terms of
unimolecular processes.
We also calculated the barriers for the processes leading to the
observed products for the case where R5 = Me, that is, the pyrolysis
of 1b. All the transition states were located on the CAS(8,7)SCF
surface, but it turned out that the barriers for cleavage of the NꢀO
bond in 1b as well as those leading to ketonitrile 8b and its enol
form, hydroxybutenenitrile 9b, were within a kcal/mol of those in
the parent compounds (see Supporting Information). When the
position 3 is also occupied by a methyl group (3,5-dimethylisox-
azole, 1c), all the reaction channels that involve migration of R3
and R5 are blocked, which leaves only azirine 3c and oxazole 5c as
possible rearrangement products. However, new (dissociative)
reaction channels appear to open up, which will, however, not be
discussed in the present paper.
Thus the most advantageous pathway for the formation of 8
depends on the experimental conditions. The fact that if R5 =
CH3, that is, if cleavage of 8b is prohibited, hydroxypropeneni-
trile 9b persists under the conditions of pulsed pyrolysis seems to
indicate that nitrile 8b may be formed (also) in other ways than
by ketonization of 9b. However, additional experiments would
be needed to settle this question. If R3 ¼ H, then 8 should not be
formed at all because it invariably involves a shift of R3. Indeed it
was found that, if R3 ¼ H then carbonylnitriles 8 were not
observed in flash vacuum pyrolysis of isoxazole derivatives.14
A third question involves the mechanism of formation of
oxazole 5, which is the only observed product if both R3 and R5
are not H-atoms. In this case, all rearrangements that involve
migrations of either of these substituents (i.e., formation of
6ꢀ11) are suppressed, so that decay to azirines 3 is the only
viable (nondissociative) deactivation pathway. Indeed, this pro-
duct was observed on flash vacuum pyrolysis of 3,4-dimethyl-5-
aminoisoxazole at moderate temperature, whereas on pyrolysis
of 3,5-dimethyl- and 3-amino-5-methylisoxazole at higher tem-
peratures, 2,5-disubstituted oxazoles were the only observed
products.14 This indicates that a pathway is available for re-
arrangement of carbonylazirines 3 to oxazoles 5.
It has been demonstrated that in photochemical reactions of
(aryl)2H-azirines, this rearrangement proceeds via a nitrile ylide
4,27 and it has been postulated that 4 is also involved in the
thermal process.10,50 Our calculations show that one can indeed
find a transition state for this CꢀC ring-opening reaction, but the
RHF wave function at this transition state turns out to be
unstable. Thus we located the same transition state by the
CASSCF method, and we found indeed that the wave function
has significant open shell character, so the calculation of the
activation barrier for this process requires a multiconfigurational
wave function. Hence we characterized the CꢀC ring-opening
pathway of azirine 3a by CASSCF with an active space that
is consistent for that process (but which is not identical to that
used for the reactions leading to or from the vinylnitrenes, see
Methods section). At the MRCISD(Q) level, the barrier for this
reaction is almost 50 kcal/mol if one starts from thermalized
azirine 3a and over 67 kcal/mol relative to isoxazole 1a. The ring
closure of singlet nitrile ylide 4a to oxazole 5a (both of which are
solidly closed-shell compounds) is very exothermic and has a
very low barrier.51 At the DFT level, one can also find a transition
state (with a stable wave function) for the 3a f 5a rearrangement,
which bypasses the nitrile ylide, but all attempts to locate this
transition state at the CASSCF level failed or led to the open-shell
transition state leading to 4. Thus we conclude that this lower-
lying “concerted” transition state is an artifact of constraining
the system to remain on the closed-shell surface throughout
this reaction.
3. METHODS
3.1. Pulsed Pyrolysis. In order to study the pyrolysis of isoxazoles
1a, 1b, and 1c, we used a pulsed pyrolyzer system, based on the design
conceived by Chen and co-workers,52 developed in the Integrated
Instrument Development Facility (IIDF) of the Cooperative Institute
for Research on Environmental Sciences (CIRES) of the University of
Colorado in Boulder.53
In this instrument, 5 ms pulses of a host gas (in our case Ar at ca.
800 mbar) seeded with ca. 0.1% of the precursor to be pyrolyzed are
passed through a resistively heated 25 mm ꢁ 1 mm ID silicon carbide
(SiC) tube, the temperature of which can be maintained at (5 ꢀC up to
ca. 1200 ꢀC. In contrast to flash vacuum pyrolysis (FVP) where the
activation occurs primarily by contact with the wall of the pyrolysis tube
(or of added material therein), the activation in this experiment occurs
almost exclusively by collision with the hot host gas atoms, which are
entirely unreactive, which largely avoids unwanted side reactions.
