The gas phase 1,2-Wittig rearrangement is an anion reaction. A joint experimental and theoretical study

Article abstract of DOI:10.1039/a805319g

The migratory aptitudes of alkyl groups in the gas phase 1,2-Wittig rearrangement have been determined experimentally as follows. An anion Ph- ^-C(OR¹)(OR²), on collisional activation, competitively rearranges to the two 1,2-Wittig ions PhC(R¹)(OR²)(O^-) and PhC(R²)(OR¹)(O^-)[R¹ and R² = alkyl and R¹ < R²]. These two ions respectively eliminate R²OH and R¹OH. The smaller alkanol is eliminated preferentially, indicating that R² (the larger alkyl group) is migrating preferentially (observed tert-Bu > iso-Pr > Et > Me): a trend generally taken to indicate a radical reaction. However, a Hammett investigation of the relative losses of MeOH from R-C₆H₄-^-C(OMe)₂ shows this loss decreases markedly as R becomes more electron withdrawing, an observation not consistent with a radical reaction. Ab initio calculations [at the CISD/6-311 + + G**//RHF (and UHF)/6-311 + + G** levels of theory] have been used to construct potential surface maps for the model 1,2-Wittig systems ^-CH₂OMe→ EtO^-, and ^-CH₂OEt→PrO^-. Each of these exothermic reactions involves migration of an alkyl anion. There are no discrete intermediates in the reaction pathways. There is no indication of a radical pathway for either rearrangement. It is proposed that the gas phase 1,2-Wittig rearrangement involves an anionic migration, and that it is not the barrier to the early saddle point but the Arrhenius A factor (or the frequency factor of the QET), which controls the rate of the rearrangement. Weak H-bonding between the alkyl anion and the oxygen of the neutral carbonyl species acts as a pivot in holding the molecular complex together during the migration process. This electrostatic interaction increases with an increase in the number of hydrogens able to H-bond to oxygen and with the number of equivalent ways this H-bonding can occur. The relative migratory aptitude of alkyl anions bound within these molecular complexes is tert-Bu^- > iso-Pr^- > Et^- ? Me^-, an order quite different from the migratory aptitudes of anions expected from thermodynamic considerations. This conclusion indicates that great care must be exercised in utilising thermodynamically derived migratory aptitudes to explain the course of a kinetically controlled reaction in the gas phase.

Full text of DOI:10.1039/a805319g

View Article Online / Journal Homepage / Table of Contents for this issue
The gas phase 1,2-Wittig rearrangement is an anion reaction.
A joint experimental and theoretical study
John C. Sheldon, Mark S. Taylor, John H. Bowie,* Suresh Dua, C. S. Brian Chia and
Peter C. H. Eichinger
Department of Chemistry, The University of Adelaide, South Australia, 5005, Australia
Received (in Cambridge) 8th July 1998, Accepted 4th December 1998
The migratory aptitudes of alkyl groups in the gas phase 1,2-Wittig rearrangement have been determined
experimentally as follows. An anion Ph–ϪC(OR1)(OR2), on collisional activation, competitively rearranges to the
two 1,2-Wittig ions PhC(R1)(OR2)(OϪ) and PhC(R2)(OR1)(OϪ) [R1 and R2 = alkyl and R1 < R2]. These two ions
respectively eliminate R2OH and R1OH. The smaller alkanol is eliminated preferentially, indicating that R2 (the
larger alkyl group) is migrating preferentially (observed tert-Bu > iso-Pr > Et > Me): a trend generally taken to
indicate a radical reaction. However, a Hammett investigation of the relative losses of MeOH from R–C6H4–
ϪC(OMe)2 shows this loss decreases markedly as R becomes more electron withdrawing, an observation not
consistent with a radical reaction. Ab initio calculations [at the CISD/6-311ϩϩG**//RHF (and UHF)/6-311ϩϩG**
levels of theory] have been used to construct potential surface maps for the model 1,2-Wittig systems ϪCH2OMe→
EtOϪ, and ϪCH2OEt→PrOϪ. Each of these exothermic reactions involves migration of an alkyl anion. There are no
discrete intermediates in the reaction pathways. There is no indication of a radical pathway for either rearrangement.
It is proposed that the gas phase 1,2-Wittig rearrangement involves an anionic migration, and that it is not the barrier
to the early saddle point but the Arrhenius A factor (or the frequency factor of the QET), which controls the rate
of the rearrangement. Weak H-bonding between the alkyl anion and the oxygen of the neutral carbonyl species acts
as a pivot in holding the molecular complex together during the migration process. This electrostatic interaction
increases with an increase in the number of hydrogens able to H-bond to oxygen and with the number of equivalent
ways this H-bonding can occur. The relative migratory aptitude of alkyl anions bound within these molecular
complexes is tert-BuϪ > iso-PrϪ > EtϪ ӷ MeϪ, an order quite different from the migratory aptitudes of anions
expected from thermodynamic considerations. This conclusion indicates that great care must be exercised in utilising
thermodynamically derived migratory aptitudes to explain the course of a kinetically controlled reaction in the gas
phase.
the gas phase (i.e. in the absence of solvent) for deprotonated
Introduction
benzyl alkyl ethers7 and a range of cognate anions,8 but
The Wittig rearrangement is one of the better known carbanion
whether these reactions involve radical or anionic rearrange-
rearrangements in solution.1 It has been proposed that the 1,2-
ments has not been determined. An ab initio study of a model
Wittig rearrangement, could, in principle, involve either of the
1,2-Wittig system (ϪCH2OMe→EtOϪ) suggests an anionic
intermediates shown in Scheme 1 (R and R1 are generally aryl
process through an early transition state with no discrete inter-
mediates in the reaction pathway: the possibility of the analo-
[R1–(RCHO)]
gous radical reaction was not considered in depth.9 There is
R1
(A)
also debate concerning the mechanism of the analogous 2,3-
CH O–
R-–CHOR1
Wittig rearrangement. For example, (i) it is not known whether
the condensed phase 2,3-Wittig rearrangement of allyl lithio-
methyl ether to form homoallyl alcohol is a stepwise or
concerted process,10 while (ii) the gas phase 2,3-(thio)Wittig
rearrangement of deprotonated allylthiomethyl ether is sug-
gested to be concerted.11
R
[R1•(RCHO)–•]
(B)
Scheme 1
and alkyl respectively),1,2 and to support such mechanistic pro-
posals, it has been shown that aldehydes are often byproducts
of the reaction.2 The migratory aptitudes of RЈ are in the order
of free radical thermodynamic stabilities,3 thus the radical pair
mechanism for the reaction is favoured.4 However the evidence
in favour of a radical reaction is not overwhelming in all cases.
