The Mechanism of Phosphonium Ylide Alcoholysis and Hydrolysis: Concerted Addition of the O?H Bond Across the P=C Bond

Article abstract of DOI:10.1002/chem.201600530

The previous work on the hydrolysis and alcoholysis reactions of phosphonium ylides is summarized and reviewed in the context of their currently accepted mechanisms. Several experimental facts relating to ylide hydrolysis and to salt and ylide alcoholysis are shown to conflict with those mechanisms. In particular, we demonstrate that the pK_avalues of water and alcohols are too high in organic media to bring about protonation of ylide. Therefore, we propose concerted addition of the water or alcohol O?H bond across the ylide P=C bond. In support of this, we provide NMR spectroscopic evidence for equilibrium between ylide and aclohol that does not require the involvement of phosphonium hydroxide. We report the first P-alkoxyphosphorane to be characterised by NMR spectroscopy that does not undergo exchange on an NMR timescale. Two-dimensional NMR spectroscopic techniques have been applied to the characterisation to P-alkoxyphosphoranes for the first time.

Full text of DOI:10.1002/chem.201600530

DOI: 10.1002/chem.201600530
Full Paper
&
Reaction Mechanisms |Hot Paper|
The Mechanism of Phosphonium Ylide Alcoholysis and
Hydrolysis: Concerted Addition of the OÀH Bond Across the P=C
Bond
Peter A. Byrne*[a, b] and Declan G. Gilheany[a]
Chem. Eur. J. 2016, 22, 9140 – 9154
9140
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Abstract: The previous work on the hydrolysis and alcoholy-
sis reactions of phosphonium ylides is summarized and re-
viewed in the context of their currently accepted mecha-
nisms. Several experimental facts relating to ylide hydrolysis
and to salt and ylide alcoholysis are shown to conflict with
those mechanisms. In particular, we demonstrate that the
pKa values of water and alcohols are too high in organic
media to bring about protonation of ylide. Therefore, we
propose concerted addition of the water or alcohol OÀH
bond across the ylide P=C bond. In support of this, we pro-
vide NMR spectroscopic evidence for equilibrium between
ylide and aclohol that does not require the involvement of
phosphonium hydroxide. We report the first P-alkoxyphos-
phorane to be characterised by NMR spectroscopy that does
not undergo exchange on an NMR timescale. Two-dimen-
sional NMR spectroscopic techniques have been applied to
the characterisation to P-alkoxyphosphoranes for the first
time.
Introduction
which expels a carbanion (probably protonated in the process
of its expulsion) to give phosphine oxide and alkane or
arene.[8] The leaving group (R4) is invariably the most stable
anion.[3,4,9,11] Reactions of phosphonium salts in which phos-
phorus is not involved in a ring appear to proceed stereospe-
cifically with inversion at phosphorus.[12,13,33] Consistent with
the above mechanism, phosphonium salt hydrolyses are usual-
ly first-order in phosphonium salt, second-order in hydroxide,
and therefore third-order overall.[2–4,14] Exceptional cases are
known in which reactions show first-order dependence on hy-
droxide concentration, for example, the hydrolysis reactions of
para-nitrobenzyltriphenylphosphonium bromide,[3,15] of methyl-
tris(pentafluorophenyl)phosphonium fluorosulfonate,[15] and of
phospholium iodides.[16]
1. Phosphonium salt and ylide hydrolysis
The alkaline hydrolysis of phosphonium salts is one of the pro-
totypical reactions in phosphorus chemistry, having been re-
ported on for the first time (to our knowledge) as early as
1857.[1] The reaction has been the subject of numerous mecha-
nistic investigations over the course of many years.[2–17] Phos-
phorane (PV) species are the only observable intermediates in
phosphonium salt and ylide hydrolysis[18] and alcoholysis reac-
tions (vide infra).[11,19–21] The growth in the number and breadth
of catalytic organophosphorus reactions that rely on the inter-
vention of a phosphorane intermediate in recent years has
been rapid.[22–26] Of particular relevance to the present study is
a recent report on the use of an organocatalytic species
formed by activation of CO2 by phosphonium ylides in reac-
tions with epoxides, alkynols and aziridines (to form cyclic car-
bonates and carbamates) in which a putative alkoxyphosphor-
ane (or aminophosphorane) is a key catalytic intermediate.[27]
In addition to being of interest for historical and mechanistic
reasons, phosphonium salt hydrolysis provides a very useful
synthetic route to tertiary phosphine oxides,[28,29] often with
control of stereochemistry that can be predicted based on the
nature of the groups attached to phosphorus. Given the
recent advent of methods for the facile reduction of phosphine
oxides to phosphines and phosphine boranes,[30–32] phosphoni-
um salt hydrolysis now has the potential to become a relatively
straightforward and stereospecific route to access phosphines.
The mechanism of phosphonium salt hydrolysis is well-es-
tablished,[2–7] and the steps are summarised in Scheme 1. Nu-
cleophilic attack of hydroxide at phosphorus gives a P-hydrox-
ytetraorganophosphorane (with apical oxygen). This is depro-
tonated by hydroxide to give an oxyanionic phosphorane,
Scheme 1. The mechanism of alkaline hydrolysis of a phosphonium salt.
In contrast, comparatively little work has been done on the
mechanism of the closely related reaction, hydrolysis of phos-
phonium ylides. It has commonly been observed that phos-
phonium ylide hydrolysis results in the cleavage of the same
(ultimately protonated) leaving group that would be expected
in the hydrolysis of the corresponding phosphonium
salt.[4,29,34,35] Thus, as identified by Johnson, “Largely on this
basis it has been concluded that hydrolysis of ylides proceeds via
initial protonation to a phosphonium salt, followed by hydrolysis
of the salt to hydrocarbon and phosphine oxide”.[5] In other
words, the currently accepted mechanism for ylide hydrolysis
is essentially the same as that shown in Scheme 1 for phos-
phonium salt hydrolysis, just involving an extra step (ylide pro-
tonation) at the start. Therefore, the two processes should
have several common intermediates. Importantly, implicit in
the above mechanism for ylide hydrolysis is that the hydroxy-
phosphorane intermediate is formed in a stepwise fashion via
phosphonium hydroxide.
[a] Dr. P. A. Byrne, Prof. D. G. Gilheany
Centre for Synthesis and Chemical Biology, University College Dublin
Belfield, Dublin 4 (Ireland)
[b] Dr. P. A. Byrne
Current address: Department Chemie
Ludwig-Maximilians-Universität München, Butenandtstr. 5–13 (Haus F)
81377 München (Germany)
E-mail: peter.byrne@cup.lmu.de
Although strong indirect evidence indicating the involve-
ment of a pentavalent intermediate in phosphonium salt and
ylide hydrolysis has existed for some time, and despite the fact
Supporting information and the ORCID identification number for the
author of this article can be found under http://dx.doi.org/10.1002/
chem.201600530.
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9141
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
that analogous hydroxyphosphoranes and related compounds
with two or more cyclic P-alkoxy groups have been report-
ed,[36,37] until recently, no P-hydroxytetraorganophosporane
(with four PÀC bonds) had ever been detected experimental-
ly.[38,39] However, very recently, we reported the spectroscopic
observation and characterisation of P-hydroxytetraorganophos-
phorane 1 at low temperature (Scheme 2, obtained by addi-
tion of H2O to the parent ylide 2), finally confirming the in-
volvement of such species in these reactions.[18]
Scheme 2. The production of P-hydroxyphosphorane 1 from ylide 2.
Finally, we note that a small number of differences have
been observed between phosphonium salt and ylide hydroly-
sis. For example, hydrolysis of the enantiopure chiral ylide (R)-
(benzylidene)ethylmethylphenylphosphorane (derived from the
enantiopure parent salt) gives racemic ethylmethylphenylphos-
phine oxide[40] (whereas the hydrolysis of the enantiopure
parent salt is stereospecific[12]). Ylide hydrolysis is faster than
salt hydrolysis,[35] which has been ascribed to the low polarity
of the medium in which the ylide must necessarily be prepared
compared to the relatively high polarity of the aqueous organ-
ic media in which salt hydrolysis is usually conducted.[35] Addi-
tionally, we have reported one specific case where different
products were obtained from hydrolysis of a salt and its de-
rived ylide.[41]
Scheme 3. Phosphonium salt and ylide alcoholysis reactions of: a) Schmid-
baur,[19] b) Hey and Ingold[10] and Eyles and Trippett,[45] c) Grayson and
Keough.[49]
oxide and alkane. In both the alcoholysis of 3 and that of 5–7,
the alkane formed was the product of CÀC bond formation be-
tween the oxygen-bearing carbon of the alkoxy group and the
phosphorus-bearing group of the phosphonium ylide or salt
that is, between groups initially attached to P and C, respec-
tively.[43] Heating of alkoxymethyltriphenylphosphoranes (de-
rived from 4; see Scheme 3a)[19] or of alkyltriphenylphosphoni-
um halides (8) and solid sodium alkoxide (Scheme 3b)[44–46] in
the absence of solvent gave mainly P-alkyldiphenylphosphine
and the phenyl ether derived from the alkyl group of the alk-
oxide/alcohol.[47] In the series of reactions of n-hexyltriphenyl-
phosphonium bromide with NaOMe, NaOEt and NaO(iPr), pro-
gressively less of the phenyl alkyl ether was produced, and
indeed none at all was produced if NaO(tBu) was employed
(instead, products arising from elimination or hydrolysis were
observed).[45] Eyles and Trippett concluded that the size of the
larger alkoxides hindered or prevented altogether the forma-
tion of the phosphorane intermediate, resulting in the ob-
served reduction in the trend of phenyl alkyl ether formation.
The reaction of benzyltriphenylphosphonium bromide with
solid NaOMe at 2408C gave Ph3P and stilbene (isomeric mix-
ture).[45]
2. Phosphonium salt and ylide alcoholysis
The alcoholysis reactions of phosphonium salts and ylides are
closely related to the hydrolysis reactions of the same species.
A large and relatively complex body of data exists on these al-
coholysis reactions, the details of which have never previously
been fully summarised. Thus we collect together the most per-
tinent details below.
Several examples exist in which the alkoxyphosphorane in-
termediates of alcoholysis reactions have been observed spec-
troscopically.[11,19–21] For example, Schmidbaur and co-workers
generated alkoxyphosphoranes by reacting ylides 3 and 4 with
various alcohols[19] (Scheme 3a) and ethene oxide,[19] and char-
1
acterised them by H, 13C, and 31P NMR spectroscopies. Alkoxy-
phosphoranes have also been proposed (although not ob-
served) numerous times as intermediates in reactions of phos-
phonium salts or ylides with alcohols (vide infra).