The pyrolyzer was flanged onto a cryostat based on an APD
Cryogenics closed-cycle helium refrigeration system with DE-202A
expander. Thus, the output of the SiC tube is directed to the CsI
window that is used as optical substrate for the matrices. During
As a conclusion of the above, our calculations have shown that,
under conditions where the NꢀO bond in isoxazoles 1 can be
cleaved, azirines 3 form spontaneously and equilibrate (enantio-
merize) via vinylnitrenes 2. As the temperature is raised, H-shifts
begin to kick in and eventually siphon off the vinylnitrenes. If R3
or R5 are methyl groups some or all of these channels are blocked
18920
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
deposition of the pyrolysis products, the temperature of the CsI window
was stabilized between 15 and 20 K.
Isoxazole 1a (99% purity) and 3,5-dimethylisoxazole 1c (98% purity)
were purchased from Aldrich. 5-Methylisoxazole 1b (>95% purity) was
purchased from TCI Europe. Authentic samples of 2-acetyl-3-methyl-
2H-azirine 3c and 2,5-dimethyloxazole 5c were prepared according to
to find stationary points (we checked that the geometries do not change
significantly on enlarging the basis set, see Figure S11 and Table S9 in the
Supporting Information). At these geometries, we carried out single-
point calculations at the multireference configuration interaction
(MRCI) level including all single and double excitations (SD) with
the cc-pVTZ basis set (we ascertained that at this level the results are
close to converged with regard to the basis set limit, see Figure S12 in the
Supporting Information). In studies of chemical reactions by these
methods, it is very important to choose an active space that contains all
the MOs and the electrons that change along the considered pathways,
to avoid discontinuities of the potential surfaces and to be able to
compare energies. Eventually we found that an active space containing
eight electrons in seven active orbitals allowed us to fulfill this condition,
provided that it contained the right orbitals (see below). Since the
reactions involve passages between open- and closed-shell singlet states,
we used state-averaged orbitals throughout all calculations.
In the case of isoxazole, those were the five π/π*-MOs containing the
six π-electrons and the NꢀO σ- and σ*-MOs. In the vinylnitrenes, the
NꢀO orbitals were replaced by the singly occupied in-plane p-AO of N
and the in-plane p-lone pair of O, while in the azirine 3 one of the π/π*
pairs turns into the σ- and σ*-MOs of the newly formed CꢀN bond. It
was not possible to find a compatible active space for the other primary
products (6, 8, 9, and 11), but we profited from the fact that the transition
states for formation of these products are very early (because the re-
actions are exothermic by over 55ꢀ70 kcal/mol), so the active space that
we used for the vinylnitrenes 2 was used to describe also those transition
states because we had to use a consistent active space to ensure com-
parability of energies. This active space does not explicitly contain the
σ- and σ*-MOs of the reactive XꢀH bonds because we were not able to
include those MOs in the vinylnitrenes. All these MOs are depicted in
Figures S6ꢀS8 of the Supporting Information.
a literature procedure.8,54 2-Acetyl-3-methyl-2H-azirine 3c, UVmax
=
254 nm (ACN). 1H NMR (400 MHz, CDCl3): δc = 2.05 (3H, s); 2.54
(3H, s); 2.84 (1H, s). 13C NMR (125 MHz, CDCl3): δc = 13.1, 27.3,
36.9, 159.5, 207.2. 2,5-Dimethyloxazole 5c, UVmax = 208 nm (ACN).
1H NMR (400 MHz, CDCl3): δ = 2.26 (3H, s); 2.39 (3H, s); 6.59
(1H, s).
Prior to usage, the samples were degassed by using the standard
freezeꢀpumpꢀthaw method, whereupon the precursor vapor was
premixed with high-purity argon in ratios ranging from 1:500 to
1:2000 in a 3 L Pyrex glass reservoir to a pressure of 800 mbar, using
standard manometric techniques. During the experiments, the flux of the
mixture was controlled by reading the drop pressure in the reservoir with
a capacitance manometer. In a typical experiment, the pulsed valve
operated with an opening time of 5 ms and pulse frequency of 12 Hz
during 1000 to 3000 s.