For example, when R1 is chiral, a radical mechanism for the 1,2-
Wittig rearrangement (i) demands racemisation of R1 and this
is not always observed,5 and (ii) should result in the loss of
stereochemical integrity in nuclear magnetic resonance CIDNP
experiments: e.g. as observed for the cognate Stevens rearrange-
ment,6 but not for the 1,2-Wittig rearrangement.5 Indeed, it has
even been suggested that some 1,2-Wittig rearrangements may
have both radical and ionic character.5
This paper reports the results of a joint experimental/
theoretical study planned to determine whether (i) the gas
phase 1,2-Wittig rearrangement is stepwise or concerted, and
(ii) this rearrangement involves migration of an anion or a
radical.
Results and discussion
Experimental studies of Ph؊C(OR1)(OR2) systems
We have used two experimental methods to attempt to deter-
mine whether the gas phase 1,2-Wittig rearrangement involves
anionic or radical migration, viz (i) the determination of the
migratory aptitude of different alkyl groups, and (ii) the way the
rearrangement is affected by changing the electrophilicity of
We have shown that facile 1,2-Wittig rearrangement occurs in
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340
333

View Article Online
Table 1 Spectra of product ions from Ph–ϪC(OMe)2. Comparison
with the spectra of ions formed by independent syntheses
the recipient carbon (of the carbonyl group), i.e. a substituent
effect study.
Daughter ion
(m/z)
Spectrum
type
Spectrum CA m/z (loss) abundance
CR m/z (abundance)
(a) The relative migratory aptitudes of alkyl groups. The
relative migratory aptitude of two migrating groups can be
determined for competitive rearrangements within the same
reactant anion. The obvious system for investigation is a depro-
tonated unsymmetrical benzaldehyde acetal PhϪC(OR1)(OR2)
(C) where R1 and R2 are different migrating groups. The two
possible Wittig product ions (D and E) of this system are shown
in Scheme 2. Assuming the migration is rate determining (and
Ϫ
ؒ
(119)
CA
CR
118(H ) 100, 77(Ph ) 10, 41(PhH) 1
105 (26), 103 (18), 102 (20), 91 (23),
89 (28), 77 (100), 74 (24), 63 (21), 51
(48), 50 (46), 42 (18), 38/39 (9)a
Ϫ
Ϫ
ؒ
PhCOCH2
CA
CR
118(H ) 100, 77(Ph ) 8, 41(PhH) 1
105 (23), 103 (15), 102 (18), 91 (21),
89 (27), 77 (100), 74 (22), 63 (23), 51
(44), 50 (43), 42 (15), 38/39 (6)a
(119)
CAb
CR
134(H ) 100, 133(H2) 25, 107(CO) 51,
ؒ
(PhCOR1-H)– + R2OH
(135)
91(CO2) 10, 77(PhϪ) 9
135 (12), 134 (8), 119 (11), 105 (85),
91 (45), 90 (61), 89 (66), 77 (100), 74
(34), 63 (35), 51 (52), 50 (48), 39 (13)
O–
Ph
[(PhCOR1) R2O– ]
(C6H4)Ϫ–CO2Me
(135)
CA
CR
134(H ) 100, 133(H2) 18, 107(CO) 45,
ؒ
OR2
R1
105(CH2O) <5a, 91(CO2) 7, 77(PhϪ)
12
(D)
135 (10), 134 (6), 119 (10), 105 (75),
91 (40), 90 (55), 89 (60), 77 (100), 74
(30), 63 (32), 51 (46), 50 (42), 39 (8)
_
OR1
OR2
Ph
a Not resolved. b Weak spectrum—peaks of abundance less than 5% are
lost in baseline noise.
(C)
Ph
O–
[(PhCOR2) R1O– ]
R2 OR1
(E)
(PhCOR2-H)– + R1OH
Scheme 2
we will discuss this later), and in the simplest case where
R1 = Me and R2 = Et, the migratory aptitude of MeϪ should be
greater than that of EtϪ, if previous work in the condensed
phase is of relevance here. In the condensed phase, rate and
bracketing experiments indicate that the migratory aptitude of
alkyl anions is MeϪ > EtϪ > iso-PrϪ > tert-BuϪ in several sys-
tems.12 Alkyl anion migratory aptitudes are not known with
certainty for gas phase reactions: thermodynamic migratory
aptitudes should, in principle, follow the ∆GЊacid values of RH13
[i.e. the smaller the value of ∆GЊacid (for the process RH→RϪ ϩ
Hϩ) the higher the migratory aptitude of RϪ], viz tert-BuϪ
MeϪ > EtϪ ~ iso-PrϪ. The trend for radicals is different, i.e.
~
ؒ
ؒ
ؒ
Me < Et < iso-Pr < tert-Bu (the migratory aptitude increases
with a decrease in the homolytic bond dissociation energy of
RH).14 For the particular scenario shown in Scheme 2, when
R1 = Me and R2 = Et, product E will be predominant for a
radical reaction, whereas D will be the major product for an
anion reaction.
Fig. 1 The collision induced tandem mass spectrum (MS/MS) of
[PhCD(OMe)2 Ϫ D]Ϫ. VG ZAB 2HF instrument. For experimental
procedures see Experimental section.
start with the simplest precursor anion, PhϪC(OMe)2. The reac-
tion between HOϪ and PhCD(OMe)2 in the ion source of the
VG ZAB 2HF mass spectrometer yields an (M Ϫ D)Ϫ ion
exclusively. No deprotonation on the phenyl ring is noted. The
spectrum of this ion (Fig. 1) shows competitive losses of CH4
and MeOH. The loss of MeOH is the expected loss from
the 1,2-Wittig product ion whereas the loss of CH4 could, in
principle, come directly from the 1,2-Wittig ion (by a process
analogous to the loss of ROH shown in Scheme 2), or from a
transient intermediate [(PhCO2Me)MeϪ] formed en route to the
1,2-Wittig ion. In order to confirm the operation of the 1,2-
Wittig rearrangement, we need to compare the decompositions
of PhϪC(OMe)2 with those of independently synthesised
PhC(Me)(OMe)(OϪ).