The nature of the products of these reactions seems to
depend on the conditions. Heating of the alkoxytetralkylphos-
phoranes generated from ylide 3 (Scheme 3a)[19] or of tetraal-
kylphosphonium alkoxide salts (5–7, generated by treatment
of phosphonium halides with NaOEt in dry ethanol,
Scheme 3b)[10,42] in the absence of solvent gave phosphine
Grayson and Keough found that treatment of either p-nitro-
benzyltriphenylphosphonium bromide (9) with sodium alkox-
ide in refluxing alcohol or of p-nitrobenzylidenetriphenylphos-
phorane (10) with refluxing alcohol gave the same products À
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9142
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Ph3PO, p-nitrotoluene,[48] and the homodialkyl ether derived
from the alcohol (Scheme 3c).[49–51] Entirely analogous products
were observed by McEwen and co-workers in the reaction of
enantioenriched benzyl phosphonium salt 11 with sodium n-
butoxide in refluxing butanol—phosphine oxide 12, toluene
and Bu2O (see Scheme 4).[52] Of particular significance is that al-
coholysis of 11 led to racemisation of the phosphine oxide
product (this phenomenon was also observed in the alcoholy-
sis of other benzyl phosphonium salts by Luckenbach,[53] and is
similar to what is observed in the hydrolysis of the derived
chiral ylide,[40] vide supra). The authors proposed that racemisa-
tion occurred as a consequence of the interconversion of
enantiomeric butoxyphosphonium salts 14a and 14b via
meso-dibutoxyphosphorane 15 (Scheme 4).[54–57]
ylide protonation to form phosphonium hydroxide or alkoxide
(exactly as for ylide hydrolysis). This is followed by nucleophilic
attack of the alkoxide anion to give alkoxyphosphorane (direct
displacement of the carbanion leaving group has also been
suggested[49]). Decomposition of the alkoxyphosphorane gives
different products depending on their structure and the reac-
tion conditions (vide supra). In contrast, hydroxyphosphoranes
(whether produced in the presence of excess base, as in phos-
phonium salt hydrolysis, or in its absence, as in ylide hydroly-
sis) decompose exclusively to phosphine oxide and protonated
alkyl/aryl leaving group, with (to our knowledge) only one ex-
ception.[41] The alcoholysis reactions of the alkoxyphosphorane
derived from 3 and 4, and those of the phosphonium ethox-
ides (or possibly ylides[61]) derived from 5–8 were carried out in
the absence of solvent or any external nucleophile at very high
temperature, and in these reactions the alkyl group of the al-
cohol ends up attached to the carbanion leaving group in
either an alkane (Scheme 5, path A) or via oxygen in a phenyl
alkyl ether (path B).[62] Alcoholysis reactions of compounds 8, 9,
10 and 11 (and related benzylides) and cyano-stabilised ylides
were carried out in refluxing alcohol[49,52] and propionitrile sol-
vents,[60] respectively. In these reactions, the alkyl group of the
alcohol ends up attached to an external nucleophile (as a thio-
ether,[60] amine,[60] or homoether of the alcohol[49,52]), separate
to the carbanion leaving group, which ends up as a protonated
alkane or arene (Scheme 5, path C). Despite the differences in
the products formed—perhaps imposed by the differing reac-
tion conditions—the two sets of reactions still share many
common features: alkoxyphosphorane intermediates are
almost certainly involved in all of these reactions; the PÀC
bond that is cleaved is that to the most stable carban-
ion;[5,11,49,52,59,60] and analogous phosphine and phosphine
oxide products are formed in each case. Furthermore, the alco-
holysis of 9 is second-order in alkoxide and third overall,[3] simi-
lar to the order observed for phosphonium salt hydrolysis.[2–4,14]
These common features are shared with phosphonium salt
and ylide hydrolysis, and since analogous intermediates and
products are formed in each process, there is a high likelihood
that closely related mechanisms operate in each. Importantly,
both hydrolysis of ylides and alcoholysis of phosphonium salts
and ylides give racemic product from enantiopure starting ma-
terial.
Scheme 4. Alcoholysis reaction of scalemic phosphonium salt 11 to give rac-
emic phosphine oxide 12.
Grayson and Keough found butyltriphenylphosphonium and
tetrabutylphosphonium bromide to be inert to treatment with
sodium alkoxide or in refluxing alcohol solvents.[49] All of the
congeners of [MenPhmP]Br (n=4Àm, mꢀ3) have been shown
to be inert to treatment with ethoxide in ethanol,[58] while
[MePh3P]I has been observed not to react with NaOMe in
methanol.[11] Thus, phosphonium salts that are closely related
to those studied by Schmidbaur and co-workers and by Hey
and Ingold do not undergo phosphorane formation or ether
formation under the conditions of Grayson and Keough. o-
Chlorophenyltriphenylphosphonium iodide underwent rapid
alcoholysis when treated with NaOMe in refluxing methanol,
with the expulsion of chlorobenzene (this was deuterated in
the 2-position if the reaction was conducted in MeOD).[59] Al-
kyltri(fur-2-yl)phosphonium and benzyltri(fur-2-yl)phosphonium
salts (and their thiophen-2-yl analogues) expel a (protonated)
heteroarene during hydrolysis and alcoholysis.[11]
Almost no examples of alcoholysis of stabilised ylides can be
found in the literature; however, there do exist two examples
in which various alcohols are reacted with cyanomethylidene-
trialkylphosphorane to give a proposed P-alkoxy-P-cyanome-
thylphosphorane intermediate.[60] The phosphorane is pro-
posed to expel MeCN (the anion of which is protonated on ex-
pulsion) to give alkoxyphosphonium salt, which can be reacted
with nucleophiles to give, for example, thioethers or amines.
3. The mechanism of ylide alcoholysis
The existing consensus on the mechanism of phosphonium
ylide alcoholysis[3,5,15,49,52,59] (Scheme 5) is that the first step is
Scheme 5. The currently accepted mechanism(s) for alcoholysis of a phos-
phonium ylide.
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9143
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Our first inkling that the mechanisms discussed above, or as-
pects thereof, for ylide hydrolysis and alcoholysis might not
operate in aprotic organic solvents (or perhaps not at all for
some types of ylide) came from the simple observation that
the pKa of water in DMSO is 31,[63] much higher than that of
even the most basic ylides (e.g., pKa =22 for deprotonation of
MePh3P+),[64] a situation which is likely to be replicated in
other organic media such as (dry) THF or acetonitrile. Similar
remarks apply to alcohols in organic solvents—for example,
potassium tert-butoxide (pKa in DMSO=32[63]) is routinely used
to irreversibly deprotonate phosphonium salts (even the most
basic ones) to give ylides for use in Wittig reactions,[65,66] while
Wittig reactions of stabilised ylides can even be carried out in
alcohol solvents.[67] Furthermore, subsequent to the work of
Grayson and Keough, it has been established that the pKa of
ethanol in water is 15.9,[68] while that of the p-nitro phosphoni-
um salt 9 used in their study is 11.0 in DMSO,[69] a value that is
likely to be lower, if anything, in protic solvents.[70,71] Conse-
quently, protonation of the benzylide (10) derived from the p-
nitro salt may not be possible even in polar protic solvents.
Indeed, in several of the studies cited above, the authors
report the formation of coloured solutions that they attributed
to the presence of ylide (the colour of which in some cases
persisted for hours in refluxing alcohol, even as products
formed).[15,49,72] Furthermore, the observation of Aksnes and
Songstad that the hydrolysis of 9 is first order in each of phos-
phonium salt and hydroxide (cf. hydrolysis of 16; 3rd order
overall) may indicate that the reaction proceeds by initial for-
mation of ylide, and hence occurs by a different mechanism to
Scheme 1.[3,73]
Results
In our previous investigations on the mechanism of the Wittig
reaction,[74,75] we had devoted significant effort to the NMR ob-
servation of oxaphosphetanes (OPAs, see 18 in Scheme 7).
Given the structural similarity of OPAs to alkoxyphosphoranes
(which differ from OPAs only in the absence of a “tether” be-
tween the ylide a-carbon and the alcohol hydroxyl carbon)
and hydroxyphosphoranes, we became interested in the latter
entities, as we supposed that spectroscopic observation of
them might shed some light on the mechanisms of ylide alco-
holysis and hydrolysis.[18]
Scheme 7. Oxaphosphetanes 18 (DBP: dibenzophosphole ring spanning
axial and equatorial positions, cf. 1, 2, 19, 20), alkoxyphosphoranes 19–24,
and ylide 25.
We have found in the course of this study that benzylides
16 and 17 (Scheme 6) are protonated in dry CD3OD solvent in
dry methanol, but not by CD3OD (or CH3OH) in [D8]THF solvent
(vide infra). 16 has a pKaH of 17.4 in DMSO,[64] substantially
higher than that of the p-nitro ylide (10). Therefore, in organic
media containing only small amounts of water or alcohol,
ylides cannot deprotonate water or alcohol to produce phos-
phonium hydroxide or alkoxide, respectively, and so the first
step of the currently accepted mechanism of ylide hydrolysis/
alcoholysis (ylide protonation) is an impossible process in
aprotic organic solvents. Even in aqueous or alcohol solvents,
the pKa of at least some benzylides may be too low for proto-
nation by alcohol (i.e., phosphonium alkoxide formation) to
occur. Indeed, in our recent report on the observation and
characterisation of a P-hydroxytetraorganophosphorane inter-
mediate (1) in the hydrolysis of ylide 2, it was notable that the
phosphorane could be seen to co-exist with ylide and water,
apparently in the absence of any phosphonium hydroxide.[18]
1. Reactions of non-stabilised ylides with alcohols
After circumventing problems arising from the sensitivity of
non-stabilised ylides (present over extended periods in an
equilibrium) to oxidation and hydrolysis,[76,77] we were able to
generate alkoxyphosphoranes 19, 20 (from ylide 2) and 21–24
1
(from ylide 25), and characterise them by H, 31P, 13C, gCOSY,
gHSQC and gHMBC NMR spectroscopy. This is the first time
that two dimensional NMR techniques have been applied to
the study of these types of compounds. The use of NMR tech-
niques that were previously not available in studies of alkoxy-
phosphoranes has allowed us to access information that is
highly significant for the mechanism of ylide alcoholysis, and,
by extension, for the closely related mechanism of ylide hy-
1
drolysis. Selected 31P, H and 13C NMR chemical shifts and cou-
pling constants that carry illustrative physical information
about phosphoranes 19–24 are collected in Table 1.
The upfield 31P chemical shifts found for each reaction are
indicative of an electron-rich, pentacoordinate environment at
phosphorus, consistent with the formation of alkoxyphosphor-
ane. In the case of alkoxyphosphoranes 21–23 the signals are
rather broad (width of ca. 5 ppm in the spectrometer used;
see Figure 1a for the case of 21).
Scheme 6. Benylides 16 and 17.
