Mysterious Decomposition of Alkoxyphosphonium Chlorides: Postulated Involvement of the HCl2 Anion and Its Capture in the Solid State

Article abstract of DOI:10.1002/chem.201604898

P-Alkoxyphosphonium (AP) chlorides were generated by reacting P-chlorophosphonium chlorides with alcohols. Their well-known spontaneous Arbuzov-type collapse leading to phosphine oxides was studied and its rate found to be dependent on a number of factors in an unexpected fashion: it is inversely proportional to the initial concentration and it shows strong dependence on the acidity of the media but is not very sensitive to the presence of base. To explain these observations, we evoke a self-inhibition model with the formation of the less nucleophilic hydrodichloride anion HCl₂ in solution. Detailed analysis of the kinetic data yields the association constant (K=3×10² m^?1) of the putative HCl₂ species in chloroform. Experimental observations for the collapse of highly enriched diastereomeric alkoxyphosphonium (DAP) chlorides are fully analogous to the achiral AP also implying the involvement of HCl₂ anions. Moreover, crystallisation of a highly enriched DAP salt derived from (?)-menthol furnished, for the first time, crystals of individual (R_P)-DAP hydrodichloride as confirmed by X-ray diffractometry. Importantly, the P-configuration and detailed conformation of the DAP moiety is in good agreement with DFT-level computational results. The thermal collapse of (R_P)-DAP?HCl₂ proceeds with complete retention of the P-configuration furnishing the phosphine oxide of exceptional enantiomeric purity.

Full text of DOI:10.1002/chem.201604898

DOI: 10.1002/chem.201604898
Full Paper
&
Phosphorus Chemistry |Hot Paper|
Mysterious Decomposition of Alkoxyphosphonium Chlorides:
Postulated Involvement of the HCl2 Anion and Its Capture in the
Solid State
Kirill Nikitin,* Helge Mꢀller-Bunz, Jimmy Muldoon, and Declan G. Gilheany*[a]
Abstract: P-Alkoxyphosphonium (AP) chlorides were gener-
ated by reacting P-chlorophosphonium chlorides with alco-
hols. Their well-known spontaneous Arbuzov-type collapse
leading to phosphine oxides was studied and its rate found
to be dependent on a number of factors in an unexpected
fashion: it is inversely proportional to the initial concentra-
tion and it shows strong dependence on the acidity of the
media but is not very sensitive to the presence of base. To
explain these observations, we evoke a self-inhibition model
with the formation of the less nucleophilic hydrodichloride
anion HCl2 in solution. Detailed analysis of the kinetic data
yields the association constant (K=3ꢁ102 mꢀ1) of the puta-
tive HCl2 species in chloroform. Experimental observations
for the collapse of highly enriched diastereomeric alkoxy-
phosphonium (DAP) chlorides are fully analogous to the
achiral AP also implying the involvement of HCl2 anions.
Moreover, crystallisation of a highly enriched DAP salt de-
rived from (ꢀ)-menthol furnished, for the first time, crystals
of individual (RP)-DAP hydrodichloride as confirmed by X-ray
diffractometry. Importantly, the P-configuration and detailed
conformation of the DAP moiety is in good agreement with
DFT-level computational results. The thermal collapse of (RP)-
DAP·HCl2 proceeds with complete retention of the P-config-
uration furnishing the phosphine oxide of exceptional enan-
tiomeric purity.
Introduction
tween the acidic and neutral routes, which lead to consequen-
ces for the Arbuzov collapse process.
Alkoxyphosphonium (AP) salts (1), also known as quasi-phos-
phonium compounds, are the reactive intermediates in several
classic applications of organophosphorus chemistry to organic
synthesis, specifically the Appel,[1] Mitsunobu[2] and Horner–
Wadsworth–Emmons (HWE) reactions.[3] In all these cases, their
significance lies in the well-known Michaelis–Arbuzov process
in which, once formed, they collapse (Scheme 1, reaction iii) to
the corresponding phosphoryl derivative, 4, and alkyl halide, 5,
typically in high yields.[4] Therefore, AP salts are kinetically un-
stable and thus, with a few exceptions,[5,6] are poorly character-
ised.
Recently, the formation of AP intermediates in the course of
Appel-type processes has been applied by Denton and co-
workers[7] to design a catalytic version. During that work, cer-
tain kinetically stabilised AP chlorides were isolated and their
structure was confirmed by X-ray crystallography.[7b] O-Alkyla-
tion[8] of 4 with alkyl triflates is also an efficient route to more
kinetically stable 1 triflate salts.
We became interested in the chemistry of APs through the
“backdoor” so to speak—due to their role in the preparation
of P-stereogenic compounds rather than as sources of alkyl
halides or esters. Over the past few years, we have reported ef-
ficient stereoselective syntheses of various chiral phosphines 3
In addition to the standard quaternisation process for HWE
purposes[3] (not shown), there are two slightly different ap-
proaches to AP halides, summarised in Scheme 1: by reaction
of alcohols with the corresponding halophosphonium salts 2
(acidic route i) or by oxidative halogenation of phosphines, 3,
in the presence of alcohols (neutral route ii). In this report, our
focus is on significant, but earlier overlooked, differences be-
[a] Dr. K. Nikitin, Dr. H. Mꢀller-Bunz, Dr. J. Muldoon, Prof. D. G. Gilheany
School of Chemistry, Centre for Synthesis and Chemical Biology
University College Dublin (Ireland)
E-mail: kirill.nikitin@ucd.ie
declan.gileany@ucd.ie
Supporting information, including experimental procedures, characterisa-
tion of materials; kinetic, X-ray diffractometry and computational data,
and the ORCID number(s) for the author(s) of this article can be found
under http://dx.doi.org/10.1002/chem.201604898.
Scheme 1. Synthetic approaches to AP chlorides 1·Cl: i) alcoholysis of chlor-
ides 2 (acidic route, when 1 equiv of HCl is released); ii) ROH and a positive
chlorine source, for example hexachloroacetone HCA (no HCl produced,
neutral route); iii) gentle heating leading to collapse.