Since C, D and E are isomeric, we need characteristic frag-
mentations of D and E to use as probes in order to determine
their relative formation from reactant C. Products D and E
should fragment by losses of R2OH and R1OH respectively, as
shown in Scheme 2: neither of these fragmentations can occur
from precursor C. In addition, the losses of these alkanols from
D and E will not be involved in the rate determining step (of the
sequences shown in Scheme 2) because D and E are energised
[the process C to D (R1 = R2 = Me) is calculated to be exo-
thermic by 175 kJ molϪ1 (using Benson’s rules15 and estimated
electron affinities16) and if the barrier to the transition state of
the 1,2-Wittig rearrangement is of the order of 100 kJ molϪ1
(cf.9), the Wittig product ion may have an excess energy of
some 275 kJ molϪ1. This energy is more than sufficient to both
effect formation of the alkoxide anion–neutral complex and
subsequent deprotonation within that complex (see Scheme 2)].
Thus the ratios of the losses of R1OH and R2OH will not be
dependent on the relative basicities of R1OϪ and R2OϪ, but will
reflect the relative amounts of D and E formed from C.
We approached the synthesis of PhC(Me)(OMe)(OϪ) by
using an SN2 displacement reaction between PhC(Me)(OMe)2
and a range of nucleophiles (including HOϪ and MeOϪ) in the
ion source of the mass spectrometer. Unfortunately, the ener-
gised ion PhC(Me)(OMe)(OϪ) is not detected in the ion source,
thus we could not compare the MS/MS data of this anion with
Can the above scenario be achieved experimentally? Let us
334
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340

View Article Online
Table 2 Ratios of R1OH/R2OH and R1H/R2H losses from Ph–
Table 3 Ratio of Me and MeOH losses in CA MS/MS data from
ؒ
ϪC(OR1)(OR2) (R1 < R2)a
RC6H4–ϪC(OMe)2
R1
R1
ϪR1OH:ϪR2OH
ϪR1H:ϪR2H
R
σ
Me
MeOH
ؒ
Me
Me
Et
Et
100:8
100:4
100:78
100:60
65:100
45:100
88:100
82:100
p-NMe2
H
p-F
m-F
p-CN
Ϫ0.58
0
0.15
0.34
68
0
100a
50
33
2
b
iso-Pr
iso-Pr
tert-Bu
100
100
100
Et
0
a Each ratio is an average of five separate determinations. The error is
ca. 2%.
a Source formed ion. No parent ion p-Me2NC6H4COCH2 is observed
in the ion source, thus MS/MS data are not available. Loss of Me
does not occur when R = H. In this case, loss of CH4 produces the base
peak.
Ϫ
b
ؒ
(b) A Hammett substituent effect study. A Hammett type
study may assist in differentiating an anionic from a radical
mechanism. Consider the summary in Scheme 3, which
1
2
[(ArCO2Me) Me– ]
O–
Ar
_
OMe
OMe
Ar
Me
OMe
4
[(ArCO2Me)–• Me•]
3
Scheme 3
assumes neither a concerted nor a stepwise process, only that
there may be dissociative (1 or 3) and/or rearrangement (2 or 4)
influences on the overall rate of reaction. How will the elec-
tronic nature (the Hammett σ value) of a substituent on the
phenyl ring affect 1 to 4? Radical reactions generally show small
substituent effects.17 For this radical reaction, electron with-
drawing substituents (> σ) on the phenyl ring will enhance the
rate of dissociation 3 and of migration 4 (electron donating
groups will make the radical anion more reactive). The scenario
concerning the anionic reaction is different. Firstly, substituent
effects for anionic reactions are generally pronounced.17 In the
current case, electron donating substituents (< σ) will aid dis-
sociation 1. The situation with regard to migration 2 is not
straightforward however. Although electron withdrawing sub-
stituents (> σ) will aid the final approach of the anion to the
electrophilic centre, electron donating substituents (< σ) may
facilitate the migration of the anion by the carbonyl moiety
‘holding’ the migrating anion in place during its circuit to the
reactive centre (see later).
Fig.
2
The collision induced mass spectrum (MS/MS) of Ph–
ϪC(OEt)(Oiso-Pr). VG ZAB 2HF instrument.
that of PhϪC(OMe)2. The ion PhC(Me)(OMe)(OϪ), on form-
ation in the ion source, decomposes by competitive losses of
MeOH and CH4 to yield pronounced peaks corresponding to
PhCOCH2 and [(C6H4) ϪCO2Me] respectively [these source
Ϫ
formed ions were identified by their collisional activation (CA)
and charge reversal (CR) mass spectra (see Table 1)]. This result
is not unexpected since the SN2 process [PhC(Me)(OMe)2 ϩ
HOϪ→PhC(Me)(OMe)(OϪ) ϩ MeOH] is calculated (using
Benson’s rules15 and estimated electron affinities16) to be
significantly exothermic (∆H = Ϫ128 kJ molϪ1), as is the
subsequent elimination of methanol, i.e. PhC(Me)(OMe)OϪ→
PhCOCH2Ϫ ϩ MeOH (∆H = Ϫ280 kJ molϪ1).† We conclude
that the loss of MeOH observed in the spectrum of PhϪ-
C(OMe)2 is a facile process occurring following rearrangement
to the energised 1,2-Wittig product anion PhC(Me)(OMe)(OϪ).
The spectrum of [PhϪC(OEt)(Oiso-Pr)] is recorded for
illustrative purposes in Fig. 2 while data for a number of depro-
tonated unsymmetrical benzaldehyde acetals are listed in Table
2. The trend from these spectra is clear: the smaller alkanol is
lost preferentially. It is also of interest that the larger alkane is
always lost preferentially [although we do not know whether
alkane loss is a fragmentation of the Wittig ion, or of an inter-
mediate preceding the formation of the Wittig ion (see earlier)].
The trend for the relative losses of alkanol (for the systems
studied) indicates that alkyl migratory aptitudes are tert-
Bu > iso-Pr > Et ӷ Me. This is the expected ratio of radical
migratory aptitudes, consistent with relative rate data for 1,2-
Wittig rearrangements in solution.2–4
The major processes of some substituted species are listed in
Table 3. There are problems in using Hammett relationships
in gas phase studies.18 Often there is no inbuilt standard
with which the studied process may be related quantitatively.