The 1H NMR spectra of 21–23 (see Table 1 for individual
1
chemical shifts, and Figure 1b for the H NMR spectrum of 21)
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9144
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Table 1. Selected NMR data for phosphoranes 19–24.
#
dP [ppm]
dC(PCH2)
1JPC
dH(PCCH3)
3JPH
([%] by integration)
[ppm]
[Hz]
[ppm]
[Hz]
19[a]
20
À72.8 (89%)
À68.6 (80%)[b]
À52.6 (95%)
À66 (93%)
À51 (7%)
À64 (85%)
À43 (broad)
8.7[e]
25.9
–
29.6
33.1
112
–
115
111
0.86
–
1.32
1.20
23.5
–
20.2
20.8
21[a]
22[c]
23[c]
24[d]
35.4
–
119
–
1.34
1.56
21.1
19.3
11.7[f]
[a] Reaction solvent: [D8]THF. [b] Reaction solvent THF. Alcohol=CH3OD;
ylide present in excess. [c] Reaction solvent: [D8]toluene. [d] Reaction sol-
vent: mixed THF/[D8]THF. [e] Two equivalents of (À)-menthol used, broad
signal. [f] One equivalent of (À)-menthol used, broad signal.
generally show one broad signal of variable chemical shift (the
variability is even between different runs of a given reaction,
and appears to be concentration dependent), which we attri-
bute to the alcohol hydroxyl proton. The relative integration of
this signal indicates that there is no contribution from the
ylide PCH or phosphorane PCH2, and thus that there is no
signal present for these protons. The P-ethyl CH3 signal is gen-
erally a sharp doublet (see Table 1 column 6 for coupling con-
1
stants), showing only coupling to phosphorus (i.e., no H–1H
coupling to this signal is observed in either the 1D spectrum
or the gCOSY).[78] This signal collapses to a singlet under broad-
band decoupling from phosphorus.[79] The signals of the P-
alkoxy group of each phosphorane appear to be coincident
with those of the alcohol, which is typically present in excess.
However, in the reaction of 25 with MeOH to give 22, separate
signals could be observed for the protons of the methanol CH3
moiety and the methoxyphosphorane PÀOÀCH3 moiety (sepa-
rate 13C signals could also be observed for these groups, vide
infra).
In the 13C NMR spectra of 21–23 (see Table 1 for individual
chemical shifts, and Figure 1c for parts of the 13C NMR spec-
trum of 21), a doublet is generally observed in the neighbour-
1
hood of dC =30 ppm, with JPC between 110 and 120 Hz (see
Figure 1. a) 31P NMR, b) 1H NMR, and c) partial 13C NMR spectra of isopropox-
yphosphorane 21 from the reaction of isopropanol with ylide 25.
Table 1 column 4 for coupling constants). This signal shows no
cross-peaks in the gHSQC spectrum (see the Supporting Infor-
mation), but couples to the PÀCÀCH3 protons according to the
gHMBC spectrum (see the gHMBC spectrum of 21 in Figure 2,
and examples from other phosphoranes in the Supporting In-
chemical shift to move to 8.7 ppm (compare Figures S10a and
1
1
formation). The magnitudes of the JPC values show clearly that
b in the Supporting Information). The H NMR spectrum, how-
1
the ethyl moiety in each of 21–23 resides primarily in an equa-
ever, showed the same characteristics that are seen in the H
torial position in the phosphorus-centred trigonal bipyramid.[80]
spectrum of 21–23, that is, a broad signal for the alcohol OH,
a doublet only for the P-ethyl CH3 at d=1.56 ppm,[78] and no
signals for ylide PCH or phosphorane PCH2. No signal could be
detected for the phosphorane PCH2 carbon in the 13C NMR
spectrum of the reaction mixture. Neither could any signals in-
dicating the presence of this moiety be observed in two-di-
1
As in the H NMR spectrum, the signals of the phosphorane P-
alkoxy group and alcohol appear to be coincident. However,
separate 13C signals could be observed for the methyl carbons
of methanol and methoxyphosphorane 22.
The reaction of 25 with (À)-menthol in mixed THF/[D8]THF
gave a 31P NMR spectrum that is somewhat different to those
of 21–23. A broad signal was observed at dP =11.7 ppm after
the addition of one equivalent of the alcohol. The addition of
a second equivalent of the menthol solution caused the 31P
1
mensional H–13C NMR experiments (ASAPHSQC and gHMBC).
A broad and very small doublet at 7.5 ppm in the 13C NMR
which is shown by the gHMBC spectrum to couple to the P-
ethyl CH3 doublet may indicate the presence of ylide 25.[81] The
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9145
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
tent with the observations of Eyles and Trippett on the reac-
tions of n-hexyltriphenylphosphonium bromide with alkoxides
of varying steric bulk (vide supra).[45]
The ylide/phosphorane equilibration process leads to broad-
ening or averaging of the PÀCH2 signals in the 31P, 1H and
13C NMR spectra of 21–24, meaning that discrete chemical
shifts and coupling constants containing valuable physical in-
2
formation either could not be readily observed (e.g., JPC for
1
the PÀOÀC moiety, JPC for the P-ethyl moiety, vide infra) or
simply did not exist (e.g., the P-ethyl group CH2 signal in the
3
1H NMR spectrum, and hence the value of JHH for that moiety).
In an attempt to circumvent this limitation, we decided to gen-
erate alkoxyphosphoranes 19 and 20 from ylide 2 to investi-
gate if this phosphorane would behave differently to uncon-
strained analogues 21–24 (see Table 1, row 1 for chemical
shifts and coupling constants from the 31P, 1H and 13C NMR
spectra of 19, and row 2 for 31P NMR chemical shift of 20). In
doing so, we hoped to take advantage of the known effect by
which constraining two of the phosphorus substituents in
a five-membered ring dramatically affects the rates of reactions
of compounds containing pentacoordinate phosphorus.[67,74,82]
Gratifyingly, one or more sharp peaks were observed in the
high field region of the 31P NMR spectra of each of 19 (dP =
À74.4 ppm, see Figure 3a) and 20 (two almost coincident
peaks for separate pseudorotamers at dP =À68.6 ppm). Since
there is also a signal at dP =À11.6 ppm for the ylide (2), these
phosphoranes undergo only slow or non-existent reversion to
ylide+alcohol on the NMR timescale.
Figure 2. Close up on a region of the gHMBC spectrum of 21 showing the
coupling between PÀCH2 and PÀCÀCH3.
31P NMR chemical shift of the adduct of the reaction of 25 and
(À)-menthol is not by itself directly indicative of a phosphorane
species. However, based on the broadness of the 31P NMR
signal (similar to 21–23), the difference in the chemical shift
(ca. 12 ppm) compared to the 31P shift of ylide 12 in the ab-
sence of (À)-menthol, and the similarity of the features of the
1H NMR spectrum to those of 21–23, we conclude that there is
menthoxyphosphorane, 24, present in the reaction mixture.
Across the spectra of 21–24, the broadness of the 31P NMR
signals, the absence of a discrete signal for the phosphorane
1
a-protons in the H NMR spectrum, the absence of coupling to
1
the P-CH2 protons in the H NMR, gCOSY, gHSQC and gHMBC,
A discrete signal for the PÀCH2 protons of phosphorane 19
is present at d=2.52 ppm, and the 1H–1H coupling of this
group to the vicinal CH3 protons is detected in the 1H NMR
(3JHH =7.7 Hz; see Figure 3b) and also in the COSY spectrum
(Figure S1c in the Supporting Information). In all the other ex-
amples given above (21–24), no discrete signal for the PÀCH2
and the absence of a two-bond coupling constant between
phosphorus and the oxygen-bearing alkoxy carbon is consis-
tent with the existence of an equilibrium in which the alkoxy-
phosphorane undergoes rapid exchange with ylide+alcohol.
Similar exchange phenomena were observed by Schmidbaur
and co-workers in their studies of alkoxyphosphoranes.[19] Re-
version of the alkoxyphosphorane to ylide+alcohol can in-
volve transfer of either of the a-protons to the departing alk-
oxide, hence both are subject to the rapid exchange process.
Entry of the alcohol to the trigonal bipyramidal alkoxyphos-
phorane should occur along a trajectory to place the alkoxy
group in an apical position.[82] In addition, the apicophilicity of
the alkoxy oxygen is such that this group is highly likely to
occupy an apical position in all stable (or metastable) phos-
phoranes,[82,83] since in general electronegative elements favour
being positioned in apical sites in hypervalent com-
pounds.[82–86] The fact that the P-ethyl group appears to be in
an equatorial position in each alkoxyphosphorane means that
it is ideally placed to swap protons with alcohol undergoing
apical entry and departure from the trigonal bipyramidal spe-
cies. In 24, the steric bulk in the vicinity of the hydroxyl group
of menthol may disfavour phosphorane formation with this al-
cohol compared with reactions of other alcohols, meaning that
the ylide remains the predominant form in the equilibrium (as
reflected by the relatively low field 31P chemical shift). Indeed,
the reaction mixture in question retains the vibrant red/orange
colour of the ylide, which is dissipated in the reactions of the
other alcohols detailed above. This interpretation is also consis-
1
is observed, and only H–31P coupling is observed for the PÀCÀ
CH3 protons. Furthermore, the gHMBC spectrum of 19 indi-
cates coupling between the methylene carbon and methyl
protons, and between the methyl carbon and methylene pro-
tons (see Figure 3c; these signals were identified using the
gHQSC spectrum in the Supporting Information). For phos-
phoranes 21–24 (vide supra), the gHMBC shows that no cou-
pling exists to the P-methylene protons.
Phase-sensitive gHSQC allowed assignment of the 1H and
13C signals of the PÀOÀiPr (Figure 4a) and PÀCH2CH3 (Fig-
2
ure 4b) moieties of 19 to be made. Hence, JPC for coupling be-
tween phosphorus and the secondary isopropoxy carbon was
established as 9.5 Hz (Figure 4c), establishing unequivocally
that the structures produced in these reactions are indeed al-
koxyphosphoranes, since there is clear physical evidence of
bonding between the isopropoxy unit and the phosphorus. In
addition, although the PÀCH2 signal in the 13C NMR spectrum
overlaps with one of the signals of the [D8]THF solvent (Fig-
ure 4c), and is also almost coincident with a signal of isopropa-
nol, we can say with reasonable confidence based on the
1
gHSQC spectrum (Figure 4b) that the value of JPC is about
112 Hz, indicating that the P-ethyl group of 19 occupies an
equatorial position. The additional structural data obtained by
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9146
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 4. a) Close-up of gHSQC spectrum of 19 showing one-bond coupling
for PÀOÀCH2 unit. b) Close-up of gHSQC spectrum of 19 showing one-bond
coupling for PÀCH2 unit. c) Close-up of gHMBC spectrum of 19 showing the
connectivity in the PÀCH2ÀCH3 moiety.