Chem. Eur. J. 2017, 23, 1 – 10
1
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
(R1 ¼R2 ¼R3), their borane complexes and their corresponding
phosphine oxides 4.[8b,9] In all these cases, the fundamental
aspect is that the R group in 1·Cl is itself chiral, which results
in a pair of diastereomeric alkoxyphosphonium (DAP) salts, for
example, 1b,c chlorides as shown in Scheme 2. Thus, after Ar-
buzov collapse, the enantiopurity (ee) of the resultant phos-
phine oxides is set by the diastereoselectivity (d.r. or de) in the
production of the DAP intermediates. The salts 1·Cl are typical-
ly generated through a dynamic kinetic resolution process by
reacting the rapidly racemising chlorophosphonium sub-
strate[10] (e.g., 2b, acidic route, Scheme 2) with a chiral alcohol,
such as (ꢀ)- menthol.
Scheme 3. Thermal collapse of AP chlorides: Some unexplained facts and
their possible rationalisation.
netic measurements and X-ray crystallography evidence, we
show that the interconversion of chloride and hydrodichloride
ꢀ
anions HCl2 in acidic media is responsible for a range of unex-
pected phenomena observed over the course of this nucleo-
philic reaction. We also believe that this acid–base equilibrium
is relevant to other reactions involving chloride anions in many
related areas of chemistry.
Scheme 2. Formation and collapse of diastereomeric 1b and c. Note that al-
though RP-1 is typically the major product, the collapse of SP-1 is faster, that
is, k(S) >k(R)
.
Results and Discussion
Recently, we expanded our knowledge and understanding
of the chemistry of P-chiral 1·Cl as follows: a) the stereoselec-
tivity, that is to say, the de of 1, does not depend on whether
it is generated from the chloride 2 with concomitant release of
HCl or from phosphine 3 in the presence of a suitable activa-
tor;[11a] b) both enantiomers of the oxide 4 can be prepared
from a single diastereomer of 1·Cl using the same chiral auxil-
iary;[11b] c) thermal collapse of the minor diastereomer for ex-
ample SP-1b·Cl is typically faster than RP-1b·Cl leading to
a gradual diastereomeric enrichment of the remaining DAP salt
RP-1b·Cl.[11c]
During these studies, we noticed that the rates of Arbuzov
collapse of the AP salts generated through the acidic route
were substantially less than those generated by the neutral
route, an observation we had no explanation for at the time.
Upon further investigation, the AP behaviour (both achiral 1a
and diastereomeric 1b,c) became even more mysterious. Spe-
cifically, as shown in Scheme 3, the Arbuzov collapse is (i) in-
hibited by acids but is only slightly faster in the presence of
bases; (ii) fast in non-polar solvents; and (iii) apparently slower
in more concentrated solutions exhibiting a nominally negative
kinetic order. Eventually, we found a solution to the mystery
that could potentially explain all the observations by making
just one additional assumption: that a generally overlooked
chemical process was intervening, an equilibrium involving the
Clꢀ counterion and HCl (Scheme 3).
As a matter of background general interest, the collapse of AP
halides has the characteristics of both bi- and monomolecular
nucleophilic reactions and, in certain cases, even a borderline
SN1 process was proposed.[4] It is characterised by inversion of
the C-configuration, although a kinetic study of sterically stabi-
lized (e.g., neopentyl-derived) AP halides has revealed a first-
order reaction. The explanation is that the halide anion of the
ion pair attacks the carbon centre of the alkyl group (attack on
the P-centre is more common for the O-nucleophiles.)[11b,12] For
secondary cyclic AP halides, a concomitant elimination process
competes with the nucleophilic substitution and can even be
dominant in some cases.[13]
We embarked on a kinetic study of this process using our
AP chlorides in hopes that it would help to reveal fine details
of the mechanism, and possibly explain the unexpected obser-
vations. Initially, our study included DAP chlorides for example,
1b,c·Cl, however, because the rates of collapse of the individu-
al diastereomers (e.g., RP- and SP-1b, Scheme 2) are differ-
ent,[11c] we designed a simplified analogous achiral model
system. The achiral AP chloride 1a·Cl (Scheme 1) is a very close
match to the diastereomeric analogues 1b,c because it also
contains a diarylalkylphosphonium centre and a cyclic alkoxy
group. The chloride 1a·Cl is soluble in most common solvents.
Its thermal collapse (Scheme 4) mostly leads to the elimination
product 7, the protonated form of oxide 4a·HCl and the alkyl-
halide 8 (ca. 30%). The reaction is very slow at 08C but at
308C, it takes place over a period of few hours and can con-
veniently be monitored by NMR spectroscopy in a range of
conditions as described below.
In this report, we focus on the ionic equilibria involving
both the achiral and chiral AP intermediates, and their nucleo-
philic thermal collapse to form the oxides 4. By combining ki-
&
&
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
2
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Paper
the reaction mixture, upon reaction completion, indicates that
the oxide 4a (dP =45 ppm) is partially protonated by the HCl
present in this system and that both the elimination product 7
(dH =5.67 ppm) and the substitution product 8 (dH =4.00 ppm)
were formed in an approximate 70:30 ratio (no change of this
ratio was observed over the subsequent 10 days).
Having established the rate and product ratio under these
conditions, we set out to investigate how the collapse process
is affected by the presence of acids, bases and nucleophiles
(we note that addition of cyclohexanol, Table 1, entry 2, did
not affect the rate). The results are presented in Table 1 and in
Figure 1. Evidently, the reaction is slowed down in the pres-
Scheme 4. Spontaneous collapse of the model chloride 1a·Cl.