This is true in the current study, and as a consequence, the
data must only be used to indicate overall trends. There are only
ؒ
two processes that can be compared, viz (i) the loss of Me , a
radical cleavage process of the unrearranged acetal [i.e. Ar–
Ϫ
Ϫ
ϪC(OMe)2→[(ArCO2Me) Me ]→(ArCO2Me) ϩ Me ], and
ؒ
ؒ
᝽
᝽
(ii) the loss of MeOH from the 1,2-Wittig product species. The
trends are clear. The loss of Me only occurs where there is
ؒ
an electron withdrawing substituent on the ring. The process
becomes more pronounced with increasing σ (i.e. the loss of
ؒ
Me increases as the electron affinity of the neutral ester
increases). In contrast, the loss of MeOH decreases dramatic-
ally as the electron withdrawing capability of the substituent
increases, indeed, there is no loss of MeOH when the substitu-
ent is p-cyano. These data do not support the proposition that
this gas phase 1,2-Wittig rearrangement is a radical process.
We now have the following situation. The generally accepted
view is that the condensed phase 1,2-Wittig rearrangement is
best explained by a radical mechanism. The published ab initio
(gas phase) study9 suggests an anion rearrangement, our
experimental (gas phase) migratory aptitudes follow the clas-
† The ion PhCH(OMe)(OϪ), formed by an analogous SN2 process, is
detected in the ion source. In this case the loss of MeOH involves
deprotonation of PhCHO, a process less favourable (by some 190 kJ
molϪ1) than deprotonation of PhCOMe.
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340
335

View Article Online
Fig. 3 Potential surface map of the decomposing system ϪCH2OMe. Ab initio calculations at the CISD/6-311ϩϩG**//RHF/6-311ϩϩG** level of
theory.
Table 4 Geometries and energies of reactant, product and transition
state for the 1,2-Wittig rearrangements of ϪCH2OMe [QCISD(T)
6-311ϩϩG(d,p)//HF 6-311ϩϩG(d,p) including zero point energy
(scaled to 0.9)]. [MeϪ = Ϫ39.6952202 hartrees; CH2O = Ϫ114.2395661.
MeϪ ϩ CH2O = Ϫ153.9347863 (nominally 0 kJ molϪ1 in Fig. 3)]
Et
–
]
[
CH2
A
[
B
O
Me
–
]
CH2
O
(+ 86)
( + 72)
( + 48.5)
CH2O + Et–
Ϫ15.4 kJ molϪ1
ϩ56.6 kJ molϪ1
Ϫ202.6 kJ molϪ1
(+ 15)
Reactant
Ϫ153.9406641 hartrees
CH2O + Me–
(0)
–CH2OMe
(0)
C1O2
O2C3
C3H
= 1.476 Å
_
–CH2OEt
CH3
3
= 1.371 Å
= 1.123 Å
C
O
2
H
1
H
O2C1C3 = 114.06Њ
Saddle point
Ϫ153.9132451 hartrees
C1O2
O2C3
C3H
= 1.257 Å
= 1.961 Å
= 1.087 Å
( -173)
PrO–
H
H
CH3
3
–
C
O
2
(-187)
EtO–
]
1
C1O2C3 = 113.9Њ
Fig. 4 Reaction coordinate diagram of (A) the CH2O–MeϪ and (B)
CH2O–EtϪ systems. Ab initio calculations at the QCISD(T)/6-311ϩϩ
G(d,p)//HF/6-311ϩϩG(d,p) level of theory.
Product
Ϫ154.0119473 hartrees
C1O2
C1C3
= 1.325 Å
= 1.550 Å
CH3CH2O–
sical trends for radicals but not anions, and the (gas phase)
Hammett study does not support a radical mechanism. We
must now find a mechanism which resolves the (apparently)
conflicting experimental evidence, and to do that we must have
access to a potential surface map for the gas phase 1,2-Wittig
rearrangement.
O2C1C3 = 114.30Њ
3
2
1
wish to compare the methyl migration (of ϪCH2OMe) with the
ethyl migration (of ϪCH2OEt) in order to ascertain whether
the rearrangements proceed by similar or different mechanisms
Ϫ
ؒ
ؒ
Potential surfaces for the model systems ؊CH2OR to RCH2O؊
[the electron of Me is bound in the free state (E.A. of Me is
7.5 kJ molϪ1),20 whereas that of Et is unbound (E.A. of Et is
Ϫ
(R ؍ Me and Et)
Ϫ30 kJ molϪ1)21].
Ab initio calculations of the rearrangement of ϪCH2OMe to
EtOϪ describing the reaction coordinate have already been
reported.9 The data suggest an anionic reaction with an early
saddle point but no discrete intermediate. We need to construct
a potential surface map for this prototypical system in order to
explore the anion mechanism more fully and also to test for the
possibility of an alternative radical pathway.‡,§ In addition, we
The potential surface map for the model 1,2-Wittig
rearrangement ϪCH2OMe→EtOϪ is shown in Fig. 3, and is
constructed to display the energy and geometry changes which
accompany methyl migration from oxygen to carbon. The
CISD/6-311ϩϩG(d,p)//RHF/6-311ϩϩG(d,p) level of theory
was used. The corresponding reaction coordinate diagram
is shown in Fig. 4: in this case, a higher level of theory,
QCISD(T)/6-311ϩϩG(d,p)//HF/6-311ϩϩG(d,p), was used to
represent the major features. Details of geometries and energies
of reactant, saddle point and product (shown in Fig. 4) are
listed in Table 4. The preferred geometries of reactant and
saddle point leading to methyl dissociation possess a plane of
symmetry (Cs) through methyl and formaldehyde groups: this
plane defines the plane of the map shown in Fig. 3. A point on
the map signifies the energy of the rearranging system (in kJ
‡ It is not feasible to construct a potential surface map of Ph–ϪCHOR
with our facilities. The potential surface map of ϪCH2OMe is con-
structed using 170 different geometries: each geometry (at the level
of theory used) needs some 7–8 h of super-computer time. The corre-
sponding time requirement for each calculation of a Ph–ϪCHOR
geometry is >48 h.
§ The radical CH2OMe is more stable than the anion: this does not
invalidate the use of ϪCH2OMe as a model system for the computation
of the potential surface for the rearrangement.