Figure 3. a) 31P NMR, b) 1H NMR, c) partial 13C NMR of phosphorane 19 from
the reaction of isopropanol with ylide 2, showing doublets for the P-isopro-
poxy CH (dC =64.2 ppm) and the PÀCH2 carbon (doublet centred at
dP =25.9 ppm, partially obscured by [D8]THF signal). Note, the small signal at
d=25.8 ppm is due to non-deuterated THF.[87]
peared (giving EtPh2PO and benzene), and a substantial
amount of diisopropyl ether had been produced.[88] In this
case, it is possible that phosphorane decomposition could
occur by an SN1-type process, ultimately resulting in ether for-
mation. Such an occurrence would be highly unlikely for phos-
phoranes derived from primary alcohols (although an SN2 pro-
cess may occur at higher temperatures based on the work of
Schmidbaur and co-workers[19]). We surmise that a similar oc-
currence (i.e., production of dimenthyl ether) does not occur
at an appreciable rate in the reaction of 25 and (À)-menthol
because the concentration of menthoxyphosphorane 24 is so
low. That acyclic 21 undergoes decomposition comparatively
quickly at 208C while 19 does not is not altogether surprising
in light of the relative stability of analogous pentavalent diben-
spectroscopic study of 19, which is simply not available from
NMR studies of unconstrained phosphoranes such as 21–24, is
in complete agreement with our interpretation of the structure
and behaviour of each of the phosphoranes 21–24.
The alkoxyphosphoranes presented in Scheme 7 appear to
be stable indefinitely at 208C under an inert atmosphere
based on repeated NMR observations of the samples that
yielded the data given above, with one exception: when the
reaction of 25 and iPrOH to give 21 was left to stand for two
days at 208C, the phosphorane was observed to have disap-
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9147
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
zophosphole-derived compounds such as 1 and 18b com-
pared to their unconstrained equivalents.[18,65,74]
stable intermediates in those reactions. We further hoped to
establish if benzylphosphonium alkoxide salts are formed in
the process of alcoholysis of semi-stabilised benzylides, as is
suggested by the existing literature on the topic.[3,5,15,49,52,59]
To these ends, salt-free benzylide 16 was generated in dry
[D8]THF under argon and characterised by 1H and 31P NMR
spectroscopy (Figure 5). One equivalent of dry CD3OD was
then added to the ylide solution (under argon). Since the
parent phosphonium salt of 16 is almost completely insoluble
in THF, the solution was diluted to five times its original
volume with dry CD3CN, in which the phosphonium salt is
Addition of water to reaction mixtures containing 22–24
(derived from ylide 25) yields EtPh2PO exclusively (or nearly
so), the product expected in the hydrolysis of 25. Addition of
water to 19 yields two products, 26 and 27 (see Scheme 8; for-
mation of 27 predominates strongly, exactly as has been ob-
served in the hydrolysis of ylide 2[18]). Examination of the
1H NMR spectrum of the crude hydrolysis product formed from
the reaction of 25 and (À)-menthol shows that the alcohol
product is (À)-menthol (dH =3.3 ppm[89]). This shows that the
addition of water to the ylide/phosphorane mixture results in
the hydrolysis of the ylide exclusively, whereas if the phosphor-
ane (24) were hydrolysed, one would expect to see evidence
of the formation of (+)-neomenthol (dH =4.10 ppm in CDCl3)[90]
or 2-menthene (dH =5.52 ppm in CDCl3;[91] see Scheme 9).[92]
The starting alcohol is also the only observed non-phosphorus-
containing product when water is added to reactions involving
19, 22 and 23 (demonstrated by NMR spectroscopy and, in the
case of 23, GC), which we again interpret to be a consequence
of hydrolysis of the starting ylide.
1
readily soluble. H and 31P NMR spectroscopy of the resulting
solution indicated that the ylide was unchanged, that is, nei-
ther alkoxyphosphorane formation nor deuteration of the ylide
had occurred (see 31P NMR spectrum in Figure 5, middle spec-
trum).[93] Identical results were obtained in similar experiments
in which benzylides 16 and 17 were generated in dry CD3CN
under argon and treated with one equivalent of dry CD3OD or
methanol. This confirms that the pKa of methanol is too high
in aprotic organic media for a single equivalent to protonate
semi-stabilised ylides.
Scheme 8. Phosphine oxides 26 and 27.
Figure 5. Stacked 31P NMR spectra of experiments involving salt-free benzy-
lide 16. Bottom spectrum: ylide (dP =7.9 ppm) generated in [D8]THF and di-
luted with CD3CN, containing a very small amount of the parent phosphoni-
um salt (dP =25.6 ppm; not visible in this figure; integrates for 1% of ylide
signal). Middle spectrum: solution of ylide in [D8]THF after addition of
1 equivalent of CD3OD and subsequent dilution with CD3CN, parent phos-
phonium salt is present and integrates for 1% of the ylide signal (i.e., identi-
cal to the control sample in the bottom spectrum). Top spectrum: formation
of parent phosphonium salt of 16 after addition of CD3CN solution of 16 to
excess dry CD3OD. A very small signal phosphine oxide can be seen at low
field, formed by reaction of the ylide with adventitious water.
Scheme 9. Possible mechanisms for hydrolysis of menthoxyphosphorane
(neither of which is observed; see the main text).
Of course, many of the experiments reported in the litera-
ture for alcoholysis of benzylides (summarised in section 2 of
the Introduction) were carried out not in polar aprotic solvents,
but in alcohol solvents in which the pKa of the hydroxyl hydro-
gen is much lower. Generation of salt-free benzylides 16 and
17 (in CD3CN and THF, respectively) under an argon atmos-
phere and subsequent addition to a large excess of dry CD3OD
solvent resulted in immediate dissipation of the orange colour
2. Reactions of semi-stabilised and stabilised ylides with
alcohols
Having observed alkoxyphosphoranes by NMR spectroscopy
that had been produced from non-stabilised ylides, we set
about an attempt to determine if similar species are formed in
reactions of semi-stabilised ylides, and if they could be ob-
served spectroscopically. The results of Grayson and Keough[49]
strongly indicate that an alkoxyphosphorane-type species is in-
volved at some point on the reaction coordinate in the reac-
tions of Scheme 9, but these species need not necessarily be
1
of the ylide. H and 31P NMR spectroscopic characterisation of
the adducts indicated that the products were the parent ben-
zylphosphonium trideuteromethoxide salts (Figure 5, top spec-
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9148
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
trum; parent phosphonium salt of 16 has dP =23.9 ppm in
CD3OD, while the parent salt of 17 has dP =18.0 ppm in
CD3OD).[94] We surmise that the pKa of the hydroxyl hydrogen/
deuterium of the alcohol becomes sufficiently low in alcohol
solvent that protonation/deuteration of benzylides (and there-
fore, by implication, non-stabilised ylides) can occur. No alkoxy-
phosphorane was observed in any of these experiments, either
in protic or aprotic media.
The solution of the phosphonium trideuteromethoxide salt
derived from 17 in CD3OD (plus a small amount of CD3CN or
THF) was heated in an oil bath (under nitrogen atmosphere,
bath temperature 1208C), resulting in the gradual formation of
methyldiphenylphosphine oxide (dP =30.0 ppm),[95] according
to 31P NMR analysis of the reaction mixture. The formation of
phosphine oxide under these conditions, which is consistent
with the observations of Grayson and Keough,[49] presumably
occurs through a transient alkoxyphosphorane or similar pen-
tacoordinate species, which may not be a stationary point on
the reaction coordinate.[83]
Scheme 11. Summary of results in ylide alcoholysis and hydrolysis. R1 =H,
Me, iPr, CH2Ph.
We also attempted to observe an alkoxyphosphorane de-
rived from a stabilised ylide, 28 (Scheme 10), using the same
method as was applied for benzylides 16 and 17 and non-sta-
bilised ylide 25. Neither the addition of one equivalent of
CH3OD nor of a large excess of the alcohol to a CD3CN solution
of 28 caused either protonation of this ylide or phosphorane
formation—the spectra of the reaction mixture after alcohol
addition merely indicated the continued presence of acetony-
lide.[96] The fact that 28 is not protonated by methanol is of
little surprise given that Wittig reactions of stabilised ylides can
be carried out in alcohol solvents,[67] and that the pKa of related
acetonylide 29 is 7.1.[64] To prove that the ylide was indeed still
present in the above reaction mixture, benzaldehyde was
added, resulting in the formation of 4-phenylbut-3-en-2-one
and methyldiphenylphosphine oxide by Wittig reaction.
the most basic phosphonium ylides (exemplified by 2 and 25)
to occur. We have also definitively established that protonation
of benzylides 16 and 17 by alcohol does not occur in aprotic
organic media (i.e., when ca. one equivalent of alcohol is
used). Given that the pKa of water is at least as high as that of
alcohol in the same media, it is reasonable to conclude that
water also cannot protonate phosphonium ylides in aprotic or-
ganic media.
Notwithstanding this impossible protonation, hydroxyphos-
phorane 1[18] and alkoxyphosphoranes 19 and 20 are evidently
formed in the reactions of ylide 2 with water, isopropanol and
methanol, respectively (Scheme 11), while in the reactions of
non-stabilised ylide 25 with alcohol, the ylide and alcohol are
in rapid equilibrium with alkoxyphosphorane (21–24; see
Figure 3 and Figure 5 and associated discussion, and the work
of Schmidbaur and co-workers for further examples).[19] Fur-
thermore, each of 2, 17 and 25 undergo exceedingly rapid hy-
drolysis when small amounts of water (2 equivalents or less)
are added to solutions of these ylides in dry aprotic organic
media.
Scheme 10. Stabilised ylides 28 and 29.