Acid-inhibition of the collapse
Stock solutions, 0.40m of 1a·Cl in deuterated chloroform, were
prepared by reacting chlorophosphonium chloride[14] 2a with
1.05 equiv of cyclohexanol at 08C for 1 h after which time
100% conversion to 1a·Cl was determined by 31P NMR. In the
1H NMR spectrum of 1a·Cl, a broad peak at 10.8 ppm corre-
sponds to the rapidly exchanging protons of the unreacted cy-
1
clohexanol and HCl formed in the system. The H NMR charac-
terisation of this sample (see the Supporting Information for
details) indicated, along with unreacted cyclohexanol (dH =
3.6 ppm), the presence of traces of dicyclohexyl oxalate (dH =
4.9 ppm), which is the product of oxalyl chloride alcoholysis.
These components do not significantly affect the kinetics of
1a·Cl collapse.
Spontaneous collapse of the model 1a·Cl in CDCl3 (initial
concentration C1a =0.050m) was monitored by 31P NMR at
308C. Under these conditions, the reaction follows first order
kinetics, k=3.6ꢁ10ꢀ5 sꢀ1 (Table 1, entry 1) as is evident from
the kinetic log plot (r2 =0.9995; see the Supporting Informa-
Figure 1. Effect of added acids and bases on the relative rate of collapse of
1a·HCl2 (308C, CDCl3).
1
tion for details) over a period of 36 h. The H NMR analysis of
ence of added acids, HA. For example, large excess of the rela-
tively weak CH3COOH reduces the rate nearly by half (entry 5).
However, in the presence of a stronger acid, CF3COOH, the
drop is rather spectacular, about 20-fold (CHA =0.4m, entry 8).
To explain these striking rate variations we considered possible
ionic equilibria involving 1a·Cl as shown in Scheme 5.
Table 1. Effect of acids and bases on the thermal collapse of 1a·HCl2.
Entry
Additive
C
[m]
k[a]
[10ꢀ5 sꢀ1
]
1
2
3
4
5
6
7
8
none
–
3.6
3.6
2.6
2.4
1.7
Cyclohexanol
CH3COOH
CH3COOH
CH3COOH
CF3COOH
CF3COOH
CF3COOH
CF3SO3H
CF3SO3H
CF3SO3H
CF3SO3H
CF3SO3H
CF3SO3H
HCl
Pyridine
Pyridine
Pyridine
Et3N
Et3N
Et3N
0.2
0.1
0.2
0.4
0.1
0.2
0.4
0.05
0.1
0.15
0.2
0.25
0.3
0.2
0.1
0.2
0.4
0.1
0.2
0.4
0.27
0.1
Scheme 5. Ionic equilibria involving 1a·Cl in the presence of acids and
bases.
0.08
0.038
0.057
0.45
7.0
20
200
1.0
9
10
11
12
13
14
15
16
17
18
19
20
21
According to the left-hand side equilibrium in Scheme 5, an
acid HA, for example, CF3COOH, added to the solution of the
chloride 1a·Cl will furnish 1a·A incorporating a less nucleophil-
ic and less basic anion Aꢀ. As the concentration and the mole
fraction of the kinetically active 1a·Cl is less, the observed rate
of collapse of 1a·Cl is lower. However, this hypothesis alone
could not explain the fact that the rate also drops dramatically
in the presence of excess HCl itself (entry 15). We interpret this
fact as an indication of the formation of kinetically inert ionic
1a·HCl2 as shown in Scheme 5 on the right. We note that the
presence of hydrogen-bonded anionic species [ClꢀHꢀCl]ꢀ in
10
10.3
10.5
13
14
16
[a] First-order rate constant (chloroform, Co =0.05m, 308C) determined by
linear approximation of log plot with relative error of ꢁ10%.
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
3
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
solutions containing alkylammonium chlorides and HCl had
been proposed many years ago by Pocker.[15] Subsequent ex-
amples of their appearance in crystalline inorganic salts,[16] as
detected by X-ray analysis, are discussed by Emsley.[17] Recently,
crystal structures containing cryptate-stabilised and non-stabi-
lised [ClHCl]ꢀ have also been reported.[18] In most cases, the
structure of the anion was found to deviate from linearity and
to be non-symmetrical.
a very strong effect; in acetic acid, the collapse is suppressed
(entry 5) and in CF3COOH, it is essentially arrested (entry 6). In
the case of DMSO, two opposing trends clash: a) deceleration
is caused by better ion-pair solvation and b) acceleration is
due to its action as a mild base. Below we show that owing to
its higher polarity, DMSO has a pronounced slowdown effect,
which can clearly be seen only in using the neutral 1a·Cl, that
is, in the absence of HCl.
The behaviour of the very strong triflic acid TfOH is more
complex as shown in Figure 1 (solid line): the addition of
a single equivalent leads to the triflate 1a·OTf (Table 1, entry 9)
and slows the reaction by a factor of 90. In contrast, a four-
fold excess of TfOH causes a dramatically increased breakdown
of 1a. This acceleration effect, we hypothesize, is due to the
O-protonation of the AP cation by TfOH and formation of
a very electron-poor bis-triflate H-1a·(TfO)2, the collapse of
which is much faster.
Neutral AP chloride: Self-inhibition
Having established the pronounced effect of acids and bases
on the collapse rate of the acidic (i.e., containing excess HCl)
1a·HCl2, we studied HCl-free samples of 1a·Cl prepared by
adding hexachloroacetone (HCA) to a solution containing
phosphine 3a and cyclohexanol (route ii, Scheme 1). When
a sample of neutral 1a·Cl was monitored by 31P NMR at 308C
in CDCl3, its resulting collapse was about three times faster
than the acidic 1a·HCl2 (Table 3, entry 1 vs. Table 1, entry 1). In
fact, the initial rate for neutral 1a·Cl is approximately the same
as for acidic samples with some pyridine added (Table 1, en-
tries 16–18).