Photochemical transformations of the matrix-isolated pyrolysis pro-
ducts were induced using the frequency-doubled signal beam provided
by a Quanta-Ray MOPO-SL optical parametric oscillator (fwhm ≈
0.2 cmꢀ1) pumped with a pulsed Nd:YAG laser (repetition rate = 10 Hz,
pulse energy ≈ 5 mJ, duration = 10 ns).
The IR spectra, in the 4000ꢀ400 cmꢀ1 range, were obtained using a
Mattson (Infinity 60AR series) Fourier transform infrared spectrometer,
equipped with a deuterated triglycine sulfate (DTGS) detector and a
Ge/KBr beam splitter with 0.5 cmꢀ1 spectral resolution. To avoid
interference from atmospheric H2O and CO2, a stream of dry air was
continuously purged through the optical path of the spectrometer.
3.2. Quantum Chemical Calculations. For the purpose of
modeling IR spectra, the geometry optimization and the frequency
calculations were done with the B3LYP/6-311++G(d,p) method. After
scaling the calculated frequencies by factors of 0.980 below 3000 cmꢀ1
and 0.950 above 3000 cmꢀ1, the resulting frequencies, together with the
calculated intensities, served to simulate the spectra shown in the figures
by convoluting each peak with a Lorentzian function with a full width at
half-maximum (fwhm) of 2 cmꢀ1, so that the integral band intensities
correspond to the calculated infrared absolute intensity.55 Note that the
peak intensities in the simulated spectra are several times less (in the
arbitrary units of “relative intensity”) than the calculated intensity (in km
molꢀ1).
It turned out that the active space needed to describe the transition
states leading from 3a to both 4a and 7a is not consistent with that
described above. Instead of the σ- and σ*-MOs of the CꢀN bond in 3a,
those of the CꢀC bond in the azirine ring were included. In order to
make all MRCI relative energies comparable, we recomputed also the
geometries and the energies of azirine 3a with this active space. These
orbitals are depicted in Figure S9 and S10 of the Supporting Information).
All transition states (TS) were characterized by intrinsic reaction
coordinate (IRC) calculations, which led to the starting vinylnitrenes on
one side and to the closed-shell products on the other side, sometimes
after optimization at the end of an IRC path. The three TS leading from
the vinylnitrenes 2a to the azirines 3a (and the anti-(Z)-vinylnitrene,
which is itself already a transition state for the interconversion of
enantiomeric azirines) were of predominant open-shell nature, that is,
they occur before the crossing to the closed-shell singlet surface of the
azirine. The wave function of the TS for the NꢀO cleavage of isoxazole
1a was found to consist of similar parts of closed- and open-shell
configurations, that is, that TS lies very close to the conical intersection
of the two crossing surfaces (cf. Scheme 5).
The wave functions for the TSs for the very exothermic decay of the
vinylnitrenes 2a to the stable products 6, 8, 9, and 11 were all
predominantly closed-shell, so they lie after the crossings of the open-
and closed-shell singlet surfaces, which therefore appear to occur quite
close to the vinylnitrenes, despite the fact that the first (predominantly
closed shell) excited state of these vinylnitrenes lies ca. 10ꢀ12 kcal/mol
above the lowest open-shell state. This can be explained by the fact that
the open-shell singlet surface, which leads to excited states of the
products (cf. Scheme 6) is, at least initially, quite flat, whereas the
closed-shell surface that cuts through it and leads to the product ground
states is very steep, because these reactions are very exothermic. The
details of this will be discussed in a forthcoming paper.
In the regions of the potential surface where open-shell singlet states
do not intervene during the reactions that interconvert different species
(see below), we resorted to the CBS-QB348 and the G4 methods49 to get
reliable relative energies. Both of these methods imply geometry
optimizations and frequency calculations by the B3LYP density func-
tional method (with differently polarized 6-31G(d,p)-type basis sets for
the two methods), followed by series of single-point calculations that are
designed to capture the effects of going to very large basis sets, either by
direct extrapolation (CBS-QB3) or in an additive fashion (G4). Even-
tually, both methods aim at estimating the result of a CCSD(T)
calculation with a very large basis set, supplemented by empirical
corrections to account for relativistic or spinꢀorbit coupling effects.
Generally, both methods are thought to be accurate to within 1 kcal/
mol. It is gratifying to note that the results obtained by both methods
rarely diverge by more than that amount.