Phosphonium alkoxides/hydroxides are not necessary inter-
mediates in the processes of formation of alkoxyphosphor-
anes/hydroxyphosphoranes from phosphonium ylides+alkox-
ide/hydroxide. The available evidence indicates that protona-
tion of ylides by alcohols or water in polar aprotic organic sol-
vents does not occur. Therefore, we propose an alternative ex-
planation that accounts for the results shown here and for the
observations described in the introduction which does not re-
quire the initial protonation of ylide by water or alcohol, but
remains consistent with the other observations: The first step
involves addition of an OÀH bond of the water or alcohol
Discussion
Protonation of benzylides 16 and 17 in solvent containing
a high proportion of methanol or CD3OD results in the produc-
tion of phosphonium trideuteromethoxide (Scheme 11). Thus,
alcoholysis of alkylides and benzylides in alcohol solvent may
occur, in at least some circumstances, by the existing mecha-
nism involving phosphonium alkoxide as an intermediate
(Scheme 5). By analogy, presumably benzylide hydrolysis in
media in which the pKa of water is lower than that of the ben-
zylide (i.e., media containing a substantial amount of water or
alcohol) proceeds through phosphonium hydroxide (mecha-
nism of Scheme 1).
across the ylide P=C bond in
a
concerted manner
(Scheme 12a), resembling in certain aspects the mechanisms
of other concerted reactions such as the Alder–Ene reaction[97]
and the Wittig reaction.[67,74,75] In this mechanism, the ylide
carbon could be considered to be acting as an internal general
base. In the trigonal bipyramidal alkoxyphosphorane or hy-
However, based on the results presented above, we con-
clude that the pKa values of the hydroxyl groups of alcohols
are too high in organic media for direct protonation of even
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9149
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
droxyphosphorane intermediate, the oxygen would occupy an
apical position.[82,83,85,86] The ethyl moiety derived from the
carbanion appears to be in an equatorial position, at least ini-
tially, based on the magnitudes of the observed one bond PÀC
coupling constants (1JPC > ca. 100 Hz), and also by analogy
with pentacoordinate spirophosphoranes.[98]
concerted four-centre apical–apical elimination of benzene
from 30a (Scheme 12c), perhaps through a square pyramidal
transition state, 30b, with the phenyl and hydroxyl groups in
basal positions (and trans relative to each other), reminiscent
of the transition state in Berry pseudorotation, or 2) by con-
certed four-centre apical–equatorial elimination from pseudor-
otamer 30c, which itself would probably be a transition
state.[83] One way or another it seems unlikely that a second
equivalent of water is necessary since no means of regenerat-
ing hydroxide is needed (cf. phosphonium salt hydrolysis).
As mentioned above, Schnell and Tebby observed that an
enantiopure chiral benzylide is racemised during hydrolysis,
while the parent benzyl phosphonium salt is hydrolysed ste-
reospecifically.[40] McEwen and co-workers observed that chiral
benzyl phosphonium salt 11 was racemised during alcoholysis
(using one equivalent of butoxide).[52] In ylide hydrolysis and al-
coholysis (and indeed in phosphonium salt alcoholysis reac-
tions), once the phosphorane intermediate is produced, there
is not necessarily any remaining external source of nucleophile
or base (hydroxide or alkoxide). Thus, the lifetime of the phos-
phorane intermediate may be sufficiently long for pseudorota-
tion to occur, giving rise to racemic product from chiral start-
ing phosphonium ylide or salt.[101–103] The case is markedly dif-
ferent in phosphonium salt hydrolysis, in which at least
a second equivalent of hydroxide is present. Racemisation
could also occur via dialkoxyphosphorane or dihydroxyphos-
phorane intermediates, but we consider it to be unlikely that
this is the case. In particular, in the case of ylide hydrolysis, rac-
emisation via dihydroxyphosphorane would require the forma-
tion of hydroxyphosphonium salt in a non-acidic medium.
Conclusions
Scheme 12. a) Mechanism for exchange process between alcohol+ylide and
alkoxyphosphorane in aprotic organic media. b) Possible mechanism for the
breakdown of isopropoxyphosphorane 21. c) Possible mechanisms for hy-
drolysis of ylide 25.
In summary, we have proposed and provided evidence for
a new mechanism for the first step of phosphonium ylide alco-
holysis and hydrolysis (Scheme 13) which applies at least to re-
actions in aprotic organic media, but may also be applicable in
all media to reactions of ylides derived from particularly acidic
phosphonium salts (pKa <14) for example, stabilised ylides and
certain semi-stabilised ylides. In certain examples of phospho-
nium salt hydrolysis or alcoholysis involving particularly acidic
phosphonium salts (e.g., p-nitrobenzyl salt 9), it may even be
the case that the generation of alkoxyphosphoranes from
phosphonium salts and sodium alkoxide goes by deprotona-
tion of the salt to give ylide, which then undergoes concerted
addition of the alcohol OÀH bond across the P=C bond—that
is, that this reaction goes from phosphonium salt to ylide and
not the other way around! This may explain why no members
of the [RnPhmP]+ family (n=4Àm, mꢀ3) undergo alcoholysis
in alcohol solvent,[11,49,58] since they cannot be deprotonated
by alkoxide in this medium. The mechanisms of hydrolysis and
alcoholysis of non-stabilised ylides and benzylides derived
from relatively non-acidic phosphonium salts in polar protic
media (solvents in which the pKa values of alcohols and water
are as low as possible) are unaffected by these considerations
and are as proposed previously.[4,49] Additional studies focusing
on the reactivity and synthetic utility of the hydroxy and alkox-
The pathway followed during decomposition of alkoxyphos-
phorane intermediates generally depends on the conditions
employed (vide supra, section 2 of Introduction). Breakdown of
isopropoxyphosphorane 21 appears to occur by an SN1 mecha-
nism at 208C (Scheme 12b) based on the formation of diiso-
propyl ether in this reaction and that phosphoranes 22 and 23
derived from primary alcohols are stable at room temperature
under the same conditions (in the absence of water). Analo-
gous decomposition products to those observed in the de-
composition of 21 have been observed previously in reactions
involving alkoxybenzylphosphoranes derived from primary al-
cohols conducted at high temperatures or in the presence of
alkoxide at high temperatures (vide supra).[49–53] It is likely that
in these cases, decomposition of alkoxyphosphoranes derived
from primary alcohols occurs by an SN2 mechanism.
It is unclear whether breakdown of the hydroxyphosphorane
produced in ylide hydrolysis is intramolecular[99] or intermolec-
ular. Assuming that equatorial departure of the P-phenyl group
from a trigonal bipyramid cannot happen,[82,100] intramolecular
decomposition of hydroxyphosphorane can occur either: 1) by
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9150
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
1H NMR (500 MHz, [D8]THF): d=8.01 (d, J=7.8 Hz, 1H), 7.71–7.66
(m, 1H), 7.48–7.39 (m, 3H), 7.34–7.25 (m, 2H), 7.25–7.12 (m, 3H),
3.89–3.71 (m, 3.3H, contains CHOH (2.3H) and CHOP (1H) signals),
2.58–2.43 (m, 2H, CH2P), 0.86 ppm (dt, J=23.5, 7.7 Hz, 3H,
CH3CH2P). Several signals were broadened to such an extent that
unambiguous assignment and accurate integration were not possi-
ble. These signals were: d=7.91–7.77 (m), 7.66–7.52 (m), 7.11–6.90,
1.20–0.90 ppm.
13C NMR (101 MHz, [D8]THF): d=145.0 (d, J=2.8 Hz), 143.9 (d, J=
3.8 Hz), 140.5 (d, J=142.1 Hz), 133.6 (d, J=20.5 Hz), 131.5 (d, J=
22.3 Hz), 130.2 (s), 129.7 (s), 129.7 (d, J=7.8 Hz), 129.1 (d, J=
14.6 Hz), 128.6 (d, J=7.3 Hz), 128.3 (d, J=3.0 Hz), 125.9 (d, J=
9.4 Hz), 122.6 (s), 64.5 (d, J=9.4 Hz, CH2OP), 26.2 (s, (CH3)2CHOP, by
[D8]THF signal and identified from 2-dimensional spectra), 25.9 (d,
PCH2, obscured by [D8]THF signal and identified from 2-dimension-
al spectra), 7.2 ppm (d, J=5.4 Hz).
Scheme 13. Summary of mechanisms of ylide alcoholysis and hydrolysis in
different solvents. In aprotic organic media (in which the pKa of the ylide is
lower than that of water/alcohol), the first step is concerted addition of ylide
and water/alcohol to give phosphorane. In solvents in which the pKa of
water/alcohol is lower than that of the ylide (generally protic media), phos-
phorane is formed in a stepwise fashion through phosphonium alkoxide or
hydroxide.
yphosphoranes discussed here are underway and will be re-
ported presently.
NMR signals assigned to isopropanol:
1H NMR (500 MHz, [D8]THF): d=3.89–3.71 (m, 3.3H, contains CHOH
(2.3H) and CHOP (1H) signals), 3.36 (d, J=4.2 Hz, 2.3H), 1.08 ppm
(t, J=6.3 Hz, signal partially obscured by broad signal).
13C NMR (101 MHz, [D8]THF): d=63.9 (s, (CH3)2CHOH), 26.3 ppm
(CHOH).
Experimental Section
General procedure for ylide alcoholysis
After obtaining NMR characterization for the phosphorane, H2O
(0.03 mL) was added to the reaction mixture in the NMR tube at
208C, resulting in formation of ylide hydrolysis products in a very
similar ratio to that observed in the hydrolysis of ylide 2.[18] Isopro-
panol (but no diisopropyl ether) was also observed to be present.
All glassware used for inert atmosphere operations was flame-
dried and cooled under vacuum. Phosphonium salt (1.0 equivalent)
and KHMDS (1.0 equivalent) were added to a flask in a glove box
under an atmosphere of dry argon. [D8]THF or [D8]toluene (1.0 mL)
was added, and the resulting ylide solution was stirred for 15 min.
Stirring was then stopped, and the KBr precipitate was allowed to
settle. The (brightly coloured) supernatant solution was carefully
withdrawn by syringe, and added to an NMR tube. The dry alcohol
(2–4 equivalents) was then added by one of two methods:
31P NMR (121 MHz, [D8]THF): d=42.8 (0.04P, phosphine oxide
EtDBPO, compound 26, lit.[105] dP =46.11 ppm (CDCl3)), 33.2 (1P,
phosphine oxide 27, lit.[18] dP ([D8]THF)=32.2 ppm), 30.7 (0.15P,
PhDBPO from ylide oxidation (lit. dP ([D8]dioxane)=30.0 ppm;[106]
dP (CDCl3)=33.5 ppm[30])), two signals at 22.9 (cumulative 0.1P),
À10.1 (PhDBP0, 0.1P, lit.[106] dP ([D8]dioxane)=À10.0 ppm).
1) Direct addition by syringe in the glove box, or
1H NMR (300 MHz, [D8]THF): d=3.92–3.73 (m, 1H, Me2CHOH),
1.10 ppm (d, J=6.2 Hz, (CH3)2CHOH).
2) The NMR tube was placed into a long Schlenk flask,[104] which
was then sealed with a greased stopper, removed from the
glove box and attached to a nitrogen supply through a nitro-
gen/vacuum manifold by the pump and fill technique. The al-
cohol could then be added to the NMR tube by syringe
through a rubber septum.
Acknowledgements
This work was supported by the Synthesis and Solid State
Pharmaceutical Centre (SSPC) and Science Foundation Ireland
(SFI) under grant number 12/RC/2275.
In either case, the NMR tube was sealed with a rubber septum
under inert atmosphere, and brought to the NMR spectrometer to
record spectra.
Keywords: alkoxyphosphorane
·
concerted additions
·
hypervalent compounds · reaction intermediates · ylides
Phosphorane 19 from ylide 17+isopropanol
The ylide was generated from P-ethyl-P-phenyl-5H-dibenzophos-
pholium bromide (31 mg, 0.080 mmol) and KHMDS (18 mg,
0.090 mmol) in [D8]THF (0.8 mL). To this was added a 3 molLÀ1 so-
lution of isopropanol in [D8]THF (0.1 mL, 0.3 mmol).