Finally, as strong acids have a pronounced inhibition effect
on the rate of 1a·Cl collapse, one would expect bases to in-
crease the rate of this reaction because the formation of
anions [HCl2]ꢀ in basic media would be minimized (see
Scheme 5). We found that in the case of pyridine, the accelera-
tion is limited to a ca. three-fold rate increase and the ob-
served rate constant is maintained around 1.0ꢁ10ꢀ4 sꢀ1 regard-
less of the amount of pyridine added (Figure 1). These findings
suggest that in the presence of pyridine, the equilibrium HCl+
ClꢀQ[HCl2]ꢀ is fully shifted to the left and no extra amount of
pyridine can produce more of the active species 1a·Cl. In the
presence of Et3N, which is a stronger base, the observed grad-
ual rate increase at higher concentrations of Et3N (Table 1 en-
tries 19–21 and Figure 1) may indicate general base-catalysed
acceleration of the elimination process.
The self-inhibition effect becomes obvious as the kinetic log
plot deviates markedly from the expected linear first-order ki-
netics (Figure 2) after about 3 h. The apparent reaction slow-
down and a gradual downfield 31P NMR shift seen for the
Table 3. Collapse of neutral 1a·Cl in different conditions.
Entry
Solvent
Additive[a]
k[b]
[10ꢀ5 sꢀ1
]
1
2
3
4
5
6
7
CDCl3
CDCl3
CDCl3
CDCl3
Benzene
DMSO
CH3CN
none
11
CF3COOH
CH3COOH
HCl
none
none
0.078
3.5
1.0
27
4.6
4.7
The effect of solvent
The effect of solvent on the rate of AP collapse (we use
1a·HCl2 notation for AP generated by the acidic route,
Scheme 1) was examined in two ways: changing solvent polari-
ty and solvent acidity. The rates were then compared to the
rate in CDCl3 (Table 2). In the less polar benzene (entry 1), the
collapse is faster presumably due to relatively poor solvation
of ion-pairs and higher ground state energy.[4d,19] For the same
reason, in more polar MeCN (entry 4), the rate is significantly
lower. The acidity of the solvent has, as discussed above,
none
[a] C=0.2m; [b] first-order rate constant determined by linear approxima-
tion of log plot, Co =0.05m, 308C, error bar is ꢁ10%.
Table 2. Collapse of acidic 1a·HCl2 in different solvents.
Entry
Solvent
k[a]
[10ꢀ5 sꢀ1
Relative
k
]
1
2
3
4
5
6
Benzene
DMSO
CDCl3
CH3CN
CH3COOD
CF3COOD
5.2
4.8
3.6
0.64
0.24
0.034
1.5
1.3
1
0.18
0.07
0.01
[a] First-order rate constant (Co =0.05m, 308C) determined by linear ap-
proximation of log plot with relative error of ꢁ10%.
Figure 2. Thermal collapse of neutral 1a·Cl showing deviation from linearity
after about 3 h due to accumulation of HCl (chloroform, 308C).
&
&
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
4
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Paper
ꢀ
phosphine oxide product 4a, due to its protonation to give
4a·HCl (dP =58 ppm, see the Supporting Information), can be
readily explained by accumulation of HCl in the reaction
system due to elimination (Scheme 4). The effect of solvent on
the collapse of neutral AP 1a·Cl is entirely similar to acidic
1a·HCl2; it is faster in less polar benzene (Table 3, entry 5) and
slower in the more polar DMSO and acetonitrile (entries 6 and
7).
Therefore, addition of non-nucleophilic ions PF6 leads to
a higher proportion of cations 1a that are bound in the form
of 1a·PF6. As a result, the overall reaction rate drops by ten-
fold (Figure 3).
In the presence of halides Bu4NX, which are a source of nu-
cleophilic anions Xꢀ, the reaction rate variations are even more
fascinating. The chloride Bu4NCl has a pronounced accelerating
effect as the ion-pair equilibrium is now shifted towards 1a·Cl;
no change in the elimination versus substitution (Scheme 4) is
observed in this case. Conversely, the rate of collapse is slower
when the iodide Bu4NI is added, which is in good agreement
with the lower nucleophilicity of iodide in aprotic media, an
effect well-known experimentally and theoretically.[20] However,
the fact that more iodo- than chlorocyclohexane is formed in
the presence of excess Bu4NI (for example, 5:1 ratio with
4 equiv of iodide) suggests that the shift of equilibrium to-
wards the iodide 1a·I outweighs the loss of reactivity versus
the chloride 1a·Cl. Remarkably, in the case of bromide, the two
effects appear to nearly cancel each other.
Finally, when neutral 1a·Cl was monitored in the presence
of added acids (Table 3, entries 2–4), the observed rates agreed
with the acidic 1a·HCl2 (Table 1, entries 4, 7, 15). It can be con-
cluded that the rate of collapse depends chiefly on the pres-
ence of acids and type of solvent, but not on the method of
generation of the AP cations 1a.
The effect of counter-ion
Our kinetic data point to the fact that the acid-free neutral
1a·Cl is kinetically more active and is in equilibrium with its
less active form 1a·HCl2. This equilibrium between ion pairs
may also be affected by other ions which are present. To test
this hypothesis, the collapse of acidic 1a·HCl2 was monitored
in the presence of soluble ammonium salts Bu4NX (X=Cl, Br, I,
PF6, TfO, ClO4). It turned out that, as expected, the collapse
was practically identically suppressed by the non-nucleophilic
anions PF6ꢀ, TfOꢀ and ClO4ꢀ, which are also very poor bases.
This can be demonstrated by comparing the relative reaction
rates (Figure 3); one can evoke a more complex equilibrium
(Scheme 6) involving the neutral chloride form 1a·Cl, the acidic
form 1a·HCl2 and the non-reactive ion pairs, such as 1a·PF6.