All calculations described above were carried out with the Gaussian
program package.56
For the processes that involve the formation or decay of open-shell
carbonylvinylnitrenes, we used the multiconfigurational complete active
space self-consistent field (CASSCF) method with the 6-31G* basis set
All CASSCF and MRCI calculations were done with the Molpro
program package.57 All energies are listed in Tables S5ꢀS8 of the
18921
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
Supporting Information, which contains also the Cartesian coordinates
of all stationary points.
QUI-QUI/111879/2009, and PTDC/QUI-QUI/118078/2010,
cofunded by QREN-COMPETE-UE. C. M. Nunes acknowl-
edges FCT for Grant No. SFRH/BD/28844/2006. The work is
also part of project No 200020-132005 of the Swiss National
Science Foundation. T.S. profited from a stipend of the SCIEX
(Scientific Exchange Programme between Switzerland and the
New Member States of the EU) program of the Swiss Confedera-
tion and of the Brno Ph.D. Talent program sponsored by Brno City
Municipality. We are very indebted to Prof. Paul Rablen
(Swarthmore College) who carried out many of the G4 calculations.
4. CONCLUSIONS
We have studied the pyrolysis of parent isoxazole, 1a, and of its
5-methyl and 3,5-dimethyl derivatives, 1b and 1c, respectively, by
the high-pressure pulsed pyrolysis method, where activation of
the precursor molecules occurs by collisions with the host gas (Ar
in our case), rather than with the walls of the pyrolysis tube,
where catalyzed processes may occur. The products were trapped
at 15 K in Ar matrices where they were characterized by
vibrational spectroscopy. Thereby we found a hitherto unob-
served primary product of pyrolysis of isoxazole, the 3-hydro-
xypropenenitrile 9a, or its methyl derivative 9b. EꢀZ photo-
isomerization could be induced in compounds 9.
On pyrolysis of 1a, ketenimine 10 and CO were observed as
decomposition products, but this process did not occur when the
5-methyl derivative, 1b, was pyrolyzed. Instead, ketonitrile 8b
was formed. In the case of 3,5-dimethylisoxazole 1c, acetyl-
azirine 3c was detected at lower temperatures, whereas at higher
temperatures, 2,5-dimethyloxazole 5c was the only observed
rearrangement product (next to products of dissociation, which
will be discussed separately).
We have rationalized the above findings on the basis of
quantum chemical calculation. Thereby it became evident that
carbonyl-vinylnitrenes 2 play a pivotal role in the observed
rearrangements, a role that had not been recognized in previous
theoretical studies because it had been assumed that vinylni-
trenes are closed-shell singlet species, whereas they are in fact
open-shell singlet biradicaloids. Thus, the primary processes had
to be modeled by the multiconfigurational CASSCF method,
followed by single-point MR-CISD calculations. The picture that
emerges from these calculations is in excellent accord with the
experimental findings, that is, they explain why some possible
products are observed while others are not.
’ REFERENCES
(1) Comprehensive Heterocyclic Chemistry; Lang, S. A., Lin Jr., Y.-I.,
Eds.; Pergamon: Oxford, 1984; Vol. 6, Part 4B.
(2) Chemistry of Heterocyclic Compounds; Gr€unanger, P., Vita-Finzi,
P., Eds.; John Wiley and Sons, 1991; Vol. 47, Part 1.
(3) Dalvie, D. K.; Kalgutkar, A. S.; Khojasteh-Bakht, S. C.; Obach,
R. S.; O’Donnell, J. P. Chem. Res. Toxicol. 2002, 15, 269.
(4) Kalgutkar, A. S.; Nguyen, H. T.; Vaz, A. D. N.; Doan, A.; Dalvie,
D. K. Drug Metab. Dispos. 2003, 31, 1240.
(5) Yu, J.; Folmer, J. J.; Hoesch, V.; Doherty, J.; Campbell, J. B.;
Burdette, D. Drug Metab. Dispos. 2011, 39, 302.
(6) Pinho e Melo, T. M. V. D. Curr. Org. Chem. 2005, 9, 925.
(7) Murature, D. A.; Perez, J. D.; Debertorello, M. M.; Bertorello,
H. E. Anal. Asoc. Quim. Argentina 1976, 64, 337.
(8) Perez, J. D.; de Diaz, R. G.; Yranzo, G. I. J. Org. Chem. 1981,
46, 3505.
(9) Perez, J. D.; Yranzo, G. I.; Wunderlin, D. A. J. Org. Chem. 1982,
47, 982.