19
ؒ
Ϫ
molϪ1 with respect to CH3 and CH2O at infinite separation)
336
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340

View Article Online
shown in Fig. 3) for the homologous system ϪCH2OEt→PrOϪ.¶
Table 5 Geometries and energies of reactant, product and transition
state for the 1,2-Wittig rearrangement of ϪCH2OEt [QCISD(T)
6-311ϩϩG(d,p)//HF 6-311ϩϩG(d,p) including zero point energy
(scaled to 0.9)]. [EtϪ = Ϫ78.8741163 hartrees; CH2O = Ϫ114.2395661.
EtϪ ϩ CH2O = Ϫ193.1136824]
However we have studied the analogous region to that bounded
by R, S, T1, T2 and P in Fig. 3. In this region, the surface is
visually very similar to that shown in Fig. 3. This reaction
involves migration of an ethyl anion bound within the molecu-
lar complex. UHF calculations again fail to uncover any indi-
cation of a radical/radical anion pathway. The geometries and
energies of reactant, saddle point and product have also been
calculated at the higher level of theory [QCISD(T)/6-311ϩϩ
(d,p)//6-311ϩϩG(d,p)]. These data are summarised in Fig. 4,
with details of relevant geometries and energies listed in Table
5. The difference in saddle point energies are 10, 10.5 and 14
kJ molϪ1 respectively for the following levels of theory:
HF/6-311ϩϩG(d,p), CISD/6-311ϩϩG(d,p) and QCISD(T)/
6-311ϩϩ G(d,p). Thus the barrier to the transition state for
the EtϪ migration is calculated to be 14 kJ molϪ1 higher than
that for the corresponding MeϪ migration at the highest level
of theory used.
Reactant
Ϫ193.1328663 hartrees
Ϫ50.4 kJ molϪ1
C1O2
= 1.478 Å
H
H6 H8
C3 C4
O2C3
C3C4
C1C5
C3H6
C4H7
C4H8
= 1.371 Å
= 1.526 Å
= 1.102 Å
= 1.094 Å
= 1.088 Å
= 1.092 Å
_
H
C1 O2
H
H7
H5
C1O2C3 = 113.9Њ
O2C3C4 = 109.4Њ
Saddle point
Ϫ193.0996784 hartrees
ϩ36.8 kJ molϪ1
C1O2
= 1.249 Å
= 2.021 Å
= 1.499 Å
H
H
O2C3
C3C4
H
C3 C4
An anionic migration
–
H
H
C1O2C3 = 116.8Њ
O2C3C4 = 150.0Њ
H
O2
C1
Cs
[
]
H
Ab initio calculations indicate that the gas phase 1,2-Wittig
reactions of two model systems are anionic rearrangements.
The experimental results of the Hammett study are not consist-
ent with a radical mechanism. How then can we explain the
experimental alkyl migratory aptitudes which are apparently
consistent with a radical process, but not with the classical
relative migratory aptitudes expected of anions. The clue to this
problem is to be found in the potential surface shown in Fig. 3,
and with the analogous, but partial surface, that has been con-
structed for the cognate system ϪCH2OEt→PrOϪ. These sur-
faces exhibit subtle differences which cannot be seen when the
maps are represented as shown in Fig. 3. The first difference
between the two processes is that the barrier for the 1,2-Wittig
rearrangement of ϪCH2OEt is 14 kJ molϪ1 higher than that for
the analogous reaction of ϪCH2OMe. If these barriers deter-
mine migratory trends, MeϪ must migrate faster than EtϪ. This
is the reverse of what is observed experimentally. Thus it cannot
be the modest barriers to the early transition states which are
causing the pronounced difference noted in the migratory apti-
tudes of Me and Et.
The rate of a gas-phase reaction is dependent on both the
barrier to the saddle point and the Arrhenius A factor22 (or the
frequency factor of the QET expression) (see23 for an illus-
tration of the importance of this parameter). If the reaction
proceeds through the saddle point S in an upward direction in
Fig. 3, there is an arc of some 60Њ where decomposition may be
effected to yield MeϪ and CH2O. In contrast, the competitive
formation of EtOϪ has a low probability relative to dissociation
to MeϪ and CH2O. In this system, it is not the barrier from R to
S which is the major factor influencing the rate of reaction, it is
the low frequency factor of the migration stage of this process,
even though migration is occurring as part of a synchronous
reaction at a stage of rapid energy descent. Any feature which
affects the frequency factor will alter the rate of this reaction.
When particular details of the two potential surfaces are
compared it is clear why the migration of EtϪ is more efficient
than that of MeϪ. We have arbitrarily chosen two transit points
on the rearrangement reaction coordinate of each potential
surface map to illustrate particular features of the molecular
rearrangements. We have called these transit points T1 and T2
(see Fig. 3). The T1 structures immediately follow the saddle
points on the rearrangement reaction coordinate on each
potential surface map and are in identical positions, i.e. they
have the same significant coordinates. The geometries of these
T1 structures are shown for comparison, in Fig. 5. There is a
weak electrostatic interaction between the electron rich oxygen
Product
Ϫ193.198872 hartrees
Ϫ223.1 kJ molϪ1
C1O2
C1C3
C3C4
= 1.325 Å
H
H
H6
= 1.550 Å
= 1.531 Å
= 1.124 Å
= 1.091 Å
= 1.091 Å
= 1.090 Å
C4 C3
H
H7
C1H5
C3H6
C4H8
C4H7
O–
C1
2
H
H5
C2C1C3 = 143.3Њ
when the methyl carbon occupies that point. The notional
origin of the 1 Å grid is fixed on the formaldehyde carbon and
the lower edge defines the formaldehyde C–O direction.
The potential map of the MeϪ/CH2O system is shown in Fig.
3, and is constructed from 170 geometries. The reactant is
represented by the shallow local minimum R (Ϫ26 kJ molϪ1
relative to MeϪ and CH2O at infinite dissociation; O–C(H3)
bond = 1.37 Å). The dissociation to MeϪ and CH2O is repre-
sented by the arrow RSD passing through saddle point S [ϩ74
kJ molϪ1, O–C(H3) distance = 1.96 Å]. The most probable con-
tinuation of such movement is into the broad flat region leading
to MeϪ plus CH2O [MeϪ will eventually eject an electron to give
ؒ
Me since the excess energy of the system is greater than the
Ϫ1
ؒ
E.A. of Me (7.5 kJ mol )]. The rearrangement channel from S
through T1 to T2 (T1 and T2 are transit structures, the relevance
of which will be described later) is not well defined: there-
after, the descending channel to the product P is well defined.
There is no discrete intermediate along the reaction coordinate.