Individual halide-free phosphonium salts
We have observed (Figure 3) a pronounced deceleration of the
collapse process in the presence of non-nucleophilic anions,
which could also indicate rather high affinity of the AP cations
towards such anions. Upon some experimentation, we isolated
individual AP salts 1a·X (X=PF6, TfO, ClO4) by using simple bi-
phasic ion-exchange procedures and characterised new salts
by standard 1D and 2D NMR techniques. When the collapse of
1a·OTf was monitored at 308C, a dramatically lower rate was
observed with k=3.8ꢁ10ꢀ7 sꢀ1 (Table 4, entry 1), which is
some 300 times slower than the chloride 1a·Cl (no cyclohex-
ane-derived products, other than alkene 7, are formed). Con-
versely, the collapse of 1a·OTf can again be made fast (en-
tries 2–4, Table 4) via the addition of Bu4NCl leading to the for-
mation of 1a·Cl (Scheme 6). As a sufficient amount of Bu4NCl is
added, the collapse rate (k=1.0ꢁ10ꢀ4 sꢀ1) matches the rate for
individual neutral 1a·Cl (see Table 3, entry 1). Evidently, the for-
mation of the triflate 1a·OTf is reversible and the chloride
anions can replace triflate anions under those conditions.
Table 4. Effect of added chloride salt on collapse of the AP triflate
1a·OTf.
Entry
Additive
C
[m]
k[a]
[10ꢀ5 sꢀ1
]
Figure 3. Effect of halide anions and non-nucleophilic anions on the rate of
collapse of 1a·HCl2.
1
2
3
4
none
–
0.038
5.3
7.4
Bu4NCl
Bu4NCl
Bu4NCl
0.05
0.1
0.2
9.9
[a] First-order rate constant determined by linear approximation of log
plot, CDCl3, Co =0.05m, 308C, error bar is ꢁ10%.
Negative-order rate law and stability of [HCl2]ꢀ
The question arises as to how stable is the anion [HCl2]ꢀ in so-
lution? To address this, let us consider what should happen if
Scheme 6. Ionic equilibria involving 1a·HCl2 in the presence of Bu4NPF6.
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
5
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
a solution of acidic AP salt 1a·HCl2 is gradually diluted. In this
case, the equilibrium shown in [Eq. (1)]:
The collapse of neutral 1c·Cl (derived from ethyl-phenyl-o-
tolylphosphine, Scheme 2) in solution was monitored by NMR
at 308C and the data are shown in Table 5 (entries 1–6). While
the rates were typically three to four times lower and the pro-
portion of the nucleophilic product, neomenthyl chloride, was
HCl þ 1 a ꢂ Cl Ð 1 a ꢂ HCl2
ð1Þ
should shift towards the chloride 1a·Cl, which is kinetically far
more active than 1a·HCl2. Assuming that the dissociation of
tight ion pairs 1a·Cl does not yet take place in this concentra-
tion range, the dilution should increase the observed rate con-
stant and this effect will depend on the association constant
(stability) of the anion [HCl2]ꢀ. When a series of sequentially di-
luted solutions of 1a·HCl2 was prepared (C=0.4, 0.2, 0.1, 0.05,
0.025m) and tested, the apparent rate constant was inversely
proportional to the initial concentration of 1a·HCl2 correspond-
ing to a negative-order rate law as shown in Figure 4. This find-
ing is in excellent agreement with the proposed ionic equilibri-
um in solution (Scheme 5). In addition, by using double recip-
rocal plot analysis,[21] we can estimate the stability constant for
the [HCl2]ꢀ anion in solution; K=3ꢁ0.5ꢁ102 mꢀ1 (chloroform,
308C) is in reasonable agreement with the literature value of
about 6.0ꢁ102 mꢀ1 in sym-tetrachloroethane.[22]
Table 5. Collapse of neutral and acidic DAP chlorides 1c·Cl.
Entry
Solvent
Additive[a]
k[b]
[10ꢀ5 sꢀ1
]
Neutral 1c·Cl
1
2
3
4
5
6
Benzene
Toluene
CH2Cl2
CDCl3
DMSO
CH3CN
–
–
–
–
10
10
3.0
2.4
1.5
0.6
–
–
Acidic 1c·HCl2
–
7
8
9
CDCl3
CDCl3
CDCl3
0.9
2.8
2.5
Pyridine
CH(OMe)3
[a] concentration 0.2m; [b] first-order rate constant determined by linear
approximation of log plot, Co =0.05m, 308C, error bar is ꢁ10%.
higher, in other respects the observations were very similar to
neutral 1a·Cl (Table 3). For 1a, the reaction is accelerated in
less polar solvents (entries 1, 2 vs. 5, 6). On the other hand, as
expected, the collapse of the acidic 1c·HCl2 (entry 7) is much
slower but the rate can be reset back to its higher value by
the addition of a base (entry 8) or an acid scavenger (entry 9).
Importantly, gradual inhibition of the collapse of 1c·Cl by accu-
mulating HCl is evident from the log plot showing reaction de-
celeration after about 10 h in this case (see Figure S1d in the
Supporting Information).
Figure 4. The rate constant of thermal collapse as a function of initial con-
centration 1a·HCl2 in chloroform at 308C.
As was the case for 1a·HCl2, a negative-rate order is ob-
served for the rate of collapse of both the acidic 1b·HCl2 and
1c·HCl2 (Scheme 2), which can be seen in Figures 5a,b, respec-
In short, the study of the model AP salts 1a has demonstrat-
ed acid inhibition, base acceleration, an apparent negative rate
law and pronounced ion-exchange effects in solution. These
findings are in excellent agreement with the proposed ion-pair
structure of AP chlorides 1a·Cl in equilibrium with hydrodi-
chlorides 1·HCl2.
P-stereogenic diastereomeric chlorides
Diastereomeric DAP pairs (R/S)-1b and (R/S)-1c (Scheme 2)
were generated by reacting (ꢀ)-menthol with chlorophospho-
nium chlorides 2b,c (acidic DAP) or, alternatively, with phos-
phines 3b,c and HCA (neutral DAP). The collapse of diastereo-
meric 1 leads to enantiomerically enriched 4 in such a fashion
that the ee of the product 4 is numerically equal to the de of
the DAP intermediate 1 (erosion of stereoselectivity sometimes
can occur such that the final ee is lower than the de).[11b,c]
Whereas the rates of thermal collapse for the minor and major
diastereomers of 1 are somewhat different,[11c] we can describe
the kinetics of this reaction as a single process with a weighted
average rate. This approximation works reasonably well and
does not affect our key observations.
Figure 5. The rate of thermal collapse as a function of the initial concentra-
tion of a) 1b·HCl2 and b) 1c·HCl2 (chloroform, 308C).
&
&
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
6
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Paper
tively. It closely replicates that of acidic 1a·HCl2 (Figure 4) and
can be ascribed to a gradual shift of the equilibrium HCl+
ClꢀQ[HCl2]ꢀ to the right at higher concentrations. It can be
concluded that the key kinetic characteristics of achiral and P-
chiral APs are essentially identical owing to participation of the
chloride and hydrodichloride anions in the equilibrium.
Scheme 7. Stereospecific transformation of individual diastereomer RP-
1b·HCl2 to highly enantiomerically enriched phosphine oxide 4b.
Crystal structure of P-chiral DAP hydrodichloride
We have recently characterised two diastereomeric forms of
DAP 1d·Cl (Scheme 2, R3 =Me) by a range of NMR techni-
ques.[11c] By using a stereochemical correlation with crystalline
phosphine oxide products, we have also established that the
major diastereomers of DAP 1b–d have RP-configuration if (ꢀ)-
menthol is used. However, no definitive direct evidence of the
structure of chiral DAP has been obtained previously. Thus, we
were excited to observe fine crystalline precipitate of highly
diastereomerically pure 1b chloride prepared by slow addition
of menthol to a solution of chlorophosphonium chloride 2b·Cl
in DCM. This material was recrystallised at ca. 08C to furnish X-
ray-quality crystals.
strates full retention of the P-configuration over the course of
the thermal Arbuzov-type collapse of P-chiral DAP chlorides,
which is an important finding that was absent in the literature
until now.
Diastereoselectivity: Computation of relative intermediate
stability
Here we report our first computational results indicating that
the major diastereomer RP-1b·HCl2 has greater stability than SP-
1b·HCl2 intermediate in full agreement with the observed RP-
diastereoselectivity. Recently, we hypothesized that the ob-
served selectivity in the dynamic kinetic resolution is driven,
through the application of the Hammond postulate, by relative
thermodynamic stability of the respective diastereomeric inter-
mediates.[11c] In this case, the energy of the RP-DAP intermedi-
ate is lower than the minor SP-isomer and this difference also
explains predominant collapse of the minor SP-isomer, which
helps to achieve additional surplus selectivity (“wait and
attack” effect).[11c] Considering that the X-ray diffractometry
analysis of the DAP intermediate RP-1b·HCl2 not only con-
firmed its P-configuration but also revealed detailed conforma-
tion of the o-tolyl, n-butyl and menthol moieties, we turned to
electron structure computational methods to obtain detailed
geometry and also relative energies of both diastereomers.
This approach, if successful, would serve as a handy predictive
method prior to carrying out numerous (and costly) stereose-
lectivity experiments.
The structure of DAP hydrodichloride 1b·HCl2 (Figure 6a) is
in excellent agreement with our hypothesis. Importantly, it has
RP-configuration, and it contains hydrodichloride anions rather
Figure 6. Molecular structure of diastereomeric 1b·HCl2. a) X-ray crystal
structure of a single diastereomer RP-1b·HCl2; b) DFT calculated structure of
RP-1b·HCl2 in DCM; c) DFT calculated structure of SP-1b·HCl2 in DCM.
than individual chloride anions. The anions [ClꢀHꢀCl]ꢀ are ap-
proximately linear (with distances (dHCl) of 1.1 and 2.0ꢃ. How-
ever, as positions of electron-stripped hydrogen atoms cannot
be reliably obtained from the diffraction data the exact values
of dHCl could not be determined with high degree of certainty.
Similar geometry of this anion has been reported in its salts
with magnesium-crown ether cationic complexes.[23]
The experimentally observed orientation of the n-butyl, o-
tolyl and (ꢀ)-menthyl groups of RP-1b·HCl2 in the solid state
(Figure 6a) corresponds to the minimum energy conformation
predicted computationally in CH2Cl2 as solvent (Figure 6b).
Computations of the equilibrium geometry further revealed an
entirely different relative orientation of the menthol moiety in
the diastereomeric SP-1b as shown in Figure 6c. Owing to
steric interaction with the (ꢀ)-menthol moiety, even the most
stable conformation of SP-1b·HCl2 is about 1 kcalmolꢀ1 less
stable than the analogous minimum conformation of the RP-
isomer (Figure 6b). This energy difference is consistent with RP-
stereoselectivity for 1b and other n-alkylphenyl-o-tolylphos-
phine RP-DAP intermediates.[11b] Work is underway on the fur-
ther refinement of this computational approach and its appli-
cation to stereoselective reactions between chiral alcohols and
halophosphonium salts.
The crystalline RP-1b·HCl2 can be stored at ꢀ208C virtually
indefinitely. The NMR characterisation of this salt carried out at
08C in CDCl3 (see the Supporting Information for details) re-
vealed, within the detection limit of this measurement, the
presence of only one diastereomer (dP =67 ppm). Considering
1
that the 13C JC-P satellite signals are clearly visible, and the
peak of the minor diastereomer is not, the de of the sample is
at least but clearly higher than 99%. The Arbuzov-type col-
lapse of this highly enriched material was carried out at 308C
in chloroform. The product, phosphine oxide 4b, was obtained
in quantitative yield and 99.5% ee as determined by chiral
HPLC (Scheme 7). This selective transformation clearly demon-
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
7
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
Conclusions
Keywords: diastereoselectivity · ion pairs · kinetic resolution ·
nucleophilic substitution · phosphonium salts
Alkoxyphosphonium halides (APs) are the key intermediates
proposed for, and indeed observed in a number of reactions
but their structure and properties were previously poorly in-
vestigated partly due to their kinetic lability at ambient condi-
tions. In this work, we have shown that the kinetics of thermal
collapse of achiral and chiral diastereomeric AP chlorides de-
pends critically on the dynamic ionic equilibria involving phos-
phonium cations and their counteranions in solution.
[1] a) R. Appel, Angew. Chem. Int. Ed. Engl. 1975, 14, 801–811; Angew.
Chem. 1975, 87, 863–874; b) R. Appel, M. Halstenberg in Organophos-
phorus Reagents in Organic Synthesis (Ed.: J. I. Cadogan), Academic
Press, London, 1979, p. 387; c) C. N. Barry, S. A. Evans, J. Org. Chem.
1983, 48, 2825–2828; d) R. Bruckner in Organic Mechanisms-Reactions,
Stereochemistry and Synthesis (Ed.: M. Harmata), Springer, Berlin, 2010,
p. 53.
We have demonstrated that non-polar solvents accelerate
and polar solvents decelerate the collapse of APs due to, pre-
sumably, the relative stabilisation of the ionic ground state in
polar media. Protonation of Cl anions strongly affects the ki-
netics of collapse and certain strong acids can be used to
arrest the collapse of AP halides. Conversely, the addition of
mild bases accelerates the reaction but up to a certain limit.
These findings are consistent with the formation of hydrogen-
bonded anions HCl2 in solution.
[2] a) O. Mitsunobu, Y. Yamada, Bull. Chem. Soc. Jpn. 1967, 40, 2380–2382;
b) O. Mitsunobu, Synthesis 1981, 1–28.
[3] a) B. J. Walker in Organophosphorus Reagents in Organic Synthesis (Ed.:
J. I. Cadogan), Academic Press, London, 1979, p. 155; b) I. Saikia, A. J.
Borah, P. Phukan, Chem. Rev. 2016, 116, 6837–7042.
[4] a) W. Gerrard, W. J. Green, J. Chem. Soc. 1951, 2550–2553; b) K. Henrick,
H. R. Hudson, A. Kow, Chem. Commun. 1980, 226–227; c) A. Streitwie-
ser, Jr., C. H. Heathcock, Introduction to Organic Chemistry, 2nd ed., Mac-
Millan, New York, 1981; d) H. R. Hudson, A. Kow, J. C. Roberts, J. Chem.
Soc. Perkin Trans. 2 1983, 1363–1368; e) E. Lewis, B. McCortney, Can. J.
Chem. 1986, 64, 1156–1160; for a kinetic study of this process, see: f)
H. R. Hudson, A. R. Qureshi, Pol. J. Chem. 2005, 79, 473–480.
[5] a) H. R. Hudson, R. G. Rees, J. E. Weekes, J. Chem. Soc. Perkin Trans.
1 1974, 982–985; b) C. Carcedo, A. Dervisi, I. A. Fallis, L. Ooi, K. M. A.
Malik, Chem. Commun. 2004, 1236–1237; c) Y. Yamamoto, K. Nakao, T.
Hashimoto, S. Matsukawa, N. Suzukawa, S. Kojima, K. Akiba, Heteroat.
Chem. 2011, 22, 523–530; d) R. M. Denton, J. An, B. Adeniran, A. J.
Blake, W. Lewis, A. M. Poulton, J. Org. Chem. 2011, 76, 6749–6767.
[6] a) J. B. Hendricksons, S. M. Schwartzman, Tetrahedron Lett. 1975, 277–
280; b) S.-L. You, H. Razavi, J. W. Kelly, Angew. Chem. Int. Ed. 2003, 42,
83–85; Angew. Chem. 2003, 115, 87–89.
[7] a) R. M. Denton, J. An, B. Adeniran, Chem. Commun. 2010, 46, 3025–
3027; b) R. M. Denton, X. Tang, A. Przeslak, Org. Lett. 2010, 12, 4678–
4681; c) R. M. Denton, J. An, P. Lindovska, W. Lewis, Tetrahedron 2012,
68, 2899–2905; d) X. Tang, J. An, R. M. Denton, Tetrahedron Lett. 2014,
55, 799–802.
[8] a) J. Omelanczuk, M. Mikolajczyk, J. Am. Chem. Soc. 1979, 101, 7292–
7295; b) K. V. Rajendran, D. G. Gilheany, Chem. Commun. 2012, 48, 817–
819; c) N. P. Kenny, K. V. Rajendran, D. G. Gilheany, Chem. Commun.
2015, 51, 16561–16564.
[9] a) E. Bergin, C. T. O’Connor, S. B. Robinson, E. M. McGarrigle, C. P.
O’Mahony, D. G. Gilheany, J. Am. Chem. Soc. 2007, 129, 9566–9567;
b) K. V. Rajendran, L. Kennedy, D. G. Gilheany, Eur. J. Org. Chem. 2010,
5642–5649; c) K. V. Rajendran, J. S. Kudavalli, K. S. Dunne, D. G. Gilhea-
ny, Eur. J. Org. Chem. 2012, 2720–2723; d) D. J. Carr, J. S. Kudavalli, K. S.
Dunne, D. G. Gilheany, J. Org. Chem. 2013, 78, 10500–10505.
[10] E. V. Jennings, K. Nikitin, Y. Ortin, D. G. Gilheany, J. Am. Chem. Soc. 2014,
136, 16217–16226.
[11] a) K. V. Rajendran, D. G. Gilheany, Chem. Commun. 2012, 48, 10040–
10042; b) K. Nikitin, K. Rajendran, H. Mꢀller-Bunz, D. G. Gilheany, Angew.
Chem. Int. Ed. 2014, 53, 1906–1909; Angew. Chem. 2014, 126, 1937–
1940; c) K. Rajendran, K. Nikitin, D. G. Gilheany, J. Am. Chem. Soc. 2015,
137, 9375–9381.
We have observed a very rare case of the reaction rate con-
stant being inversely proportional to the initial concentration
of starting substrate. This key finding is again consistent with
the presence of HCl2 anions. Careful measurements in the pres-
ence of added anions (halides, triflate, PF6) show that AP hy-
drochlorides engage in ion-exchange equilibria leading to
more reactive or less reactive ion pairs. Due to the propensity
of APs to form these species, the individual AP, free from
halide anions, can easily be prepared by ion metathesis.
For the first time, we have isolated an individual diastereo-
mer of stereochemically enriched P-stereogenic AP and deter-
mined its molecular structure. Crucially, we have made two key
findings: i) the counter-ion is indeed the proposed HCl2; and ii)
the phosphorus centre has the expected absolute R-configura-
tion. The first finding concurs with our initial HCl2 hypothesis
and the second is in excellent agreement with the earlier ob-
served R-stereoselectivity of phosphine oxides (using (ꢀ)-men-
thol). Under carefully controlled conditions, collapse of the
DAP salt furnished the respective phosphine oxide with superb
enantiomeric purity, 99.5% ee.
Whereas the role of ion pairs in complex reaction mecha-
nisms is widely accepted, the individual properties and behav-
iour of ions are perhaps still poorly understood. Our study
demonstrates that the chemistry of even such an unremarka-
ble species as chloride anion Clꢀ could amalgamate a plethora
of dynamic processes involving unexpected strong binding to
its own acid, HCl. These observations may be crucial to our un-
derstanding of chemical processes involving halide anions
under acidic conditions, for example organic and inorganic nu-
cleophilic substitution reactions.
[12] R. A. McClelland, G. H. McGall, G. Patel, J. Am. Chem. Soc. 1985, 107,
5204–5209.
[13] K. E. Elson, I. D. Jenkins, W. A. Loughlin, Org. Biomol. Chem. 2003, 1,
2958–2965.
[14] The resulting 2a·Cl is a glassy solid, which, upon careful drying under
vacuum, contains about 1–2 mol% residual oxalyl chloride.
[15] a) Y. Pocker, J. Chem. Soc. 1960, 1292–1297.
[16] a) L. W. Schroeder, J. A. Ibers, J. Am. Chem. Soc. 1966, 88, 2601–2602;
b) J. S. Swanson, J. M. Williams, Inorg. Nucl. Chem. Lett. 1970, 6, 271–
276; c) J. W. Nibler, G. C. Pimentel, J. Chem. Phys. 1967, 47, 710–717.
[17] J. Emsley, Chem. Soc. Rev. 1980, 9, 91–124.
[18] a) B. H. Ward, G. E. Granroth, K. A. Abboud, M. W. Meisel, D. R. Talham,
Chem. Mater. 1998, 10, 1102–1108; b) I. Ravikumar, P. S. Lakshminar-
ayanan, E. Suresh, P. Gosh, Inorg. Chem. 2008, 47, 7992–7999; c) W. Le-
vason, D. Pugh, G. Reid, Acta Crystallogr. Sect. C 2013, 69, 560–564.
Acknowledgements
We are very grateful to UCD Centre for Synthesis and Chemical
Biology (CSCB) and the UCD School of Chemistry for access to
their extensive analysis facilities and to Science Foundation Ire-
land for funding (Grant 09/IN.1/B2627 to D.G.G.).
&
&
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
8
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Paper
[19] a) M. Szwarc, Acc. Chem. Res. 1969, 2, 87–96; b) M. H. Abraham, Prog.
Phys. Org. Chem. 1974, 11, 1–87; c) Y. Marcus, G. Hefter, Chem. Rev.
2006, 106, 4585–4621.
[21] a) H. A. Benesi, J. H. Hildebrand, J. Am. Chem. Soc. 1949, 71, 2703–2707;
b) K. Nikitin, H. Mꢀller-Bunz, J. Muldoon, ChemPhysChem 2014, 15, 139–
150.
[22] a) F. Y. Fujiwara, J. S. Martin, J. Am. Chem. Soc. 1974, 96, 7625–7631;
b) F. Y. Fujiwara, J. S. Martin, J. Chem. Phys. 1972, 56, 4091–4097.
[23] J. L. Atwood, S. G. Bott, C. M. Means, A. W. Coleman, H. Zhang, M. T.
May, Inorg. Chem. 1990, 29, 467–470.
[20] a) J. F. Bunnett, Annu. Rev. Phys. Chem. 1963, 14, 271; b) W. N. Olmstead,
J. Brauman, J. Am. Chem. Soc. 1977, 99, 4219–4228; c) J. Yang, Y. Ren,
H.-J. Zhu, S.-Y. Chu, Int. J. Mass Spectrom. 2003, 229, 199–208; d) Y. Ren,
H.-J. Zhu, J. Am. Soc. Mass Spectrom. 2004, 15, 673–680; e) E. V. Anslyn,
D. A. Dougherty, Modern Physical Organic Chemistry, University Science,
Sausalito, 2006; f) P. Bento, F. M. Bickelhaupt, J. Org. Chem. 2008, 73,
7290–7299.
Manuscript received: October 19, 2016
Accepted Article published: December 19, 2016
Final Article published: && &&, 0000
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
9
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
&
These are not the final page numbers! ÞÞ

Full Paper
FULL PAPER
&
Phosphorus Chemistry
K. Nikitin,* H. Mꢀller-Bunz, J. Muldoon,
D. G. Gilheany*
&& – &&
Mysterious Decomposition of
Alkoxyphosphonium Chlorides:
Postulated Involvement of the HCl2
Anion and Its Capture in the Solid
State
ꢀ
A tough watermelon to crack: The rate
of Arbuzov (“Watermelonman” in Russi-
an) collapse of P-Alkoxyphosphonium
(AP) chlorides varies in an unexpected
fashion as shown. This mystery was
solved by postulating the involvement
of hydrodichloride anions HCl2 in equi-
librium with Clꢀ and HCl. The X-ray crys-
tal structure of an individual AP hydro-
dichloride salt derived from (ꢀ)-menthol
ꢀ
revealed that the HCl2 anion is also
present in the solid state.
&
&
Chem. Eur. J. 2017, 23, 1 – 10
www.chemeurj.org
10
ꢂ 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!