(10) Perez, J. D. Anal. Asoc. Quim. Argentina 1983, 71, 99.
(11) Perez, J. D.; Wunderlin, D. A. Int. J. Chem. Kinet. 1986, 18, 1333.
(12) Perez, J. D.; Wunderlin, D. A. J. Org. Chem. 1987, 52, 3637.
(13) Wunderlin, D. A.; Davico, G. E.; Perez, J. D. Int. J. Chem. Kinet.
1992, 24, 31.
(14) Yranzo, G. I.; Moyano, E. L. Curr. Org. Chem. 2004, 8, 1071.
(15) Davico, G. E. J. Phys. Org. Chem. 2005, 18, 434.
(16) Lifshitz, A.; Wohlfeiler, D. J. Phys. Chem. 1992, 96, 4505.
(17) Lifshitz, A.; Wohlfeiler, D. J. Phys. Chem. 1992, 96, 7367.
(18) Okada, K.; Saito, K. J. Phys. Chem. 1996, 100, 9365.
(19) Higgins, J.; Zhou, X. F.; Liu, R. F. J. Phys. Chem. A 1997,
101, 7231.
(20) Parasuk, V.; Cramer, C. J. Chem. Phys. Lett. 1996, 260, 7.
(21) Karney, W. L.; Borden, W. T. J. Am. Chem. Soc. 1997, 119, 1378.
(22) Rajam, S.; Murthy, R. S.; Jadhav, A. V.; Li, Q.; Keller, C.; Carra,
C.; Pace, T. C. S.; Bohne, C.; Ault, B. S.; Gudmundsdottir, A. D. J. Org.
Chem. 2011, DOI: 10.1021/jo200877k.
’ ASSOCIATED CONTENT
S
Supporting Information. Tables listing the observed IR
b
peaks and their assignment based on B3LYP calculations, full IR
spectra for the pyrolysis of 1b and 1c, energies and structures of
the stationary points found on the CASSCF and B3LYP (G4)
potential surfaces, orbitals contained in the active space for
compounds subjected to MRCISD//CASSCF calculations,
basis-set dependence of the MRCISD results, Cartesian coordinates
of all stationary points discussed in this study, full references for
the Gaussian56 and Molpro57 program packages. This material
is available free of charge via the Internet at http://pubs.acs.org.
(23) Palmer, M. H.; Larsen, R. W.; Hegelund, F. Mol. Phys. 2004,
102, 1569.
(24) Hegelund, F.; Larsen, R. W.; Nicolaisen, F. M.; Palmer, M. H.
J. Mol. Spectrosc. 2005, 229, 244.
(25) Robertson, E. G. J. Mol. Spectrosc. 2005, 231, 50.
(26) Abe, H.; Takeo, H.; Yamada, K. M. T. Chem. Phys. Lett. 1999,
311, 153.
(27) Inui, H.; Murata, S. J. Am. Chem. Soc. 2005, 127, 2628 and
references cited therein.
(28) Finnerty, J.; Mitschke, U.; Wentrup, C. J. Org. Chem. 2002, 67,
1084.
’ AUTHOR INFORMATION
Corresponding Author
reva@qui.uc.pt; thomas.bally@unifr.ch
(29) Kurtz, D. W.; Shechter, H. Chem. Commun. 1966, 689.
(30) Cumulated double bond stretches often show pronounced site
splitting in cryogenic matrices, which is why this band appears as a pair
separated by 2 cmꢀ1 (2039 and 2037 cmꢀ1).
Note
^On leave from Masaryk University, Brno, Czech Republic.
(31) The product absorption band at 1632/1627 cmꢀ1 is at too low
frequency to be attributed to the aldehyde ν(CdO) mode of formyl-
ketenimine 6a, which is predicted by the B3LYP/6-311++G(d,p) calcula-
tion as a strong band at 1711 and 1702 cmꢀ1 (originating from two
possible conformations).
’ ACKNOWLEDGMENT
These studies were partially funded by the Portuguese
“Fundac-~ao para a Ci^enciae a Tecnologia” (FCT) Projects PTDC/
QUI/71203/2006-No. FCOMP-01-0124-FEDER-007458, PTDC/
18922
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923

Journal of the American Chemical Society
ARTICLE
(32) Guennoun, Z.; Couturier-Tamburelli, I.; Combes, S.; Aycard,
J. P.; Pietri, N. J. Phys. Chem. A 2005, 109, 11733.
(33) Jacox, M. E. Chem. Phys. 1979, 43, 157.