The map demonstrates that the 1,2-Wittig rearrangement of
ϪCH2OMe is inefficient (with respect to dissociation to MeϪ
and CH2O) since few methyl groups departing from saddle
point S will enter channel T1T2P unless the forces of association
within the anion complex are favourable in directing the course
of that reaction (see later).
Migration of a methyl radical also needs to be considered,
as well as the possibility of a rearrangement that might
have both anion and radical character and different stages of
the migration (cf.5). These possibilities were explored by re-
examining structures in the vicinity of the reaction coordinate
using unrestricted Hartree–Fock calculations (the electronic
state here is still a singlet, but electrons can be assigned to sep-
arate molecular orbitals to model a homolytic bond breaking
and the emergence of diradical character). The results from the
RHF and UHF calculations are identical: there is no indication
of any radical/radical anion pathway either preceding or follow-
ing saddle point S.
¶ Calculation of each geometry for this system required some 24–26 h
of super-computer time.
We have not computed a full potential surface map (like that
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340
337

View Article Online
Table 6 Mulliken charge analysis of HF/6-311ϩϩG(d,p) wavefunctions of transit structures T2 shown in Figs. 6 and 7
T2
H-bonding
ϩ0.068
ϩ0.091
ϩ0.068a
ϩ0.079c
ϩ0.133
Carbanion C
Ϫ1.008
Ϫ0.752
Ϫ0.117
β Carbon
C(CH2O)
ϩ0.246
ϩ0.226
ϩ0.156
C(CH2O)
Ϫ0.337
Ϫ0.280
Ϫ0.284
Methyl
Ethyl
iso-Propyl
Ϫ0.418
Ϫ0.587b
Ϫ0.442d
Ϫ0.563e
Ϫ0.744f
tert-Bu
0.555
ϩ0.143
Ϫ0.278
a For bonding H that has shorter distance to O (see Fig. 7). b For bonding H that has longer distance to O. c For the carbon of the methyl group that is
involved in H-bonding. d For the carbon of the methyl group that is not involved in H-bonding. e For the two equivalent methyl groups not involved
in H-bonding. f For the methyl group involved in H-bonding.
Fig. 5 Comparison of transit structures T1 for the migrating systems
(A) [MeϪ/CH2O] and (B) [EtϪ/CH2O]. Ab initio calculations at the
QCISD(T)/6-311ϩϩG(d,p)//HF/6-311ϩϩG(d,p) level of theory.
Fig. 7 Transit structures T2 for the migrating systems [iso-PrϪ/CH2O]
(left) and [tert-BuϪ/CH2O]. Ab initio calculations at the QCISD(T)/6-
311ϩϩG(d,p)//HF/6-311ϩϩG(d,p) level of theory.
modelled using the same significant coordinates as those of the
EtϪ/CH2O transit structure T2 and are shown in Fig. 7. The
lowest energy structure in both cases is one where H-bonding is
initiated from only one methyl group, but in the iso-Pr case the
two hydrogen bonds are not equivalent (see Fig. 7).|| We pro-
pose that migrating iso-propyl and tert-butyl anions are held in
place by an H-bonding pivot occurring between two hydrogens
of one methyl group and the oxygen of formaldehyde. The
migratory aptitudes in these cases are also influenced by the
reaction path degeneracy, i.e. the number of degenerate struc-
Fig. 6 Comparison of transit structures T2 for the migrating systems
(A) [MeϪ/CH2O] and (B) [EtϪ/CH2O]. Ab initio calculations at the
QCISD(T)/6-311ϩϩG(d,p)//HF/6-311ϩϩG(d,p) level of theory.
tures that can be drawn for each transit structure (i.e. in the
ratio 3:2:1 for tert-Bu, iso-Pr and Et respectively) (for a discus-
sion of reaction path degeneracy see24). The migratory apti-
tudes are thus tert-BuϪ > iso-PrϪ > EtϪ.
(of formaldehyde) and one hydrogen of the methyl anion. For
the ethyl anion complex, the hydrogen bonding is to two of the
methyl hydrogens of the substituent α to the carbanion centre:
an electrostatic interaction which is greater than that for the
corresponding methyl anion migration.
Conclusions
We propose that the gas phase 1,2-Wittig rearrangement is an
anionic reaction. The saddle point is reactant like (i.e. only
minor dissociation of the O–C bond has occurred in attaining
the saddle point), and there is no discrete intermediate formed
during the reaction. The migration which proceeds follow-
ing formation of the saddle point has a low frequency factor:
this is the major influence on the rate of reaction. During the
migration process, the migrating anion is held, by weak hydro-
gen bonding, to the neutral to which the anion is migrating. The
efficiencies of these reactions are a function of the extent of the
H-bonding that holds the anion and neutral together during
the initial phase of the migration, together with the reaction
degeneracy of the migration pathway. Once the molecular
complex has reached T2 (see Fig. 3), the subsequent nucleo-
philic substitution should be efficient for all the systems
Transit structure T2 (for the MeϪ/CH2O system) is chosen to
illustrate the geometry of the rearranging system en route to
products (see T2 on the potential surface map shown in Fig. 3).
This structure is shown in Fig. 6. The point with the same sig-
nificant coordinates on the EtϪ/CH2O surface is also shown
in Fig. 6. Hydrogen bonding between the anion and neutral
essentially acts as a pivot during the anion migration towards
the electrophilic carbon terminus. The acidic nature of the
hydrogens involved in the H-bonding are substantiated by the
Mulliken electron density analyses of transit structures T2 listed
in Table 6. Thus EtϪ migrates more efficiently than MeϪ in these
systems, a result opposite to that expected using thermo-
dynamic anionic migratory aptitudes. This enables the rational-
isation of the relative migratory aptitudes of EtϪ and MeϪ in
this reaction, but what accounts for the further trend tert-
BuϪ > iso-PrϪ > EtϪ? The answer to this is also provided by a
knowledge of the appropriate transit structures. In order to
answer this question, we have computed the transit structures
T2 for the iso-PrϪ/CH2O and tert-BuϪ/CH2O systems: these are
|| The structure shown for the iso-PrϪ/CH2O system in Fig. 7 is 32 kJ
molϪ1 more negative in energy [at the QCISD(T)6-311ϩϩG(d,p)//HF/
6-311ϩϩG(d,p) level of theory] than the analogous transit structure in
which both methyl groups are involved equally in the H-bonding.
338
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340

View Article Online
studied. The relative kinetic migratory aptitudes of alkyl groups
imaginary frequency and its Hessian matrix possesses only
in this rearrangement are tert-BuϪ > iso-PrϪ > EtϪ ӷ MeϪ.
one negative eigen value. The remaining points (other than the
stationary points) used to construct the potential surface are
selected and fixed positions of the carbanion carbon relative to
the C–O bond. These selected points on the surface effectively
freeze just two of the molecular coordinates, that is, the vertical
and horizontal displacement of the carbanion carbon from the
carbonyl carbon. Final energies were determined at the CISD/
6-311ϩϩG**//RHF/6-311ϩϩG** level of theory. A number
of key structures on the surface were also computed at the
unrestricted Hartree–Fock (UHF) level of theory.
Ab initio calculations of stable species and transition states
shown in Fig. 4 were determined with GAUSSIAN9433 using
the QCISD(T)/6-311ϩϩG(d,p)//HF/6-311ϩϩG(d,p) level of
theory. Vibrational zero point energy corrections have been
made.
Experimental
Mass spectrometric methods
Collisional activation (CID) mass spectra (MS/MS) were
determined with a VG ZAB 2HF mass spectrometer.25 Full
operating details have been reported.26 Specific details were as
follows: the chemical ionisation slit was used in the chemical
ionisation source, the ionising energy was 70 eV, the ion source
temperature was 100 ЊC, and the accelerating voltage was 7 kV.
The liquid samples were introduced through the septum inlet
with no heating [measured pressure of sample 1 × 10Ϫ6 Torr
(1 Torr = 133.322 Pa)]. Deprotonation was effected using HOϪ
(from H2O: measured pressure 1 × 10Ϫ5 Torr). The estimated
source pressure was 10Ϫ1 Torr. CID MS/MS data were obtained
by selecting the particular anion under study with the magnetic
sector, passing it through the collision cell, and using the elec-
tric sector to separate and monitor the product ions. Argon was
used as collision gas in the second collision cell (measured pres-
sure, outside the cell, 2 × 10Ϫ7 Torr), giving a 10% reduction
in the main beam, equivalent to single collision conditions.
Charge reversal (CR) mass spectra were measured as for CID
MS/MS (above) except that the voltages of the electric sector
were reversed to allow transmission of positive ions.27
Acknowledgements
We thank the Australian Research Council for financial sup-
port. One of us (S. D.) thanks the ARC for a research associate
position.
References
1 W. Schlenk and E. Bergmann, Justin Liebigs Ann. Chem., 1928, 464,
22; G. Wittig and L. Lohmann, Justin Liebigs Ann. Chem., 1942,
550, 260; G. Wittig, Angew. Chem., 1954, 66, 10; H. E. Zimmerman,
in Molecular Rearrangements, ed. P. DeMayo, Interscience, Vol. 1,
p. 345, 1963; J. E. Baldwin, J. DeBernardis and J. E. Patrick,
Tetrahedron Lett., 1970, 353.
2 C. R. Hauser and S. W. Kantor, J. Am. Chem. Soc., 1951, 73, 1437;
J. Cast, T. S. Stevens and J. Holmes, J. Chem. Soc., 1960, 3521.
3 P. T. Lansbury, V. A. Pattison, J. D. Sidler and J. B. Bierber, J. Am.
Chem. Soc., 1966, 88, 78; H. Schafer, U. Schöllkopf and D. Walter,
Tetrahedron Lett., 1968, 2809; T. Weiske and H. Schwarz, Tetra-
hedron, 1986, 42, 6245, and references cited therein.
4 U. Schöllkopf, Angew. Chem., Int. Ed. Engl., 1970, 9, 763.
5 J. F. Garst and C. D. Smith, J. Am. Chem. Soc., 1976, 98, 1526.
6 R. A. W. Johnstone and T. S. Stevens, J. Chem. Soc., 1960, 3340;
J. J. Eisch, S. K. Dua and C. A. Kovacs, J. Org. Chem., 1987, 52,
4437, and references cited therein.
7 P. C. H. Eichinger, J. H. Bowie and T. Blumenthal, J. Org. Chem.,
1986, 51, 5078.
8 P. C. H. Eichinger and J. H. Bowie, J. Chem. Soc., Perkin Trans. 2,
1987, 1499; P. C. H. Eichinger and J. H. Bowie, J. Chem. Soc., Perkin
Trans. 2, 1988, 497; P. C. H. Eichinger and J. H. Bowie, J. Chem.
Soc., Perkin Trans. 2, 1990, 1763.
9 P. Antoniotti and G. Tonachini, J. Org. Chem., 1993, 58, 3622.
10 Y.-D. Wu, K. N. Houk and J. A. Marshall, J. Org. Chem., 1990, 55,
1421.
11 M. R. Ahmad, G. D. Dahlke and S. R. Kass, J. Am. Chem. Soc.,
1996, 118, 1398.
Syntheses of labelled and unlabelled compounds
The dimethyl and diethyl acetals of benzaldehyde were com-
mercial samples. The following acetals were prepared by
reported procedures: RC6H4(OMe)2: m-F,28 p-F,29 p-CN29 and
p-NMe2.30 The dimethyl ketal of acetophenone was prepared by
a reported method.30
3-Cyanobenzaldehyde dimethyl acetal was prepared by a
standard method29 as a colourless liquid (bp 154–155 ЊC/17
mm Hg, 64% yield). Found: C, 67.87; H, 6.07; N, 7.96%; C10H11-
1
NO2 requires C, 67.78; H, 6.25; N, 7.90%. H NMR [δ ppm,
300 MHz (in CDCl3)] 3.33 (s, 6H, OCH3), 5.43 (s, 1H, C–H),
ϩ
ؒ
7.4–7.8 (m, 4H, aromatic H). m/z 182 (M , 6%), 151 (ϪMeO ,
᝽
100%), 135 (MeOC6H4COϩ, 32%).
Benzaldehyde-α-2H1 dimethyl acetal was prepared from
benzaldehyde-α-2H1 by a reported method31 in 77% yield
(2H1 = 99%).
The mixed acetals (see Table 3). General procedure. Benz-
aldehyde dimethyl (or diethyl) acetal (2.5 g) was added to the
appropriate alcohol (0.75 g) containing sulfuric acid (concen-
trated, 2 drops). The mixture was allowed to stir at 100 ЊC
under nitrogen for 30 min, quenched with sodium metal,
and extracted with diethyl ether (2 × 15 cm3). Removal of the
solvent in vacuo followed by vacuum distillation of the residue
yielded a mixture of PhCH(OR1)2, PhCH(OR1)(OR2) and
PhCH(OR2)2 which was not separated. Injection of a sample of
this mixture into the mass spectrometer, followed by reaction
with HOϪ in the ion source of the VG ZAB 2HF yielded
the three (M Ϫ H)Ϫ ions. The required ion Ph–ϪC(OR1)(OR2)
was selected using the magnetic sector: collisional activation
MS/MS data were obtained as described above.
12 G. B. Ellison and P. C. Engelking, J. Chem. Phys., 1978, 69, 1826.
13 C. H. DePuy, V. M. Bierbaum and R. Damrauer, J. Am. Chem. Soc.,
1984, 106, 4051.
14 J. Fossey, D. Lefort and J. Sorba, Free Radicals in Organic
Chemistry, Wiley, New York, 1995, pp. 31–38.
15 S. W. Benson, Thermochemical Kinetics, Wiley, New York, London
and Sydney, 1968.
16 NIST Standard Reference Database
- March 1998. Electron
Affinities.
17 F. Ruff and I. G. Csizmadia, Organic Reactions, Equilibria, Kinetics
and Mechanism, Elsevier, Amsterdam, 1994, pp. 161–210.
18 K. Levsen, Fundamental Aspects of Organic Mass Spectrometry,
Verlag Chemie, Weinheim and New York, 1978, p. 119.
19 C. H. DePuy, V. M. Bierbaum and R. Damrauer, J. Am. Chem. Soc.,
1984, 106, 4051.
Ab initio calculations
Ab initio calculations of all stable species (local minima) and
the transition state shown in the potential surface map (Fig. 3)
were determined with GAUSSIAN9232 with restricted Hartree–
Fock wave functions [RHF/6-311ϩϩG**]. The geometries
of stable species were optimised with initial force constants
analytically computed. The final geometries were found to have
no imaginary frequencies. The optimisations of transition
states were computed with analytical force constants through-
out. A transition state is characterised by possessing only one
20 D. E. Applequist and D. F. O’Brien, J. Am. Chem. Soc., 1963, 85,
743; R. E. Dessy, W. Kitching, T. Psarras, R. Salinger, A. Chen and
T. Chivers, J. Am. Chem. Soc., 1966, 88, 460; A. I. Shatenshtein, Adv.
Phys. Org. Chem., 1963, 1, 156.
21 G. B. Ellison, P. C. Engelking and W. C. Lineberger, J. Am. Chem.
Soc., 1978, 100, 2556; C. H. DePuy, V. M. Bierbaum and R.
Damrauer, J. Am. Chem. Soc., 1984, 106, 4051; see also S. G. Lias,
J. E. Bartmess, J. F. Liebman, J. L. Holmes, R. D. Levin and W. G.
Mallard, Gas Phase Ion and Neutral Thermochemistry, J. Phys.
Chem. Ref. Data 17, 1988, Suppl. 1 (the computer version was used).
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340
339

View Article Online
22 G. W. Castellan, Physical Chemistry, Addison-Wesley Publ. Co.,
Amsterdam, 1983, 3rd edn., pp. 813, 858.
23 S. Okada, Y. Abe, S. Tanaguchi and S. Yamabe, J. Am. Chem. Soc.,
1987, 109, 295.
24 For discussions of reaction path degeneracy see: P. J. Robinson
and K. A. Holbrook, Unimolecular Reactions, Wiley Interscience,
New York, 1972, pp. 80–85; R. G. Gilbert and S. C. Smith, Theory of
Unimolecular and Recombination Reactions, Blackwell Scientific
Publications, Oxford, UK, 1990, pp. 87–98.
25 V. G. Instruments, Wythenshawe, Manchester, UK, Model ZAB
2HF.
26 M. B. Stringer, J. H. Bowie and J. L. Holmes, J. Am. Chem. Soc.,
1986, 108, 3888.
27 J. H. Bowie and T. Blumenthal, J. Am. Chem. Soc., 1975, 97, 2959;
J. E. Szulejko, J. H. Bowie, I. Howe and J. H. Beynon, Int. J. Mass
Spectrom. Ion Phys., 1980, 13, 76.
28 J. N. Frescos, G. W. Morrow and J. S. Swenton, J. Org. Chem., 1985,
6, 805.
30 I. Fleming, J. Iqbal and E. P. Krebs, Tetrahedron, 1983, 39, 841.
31 J. Klein and E. D. Bergmann, J. Am. Chem. Soc., 1957, 79, 3452.
32 M. J. Frisch, G. W. Trucks, M. Head-Gordon, P. M. W. Gill, M. W.
Wong, J. B. Foresman, B. J. Johnson, H. B. Schlegel, M. A. Robb,
E. S. Replogle, R. Gomperts, J. L. Andres, K. Raghavachari, J. S.
Binkley, C. Gonzalez, R. L. Martin, D. J. Fox, D. J. Defrees,
J. Baker, J. J. P. Stewart and J. A. Pople, GAUSSIAN92, Revision
G3, Gaussian Inc., Pittsburgh, PA, 1992.
33 M. J. Frisch, G. W. Trucks, H. B. Schlegel, P. M. W. Gill, B. G.
Johnson, M. A. Robb, J. R. Cheeseman, T. Keith, G. A. Petersson,
J. A. Montgomery, K. Raghavachari, M. A. Al-Latham, V. G.
Zakrewski, J. V. Ortiz, J. B. Foresman, J. Cioslowski, B. B. Stefanov,
A. Nanayakkara, M. Challacombe, C. Y. Peng, P. V. Ayala,
W. Chem, M. W. Wong, J. L. Andres, E. S. Replogle, R. Gomperts,
R. L. Martin, D. J. Fow, J. S. Binkley, D. J. Defrees, J. Baker,
J. P. Stewart, M. Head-Gordon, C. Gonzales and J. A. Pople,
GAUSSIAN94, Revision C3, Gaussian Inc., Pittsburgh, PA, 1995.
29 T. Sigeru, I. Tsutoma, T. Sadahito, H. Hirofimi, K. Hideki and
U. Kenji, Bull. Chem. Soc. Jpn., 1987, 60, 2173.
Paper 8/05319G
340
J. Chem. Soc., Perkin Trans. 2, 1999, 333–340