[1] A. Cahours, A. W. Hoffmann, Justus Liebigs Ann. Chem. 1857, 104, 1–
39.
[2] C. A. Vander Werf, W. E. McEwen, M. Zanger, J. Am. Chem. Soc. 1959,
81, 3806–3807.
[3] G. Asknes, J. Songstad, Acta. Chem. Scand. 1962, 16, 1426–1432.
[4] W. E. McEwen, G. Axelrad, M. Zanger, C. A. VanderWerf, J. Am. Chem.
Soc. 1965, 87, 3948–3952; J. G. Dawber, R. G. Skerratt, J. C. Tebby,
A. A. C. Waite, Phosphorus Sulfur Silicon Relat. Elem. 2012, 187, 1261;
NMR signals assigned to alkoxyphosphorane 19:
31P NMR (121 MHz, [D8]THF): d=À11.6 (0.15P), À74.4 (1.0P),
À78.9 ppm (0.01P).
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9151
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
W. E. McEwen, in Topics in Phosphorus Chemistry (Eds.: M. Grayson, E. J.
Griffith), Interscience, New York, 1965, Vol. 2, pp. 5–9.
ajan, J. W. Chang, H. P. Yennawar, R. M. Rioux, D. H. Ess, A. T. Radose-
vich, J. Am. Chem. Soc. 2014, 136, 4640–4650.
[5] A. W. Johnson, Ylides and Imines of Phosphorus, 1993, Wiley, New York,
chapter 5, pp. 129–151.
[27] H. Zhou, G.-X. Wang, W.-Z. Zhang, X.-B. Lu, ACS Catal. 2015, 5, 6773–
6779.
[6] R. R. Holmes, Pentacoordinated Phosphorus, Vol II, 1980, ACS, Washing-
ton DC, pp. 115–129.
[7] H. J. Cristau, F. PlØnat, in The Chemistry of Organophosphorus Com-
pounds, Vol. 3 (Ed.: F. R. Hartley), Wiley, Chichester, 1994, Chapter 2,
pp. 111–138.
[28] P. Wyatt, S. Warren, M. McPartlin, T. Woodroffe, J. Chem. Soc. Perkin
Trans. 1 2001, 279–297; P. Calcagno, B. M. Kariuki, S. J. Kitchin, J. M. A.
Robinson, D. Philp, K. D. M. Harris, Chem. Eur. J. 2000, 6, 2338–2349;
D. W. Allen, J. P. Mifflin, P. J. Skabara, J. Organomet. Chem. 2000, 601,
293–298.
[8] J. R. Corfield, S. Trippett, J. Chem. Soc. Chem. Commun. 1970, 1267a.
[9] G. W. Fenton, C. K. Ingold, J. Chem. Soc. 1929, 2342–2357; M. Schloss-
er, Angew. Chem. Int. Ed. Engl. 1962, 4, 287; Angew. Chem. 1962, 74,
291; D. W. Allen, B. G. Hutley, M. T. J. Mellor, J. Chem. Soc. Perkin Trans.
2 1974, 1690–1694.
[10] L. Hey, C. K. Ingold, J. Chem. Soc. 1933, 531–533.
[11] D. W. Allen, B. G. Huntley, M. T. J. Mellor, J. Chem. Soc. Perkin Trans. 2
1972, 63–67.
[29] D. Valentine, Jr., J. F. Blount, K. Toth, J. Org. Chem. 1980, 45, 3691–
3698.
[30] K. V. Rajendran, D. G. Gilheany, Chem. Commun. 2012, 48, 817–819.
[31] From our laboratory: P. A. Byrne, J. Muldoon, D. G. Gilheany, Org.
Biomol. Chem. 2012, 10, 3531–3537; N. P. Kenny, K. V. Rajendran, E. V.
Jennings, D. G. Gilheany, Chem. Eur. J. 2013, 19, 14210–14214; S. Al
Sulaimi, K. V. Rajendran, D. G. Gilheany, Eur. J. Org. Chem. 2015, 5959–
5965; see also reference [30].
[12] W. E. McEwen, K. F. Kumli, A. Blade-Font, M. Zanger, C. A. VanderWerf,
J. Am. Chem. Soc. 1964, 86, 2378–2384.
[13] N. J. De’Ath, S. Trippett, J. Chem. Soc. D 1969, 172–173; W. Hawes, S.
Trippett, Chem. Commun. (London) 1968, 295–296; K. L. Marsi, Chem.
Commun. (London) 1968, 846–847; R. L. Burwell, R. G. Pearson, J. Phys.
Chem. 1966, 70, 300–302.
[32] By others: D. HØrault, D. H. Nguyen, D. Nuel, G. Buono, Chem. Soc. Rev.
2015, 44, 2508–2528; Y. Li, S. Das, S. Zhou, K. Junge, M. Beller, J. Am.
Chem. Soc. 2012, 134, 9727–9732; Y. Li, L.-Q. Lu, S. Das, S. Pisiewicz, K.
Junge, M. Beller, J. Am. Chem. Soc. 2012, 134, 18325–18329; S. Sowa,
´
M. Stankevic, A. Szmigielska, H. Małuszynska, A. E. Kozioł, K. M. Pietru-
siewicz, J. Org. Chem. 2015, 80, 1672–1688.
[14] J. G. Dawber, J. C. Tebby, A. A. C. Waite, J. Chem. Soc. Perkin Trans. 2
1983, 1923–1925.
[33] Predominant or complete retention of configuration at phosphorus
can be observed if the phosphorus is in a small ring or bears a tert-
butyl group; see reference [13].
[34] F. Ramirez, S. Dershowitz, J. Org. Chem. 1957, 22, 41–45; A. W. John-
son, J. Org. Chem. 1959, 24, 282–284; F. Ramirez, O. P. Madan, C. P.
Smith, Tetrahedron Lett. 1965, 6, 201–205; G. Wittig, G. Laib, Justus Lie-
bigs Ann. Chem. 1953, 580, 57–68; A. W. Johnson, R. B. Lacount, Tetra-
hedron 1960, 16, 130–138; M. P. Cooke, J. Org. Chem. 1973, 38, 4082–
4084.
[15] B. Siegel, J. Am. Chem. Soc. 1979, 101, 2265–2268.
[16] K. Bergesen, Acta Chem. Scand. 1966, 20, 899–900.
[17] G. Aksnes, K. Bergesen, Acta Chem. Scand. 1965, 19, 931–934.
[18] P. A. Byrne, Y. Ortin, D. G. Gilheany, Chem. Commun. 2015, 51, 1147–
1150.
[19] H. Schmidbaur, H. Stühler, W. Buchner, Chem. Ber. 1973, 106, 1238–
1250; H. Schmidbaur, W. Buchner, D. Scheutzow, Chem. Ber. 1973, 106,
1251–1255; H. Schmidbaur, W. Buchner, F. H. Kçhler, J. Am. Chem. Soc.
1974, 96, 6208–6210; H. Schmidbaur, H. Stühler, Chem. Ber. 1974, 107,
1420–1427. In the final reference, it is suggested that in the presence
of excess alcohol, phosphonium cations with complex alkoxide/alco-
hol counter-ions may form.
[35] A. Schnell, J. G. Dawber, J. C. Tebby, J. Chem. Soc. Perkin Trans. 2 1976,
633–636; A. Schnell, J. C. Tebby, J. Chem. Soc. Chem. Commun. 1975,
134–135.
[36] Y. Segall, I. Granoth, J. Am. Chem. Soc. 1978, 100, 5130–5134; Y. Segall,
I. Granoth, J. Am. Chem. Soc. 1979, 101, 3687–3688; I. Granoth, J. C.
Martin J. Am. Chem. Soc. 1978, 100, 5229–5230; F. Ramirez, M. Nowa-
kowski, J. F. Maracek, J. Am. Chem. Soc. 1977, 99, 4515–4517.
[37] I. Granoth, J. C. Martin, J. Am. Chem. Soc. 1981, 103, 2711–2715. Alkox-
yhalophosphorane structures are assigned to the products of the reac-
tions of phosphinites with alkyl halides; these entities appear in fact
to be alkoxyphosphonium halides based on their low field 31P NMR
chemical shifts.
[38] I. Yavari, A. Alizadeh, Tetrahedron 2001, 57, 9873–9875. A hydroxy-
phosphorane structure is proposed for the product of the reaction of
arylsulfonylglycyl chlorides with Ph3P; however, as pointed out by
Chesnut and Quin (ref. [39]), on the basis of the chemical shift of d=
18.7 ppm in the 31P NMR spectrum, it seems likely that the entity in
question is in fact a phosphonium salt.
[39] D. B. Chesnut, L. D. Quin, Tetrahedron 2005, 61, 12343–12349.
[40] A. Schnell, J. C. Tebby, J. Chem. Soc. Perkin Trans. 1 1977, 1883–1886.
[41] D. G. Gilheany, P. G. Kelly, C. A. Mitchell, B. J. Walker, J. F. Malone, J.
Blaney, R. Wilson, Phos. Sulf. Sil. Rel. Elem. 1993, 75, 11–14.
[42] Eyles and Trippett reported that they had not been able to reproduce
the results of Hey and Ingold; see reference [45].
[43] An analogous set of products was obtained in the pyrolysis of tetra-n-
butylphosphonium acrylates, that is, phosphine oxide and the ketone
derived from addition of one of the P-butyl groups to the acyl group
of the acylate. This process is likely to have occurred via an acyloxy-
phosphorane species. See D. B. Denney, H. A. Kindsgrab, J. Org. Chem.
1963, 28, 1133. Note: phosphine and butyl ester were also produced;
the origin of these is likely to have been nucleophilic attack of the
acyloxy anion at C-1 of a P-butyl group.
[20] H. Schmidbaur, P. Holl, Chem. Ber. 1976, 109, 3151–3158. The product
in this case is a cyclic 1,2-oxaphospholane.
[21] H. J. Bestmann, K. Roth, R. W. Saalfrank, Angew. Chem. Int. Ed. Engl.
1977, 16, 877; Angew. Chem. 1977, 89, 915–916.
[22] Catalytic Wittig reactions: E. E. Coyle, B. J. Doonan, A. J. Holohan, K. A.
Walsh, F. Lavigne, E. H. Krenske, C. J. O’Brien, Angew. Chem. Int. Ed.
2014, 53, 12907–12911; Angew. Chem. 2014, 126, 13121–13125; C. J.
O’Brien, J. L. Tellez, Z. S. Nixon, L. J. Kang, A. L. Carter, S. R. Kunkel, K. C.
Przeworski, G. A. Chass, Angew. Chem. Int. Ed. 2009, 48, 6836–6839;
Angew. Chem. 2009, 121, 6968–6971; C. J. O’Brien, F. Lavigne, E. E.