(34) Jacox, M. E.; Milligan, D. E. J. Am. Chem. Soc. 1963, 85, 278.
(35) Bernstein, M. P.; Sandford, S. A.; Allamandola, L. J. Astrophys. J.
1997, 476, 932.
(36) Reisenauer, H. P.; Romanski, J.; Mloston, G.; Schreiner, P. R.
Eur. J. Org. Chem. 2006, 4813.
(37) Schreiner, P. R.; Reisenauer, H. P. Angew. Chem., Int. Ed. 2008,
47, 7071.
(38) Schreiner, P. R.; Reisenauer, H. P.; Pickard, F. C.; Simmonett,
A. C.; Allen, W. D.; Matyus, E.; Csaszar, A. G. Nature 2008, 453, 906.
(39) Schreiner, P. R.; Reisenauer, H. P.; Romanski, J.; Mloston, G.
Angew. Chem., Int. Ed. 2009, 48, 8133.
(40) Gerbig, D.; Reisenauer, H. P.; Wu, C. H.; Ley, D.; Allen, W. D.;
Schreiner, P. R. J. Am. Chem. Soc. 2010, 132, 7273.
(41) ΔG calculated at the CBS-QB3 level at temperatures between
300 and 700 ꢀC; for details, see Supporting Information
(42) Assuming a barrier of 65 kcal/mol for the rotation around the
double bond in ethylene, and radical stabilization energies of 8.7 kcal/
mol for an OH and 7.8 kcal/mol for a CN group acting at the
perpendicular transition state, the barrier for rotation around the double
bond in 9a is estimated to be 48.5 kcal/mol, very close to the calculated
barrier for cleavage of the NꢀO bond in 1a (see section on mechanistic
modeling):Menon, A. S.; Wood, G. P. F.; Moran, D.; Radom, L. J. Phys.
Chem. A 2007, 111, 13638.
(43) Felder, P.; G€unthard, H. H. Chem. Phys. Lett. 1979, 66, 283.
(44) Felder, P.; G€unthard, H. H. Chem. Phys. 1982, 71, 9.
(45) Kudoh, S.; Takayanagi, M.; Nakata, M. Chem. Phys. Lett. 1998,
296, 329.
(46) Vidya, V.; Sankaran, K.; Viswanathan, K. S. Chem. Phys. Lett.
1996, 258, 113.
(47) According to B3LYP calculations, the intensities of these bands
(which, after scaling by 0.98, are predicted at 2068 cmꢀ1 for 6b and at
2172 cmꢀ1 for 7b) should be equal to or stronger than that of the CdO
stretching vibration.
(48) Montgomery, J. A.; Frisch, M. J.; Ochterski, J. W.; Petersson,
G. A. J. Chem. Phys. 1999, 110, 2822.
(49) Curtiss, L. A.; Redfern, P. C.; Raghavachari, K. J. Chem. Phys.
2007, 126, No. 084108.
(50) Davico, G. E.; Perez, J. D. J. Phys. Org. Chem. 1990, 3, 611.
(51) The barrier might be higher in aryl derivatives of ylide 4, which
have been intercepted by dienophiles and have been trapped and
observed in Ar matrices.
(52) Kohn, D. W.; Clauberg, H.; Chen, P. Rev. Sci. Instrum. 1992,
63, 4003.
(53) http://cires.colorado.edu/iidf/machine/pyrolyzer.html
(accessed August 15, 2011).
(54) Sato, T.; Yamamoto, K.; Fukui, K.; Saito, K.; Hayakawa, K.;
Yoshiie, S. J. Chem. Soc., Perkin Trans. 1 1976, 783.
(55) Irikura, K. K. Program SYNSPEC. Natl. Inst. Standards and
Technol., Gaithersburg, MD 20899, USA. http://www.ccl.net/cca/soft-
ware/MS-DOS/synthetic-spectrum/SYNSPEC.TXT.shtml (accessed
August 15, 2011).
(56) Frisch, M. J. et al. Gaussian 03 and Gaussian 09; Gaussian, Inc.:
Wallingford, CT, http://www.gaussian.com (full citation see page S34
of the Supporting Information).
(57) Werner, H. J. et al. MOLPRO, version 2010.1, Cardiff, UK.,
http://www.molpro.net (full citation, see page S35 of the Supporting
Information).
18923
dx.doi.org/10.1021/ja207717k |J. Am. Chem. Soc. 2011, 133, 18911–18923