Coyle, A. J. Holohan, B. J. Doonan, Chem. Eur. J. 2013, 19, 5854–5858;
C. J. O’Brien, Z. S. Nixon, A. J. Holohan, S. R. Kunkel, J. L. Tellez, B. J.
Doonan, E. E. Coyle, F. Lavigne, L. J. Kang, K. C. Przeworski, Chem. Eur.
J. 2013, 19, 15281–15289; I. J. Fairlamb, ChemSusChem 2009, 2, 1021–
1024.
[23] Catalytic Wittig reactions: T. Werner, M. Hoffmann, S. Deshmukh, Eur. J.
Org. Chem. 2014, 6873–6876; T. Werner, M. Hoffmann, S. Deshmukh,
Eur. J. Org. Chem. 2014, 6630–6633; M. Hoffmann, S. Deshmukh, T.
Werner, Eur. J. Org. Chem. 2015, 4532–4543; d) M.-L. Schirmer, S. Ado-
meit, T. Werner, Org. Lett. 2015, 17, 3078–3081.
[24] Catalytic aza-Wittig reactions: S. P. Marsden, A. E. McGonagle, B. McKe-
ever-Abbas, Org. Lett. 2008, 10, 2589–2591; H. A. van Kalkeren, C. Gro-
tenhuis, F. S. Haasjes, C. A. Hommersom, F. P. J. T. Rutjes, F. L. van Delft,
Eur. J. Org. Chem. 2013, 7059–7066; Y.-M. Yan, Y. Rao, M.-W. Ding, J.
Org. Chem. 2016, 81, 1263–1268; L. Wang, Y. Wang, M. Chen, M.-W.
Ding, Adv. Synth. Catal. 2014, 356, 1098–1104.
[25] Reductive catalytic transposition of allyl bromides through hydrido-
phosphorane: K. D. Reichl, N. L. Dunn, N. J. Fastuca, A. T. Radosevich, J.
Am. Chem. Soc. 2015, 137, 5292–5295.
[44] S. Trippett, Proc. Chem. Soc. 1963, 19–20.
[45] C. Eyles, S. Trippett, J. Chem. Soc. C 1966, 67–71.
[26] Dihydridophosphorane in catalytic transfer hydrogenation
:
N. L.
[46] Phosphines, arising from nucleophilic attack at carbon or from elimi-
nation, were also observed, along with alkenes expected from elimina-
tion processes.
[47] Hamid and Trippett later confirmed that the alkyl group of the alcohol
becomes appended to the carbon of the phenyl group that was
Dunn, M. Ha, A. T. Radosevich, J. Am. Chem. Soc. 2012, 134, 11330–
11333. Deoxygenative condensation of a-keto esters and carboxylic
acids: W. Zhao, P. K. Yan, A. T. Radosevich, J. Am. Chem. Soc. 2015, 137,
616–619. Activation of NÀH bonds: S. M. McCarthy, Y.-C. Lin, D. Devar-
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9152
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
bound to phosphorus in the starting phosphonium salt. See A. M.
Hamid, S. Trippett, J. Chem. Soc. C 1967, 2625.
solvent had been distilled away) that the reacting entity was an ylide
and not a phosphonium ethoxide.
[62] According to reference [85], the least-motion pathway for eliminations
such as the one in path B, 3-centre apical-equatorial elimination, is for-
bidden based on the symmetry of the molecular orbital of the phos-
phorane (whether trigonal bipyramidal or square-based pyramidal;
see ref. [84]) that gives rise to the new s-bond. On this basis, the pro-
cess must occur by apical–apical or equatorial–equatorial elimination.
Formally, this reaction is the reverse of biphilic additions of, for exam-
ple, peroxides to phosphines (see ref. [55]). The process of path A, if it
is an intramolecular process, may occur by analogous 4-centre apical–
apical or equatorial–equatorial elimination mechanisms, or perhaps
through a trigonal bipyramidal transition state (see ref. [83]) in which
the alkoxy group (undergoing dealkylation) is in an equatorial posi-
tion. See however references [25] and [26].
[48] The reaction conducted with C2H5OD gave di- and trideuterated p-ni-
trotoluene, while that with CD3OH did not give deuterated p-nitroto-
luene. Thus, the hydrogen atoms that become appended to the carb-
anion leaving group originate from the alcohol hydroxyl group.
[49] M. Grayson, P. T. Keough, J. Am. Chem. Soc. 1960, 82, 3919–3924.
Ether formation was proposed to result from nucleophilic attack of
ethoxide on ethoxyphosphonium cation.
[50] That the two reactions give the same products is of little surprise,
given that the same reactants are present in each; one reaction in-
volves in situ ylide generation, while the other involves a pre-generat-
ed and pre-purified ylide. Similar results were obtained in the corre-
sponding reactions of other para-nitrobenzylides and the related un-
substituted benzylide.
[63] W. N. Olmstead, Z. Margolin, F. G. Bordwell, J. Org. Chem. 1980, 45,
3295–3299; F. G. Bordwell, Acc. Chem. Res. 1988, 21, 456–463.
[64] X.-M. Zhang, F. G. Bordwell, J. Am. Chem. Soc. 1994, 116, 968–972.
[65] E. Vedejs, C. F. Marth, R. Ruggeri, J. Am. Chem. Soc. 1988, 110, 3940–
3948.
[66] See reference [19]; Schmidbaur and co-workers observed no formation
of tert-butoxyphosphorane in the reaction of methylidenetrimethyl-
phosphorane with tert-butanol, which is the conjugate base of KOtBu
deprotonation.
[51] An analogous set of products—phosphine oxide (nBu3PO), benzoic an-
hydride and alkane (the alkane derived that would be expected from
the hydrolysis of the ylide; the best leaving group attached to P)—
was obtained from the reaction of a heavily conjugated stabilised
ylide with benzoic acid in benzene; see Y. Kawamura, Y. Sato, T. Horle,
M. Tsukayama, Tetrahedron Lett. 1997, 38, 7893–7896. Analogous re-
sults were not obtained with more typical stabilised ylides, for exam-
ple, Ph3P=C(Me)CO2Me.
[52] C. B. Parisek, W. E. McEwen, C. A. Vanderwerf, J. Am. Chem. Soc. 1960,
82, 5503–5504.
[53] R. Luckenbach, Chem. Ber. 1975, 108, 803–812.
[67] E. Vedejs, M. J. Peterson, in Topics in Stereochemistry, Vol. 21 (Eds.: E. L.
Eliel, S. H. Wilen), Wiley, Hoboken, 1994; E. Vedejs, T. Fleck, J. Am.
Chem. Soc. 1989, 111, 5861–5871.
[54] This proposal was based on the observation that the quantity of tolu-
ene produced at intermediate points in the reaction was much greater
than the amount of Bu2O (shown by GC analysis of samples removed
from the same reaction run at room temperature). Whilst the forma-
tion of alkoxyphosphonium salt seems highly likely, and the formation
of dialkoxyphosphorane under these conditions is certainly a reasona-
ble proposal (see refs. [55–57]) we suggest that an alternative means
by which this reaction may have yielded racemic product was by
pseudorotation of the initial butoxyphosphorane intermediate (e.g.,
13a to 13b, which could expel toluene to give 14b) that is, the race-
misation may have already occurred before butoxyphosphonium salt
is formed.
[55] P. L. Robinson, C. N. Barry, J. W. Kelly, S. A. Evans, Jr., J. Am. Chem. Soc.
1985, 107, 5210–5219; D. B. Denney, H. M. Relles, A. K. Tsolis, J. Am.
Chem. Soc. 1964, 86, 4487–4488; D. B. Denney, D. Z. Denney, B. C.
Chang, K. L. Marsi, J. Am. Chem. Soc. 1969, 91, 5243–5246; T. Kubota,
S. Miyashita, T. Kitazume, N. Ishikawa, J. Org. Chem. 1980, 45, 5052–
5057; I. Mathieu-Pelta, S. A. Evans, Jr., J. Org. Chem. 1994, 59, 2234–
2237; Z.-Y. Yang, J. Org. Chem. 1995, 60, 5696–5698; P. J. Hammond,
G. Scott, C. D. Hall, J. Chem. Soc. Perkin Trans. 2 1982, 205–210; G.
Scott, P. J. Hammond, C. D. Hall, J. Chem. Soc. Perkin Trans. 2 1977,
882–888; D. B. Denney, D. Z. Denney, P. J. Hammond, Y.-P. Wang, J.
Am. Chem. Soc. 1981, 103, 1785–1789; L. L. Chang, D. B. Denney, D. Z.
Denney, R. J. Kazior, J. Am. Chem. Soc. 1977, 99, 2293–2297; C. D. Hall,
J. D. Bramblett, F. F. S. Lin, J. Am. Chem. Soc. 1972, 94, 9264–9266;
B. C. Chang, W. E. Conrad, D. B. Denney, D. Z. Denney, R. Edelman, R. L.
Powell, D. W. White, J. Am. Chem. Soc. 1971, 93, 4004–4009; P. D. Beer,
R. C. Edwards, C. D. Hall, J. Chem. Soc. Chem. Commun. 1980, 351–
352; M. Betou, L. Male, J. W. Steed, R. S. Grainger, Chem. Eur. J. 2014,
20, 6505–6517.
[56] M. von Itzstein, I. D. Jenkins, J. Chem. Soc. Perkin Trans. 1 1986, 437–
445.
[57] C. L. Lerman, F. H. Westheimer, J. Am. Chem. Soc. 1976, 98, 179–184;
D. I. Phillips, I. Szele, F. H. Westheimer, J. Am. Chem. Soc. 1976, 98,
184–189.
[58] G. Aksnes, L. J. Brudvik, Acta Chem. Scand. 1963, 17, 1616–1622.
[59] J. F. Bunnett, D. A. R. Harper, J. Org. Chem. 1967, 32, 2701–2704.
[60] F. Zaragoza, H. Stephenson, J. Org. Chem. 2001, 66, 2518–2521; F. Zar-
agoza, Tetrahedron 2001, 57, 5451–5454.
[61] These entities were prepared from the parent phosphonium halides
by treatment with NaOEt in dry EtOH (see ref. [10]); although the pKa
of this base in EtOH is probably too low to deprotonate a phosphoni-
um alkylide at room temperature, it is possible that at the high tem-
peratures at which the reactions were carried out (and after the EtOH
[68] P. Ballinger, F. A. Long, J. Am. Chem. Soc. 1960, 82, 795–798.
[69] X. M. Zhang, A. J. Fry, F. G. Bordwell, J. Org. Chem. 1996, 61, 4101–
4106.
[70] G. Aksnes, J. Songstad, Acta Chem. Scand. 1964, 18, 655–661; S.
Fliszµr, R. F. Hudson, G. Salvadori, Helv. Chim. Acta 1963, 46, 1580–
1588.
[71] The pKa of the alcohol is much lower in alcohol solvent than in aprotic
organic media. In principle, this might mean that alcohol could pro-
tonate non-stabilised ylides in alcohol solvents. However, the pKaH
values of semi-stabilised and stabilised ylides in alcohol solvents may
be too low (and indeed may themselves be lowered in alcohol sol-
vents) for protonation of ylide by alcohol to occur. For example, the
pKa values of P-phenacyltriphenylphosphonium cation and P-acetonyl-
triphenylphosphonium cation are the same (or slightly lower) in
mixed ethanol/water solvents (see ref. [70]) as in DMSO (see ref. [63]).
[72] The experiments of Eyles and Trippett involved neat mixtures of phos-
phonium salt and sodium alkoxide (see ref. [45]). To investigate the
formation of ylide in their reactions, the authors reacted (a-deuteroi-
sopropyl)triphenylphosphonium iodide and NaOMe under the above
conditions and recovered substantial amounts of MeOD, indicating
that phosphonium salt deprotonation by alkoxide to give ylide had
indeed occurred. However, the other products obtained in this case
were propene (with diminished deuterium-labelling vs. starting phos-
phonium salt) and Ph3P that is, reaction that led to scission of the
phosphonium PÀC bond was an elimination reaction, mediated either
by ylide or methoxide, or perhaps by both.
[73] The possibility that benzyl phosphonium salt hydrolysis could occur
via benzylide was suggested in reference [3], but discounted on the
basis that the hydrolysis of tetraphenylphosphonium bromide, which
cannot form ylide, shows the same second-order dependence on hy-
droxide concentration that is exhibited by benzyl salt 17. The authors
concluded that the first-order dependence on hydroxide concentra-
tion exhibited by 9 was due to a change in the rate determining step
from phosphorane decomposition to phosphorane formation.
[74] P. A. Byrne, D. G. Gilheany, J. Am. Chem. Soc. 2012, 134, 9225–9239.
[75] P. A. Byrne, J. Muldoon, Y. Ortin, D. G. Gilheany, Eur. J. Org. Chem. 2014,
86–98; P. A. Byrne, D. G. Gilheany, Chem. Soc. Rev. 2013, 42, 6670–
6696.
[76] Even after alkoxyphosphorane formation had occurred, due to the ex-
istence of a phosphorane/ylide equilibrium, there was still ylide pres-
ent, and standard Schlenk techniques proved insufficient to prevent
this reactive species from undergoing oxidation and/or hydrolysis over
the time course of our experiments. Most noticeably, the process of
fitting a rubber septum (to facilitate addition of a reagent) to a Schlenk
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9153
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
flask under a flow of nitrogen allowed a sufficient quantity of water
and oxygen into the flask to cause hydrolysis and/or oxidation over
time of significant quantities of the ylide in the ylide/alkoxyphosphor-
ane mixture. Furthermore, the act of transferring the reaction mixture
to an NMR tube (under nitrogen atmosphere in a long Schlenk flask)
by cannula invariably led to some oxidation and hydrolysis of the
ylide despite our best efforts.
[94] Identical results were obtained in separate experiments with ylide 17:
1) in the presence of KBr salt, and 2) using non-deuterated dry metha-
nol. In a further experiment conducted by the same procedure (ylide
quenched with an excess of dry methanol) using dry CD3CN as sol-
vent, benzaldehyde was added after the addition of methanol. No
alkene or phosphine oxide was formed by Wittig reaction, showing
that the ylide was indeed no longer present.
[77] See the Supporting Information for description of air-sensitive ylide
generation.
[95] H. R. Hudson, A. Kow, J. C. Roberts, J. Chem. Soc. Perkin Trans. 2 1983,
1363–1368.
[78] The CH3 signal of ylide 25 in the 1H NMR spectrum is reported to
appear as a double doublet (J=20, 7 Hz) at d=1.47 ppm in: E. Vedejs,
K. A. Snoble, J. Am. Chem. Soc. 1973, 95, 5778–5780.
[96] The chemical shift of the composite signal of the two isomers that
arise by restricted rotation about the alpha-beta bond of the ylide
changed a little, from dP =8.0–7.5 ppm in neat CD3CN to dP =3.5–
2.9 ppm in CD3CN with added methanol.
[97] H. M. R. Hoffmann, Angew. Chem. Int. Ed. Engl. 1969, 4, 556–577;
Angew. Chem. 1969, 81, 597–618; L. M. Stephenson, D. L. Mattern, J.
Org. Chem. 1976, 41, 3614–3619; W. Oppolzer, V. Snieckus, Angew.
Chem. Int. Ed. Engl. 1978, 17, 476–486; Angew. Chem. 1978, 90, 506–
516.
[98] S. Kojima, R. Takagi, K.-Y. Akiba, J. Am. Chem. Soc. 1997, 119, 5970–
5971.
[99] See the discussion in reference [85] on molecular orbital symmetry
considerations in the elimination of H2 from PH5, which are certainly
of relevance to this discussion.
[100] See however reference [25] (and discussion in note [62]) in which PV
to PIII reductive elimination appears highly likely to occur by an axial–
equatorial elimination pathway.
[101] Compare with racemisation of chiral phosphine oxide in the presence
of HCl (see ref. [102]), and racemisation in alkaline hydrolysis of chiral
dialkoxyphosphonium salts (see ref. [103]), both of which are pro-
posed to occur by pseudorotation of a pentavalent entity.
[102] D. B. Denney, A. K. Tsolis, K. Mislow, J. Am. Chem. Soc. 1964, 86, 4486–
4487.
[79] See Figure S4 in the Supporting Information for the 1H{31P} spectrum
of 21.
[80] P. B. Kay, S. Trippett, J. Chem. Res. (S) 1986, 62–63; R. G. Cavell, J. A.
Gibson, K. I. The, Inorg. Chem. 1978, 17, 2880–2885; R. G. Cavell, J. A.
Gibson, K. I. The, J. Am. Chem. Soc. 1977, 99, 7841–7847; K. I. The, R. G.
Cavell, Inorg. Chem. 1977, 16, 2887–2894; K. I. The, R. G. Cavell, J.
Chem. Soc. Chem. Commun. 1975, 279–280; K. I. The, R. G. Cavell, J.
Chem. Soc. Chem. Commun. 1975, 716–717; R. G. Cavell, D. D. Poulin,
K. I. The, A. J. Tomlinson, J. Chem. Soc. Chem. Commun. 1974, 19–21;
E. Vedejs, C. F. Marth, J. Am. Chem. Soc. 1989, 111, 1519; E. Vedejs, C. F.
Marth, J. Am. Chem. Soc. 1990, 112, 3905–3909.
[81] See Figure S10 in the Supporting Information.
[82] F. H. Westheimer, Acc. Chem. Res. 1968, 1, 70–78; K. Mislow, Acc.
Chem. Res. 1970, 3, 321–331; R. K. Oram, S. Trippett, J. Chem. Soc.
Perkin Trans. 1 1973, 1300–1310.
[83] E. P. A. Couzijn, J. C. Slootweg, A. W. Ehlers, K. Lamertsma, J. Am. Chem.
Soc. 2010, 132, 18127–18140; J. García López, A. Morµn Ramallal, J.
Gonzµlez, L. Roces, S. García-Granda, M. J. Iglesias, P. OÇa-Burgos, F. Ló-
pez Ortiz, J. Am. Chem. Soc. 2012, 134, 19504–19507.
[84] T. A. Albright, J. K. Burdett, M.-H. Whangbo, Orbital Interactions in
Chemistry, 2nd ed., Wiley, Hoboken, 2013, pp. 388–390.
[103] K. DeBruin, J. R. Peterson, J. Org. Chem. 1972, 37, 2272–2278.
[104] Standard Schlenk techniques could then be applied when using the
long Schlenk flask without the occurrence of oxidation or hydrolysis.
We found that diffusion of adventitious oxygen or water down into
the NMR tube was extremely slow due to the narrow diameter of the
tube and the fact that the solution in the NMR tube was not being
stirred. Ylide hydrolysis and oxidation could also be avoided in reac-
tions where NMR observation of the intermediate was not required by
generating the ylide in the glove box (under dry argon) in a Schlenk
flask fitted with a second tap. When the Schlenk flask was removed
from the glove box and fitted to the nitrogen supply, the second tap
could be fitted with a septum and the region between the septum
and tap then purged with nitrogen gas before opening the tap to
permit addition of reagents. This prevented the entry of all but the ti-
niest quantities of oxygen and water to the reaction flask.
[105] E. Durµn, D. Velasco, F. López-Calahorra, J. Chem. Soc. Perkin Trans.
1 2000, 591–594.
[85] R. Hoffmann, J. M. Howell, E. L. Muetterties, J. Am. Chem. Soc. 1972, 94,
3047–3058.
[86] E. L. Muetterties, W. Mahler, R. Schmutzler, Inorg. Chem. 1963, 2, 613–
618; E. L. Muetterties, W. Mahler, K. J. Packer, R. Schmutzler, Inorg.
Chem. 1964, 3, 1298–1303; E. L. Muetterties, Acc. Chem. Res. 1970, 3,
266–273; R. J. Hach, R. E. Rundle, J. Am. Chem. Soc. 1951, 73, 4321–
4324; R. E. Rundle, Acta Crystallogr. 1961, 14, 585–588; R. E. Rundle, J.
Am. Chem. Soc. 1963, 85, 112–113.
[87] G. R. Fulmer, A. J. M. Miller, N. H. Sherden, H. E. Gottlieb, A. Nudelman,
B. M. Stoltz, J. E. Bercaw, K. I. Goldberg, Organometallics 2010, 29,
2176–2179.
[88] See Figure S5a–f in the Supporting Information.
[89] See Figure S11b in the Supporting Information.
[90] B. Nandi, S. Sinha, Tetrahedron 2011, 67, 106–113.
[91] P. Demel, M. Keller, B. Brei, Chem. Eur. J. 2006, 12, 6669–6683.
[92] An alternative possibility involves attack of water at phosphorus to
give a species that is hexavalent at phosphorus (cf. reference [57]). Al-
though we do not unequivocally rule out this possibility, we have
seen no evidence in our 31P NMR spectroscopy studies of metastable
hydroxyphosphorane 118 or alkoxyphosphoranes 19 and 20 for the
existence of such hexavalent species.
[106] H. A. Van Kalkeren, J. J. Bruins, F. R. J. T. Rutjes, F. L. Van Delft, Adv.
Synth. Catal. 2012, 354, 1417–1421.
Received: March 14, 2016
[93] Significantly, the red/orange colour of the reaction mixture remained
after the addition of the CD3OD.
Published online on June 6, 2016
Chem. Eur. J. 2016, 22, 9140 – 9154
www.chemeurj.org
9154
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim