Entropic strain and conformational preference in the hydrolysis of some N-alkyl-6-acetylaminotriazinediones

Article abstract of DOI:10.1039/a700736a

The rate-pH profiles for hydrolysis of the title compounds 1 show four distinct regions: k_A, k_B and k_C for rate plateaux corresponding to cationic, neutral and anionic species, plus k_D for attack of OH^- on the anion. At the ends of the pH scale the reaction is much slower than for model amides of comparable pK_lg, due to exceptional charge dispersal in reactant and leaving group. The plateau rates k_B and k_C are due to hydrolysis by water, not to some kinetically equivalent process, and are much faster than model calculations would predict. This is traced to intramolecular general base catalysis via solvent bridges, and leads to remarkable rate enhancements in aqueous alcohols. The considerable, and quite independent, variations in k_B, k_C and acid pK_a with only alkyl substitution in the amide moiety points to a dominant effect of conformation which has been explored using a number of techniques, notably octanol-water partitioning, and appears best rationalised in terms of Taft's 'steric hindrance of motions' or Tillett's 'entropic strain'. The overall picture for the effect of pH is of successively increasing C-O bond formation in the transition state along the sequence k_A → k_B → k_C but with C-N bond breaking quite out of line and largely dependent on conformational factors. Given pK_cat < 0, the presence of effective intramolecular general base catalysis in k_B is unexpected. We explain this as being due to a unique feature of 1 whereby catalyst and leaving group are part of the same conjugated structure, leading to pK_cat → pK_lg as C-N bond-breaking proceeds. Further light on k_B comes from the ring-N-methylated analogue 3d, which cannot form the intramolecular hydrogen bond found elsewhere and whose otherwise similar rate-pH profile shows an anomalous 'apparent pK_a' that can be explained as a consequence of this.

Full text of DOI:10.1039/a700736a

View Article Online / Journal Homepage / Table of Contents for this issue
Entropic strain and conformational preference in the hydrolysis of
some N-alkyl-6-acetylaminotriazinediones
Stuart Nicholson and Peter J. Taylor *
Zeneca Pharmaceuticals, Alderley Park, Macclesfield, Cheshire, UK SK10 4TG
The rate–pH profiles for hydrolysis of the title compounds 1 show four distinct regions: kA, kB and kC for
rate plateaux corresponding to cationic, neutral and anionic species, plus kD for attack of OH Ϫ on the
anion. At the ends of the pH scale the reaction is much slower than for model amides of comparable pKlg,
due to exceptional charge dispersal in reactant and leaving group. The plateau rates kB and kC are due to
hydrolysis by water, not to some kinetically equivalent process, and are much faster than model
calculations would predict. This is traced to intramolecular general base catalysis via solvent bridges, and
leads to remarkable rate enhancements in aqueous alcohols. The considerable, and quite independent,
variations in kB, kC and acid pKa with only alkyl substitution in the amide moiety points to a dominant
effect of conformation which has been explored using a number of techniques, notably octanol–water
partitioning, and appears best rationalised in terms of Taft’s ‘steric hindrance of motions’ or Tillett’s
‘entropic strain’. The overall picture for the effect of pH is of successively increasing C᎐O bond formation
in the transition state along the sequence kA → kB → kC but with C᎐N bond breaking quite out of line
and largely dependent on conformational factors.
G iven pKcat < 0, the presence of effective intramolecular general base catalysis in kB is unexpected.
We explain this as being due to a unique feature of 1 whereby catalyst and leaving group are part of the
same conjugated structure, leading to pKcat → pKlg as C᎐N bond-breaking proceeds. F urther light on kB
comes from the ring-N-methylated analogue 3d, which cannot form the intramolecular hydrogen bond
found elsewhere and whose otherwise similar rate–pH profile shows an anomalous ‘apparent pKa’ that
can be explained as a consequence of this.
The title compounds 1, originally investigated as potential
Our results point to conformation of the reacting species as a
major determinant of its reactivity. The rules appear to be dif-
ferent for neutral species and anion, so that the same structural
change has different consequences for each. Evidence for con-
formation has been obtained not only from expected sources
such as spectroscopy, but also from less usual sources such as
the partition coefficient. On the hypothesis that analgesia and
gastric irritation may relate, respectively, to the reactivities of
anion and neutral species, we might then hope to elucidate the
structural factors that underpin therapeutic ratio. A prelimin-
ary account of this work has appeared.7
R1
N
R1
N
O
O
O
O
O
N
NH
Me
N
NH
N
NH
R2
R2
1
2
herbicides,1 were later found to have analgesic and in some
cases anti-inflammatory activity.2–4 At the same time, they
proved to be gastric irritants.4 This behaviour is reminiscent of
aspirin, an irreversible inhibitor of cyclooxygenase5 whose pro-
pensity for irritating the gastric mucosa 6 may depend on the
same acetylation process that mediates its analgesic action.5 As
potential acetylating agents, the triazinediones 1 may therefore
mimic the behaviour of aspirin in both respects. Hence the
present investigation, while designed in part as a check on their
chemical stability in vitro, was also intended to throw light on
their possible mode of action in vivo.
The result of this investigation has been not only to illumin-
ate the above parallelism, but to uncover a mechanistic situ-
ation of some complexity. Despite the fact that the substituents
R2 in 1 are simple alkyl groups, variation in this respect pro-
duces considerable variation in pKa and in the rate of the reac-
tion to give 2, and these variations moreover appear to bear
little relation to one another. Furthermore, their hydrolytic
behaviour is not that of typical amides. While the expected
dependence on [Hϩ] and [OHϪ] both appear, there are rate plat-
eaux in the middle pH region, corresponding formally to the
concentrations of neutral and anionic species, that clearly point
to the presence of intramolecular catalysis. The nature of that
catalysis is the principal subject of this paper.
Experimental
Materials and methods
All compounds were supplied by colleagues; the few not
explicitly covered by the relevant patents,1–3 i.e. 3d, 4d and 5,
R1
N
R1
N
O
O
NMe
O
O
O
N
N
NMe
N
NH
R2
R2
Me
3
4
Pri
N
Me
N
O
O
O
O
N
NH
N
NH
N
O
5
6
J. Chem. Soc., Perkin Trans. 2, 1997
1771

View Article Online
Table 1 UV, pKa and log P data for some triazinediones
λmax/nm (log ε)e
R1
Pri
R2
log P a,b
pKa1
pKa2
C6H12
A
B
C
b,c
b,d
1a
Me
0.06
5.60
240
(3.79)
244
246
234
238
1b
1c
Pri
Pri
Et
0.49
0.96
4.87
5.16
Prn
233
(3.81)
241
(4.05)
240
(4.00)
243
(4.08)
1d
1e
1f
Pri
Pri
Pri
Pri
Pri
Pri
0.45
0.94
1.11
1.44
1.78
Ϫ0.95
Ϫ1.01
4.53
4.65
4.72
4.76
4.27
248
(4.03)
243
(3.90)
247
253
(3.94)
250
Bus
Bui
248
244
1g
1h
CHEt2
CH2But
248
(3.98)
250
248
(4.01)
244
243
(4.04)
1i
1j
1k
1l
Pri
CH(Et)Prn
Cyclohexyl
1-Piperidyl
CHBu2
Prn
1.96
1.59
2.33
3.22
0.88
4.75
4.38
5.26
4.70
5.04
4.99
4.46
4.11
Pri
Pri
Pri
1m
1n
1o
1p
Prn
Bun
Bun
C6H4-4-Me
C6H4-4-Me
Bun
249
248
(4.31)
250
248
(4.31)
229
(4.29)
246
(3.83)
229
(4.20)
245
244
Bui
1.65
Ϫ3.67
232
(4.29)
1q
1r
C6H4-4-Me
C6H4-4-OMe
CH2But
Bui
3.61
4.18
1.16
0.52
0.18
0.53
Ϫ3.65
1.69
<0
228
(4.37)
205
2d
3d
4d
5
Pri
Pri
Pri
Pri
Pri
Pri
7.95
223
(4.33)
(4.34)
—
1.59
13.70
6.91
<200
(>4.3)
241
(3.98)
a
Octanol–water. b At 25 ЊC. c Basic pKa. d Acid pKa. e A = cation, B = neutral species, C = anion (cf. Scheme 1); data for aqueous solution at 25 ЊC.
had been prepared by one of the methodologies described
therein. Table 1 lists salient physical properties for the 22 tri-
azinediones here considered. 13C NMR spectra were obtained
on a JEOL FX 90Q NMR spectrometer in CDCl3 solution using
tetramethylsilane as internal lock, and IR spectra in CHCl3
solution on a Perkin-Elmer 580B IR spectrophotometer. Either
a Unicam SP 8000 or a Unicam SP 1800 UV spectrometer
was used for UV spectra (Table 1) and for reaction following.
Product analysis by reversed-phase HPLC employed a Cecil
constant-pressure system using an octadecylsilane column with
90% aqueous acetonitrile as eluent, and with peak monitoring
by UV. A Radiometer PHM 28 pH meter was employed for pH
measurement at 25 ЊC using Radiometer type B electrodes,
which require correction of only 0.03 pH units even at pH 13.
pKa Values in the range pKa 4–8 were sometimes measured by
potentiometry, elsewhere and more generally by UV spectro-
photometry, using standard techniques.8 Water was glass dis-
tilled, while alcohols and all inorganic reagents were of AnalaR
grade. Temperatures, except at 100 ЊC (i.e. usually under reflux),
were measured to ±0.1 ЊC.
spot checks to have no appreciable effect on the reaction rate.
Some experiments at very low or very high pH were conducted
in HClO4–NaClO4 or NaOH–NaClO4 mixtures at I = 5.0 mol
dmϪ3. Aliquot samples were cooled rapidly to ambient tempera-
ture and their UV spectra examined in 1 cm quartz cells. Suc-
cessive UV traces were overlaid as a check on possible isosbestic
drift; this was never detected. The progress of the reaction was
then followed by the decline in the absorbance of 1 at some
wavelength in the range 220–260 nm; since reactant 1 and prod-
uct 2 both possess UV spectra, and both ionise, a variety of
wavelengths had to be employed. Only for 3d in NaOH–
NaClO4 mixtures at I = 1.0 mol dmϪ3 was the reaction followed
in situ, at 25 and 37 ЊC in a thermostatted UV cell, as well as at
83 ЊC by the method outlined above.
With the above exception, all work with 3d was carried out at
100 ЊC. Rates for the only other compound for which the full
rate–pH profile has been determined, 1d, were followed at 83 ЊC
at pH р 2 and pH > 11 and at 100 ЊC between these limits.
Temperature coefficients were measured under four conditions:
at 83, 90 and 100 ЊC for pH ca. 3.5 and ca. 6.5 at I = 1.0 mol
dmϪ3; and at 64, 72 and 83 ЊC for HClO4 and NaOH at 1.0 mol
dmϪ3 with I = 5.0 mol dmϪ3. These conditions define either rate
plateaux or constant {Hϩ} or {OHϪ}, so could be used for con-
structing the full rate–pH profile of Fig. 1. In this, pH values
are adjusted to 100 ЊC from their measured values at 25 ЊC with
corrections at pH р 2 according to Day and Wyatt 9 and at
pH > 2.5 from the temperature coefficients of Dempsey and
Perrin.10 pH Values at [Hϩ] and [OHϪ] у 1.0 mol dmϪ3 and
I = 1.0 or 5.0 mol dmϪ3 are those given by Rochester;11 the latter
incorporate the appropriate correction for pKw.8 Temperature
coefficients for 1d and 3d appear in Table 2 while those for a few
other compounds were measured on the plateaux at pH ca. 3.5
Kinetics
Because of the high temperatures (>60 ЊC) mostly employed,
sampling at appropriate intervals of time was mainly used for
reaction following. Buffer solutions, usually of 99 cm3 in a 100
cm3 volumetric flask, were equilibrated in a thermostatted bath
at the required temperature and the reaction was initiated by
adding 1 cm3 of a stock reagent solution in acetonitrile, also at
the required temperature. The reaction solution would typically
contain reagent at ca. 10Ϫ4 mol dmϪ3 and acetate or phosphate
buffer at ca. 10Ϫ2 mol dmϪ3 adjusted to I = 1.0 mol dmϪ3 with
sodium perchlorate. These buffer concentrations were shown by
1772
J. Chem. Soc., Perkin Trans. 2, 1997

View Article Online
conversion of 1 to 2. This demonstration is important, in that
the triazinedione 6 is known to hydrolyse to a ring-opened spe-
cies via covalent hydration under mild conditions.15 No such
reaction appears to take place here.
Identification of the reactive sub-species
In most studies of acyl transfer, there is no ambiguity concern-
ing the nature of the reactive species. Here however there are
tautomeric and conformational complexities which carry
mechanistic implications and so require prior discussion.
Fig. 1 Rate–pH profiles in water at 100 ЊC for 1d, circles, and 3d,
squares; open symbols, at I = 1.0 mol dmϪ3; full symbols, at I = 5.0 mol
dmϪ3
Ionisation and tautomeric form
Interrelationships between sub-species are set out in Scheme 1,
Table 2 Fitting constants at 100 ЊC,a and activation parameters, for
the hydrolysis of 1d and 3d
R1
N
R1
N
R1
N
Constant b
1d
3d
O
O
O
O
O
O
O
O
O
kA/sϪ1
2.96(0.35) × 10Ϫ2
20 320
1.14(0.22) × 10Ϫ2
Ka1
Ka2
+
–
Ea/cal molϪ1
∆S25 ‡/cal molϪ1 KϪ1
kB/sϪ1
HN
R2
NH
Me
N
NH
Me
N
N
– 0.95
4.53
Ϫ13.1
1.02(0.09) × 10Ϫ4
22 270
0.831(0.155) × 10Ϫ4
—
N
N
N
R2
R2
Ea/cal molϪ1
∆S25 ‡/cal molϪ1 KϪ1
kC/sϪ1
Me
Ϫ19.0
0.285(0.025) × 10Ϫ4
19 810
1A
1B
1C
Ea/cal molϪ1
∆S25 ‡/cal molϪ1 KϪ1
kD /dm3 molϪ1 sϪ1
Ea/cal molϪ1
∆S25 ‡/cal molϪ1 KϪ1
pKI
R1
N
R1
N
R1
N
Ϫ28.2
2.63(0.18) × 10Ϫ3
16 700
2.66(0.13)
11 250
Ϫ28.5
Ϫ0.74(0.14)
4.75(0.13)
O
O
O
O
O
O
Ka1
Ka2
+
–
HN
NH
N
NH
N
N
Ϫ27.5
Ϫ0.67(0.80)
4.69(0.21)
1.69
7.95
NHR2
NHR2
NHR2
pKII
2A
2B
2C
a
1 cal = 4.184 J. b Eqn. (1).
KT2
5.75
Ka4
13.7
Ka3
( 4.9)
KT1
( 3.2)
and ca. 6.5 and appear in the text. Rates were followed for 2–3
half-lives and pseudo-first-order rate constants (kobs) were
obtained by a least-squares procedure according to Swain’s
method.12
R1
N
R1
R1
N
O
O
O
N
O
HO
O
O
Data analysis
Ka5
HN
NH
N
–
NH
N
NH
The full rate–pH profile for 1d was analysed by ENZFITTER 13
according to eqn. (1). This contains as adjustable parameters
+
( 10.5)
NR2
NR2
N
R2
kA[Hϩ]2 ϩ kBKI[Hϩ] ϩ kCKIKII kD Kw
Me
(1)
kobs
=
ϩ
[H]2 ϩ KI[Hϩ] ϩ KIKII
[Hϩ]
2B′
2C′
1A′
Scheme 1 Measured and estimated values are for 1d, 2d and 4d; KT1,
Ka3 and Ka5 incorporate statistical factors
the plateau rate constants kA, kB and kC, the second-order con-
stant kD for attack of hydroxide ion, and the apparent ionis-
ation constants KI and KII, with pKw at 100 ЊC (cf. Fig. 1) taken
as 12.12. At the extremes of the pH scale, kobs is accelerated by
whose pKa and pKT values relate to compounds 1d, 2d and 4d
(cf. Table 1); the bracketed values are estimates, as derived
below. While the structure of 1 is unambiguous, 2 could exist as
2B or 2BЈ in the neutral form and as 2C or 2CЈ in the anion.
Hence there are problems in determining leaving group pKa
and these must be resolved.
We attempt their resolution as follows. We have previously
estimated basic pKa values of 7.05 and 9.9 for the tautomeric
acylguanidines 7 and 8, respectively.16 Further acylation of 8
to 9 is analogous to that of acetylguanidine (10a, X = COMe,
Y = H), pKa 8.20,17 to diacetylguanidine (10a, X = Y = COMe),
pKa 4.93,18 i.e. ∆pKa ca. Ϫ3.3; this analogy predicts pKa ca. 3.8
for 9. However, a closer model for 2 is the hypothetical species
11,‡ which differs from 9 in the alignment of its added carbonyl
group; we have estimated ∆pKa у Ϫ1.2 for this change in
geometry,16 hence pKa р 2.6 for 11. This is close enough to pKa
1.8 for ammelide 1220 to suggest the same tautomeric form for
the latter, given minor imponderables such as its extra ring NH
and the tendency of aromatisation to reduce basicity. Hence
pKa1 for 2A represents deprotonation to 2B, as written (Scheme
the change in ionic strength from I = 1.0 to I = 5.0 mol dmϪ3
;
this was factored out before analysis by applying corrections of
∆ log k = Ϫ0.16 at low pH and Ϫ0.20 at high pH so that the
derived fitting constants are all for I = 1.0 mol dmϪ3. A similar
procedure at low pH (∆ log k = Ϫ0.05) was used for 3d, whose
analysis employed the same equation except that the term in
kC is missing since 3d cannot form an anion. These fitting
constants, with their error limits, appear in Table 2. Activation
parameters were obtained from the temperature coefficients by
use of the Eyring equation 14 and are believed accurate to ±300
cal molϪ1 in ∆H ‡ and ±1 cal molϪ1 KϪ1 in ∆S ‡.†
Reaction products
The course of the reaction was followed for 1d in water by
HPLC for several half-lives at pH values corresponding to each
of kA–kD . Reactant 1d and hydrolysis product 2d produced well
separated peaks, at 1.6 and 2.8 min under the conditions
employed, and only these species were detectable. In this case
and others, UV spectroscopy also demonstrated quantitative
‡ The tautomeric form of 11 is likely to be disfavoured, making its pKa
value inaccessible; cf. pp. 132–133 (formula 211f) of ref. 19.
† 1 cal = 4.184 J.
J. Chem. Soc., Perkin Trans. 2, 1997
1773

View Article Online
O
O
O
Conformation in the neutral species
For the acyltriazinediones 1, the obvious conformation in non-
polar solvents is shown as I in Scheme 2. This not only contains
N
NH
NH2
HN
N
N
N
Me
NH2
NH2
O
7
8
9
N
N
N
N
XN
XNH
YNH
H
O
H
NH2
NH
θ1
θ1
YNH
N
N
Me
R2
R2
θ2
θ2
10a
10b
Me
N
O
H
N
I
II
O
N
O
O
O
O
O
NR
N
NH
N
NH
N
N
N
H
O
H
O
NH2
11
NH2
N
N
N
13
12
H
H
H
H
Table 3 IR and 13C NMR data for triazinediones
ν/cmϪ1 (CHCl3)
H
N
H
III
IV
δ/ppm (CDCl3)
R1
R2
NH
C᎐O a
C᎐N
᎐
C-2
C-4
N
N
N
N
N
O
᎐
H
O
H
H
5
3225
3180
3170
3145
3180
1683
1677
1673
1672
1672
N
N
H
1a
1f
1e
1h
1c
1d
2d
Pri
Pri
Pri
Pri
Pri
Pri
Pri
Me
Bui
Bus
But
Prn
Pri
H
H
H
H
1616
1612
H
H
V
VI
148.9
155.1
Scheme 2
151.9
Pri
a stabilising intramolecular hydrogen bond, but also avoids the
potentially destabilising clash between NH and R2 in II, its only
plausible alternative. The latter is shown in the E-conformation
so as to avoid the otherwise severe lone pair clash between
carbonyl and aza-nitrogen.
a
Ring plus exocyclic carbonyl.
1). Further evidence for 2B as the dominant species comes from
the near-identity of νC᎐N for 1e and 2d (Table 3); the switch in
structure between compound classes 7 and 8 raises νC᎐N by 50–
While the planarity expected for maximum resonance stabil-
isation is unlikely to be threatened where R2 is a linear alkyl
group, increasing rotundity might have one of three effects:
twisting away from θ1 = 0, the same for θ2, or an in-plane com-
pression of the relatively polarisable hydrogen bond. The first
two will weaken this bond, whereas the third will strengthen it.
While the second is improbable on resonance grounds, there is
evidence for both the other phenomena (Table 3). Compounds
1a, 1f and 1e represent R2 = linear, β-branched and α-branched
respectively; here νN H falls in sequence as the hydrogen bond
strengthens. Only for 1h with its very bulky neopentyl substitu-
ent does the effect reverse, pointing to some degree of twisting
out of plane. Even so, νN H is lower in 1h than in 5, where bond-
ing is weakened not by steric hindrance but by the different
bond angle at amide nitrogen dictated by the five-membered
ring. Nevertheless, even though α-branching may forcibly
strengthen this hydrogen bond, it also reduces the energy bar-
rier to rotation, as is shown by the collapse of the two ring
carbonyl 13C NMR signals for 1c to a single time-averaged
signal in 1d.
What counts for reactivity is how far this hydrogen bond may
survive in aqueous solution. Compounds 1 possess a strong
band at ca. 240 nm in cyclohexane which typically shows a
small bathochromic shift in water (Table 3) where a hypso-
chromic shift might have resulted through loss of conjugation
had twisting been severe; the deacyl reaction products 2d and 4d
do indeed show a shift to 229 nm. However, this criterion does
not distinguish between I and II if planar or near-planar, and
indeed 3d in which NMe replaces ring NH cannot form an
intramolecular hydrogen bond. Nevertheless, it may still prefer
some approximation to I as its dominant conformer, as afford-
ing the best compromise between planarity and steric strain. In
terms both of λmax and of reactivity at pH < 5, 1d and 3d are
almost identical.
21
80 cmϪ1
.
Calculation of pKa3 requires an estimate for pKT1. Species 2B
and 2BЈ are analogous to 7 and 8 in that C᎐N is deconjugated
᎐
from carbonyl in the second of each pair. If this analogy holds,
pKT1 = ca. 2.9, hence pKa3 = ca. 4.6. In fact there is evidence, for
the equilibrium 10a
10b, that 10a becomes more dominant
as both tautomers become less basic,17 so that these estimates
are probably minima. Statistically corrected, they appear in
Scheme 1.
The large reduction in the acidity of 2d brought about by ring
N-methylation to 4d shows that these do not share a common
anion. Since 4d can only give 2CЈ, its tautomer 2C must be
formed by 2d, leading to pKT2 = ca. 5.75 (Scheme 1). This is
large, presumably because the anionic charge in 2CЈ is much less
well distributed than in 2C. The unsubstituted triazinedione 6
possesses22 pKa 6.58, hence amino-substitution (relative to 2B)
is not greatly acid-weakening. Contrariwise, acylation is more
acid-strengthening than base-weakening (Scheme 1). Since
twisting out of plane of the whole acetylamino unit will reduce
its resonance donor power without much affecting its inductive
ability, this is part of the evidence, to be discussed below, that
twisting is much more severe in the anion 1C than in the cation
or neutral species.
With basic pKa values of Ϫ3.67 and Ϫ3.65 respectively, com-
pounds 1p and 1r for which R1 = aryl do not fit the above
pattern. Since pKa2 is not much affected, and neither is the
hydrolysis rate (vide infra), we can only suppose their cations to
possess the alternative O-protonated structure 1AЈ. Though
normal for amides, as, e.g., the related uracils 13,23 this is highly
unusual for guanidines16,17 and there seems no obvious explan-
ation for it. UV spectroscopy (Table 1) provides little help: while
1d shows a bathochromic shift on protonation but 1p and 1r
the reverse, a hypsochromic shift is also shown by 2d and by
compounds of type 7, which certainly N-protonate.16 Kinetic
consequences are considered below.
The partition coefficients of Table 1 provide unexpected evi-
dence as to the conformation. For R1 = Pri or Prn, and omitting
1774
J. Chem. Soc., Perkin Trans. 2, 1997

View Article Online
1j and 1k where R2 is cyclic, we obtain eqn. (2) where C, α and β
applies in all other cases, allowing for a variable, but perhaps
never very great, departure from planarity.
log P = Ϫ0.45(0.06) ϩ 0.46(0.02)C Ϫ
0.41(0.09)α Ϫ 0.07(0.06)β (2)
Leaving aside till later the precise mechanism of hydrolysis
for the neutral species, its variation with structure may be con-
sidered in this light. Available data are assembled in Table 4.
Use of the same three parameters employed for log P leads to
eqn. (3) where kB is expressed in units of 105 sϪ1 and the
(n = 11, r2 = 0.992, s = 0.10, F = 274)
are indicator variables that sum the total number of carbon
atoms, and of α- and β-branches, respectively. Here the coef-
ficient of C is close to the expected 0.53 for the value of the
methylene increment 24 while those of α and β should be com-
pared with Leo’s24 ‘branching factor’ Fcbr of Ϫ0.13. Clearly, the
coefficient of α is much greater than expected; it approaches,
in the opposite sense, the value of ϩ0.63 estimated by Leo 25
for the effect of an intramolecular hydrogen bond in 1,2-
disubstituted benzenes. If α-branching were to restrict solvation
of aza-nitrogen (cf. I), a rise in log P should result; its fall
suggests, contrariwise, that this solvation is maintained at the
expense of planarity, with some weakening of the intra-
molecular bond. Allowing for Fcbr, this appears equivalent to
about half the predicted full effect on log P, though we have no
means of knowing whether a full bond has been weakened by
one-half, a half-strength bond has wholly disappeared, or the
truth lies in between.
log kB = 0.05(0.04) ϩ 0.09(0.02)C Ϫ
0.36(0.05)α Ϫ 0.29(0.06)β (3)
(n = 6, r2 = 0.964, s = 0.04, F = 18)
above exclusions continue to apply. This is not a very good
equation but the best we can find, and far better than that
in the single parameter Es,27 for which r2 = 0.22. According to
eqn. (3), α- and β-branching both slow the rate about twofold,
whereas chain length has little effect. We consider kB for 1k
later.
The steric hindrance parameter Es was derived from the
hydrolysis of esters, and strictly applies only to the acyl group;
there is little evidence as to whether it holds even for the alkyl
moiety. Tertiary amides are necessarily different. The cases
compare as 18 and 19. Baldwin 28 has pointed out that, in
Important further evidence comes from log P for 1j and 1k.
The former could not be included in the regression since elision
of two hydrogen atoms necessarily lowers log P. As between
hexane and cyclohexane, ∆ log P = Ϫ0.44;26 ∆ log P = Ϫ0.37
between 1i and 1j is in line with this. Calculation for the piperi-
dyl derivative 1k is more complex. We start from the model
amides 14 and 15, for which log P may be reliably calculated 26
R
O
R3
N
R2
N
O
O
O
R2
R3
Me
18
Me
19a
Me
19b
amides, the expected trajectory of the approaching nucleophile
lies close to the acyl moiety as shown, and not bisecting the
angle, the result of considerable double bond character in the
nominal C᎐N linkage. So less variation in rate might be
expected here than for the defining reaction, hence the much
below unity coefficients in eqn. (3); and the structural factors
behind that variation could also be different.
NHCOMe
N
NHCOMe
14
15
as 0.95 for 14, and has been measured 18 as Ϫ0.05 for 15. Apply-
ing the difference to log P for 1j, that for 1k calculates as
log P ca. 0.6; the measured value is 2.33 (Table 1).
We believe the variation in rate for kB (and kC) depends on
‘entropic strain’;29 the ‘steric hindrance of motions’30 term RT
ln (ΠQ ‡) described by Taft as the entropic component in Es.
Neither kB nor kC in fact correlates with this, any more than
with the global term Es itself, but that is understandable given
the stereochemistry of 19. Any alkyl substituent R2 sweeps out
a volume which in part will be constricted by solvational freez-
ing in the highly structured transition state (vide infra). The
degree of this constriction will depend on distance from the
reaction centre, so that only the first few carbon atoms matter,
and branching has more effect than linearity. The equal weight
carried by α- and β-branching is just what would be expected
if all alkyl groups, not only cyclohexyl, are constrained into
roughly the conformation shown as III. This will help to
explain the very similar rates (Table 4) for R2 = isopropyl
(Es = Ϫ0.47) and cyclohexyl (Es = Ϫ0.79), and also why no
special penalty attaches to the bulky isobutyl substituent
(Es = Ϫ0.93), since if full rotation is not possible, its γ-carbon
atom can restrict attack at only one face. If conformer V or VI
dominates for 1k, its tenfold faster reaction rate than that of 1j
becomes understandable.
We believe the explanation to lie in a major shift in conform-
ation between 1j and 1k. For 1j, the conformation shown in
Scheme 2 as III is that which best ‘tucks away’ the cyclohexyl
moiety (or any other) in such a way as to maximise the disper-
sion interaction with methyl while also allowing maximum
access of water to the hydrophilic portions of the molecule.
However, the same conformation for 1k, as IV, leads to a severe
lone pair clash between hydrazine nitrogen and that of the ring.
This can be relieved by the reversal in conformation shown as
V, but only at the cost of shielding both lone pairs. We may
calculate this cost in terms of log P. Uracil 16 and 5-azauracil
17 possess26 log P = Ϫ1.05 (average) and Ϫ1.87 respectively; if
H
N
H
N
O
O
O
O
NH
N
NH
16
17
this difference, ∆ log P Ϫ0.82, is attributed to aza-nitrogen, its
subtraction from the above estimate leads to a new calculated
value for 1k of log P ca. 1.4. The difference remaining, of ∆ log
P ca. 0.9, is similar to that between 14 and 15 and provides
evidence that the hydrazine nitrogen atom is also shielded from
the solvent. An alternative possibility which may better
accommodate the kinetic results (see later) is the variant on II
shown as VI; with, in each case, one carbonyl and no nitrogen
lone pairs accessible to the solvent, a similar effect on log P
might be expected. If V or VI belongs uniquely to 1k, then it
seems probable that III or something closely approaching it
Conformation in the anion
No intramolecular hydrogen bond is possible in the anion,
while heavy solvation of the anionic charge, plus the inability
of the amide moiety to share it, both militate strongly against
planarity. The dominant conformation may not be fully
orthogonal (θ1 ca. 90Њ) but will certainly be such that no special
conformation for 1k is now required. It will be seen (Table 4)
that kC for 1k (cf. kB) is now quite normal.
On the same lines as eqn. (3), eqn. (4) for kC may be formu-
J. Chem. Soc., Perkin Trans. 2, 1997
1775

View Article Online
Table 4 Values of kB and kC of some triazinediones
At 72 ЊC
At 62 ЊC
kC/10Ϫ6 sϪ1
8.4b
R1
R2
ϪEs
kB/10Ϫ5 sϪ1
kC/10Ϫ5 sϪ1
a
1a
1c
1d
1j
1k
1f
1e
1g
1o
1p
1q
Pri
Me
0.00
0.36
0.47
0.79
c
0.93
1.13
1.98
0.39
0.93
1.74
1.4b
2.0b
Pri
Prn
1.93
3.05
Pri
Pri
0.89b
0.932
9.78
0.33b
0.703
1.01
1.38b
Pri
Cyclohexyl
1-Piperidyl
Bui
Pri
Pri
1.28
1.61
Pri
Bus
1.02b
1.45b
0.407b
0.537b
1.55b
1.88b
13.0
8.8
Pri
CHEt2
C6H4-4-Me
C6H4-4-Me
C6H4-4-Me
Bun
Bui
2.40
2.04
0.257
CH2But
0.89
a
Re-scaled to Me = 0 from the compilation of S. H. Unger and C. Hansch, Prog. Phys. Org. Chem., 1976, 12, 91. b Calculated from results at other
temperatures. c Unknown; presumably similar to cyclohexyl.
log kC = 0.20(0.02) ϩ 0.10(0.01)C Ϫ
0.97(0.02)α Ϫ 0.38(0.02)β (4)
constrained to be planar. Possibly therefore the angle of twist
(θ1) in the anion, whatever it may be, is substantially the same
through the series. The kinetic consequences of this and other
conclusions are explored below.
(n = 6, r2 = 0.999, s = 0.14, F = 1450)
lated. This differs from eqn. (3) in showing much greater sensi-
tivity to α-substitution. The reason for this may lie in the much
greater ‘sweep volume’ possible when the amide moiety is twist-
ed out of plane, since relatively free rotation of R2 is now per-
mitted. This matters most for the α-branched substituents since
these are the most rotund. The continued importance of both
forms of branching, however, demonstrates that the carbonyl
alignment shown for I still dominates, not that for II. This has
important stereochemical implications for intramolecular cata-
lysis (see later). A reversal in the amide conformation—i.e. 19b
replacing 19a—would result in the triazine ring, as R3, now
lying in the path of the nucleophile, in which case the nature of
R2 would scarcely matter. This preference for I even when non-
planar is presumably due to solvational factors: the hydrophilic
portions of ring and substituent can be serviced by a common
solvation shell, while dispersion interactions can mutually sta-
bilise methyl and R2. There is a sense in which these two hydro-
phobic moieties form a single giant substituent. In the self-
association of peptides in solution, it is known that like attracts
like.31
Kinetics and mechanism
Amide hydrolysis typically shows a V-shaped rate–pH profile
with roughly equal sensitivity to acid and alkali (kH ≈ kOH ).32 At
a pH low enough for full protonation of the substrate, a rate–
pH plateau appears for water attack on the amide cation,32 here
designated kA. Fortuitously, the actual values of kH , kA and kD
for 1d closely resemble kH ,32,33 kA32,33 and kOH 32,34 for acetamide,
though kD differs from kOH in involving hydroxide ion attack on
the anion. Only activated amides show a ‘water rate’, i.e. a cen-
tral rate–pH plateau that appears not to involve Hϩ or OHϪ,
and this is always accompanied by much greater sensitivity to
alkali; for acetylimidazole 20,35–37 kOH is of the order of 107
times greater than for acetamide.
The acetyltriazinediones 1 are possibly unique in showing
two such ‘water rates’, kB and kC (Fig. 1), and this despite their
evident lack of activation. These rate plateaux correspond to
nominal water attack on neutral species and anion respectively;
pKII = 4.69 at 100 ЊC for 1d (Table 2) corresponds closely
enough to pKa2 = 4.53 at 25 ЊC (Table 1) to demonstrate their
identity. However, the large attenuation that might be expected
for kC relative to kB is absent (Fig. 1 and Table 4). And, while
the ring-methylated derivative 3d behaves similarly to 1d at
pH < 5, it too shows an apparent pKII = 4.75, which here relates
to no change in its UV spectrum and no possible ionisation
process.
log (kC/kB) = 0.15 ϩ 0.01C Ϫ 0.61α Ϫ 0.09β
(5)
Subtraction of eqn. (3) from eqn. (4) gives eqn. (5). This
shows what simple inspection of Table 4 will also reveal: that α-
branching is the main factor in reducing kC relative to kB. These
equations may be compared with eqn. (6) for pKa; a very poor
There are, therefore, unexpected features in the kinetics of
this reaction that point to mechanistic complexities of an
unusual sort. In the following sections we attempt to unravel
them. Some data useful for the purpose of cross-comparison
are assembled in Table 5.
pKa = 5.17(0.14) Ϫ 0.01(0.04)C Ϫ
0.42(0.17)α Ϫ 0.41(0.12)β (6)
(n = 12, r2 = 0.72, s = 0.21, F = 6.9)
The rate plateau in acid: kA
The amino heterocycle 2 bears some structural resemblance to
the leaving groups of the acylazoles 20–22 studied by Jencks
equation, but the only one we can find. Given R1 = alkyl, vari-
ation in pKa can have nothing to do with electronic factors and,
indeed, most closely reflects the same steric factors as kB. To
some extent the resemblance between eqns. (3) and (6) must be
fortuitous, since their negative coefficients reflect different fac-
tors: steric hindrance to nucleophilic attack for kB, weakening
of the intramolecular hydrogen bond in the case of pKa. Never-
theless, since pKa reflects the difference in ∆G between neutral
species and anion, this does seem to suggest that structural
variation affects the free energy of the former much more. If
indeed that of the ionised species is nearly a constant, this may
reflect release of steric strain once the amide group is no longer
+
N
NCOMe
20
MeN
NCOMe
21
N
NCOMe
N
22
and co-workers,35–38 so that mechanistic parallels might be
anticipated. The near-equivalence in hydrolysis rate of 21,36
and of 20 as its cation,35,37 was used to demonstrate that the
observed general base (GB) catalysis involves partial proton
1776
J. Chem. Soc., Perkin Trans. 2, 1997

View Article Online
Table 5 Microscopic rate constants (dm3 molϪ1 sϪ1) calculated for 25 ЊC
Constant
Status
1d
3d
1e
1p
20a
21b
22c
kH
[A][Hϩ]
5.5 × 10Ϫ6
5.3 × 10Ϫ7
9.6 × 10Ϫ10
4.5 × 107
6.3 × 108
6.3 × 10Ϫ10
1 × 102
4.1 × 10Ϫ6
1.5 × 10Ϫ5
k1 = kA/[H2O]
k2 = kB/[H2O]
k2Ј = kBKa1/Kw
k2Ј = kBKa1/Kw
k3 = kC/[H2O]
k3Ј = kCKa2/Kw
kOH = kD
[HAϩ][H2O]
[A][H2O]
6.8 × 10Ϫ4 d
1.8 × 10Ϫ6 e
1.2 × 106
8.4 × 10Ϫ4
1.5 × 105
6.8 × 10Ϫ3
3.1 × 10Ϫ5
2.6 × 1011
6.3 × 10Ϫ10
3.5 × 107
2.8 × 10Ϫ9
7.4 × 1010
1.5 × 1012 g
3.5 × 10Ϫ9
1.5 × 103
[HAϩ][OHϪ]
[HAϩ][OHϪ] f
[AϪ][H2O]
[A][OHϪ]
6.0 × 108
—
3.8 × 10Ϫ10
48
[A][OHϪ]
5.8 × 10Ϫ2
2.7 × 102 h
2.7 × 103
kD
[AϪ][OHϪ]
9.1 × 10Ϫ6
8.1 × 10Ϫ4
a
Ref. 37 (I = 1.0 mol dmϪ3). b Ref. 36 (I = 0.2 mol dmϪ3). c Ref. 38 (I = 1.0 mol dmϪ3). d k = 8.4 × 10Ϫ4 at I = 0.2 mol dmϪ3 (ref. 35). e k = 1.5 × 10Ϫ6 at
I = 0.2 mol dmϪ3 (ref. 35). f At 100 ЊC. g Assumes pKa1 unchanged at 100 ЊC. h k = 3.2 × 102 dm3 molϪ1 sϪ1 at I = 0.2 mol dmϪ3 (ref. 35).
removal from water as nucleophile, not from the imidazolium
cation.37b Values of βcat = 0.34 for the hydrolysis of 20 cation,37a
and of βlg ca. Ϫ0.35 for 20 and 22 as neutral and cationic spe-
cies,38 point to an early, nearly symmetrical transition state for
addition of nucleophile, with most of the cationic charge still
on the leaving group. The equivalent transition state in the pres-
ent case may be written as 23 and, again, will explain the near-
equivalence of kA (Table 2) for 1d (23, R = H) and 3d (23,
R = Me).
+
N
N
N
N
C
R
H
R
H
N
N
O δ–
Me
R2
R2
O +
H
H
C
O
δ+
H
O
Me
24
H
Fig. 2 Hydrolysis rate at 25 ЊC for the cationic and neutral forms of
20, open circles; 22, filled circles; and 1d, squares
23
Jencks and Carriuolo 35 rejected the alternative A1 fission of
20 cation (cf. 24) on a number of grounds which included the
very large negative ionic strength effect and ∆S ‡ = Ϫ30.2 cal
is probably still present in the neutral species 1B, no such bond
is possible for 3d, yet kA and kB are little changed (Fig. 1). So
this hydrogen bond has no kinetic consequences. Jencks and co-
workers37 have demonstrated that GA catalysis by partial pro-
ton transfer to carbonyl in the hydrolysis of 20 does not occur.
Others40,41 have noted its inefficiency in this context, the reason
molϪ1
K
Ϫ1. For 1d, ∆S = Ϫ13.1 kcal molϪ1 KϪ1 while a rate
increase of ca. 45% between I = 1.0 and I = 5.0 mol dmϪ3 (ca.
10% for 3d) compares with a 16-fold fall for 20 cation between
I = 0.2 and I = 4.8 mol dmϪ3. Nevertheless, we believe that 23
continues to represent the transition state. Most of the charge
in 21 or 20 cation will be concentrated on the distal nitrogen
atom, whose charge density will then fall sharply in the transi-
tion state as the remaining charge spreads itself over both
nitrogen atoms. Here, with an extra nitrogen atom to share the
charge, an already low charge density will be relatively little
affected at the small degree of C᎐N bond fission likely to be
attained.
being that it will tend to stabilise reactant ( C᎐O ؒ ؒ ؒ HNϩ)
᎐
and transition state (уC᎐OϪ ؒ ؒ ؒ HNϩ) to much the same
extent. Hence intramolecular GA catalysis in series 1 is likely to
be minimal.
The reaction of 1d and 3d with hydroxide ion: kD
The generally accepted mechanism of alkali-catalysed amide
hydrolysis is shown as Scheme 3. The anions of simple amines
Second-order rate constants for the attack of water on 1d, 20,
22 and their cations (k1 and k2 of Table 5) are shown in Fig. 2
as a function of pKlg. Following Fox and Jencks,38 pKlg for 22
cation is taken as that of the disfavoured tautomer, pKa ca. 3.6,
while pKa = 4.9 is similarly adopted for 1d (cf. 2BЈ in Scheme 1).
As stated,38 βlg ca. Ϫ0.35 is found for imidazole and triazole.
The results for 1d fall below this line at a separation of
approaching 104. Gravitz and Jencks39 have demonstrated that
imidazole is a poorer leaving group by a factor of ca. 104 rela-
tive to a ‘point-charge’ amine of equal pKa. This further drop in
rate by a similar margin when three nitrogens are available to
share the charge reinforces their argument, and explains why
the triazinediones are far less susceptible to acid or alkaline
hydrolysis than the electronegativity of the ring might have
suggested.
A planar conformation with carbonyl locked tightly into an
intramolecular hydrogen bond is almost obligatory for the cat-
ion 1A (Scheme 1), and little difference is expected for 23 if, as
suggested above, the three nitrogen atoms are well conjugated.
Even for 3d (23, R = Me) this state of affairs may not be greatly
affected. Intramolecular general acid (GA) catalysis is therefore
a possibility. However, while as seen above this hydrogen bond
+
N
H
OH –
O
O –
O –
N
N
R
O –
–
NH + RCO2
R
OH
R
25
N – + RCO2H
Scheme 3
are very poor leaving groups and require prior protonation;42
a
proton switch to give 25 as the crucial intermediate was first
suggested by Bender and Thomas.43 However, since imidazole
and 1,2,4-triazole possess acid pKa values (14.2 and 10.1,
respectively)38 of the same order as alcohols, these are excep-
tions to the rule, and are able to leave as their anions. The same
should be true of series 1 and 3d, given pKa ca. 13.7 for the
disfavoured tautomer 2CЈ (Scheme 1) which is the actual leav-
ing group. It is notable that kOH for 3d is almost 5000 times less
than that for 2037 (Table 5) despite its very similar pKlg, so that
the cationic (see above) and anionic forms of the acyltriazinedi-
ones appear to show about the same gain in stability.
J. Chem. Soc., Perkin Trans. 2, 1997
1777

View Article Online
A similar mechanism for hydroxide ion attack on the anion
of 1 would require that 2 depart as its dianion. This can be
avoided by a mechanism (Scheme 4) analogous to that of
here, allows an alternative approach. Applied to 20 it leads to a
predicted k2Ј = 2.2 × 105 dm3 molϪ1 Ϫ1, not far from the value
s
for 21 quoted above. Applied to 1d, k2Ј = 535 dm3 molϪ1 sϪ1
results. That required to account for kB is k2Ј = 4.5 × 107 dm3
molϪ1 Ϫ1, which decisively rebuts the cationic rate alternative.
s
–
–
Even more decisively, for 1p, re-interpretation of k2 as attack by
OHϪ gives k2Ј = 7.4 × 1010 dm3 molϪ1 sϪ1 which lies beyond the
encounter rate limit, the result of this compound’s very low
basic pKa. A similar calculation for the acetyltriazole 22 would
OH –
N
N
N
N
H
O
N
O
N
R2
R2
Me
O –
Me
give k2Ј = 2.6 × 1011 dm3 molϪ1 Ϫ1, with similar mechanistic
s
consequences.
Keq
The interpretation of kC for 1d may be addressed directly.
Replacement of k3 for [AϪ][H2O] by k3Ј for [A][OHϪ] gives
k3Ј = 1.0 × 102 dm3 molϪ1 Ϫ1, but the actual kOH value for
s
N
N
N
N
3d, which cannot ionise, is 5.8 × 10Ϫ2 dm3 molϪ1 Ϫ1, so falls
s
H
H
–
short of the required rate by a factor of ca. 1700. Hence kC
also is what it appears to be. Applied to the calculation of
kOH from kO for 20 and 22, where intramolecular catalysis is
unlikely, eqn. (7) gives results that fall short of the actual value
by factors of 5.5 and 5.9, respectively; quite possibly therefore,
a small adaptation of this equation would allow its generalis-
ation from esters to activated amides, and perhaps to amides
generally.
N
N
O –
R2
R2
Me
O –
–
+ MeCO2
26
Scheme 4
Scheme 3. Evidence of this route for 1d comes from the slowing
of this reaction, by ca. 20%, in 50% aqueous methanol; every-
where else in the rate–pH profile, methanol accelerates the rate
(see below). Formation of the dianion 26 (or 25) is specific to
water. Relative to 3d, kD is lower by ca. 6.4 × 103. This margin
is much greater than the ca. 102 that is normal for the effect of a
negative charge in otherwise comparable systems44 and in part
reflects the reduced efficiency entailed in the need for an extra
reaction step; using our recent methodology,45 for the proton
switch step of Scheme 4 we estimate Keq ca. 10Ϫ5. Necessarily,
the transition state for 1d is much later than that for 3d. Never-
theless, despite the unit difference in charge between them, their
∆S ‡ values are similar, and not abnormally negative (Table 2);
most of the difference in rate is contained in ∆H ‡, presumably
reflecting the electron deficiency at the carbonyl. Possibly the
strong intramolecular hydrogen bond expected for 26 acts as
a form of internal solvation. Consistently with this theory, a
rise in ionic strength for 1d from I = 1.0 to I = 5.0 mol dmϪ3
increases kD by only ca. 60%, an acceleration scarcely more
than that found by Jencks and co-workers35,37 for the mono-
anionic process involved in the alkaline hydrolysis of 20.
The extent of intramolecular catalysis
Perhaps the most surprising feature of these rate–pH profiles (cf.
Fig. 1) is the very small effect of anion formation on kC relative
to kB (Table 4), bearing in mind the deactivation that might have
been anticipated. It is difficult to quantify this discrepancy. One
possible approach starts from the use of eqn. (7) in reverse. For
1d, kD is not as it stands a reasonable model for kOH , since elec-
trostatic repulsion must depress the rate and perturbations may
also be introduced by the special mechanism of Scheme 4. Cor-
recting for the first using the factor of ca. 102 noted above44 gives
kOH ca. 10Ϫ3 dm3 molϪ1 Ϫ1, leading to an ‘expected’ (uncata-
s
lysed) k3Ј ca. 10Ϫ13 dm3 molϪ1 Ϫ1. This falls short of the
s
observed k3Ј (Table 5) by ca. 6 × 103. A similar calculation for
aspirin, starting from kOH = 0.11 dm3 molϪ1 sϪ1 at 25 ЊC,47 leads
to kO ca. 2.2 × 10Ϫ11 dm3 molϪ1 Ϫ1; given an observed plateau
s
rate of ca. 3.7 × 10Ϫ6 sϪ1 47 which translates to k3Ј ca. 6.7 × 10Ϫ8
,
dm3 molϪ1 sϪ1, a rate enhancement of ca. 3 × 103 results. Even if
this calculation somewhat exaggerates the true rate enhance-
ment for the triazinediones, their kC values may still compare
with aspirin in the degree of intramolecular GB catalysis they
show, a point very relevant to our interest in them.§
The ‘water rates’ kB and kC: kinetic ambiguity
Before the question of catalysis in the middle pH region can be
addressed, we have to establish the nature of the reactive sub-
species. While kB and kC nominally represent attack of water on
neutral and anionic substrates respectively, each is capable of
another interpretation, as attack of OHϪ on cation or neutral
molecule. These alternatives, as k2 or k2Ј for kB and as k3 or k3Ј
for kC, are set out in Table 5 for some triazinediones and refer-
ence compounds. We shall restrict discussion so far as the data
allow to 25 ЊC and I = 1.0 mol dmϪ3, so as to permit the most
level comparison possible.
From kOH = 5.8 × 10Ϫ2 dm3 molϪ1 sϪ1 for 3d, eqn. (7) gives kO
ca. 1.0 × 10Ϫ11 dm3 molϪ1 sϪ1. Since 1d and 3d possess virtually
identical kB values, this prediction for the uncatalysed rate may
be transposed to 1d. Hence the extent of catalysis for kB = k2/
kO р 102, a notably smaller enhancement than that for kC. This
combination of rate enhancements of р102 for kB, у103 for kC,
and electrostatic repulsion of the order of 102 for the latter, will
explain why kB and kC are in practice so similar, with kB > kC
common but not universal (Table 4).
Wolfenden and Jencks36 interpret the ‘water rate’ for the
hydrolysis of acetylimidazole 20 in the following way. Re-
calculated from [A][H2O] to [AHϩ][OHϪ], k2Ј = 1.2 × 106 dm3
The nature of intramolecular catalysis
Part of the evidence for intramolecular GB catalysis in aspirin
hydrolysis comes from the effect of aqueous alcohols on the
plateau rate. In 50% aqueous alcohols, Fersht and Kirby49
found rate accelerations in the order MeOH > EtOH > PriOH >
ButOH, and we find similarly for the triazinediones (Table 6).
The rationale for this phenomenon lies in the known 50,51 non-
linear relation, for highly basic alkoxides, between basicity and
molϪ1 Ϫ1. This compares with an observed k2Ј = 1.5 × 105 dm3
s
molϪ1 sϪ1 for the cation 21. If 21 is a good model for 20 cation,
as on a number of criteria it seems to be,35–37 then hydrolysis of
neutral 20 is substantially faster than attack of OHϪ on its
cation would permit; hence the ‘water rate’ is largely what it
claims to be.
Such a simple and elegant demonstration is not possible for
1d. However, Bruice and Holmquist 46 have derived eqn. (7) for
§ These rate enhancements are not of course EMs, which measure
intramolecular vs. intermolecular catalytic efficiency; that for aspirin is
only ca. 13.48 Nevertheless, they are of interest as very visible evidence
of intramolecular catalysis relative to no catalysis at all. The present
reaction does not show catalysis by those external buffers studied at
log kOH = 0.84 log kO ϩ 8.0
(7)
the relation between the second-order rate constants for water
(kO) and hydroxide ion (kOH ) attack on esters, which if valid
concentrations of the order of 10Ϫ2 mol dmϪ3
.
1778
J. Chem. Soc., Perkin Trans. 2, 1997

View Article Online
Table 6 Rate enhancements in 50% aqueous alcohola
r
R1
R2
pKa
k
MeOH
EtOH
8.5
PriOH
1.4
ButOH
0.6
T/ЊC
1a
1d
Pri
Me
5.60
Ϫ0.95
4.53
kC
kA
kC
kD
kC
kC
kC
kC
kH
kB
kC
kC
kC
kB
49
3.7
53
0.8
49
32
20
46
3.0
12.5
50
62
37
76
37
76
76
76
62
37
62
62
76
39
25
Pri
Pri
Pri
Pri
11.6
1.8
Pri
Pri
1f
1j
1k
1o
1p
Pri
Bui
4.72
4.38
5.26
4.46
Ϫ3.67
8.5
10.6
4.2
1.4
1.2
1.6
1.6
0.6
1.2
Pri
Cyclohexyl
1-Piperidyl
Bun
Pri
C6H4-p-Me
C6H4-p-Me
C6H4-p-Me
C6H4-p-Me
C6H4-p-Me
10
Bui
Bui
4.3
8.4
8.2
3.5
1.4
2.1
1.4
1.3
2.3
Bui
4.11
3.61
3.59
0.6
0.8
1q
CH2But
40
10.2
2.6
Aspirin b
p-Nitrobenzoyl chloridec
a
Reaction at temperature T/ ЊC specified; for definition of r, see eqn. (8). b Ref. 49. c Ref. 52.
nucleophilicity towards carbonyl: the latter, while much greater
than for hydroxide, is not far from constant at pKnuc > 13,50a
and this ‘break-point’ is little affected by pKlg.50b Hence, at con-
stant pH (or, as here, at constant catalyst pKa), reactivity is
largely a function of [ORϪ] and does not cancel exactly against
this as would happen for βnuc = 1.0. This almost level nucle-
ophilicity has been attributed to the heavy solvation of these
very basic, point-charge anions, which will also explain why
OHϪ is still less reactive.51 It is not known, however, how PriOϪ
or ButOϪ compare in this respect, and whether steric hindrance
or solvational differences come into play.
In order to calibrate these results, it would help to be able to
quote neutral uncatalysed rates for the attack of aqueous alco-
hols at carbonyl. Information on this topic is sparse. Bentley
and Jones52 have studied the neutral hydrolysis in aqueous
methanol and aqueous ethanol of 4-nitrobenzoyl chloride,
whose solvolysis follows exclusively the addition–elimination
pathway so is pertinent here. In each solvent, the dominant
process is water-catalysed addition of alcohol; overall acceler-
ation in 50% methanol and 50% ethanol is 2.6- and 1.4-fold
respectively [r of eqn. (8)]. This compares with corresponding
O
δ–
O
H
O
R
O
O
δ–
Me
27
approximated as 27.49 This is written to imply partial bond
formation and little or no bond fission, as is normal for ester
hydrolysis.41,50,55 If r is some guide to the relative extent of C᎐O
bond formation, then kB for 1p is comparable in this respect to
aspirin, whereas kC throughout implies considerably more. Yet
this result is achieved, as it were, by remote control; at least for
kB, a single solvent molecule cannot be involved. Structure 28
δ+
N
N
H
H
O
H
O
N
R2
C
H
δ–
O
r = k50% ROH /kwater
(8)
Me
R
ratios of 10.2 and 3.5 for aspirin,49 consistent with the incursion
there of intramolecular GB catalysis. There are few data for
other alcohols, and none in water. In the direct reaction of ROH
with MeCO2Me, the reactivity ratio for R = Me, Et and Pri is
about 3:2:1,53 while the same three alcohols span a reactivity
range of about two in their reaction, as cyclic trimers, with p-
chloroisocyanate in diethyl ether.54 Except possibly for tert-butyl
alcohol, on which there is no information, it therefore seems
improbable that steric factors are of compelling importance.
The r-factors of 3.7 on kA for 1d in 50% aqueous methanol,
and of 3.0 on kH for 1p (Table 6), coming close as they do
to r = 2.6 as reported by Bentley and Jones,52 provide further
evidence for the absence of intramolecular GA catalysis in
the cation and, by implication, in the neutral species also. As
noted above, the near-identical kB values for 1d and 3d show
that the intramolecular hydrogen bond of which series 1 is cap-
able, and which we show above probably persists in water, has
no catalytic consequences. (It may have structural con-
sequences; see below.) Hence the rate enhancement shown by kB
for 1p, and which possesses a sensitivity to aqueous alcohol
comparable to aspirin (Table 6), must result from general base
catalysis. This is a remarkable achievement for a base of
pKa < 0. It is further compounded by the probable stereo-
chemistry of the process.
28
shows the smallest feasible solvent bridge for the neutral species
and is drawn just prior to attack, without prejudice to the tim-
ing of the transition state. There are precedents of a sort. Neu-
tral ester hydrolysis is thought to involve two water molecules;41
that of neutral 20 is supposed to go the same way;37b while
Engberts and co-workers56 provide evidence for a similar pro-
cess in the hydrolysis of an acyltriazole related to 22. However,
none of these involve the steric constraints dictated by an
intramolecular process. A multiple solvent bridge must cer-
δ–
Me
O
N
H
N
O
H
O
R
H
O
Me
O
δ+
H
H
N
O
H
30
29
N
NH
N
N
H
N –
N
O –
R2
R2
The stereochemistry of aspirin is almost ideal for incorporat-
ing one molecule of solvent in a transition state which may be
RO
32
Me
31
J. Chem. Soc., Perkin Trans. 2, 1997
1779

View Article Online
tainly be involved in the intramolecular GB catalysed meth-
anolysis, in methanol, of 29,57 while the structure shown as 30 is
considered by Fife et al.58 to account for the almost zero effect
of substituents on acetylimidazole’s neutral hydrolysis rate. The
concerted solvolysis of p-chloroisocyanate by an alcohol tri-
mer 54 takes place in an otherwise aprotic solvent where this is
not difficult to envisage.
The purpose of GB catalysis in the addition of water or an
alcohol to carbonyl is to avoid the production of an inter-
mediate whose very low pKa value, of <Ϫ2, would make it too
unstable to exist.59 Full or partial removal of a proton by the
catalyst converts the nucleophile all or part of the way to a
species of pKa > 15 which will give rise to no such problem.
Hence the catalyst must possess a pKa value somewhere
between these extremes.59 The carboxylate group of aspirin,
pKa = 3.59,41 falls in this range, as do the triazinedione anions,
pKa 3.6–5.6, as relevant to kC. However, it is difficult on the
face of it to see how efficient catalysis can result for the neu-
tral species (kB) given pKcat < 0, especially with the stereo-
chemical constraints described above. It is impossible to under-
stand it for 1p, pKcat Ϫ3.67,¶ where proton removal by water
should be more efficient than that by the ‘catalyst’, yet where
catalysis seems in practice to be of similar efficiency to that in
aspirin.
We believe the explanation, at least for kB, to lie in a struc-
tural feature of the triazinediones for which we are aware of no
precedent. Here the catalyst is not merely part of the substrate,
as for aspirin, but part of the conjugated unit which comprises
the leaving group itself. As the leaving group starts to leave,
pKcat also and of necessity changes with it: from a value, for 1d,
of Ϫ0.95, towards a limiting value of perhaps ca. 5|| (cf. struc-
ture 2BЈ). Somewhere between these limits, according to the
degree of C᎐N bond breaking in the transition state, lies the
effective pKa value of the catalyst. That is, pKcat will tend to rise
towards pKlg as the addition step proceeds; a prime example of
the principle that ‘catalysis occurs where it is most needed’.59a
For aspirin, in contrast, pKa = 3.69 becomes pKa = 3.02 in the
product salicylic acid, a change actually in the opposite direc-
tion. We may possibly represent the transition state for kB as
resembling that for kA in that C᎐N bond breaking is more or
less in line with C᎐O bond formation, but differing from it in
that both steps are much more advanced, as is necessary if
catalysis is to be efficient enough to overcome the stereochemi-
cal problems discussed above. A transition state involving a
high degree of C᎐N bond breaking accompanied by an equally
high degree of C᎐O bond formation, the latter but not the for-
mer signalled by r, might explain why r for kB is so very similar
to that for aspirin.
adopt a conformation approaching that of 31. This represents
an unconjugated, point-charge anion of high charge density,
which probably could not leave at all except via the mechanism
of Scheme 3; a route ruled out by the accelerating effect on kC of
alcohols, as discussed above in connection with kD . MO calcul-
ation at the 3-21G level of refinement suggests θ1 ca. 78Њ in the
(unsolvated) anion 1C which, as water adds to carbonyl, falls
towards θ1 ca. 26Њ even for R2 = But with concomitant transfer
of a proton to ring nitrogen. Exocyclic nitrogen remains almost
trigonal, implying minimal C᎐N bond fission. This progressive
collapse in dihedral angle suggests the possibility that, for kC,
only a single water (or alcohol) molecule may be required as a
bridge; i.e. as for aspirin 27 and contra 28 (kB). If so this helps
to explain why, allowing for electrostatic repulsion, catalysis is
more efficient in kC than in kB, as noted above. While tetra-
hedral intermediates tend to present as potential minima in cal-
culations of this sort,62 whether or not the transition state in
solution is approximated by TϪ, the much increased r value for
kC relative to kB also provides evidence that C᎐O bond form-
ation is essentially complete. (There is one exception: the appre-
ciably lower value of r for 1k than that found elsewhere may
reflect some repulsion between the nitrogen lone pair of the
1-piperidyl substituent and the attacking reagent.) This fully
formed TϪ can then rotate to give the favourable, near-planar
conformation 32, after which rate-limiting loss of 2CЈ follows.
Interestingly, and despite the longer solvent bridge for kB, ∆S ‡
on our limited evidence is always more negative for kC: Ϫ19.0
and Ϫ28.2 cal molϪ1 KϪ1 respectively for 1d, Ϫ26.7 and Ϫ31.6
cal molϪ1 KϪ1 for 1p. This again implies a late transition state
for kC, in which electrostriction of the solvent by the anionic
charge can exert its maximum effect.
The hydrolysis of 3d: further light on kB
An unexpected feature of this reaction was the discovery of an
inflection in the rate–pH profile for 3d at a position almost
identical to that for 1d (pKII of Table 2). Yet 3d cannot ionise, so
that this must represent some change in rate-determining step.
The problem is two-fold: what the nature of this step may be;
and whether, or how far, it may affect the interpretation we have
placed on the break-point between kB and kC for 1d and its
analogues.
Our postulate for 3d** is sketched in Scheme 5, which starts
δ+
+
N
N
N
N
O
N
N
H
H+
H
Me
H
Me
H
Me
δ–
O
N
O
N
H
N
R2
C
R2
R2
HOAc
HO
Me
Me
As seen above, the anomalously fast kB for 1k may be
explained in terms of conformer VI; here twisting of the acyl
moiety might allow a transition state analogous to 27. While
the alternative conformer V would remove the steric hin-
drance present in III or IV, it is difficult to see how intra-
molecular catalysis could survive shielding of the nitrogen lone
pair. An alternative explanation in terms of the ‘α-effect’60 is
discounted because this does not seem to enhance leaving
group ability,39 and since no such increase in rate is found
for kC.
The anion 1C (Scheme 1) lacks the intramolecular hydrogen
bond of 1A or 1B, and in addition, electrostatic repulsion may
produce a high degree of twist (θ1 ӷ 0 in Scheme 2); cf. above
discussion of eqn. (4). If the synchronous mechanism adduced
for kB were to prevail, then by the principle of least motion,61
the amine anion at its moment of departure would have to
33
34
35
Scheme 5
from 33, the transition state for kB delineated above. Fission of
33 as shown in 34 will generate the tautomerically disfavoured
amine 2BЈ (cf. 24), which if sufficiently nucleophilic, may simply
recapture the acyl residue before it can diffuse away. Proton-
ation of 2BЈ may be required for the reaction to proceed—
though note that the fission step need not be very disfavoured to
account for what is visible on Fig. 1. This process may be repre-
sented by eqn. (9), which gives rise to the steady-state eqn. (10).
From eqn. (11), pKII is composite of pKlg and the rate constants
kb
Hϩ
kf
(9)
33
34
35
Products
kϪb
k = kbkf[Hϩ]/(kf[Hϩ] ϩ kϪbKlg)
pKII = pKlg ϩ log kf/kϪb
(10)
(11)
¶ If as postulated above this pKa value involves O-protonation, the
N-protonated species involved in catalysis must be a still weaker base.
|| If C᎐N bond breaking were to precede N᎐H bond formation, this limit
could lie even higher, but we see no reason why the former process
should be fast enough for this to happen.
** We thank Professor M. I. Page for the germ of this idea.
1780
J. Chem. Soc., Perkin Trans. 2, 1997

View Article Online
kϪb and kf, both of which are expected to be very fast. It is
possible that, fortuitously, pKII ≈ pKlg. There is also the curious
circumstance that pKII ≈ pKa of acetic acid, so that at pH <
pKII, protonation of the leaving group might be accomplished
by the acetic acid moiety of 34. In view of this and other
imponderables, we view it as profitless to extend the analysis.
The net effect of Scheme 5 is that, at pH > pKII, the transition
state moves from 33 to some point beyond 35, i.e. it now
involves diffusion apart of the reagents, a not uncommon situ-
ation in reactions of this sort.63
Acknowledgements
We thank Drs N. H. Anderson, E. D. Brown and I. T. Kay for
the supply of compounds, Dr P. W. Kenny for the MO calcu-
lations, and Professors W. P. Jencks, A. J. Kirby, M. I. Page and
A. Williams for helpful discussions.
References
1 I. T. Kay, Brit. Appl., 1973, 50 827/73; Ger. Offen., 1975, 2 451 899.
2 I. T. Kay, W. Hepworth and E. D. Brown, Brit. Appl., 1977, 77/7183;
Ger. Offen., 1978, 2 807 381.
The situation for 1d (Scheme 6) is different in that the depart-
3 E. D. Brown, Brit. Appl., 1978, 78/22 938; Eur. Pat. Appl., 1979,
5911.
4 M. J. Billingham, unpublished observations.
5 G. J. Roth and P. J. Majerus, J. Clin. Invest., 1975, 56, 24; G. J. Roth
and C. J. Siok, J. Biol. Chem., 1978, 253, 3782.
6 J. B. Dressman, G. Ridout and R. H. Guy, in Comprehensive Medi-
cinal Chemistry, ed. C. Hansch, vol. 5, ed. J. B. Taylor, Pergamon,
Oxford, 1990, p. 615.
δ+
N
N
N
N
H
N
N
H
H
H
O
H
H
O
H
H
O
H
O
O
δ–
N
N
N
R2
C
R2
R2
Me
Me
O
Me
36
37
38
7 P. J. Taylor, paper presented at Organic Reaction Mechanisms
Conference on Reactivity, Selectivity and Structure, Maynooth, July
1983.
8 A. Albert and E. P. Serjeant, Ionisation Constants of Acids and
Bases, Methuen, London, 1962.
9 J. S. Day and P. A. H. Wyatt, J. Chem. Soc. B, 1966, 343.
10 B. Dempsey and D. D. Perrin, Buffer for pH and Metal Ion Control,
Chapman & Hall, London, 1974.
11 C. H. Rochester, Acidity Functions, Academic Press, London, 1970.
12 C. G. Swain, M. S. Swain and L. F. Berg, J. Chem. Inf. Comput. Sci.,
1980, 20, 47.
13 R. J. Leatherbarrow, ENZFITTER, Elsevier, Amsterdam, 1987.
14 K. J. Laidler and H. Eyring, Ann. N. Y. Acad. Sci., 1940, 39, 302.
15 P. Pithova, A. Piskala, J. Pitha and F. Sorm, Coll. Czech. Chem.
Commun., 1965, 30, 90.
16 J. V. Greenhill, M. J. Ismail, P. N. Edwards and P. J. Taylor, J. Chem.
Soc., Perkin Trans. 2, 1985, 1255.
17 P. J. Taylor and A. R. Wait, J. Chem. Soc., Perkin Trans. 2, 1986,
1765.
18 P. J. Taylor, unpublished data.
19 J. Elguero, C. Marzin, A. R. Katritzky and P. Linda, The Tautomer-
ism of Heterocycles, Academic Press, New York, 1976.
20 R. C. Hirt and R. G. Schmitt, Spectrochim. Acta, 1958, 12, 127.
21 J. V. Greenhill, M. J. Ismail, G. R. Bedford, P. N. Edwards and
P. J. Taylor, J. Chem. Soc., Perkin Trans. 2, 1985, 1265.
22 J. Jonásˇ and J. Gut, Coll. Czech. Chem. Commun., 1962, 27, 716.
23 D. Cook, Can. J. Chem., 1963, 41, 2575; H. M. Sobell and K. Tomita,
Acta Crystallogr., 1964, 17, 122.
24 C. Hansch and A. J. Leo, Substituent Constants for Correlation
Analysis in Chemistry and Biology, Wiley, New York, 1979.
25 A. Leo, J. Chem. Soc., Perkin Trans. 2, 1983, 825.
26 THOR Masterfile version 3.64, CLOGP version 3.64, Daylight
Chemical Information Systems, Claremont, CA 91711, USA.
27 R. W. Taft, in Steric Effects in Organic Chemistry, ed. M. S.
Newman, Wiley, New York, 1956, p. 556.
Scheme 6
ing acyl group of 36 will be tightly bonded to NH. An equiva-
lent conformation for 33 is precluded by the severe steric clash
that would result between R2 and ring N᎐Me. If this hydrogen
bond can persist through the course of the C᎐N fission process,
there is the opportunity for rotation of the acyl moiety to occur
inside the solvation shell. This is shown as just beginning in 36
and could lead to the hydrogen bonded complex 37. An ener-
getically favourable double proton switch to 38 inside this com-
plex could then generate the stable tautomer 2B, pKa ca. 1.7 and
a much poorer nucleophile. This places 37 beyond the transition
state and circumvents the problem posed by 34.
Lacunae remain, one of them being that the transformations
36
38 cannot take place as shown if the reaction product
is an ester, so that a similar phenomenon may exist for series 1
in aqueous alcohols. Our rate measurements were carried out at
the centres of the rate–pH plateaux, so shed no light on this
possibility. In view of this, our analysis has to be tentative.
Conclusions
The three plateau regions kA to kC form an interesting sequence
in which Tϩ (23), T0 (36) and TϪ (32) appear to involve succes-
sively increasing C᎐O bond formation, a reflection of increas-
ing difficulty in expelling the amine leaving group, accom-
panied by an apparently irregular pattern of C᎐N bond fission:
small, large, and probably nil, as the pH range is ascended.
What this last reflects is conformation, tempered by the special
consequences of a situation in which catalyst and leaving
group, for kB, progressively modify one another’s behaviour.
These conformational peculiarities add complexity to the
mechanism but, by a fortunate exercise in serendipity, have
helped to elucidate it. Overall, we know of no parallel for the
maze of criss-crossing factors that this investigation has
unearthed.
Nevertheless, we are left with a paradox. The analgesic
behaviour of the triazinediones, and their effects on the gas-
tric mucosa, take place at pH values which, if acyl transfer is
involved, correspond to kC and kB respectively. Hence struc-
tural changes which favour kC over kB should increase their
therapeutic ratio. Exactly the reverse is found: the α-
branching which specifically inhibits kC [eqn. (5)] has proved
therapeutically desirable,4 and all those compounds that came
closest to clinical trial possessed this feature. Either therefore
one biological response or both has nothing to do with acyl
transfer, or some special feature attaches to one or other site
of action that over-rides the behaviour expected in free solu-
tion. In the absence of a detailed biochemical picture, we for-
bear to speculate. ‘Whereof one cannot speak, thereof one
must be silent.’64
28 J. E. Baldwin, J. Chem. Soc., Chem. Commun., 1976, 738.
29 P. A. Bristow and J. G. Tillett, Chem. Commun., 1967, 1010.
30 Ref. 27, p. 670.
31 G. M. Blackburn, T. H. Lilley and E. Walmesley, J. Chem. Soc.,
Chem. Commun., 1980, 1091.
32 T. C. Bruice and F.-H. Marquardt, J. Am. Chem. Soc., 1962, 84, 365.
33 P. D. Bolton, Aust. J. Chem., 1966, 19, 1013.
34 J. Packer, A. L. Thomson and J. Vaughan, J. Chem. Soc., 1955, 2601.
35 W. P. Jencks and J. Carriuolo, J. Biol. Chem., 1959, 234, 1272; 1280.
36 R. Wolfenden and W. P. Jencks, J. Am. Chem. Soc., 1961, 83, 4390.
37 (a) D. G. Oakenfull and W. P. Jencks, J. Am. Chem. Soc., 1971, 93,
178; (b) D. G. Oakenfull, K. Salvesen and W. P. Jencks, J. Am. Chem.
Soc., 1971, 93, 188.
38 J. P. Fox and W. P. Jencks, J. Am. Chem. Soc., 1974, 96, 1436.
39 N. Gravitz and W. P. Jencks, J. Am. Chem. Soc., 1974, 96, 499.
40 S. L. Johnson, Adv. Phys. Org. Chem., 1967, 5, 273.
41 A. J. Kirby, in Comprehensive Chemical Kinetics, ed. C. H. Bamford
and C. F. H. Tipper, vol. 10, Elsevier, Amsterdam, 1972, p. 57.
42 W. P. Jencks, Catalysis in Chemistry and Enzymology, McGraw-Hill,
New York, 1969.
43 M. L. Bender and R. J. Thomas, J. Am. Chem. Soc., 1961, 83, 4183.
44 R. G. Button and P. J. Taylor, J. Chem. Res., 1996, (S) 218; (M)
1162.
45 P. J. Taylor, J. Chem. Soc., Perkin Trans. 2, 1993, 1423.
46 T. C. Bruice and B. Holmquist, J. Am. Chem. Soc., 1968, 90, 7136.
J. Chem. Soc., Perkin Trans. 2, 1997
1781

View Article Online
47 E. R. Garrett, J. Am. Chem. Soc., 1957, 79, 3401.
48 A. J. Kirby, Adv. Phys. Org. Chem., 1980, 17, 183.
49 A. R. Fersht and A. J. Kirby, J. Am. Chem. Soc., 1967, 89, 4857.
50 W. P. Jencks and M. Gilchrist, J. Am. Chem. Soc., (a) 1962, 84, 2910;
(b) 1968, 90, 2622.
51 D. J. Hupe and D. Wu, J. Am. Chem. Soc., 1977, 99, 7653;
D. J. Hupe, D. Wu and P. Shepperd, J. Am. Chem. Soc., 1977, 99,
7659.
58 T. H. Fife, H. Natarajan and M. H. Werner, J. Org. Chem., 1987, 52,
740.
59 W. P. Jencks, (a) Chem. Rev., 1972, 72, 705; (b) J. Am. Chem. Soc.,
1972, 94, 4731.
60 J. D. Aubort and R. F. Hudson, Chem. Commun., 1970, 937.
61 J. Hine, J. Org. Chem., 1966, 31, 1236.
62 See, e.g., I. H. Williams, G. M. Maggiora and R. L. Schowen, J. Am.
Chem. Soc., 1980, 102, 7831; J. D. Madura and W. L. Jorgensen,
J. Am. Chem. Soc., 1986, 108, 2517.
63 M. I. Page and W. P. Jencks, J. Am. Chem. Soc., 1972, 94, 3263;
W. P. Jencks, Acc. Chem. Res., 1980, 13, 161.
52 T. W. Bentley and R. O. Jones, J. Chem. Soc., Perkin Trans. 2, 1993,
2351.
53 G. B. Hatch and H. Adkins, J. Am. Chem. Soc., 1937, 59, 1694.
54 D. P. N. Satchell and R. S. Satchell, Chem. Soc. Rev., 1975, 4, 231.
55 D. J. Hupe and W. P. Jencks, J. Am. Chem. Soc., 1977, 99, 451.
56 W. Karzijn and J. B. F. N. Engberts, Tetrahedron Lett., 1978, 1787;
H. Benak, J. B. F. N. Engberts and M. J. Blandamer, J. Chem. Soc.,
Perkin Trans. 2, 1992, 2035.
64 L. J. J. Wittgenstein, Tractatus Logico-Philosophicus, Vienna, 1921.
Paper 7/00736A
Received 31st January 1997
Accepted 21st March 1997
57 S. M. Kupchan, S. P. Eriksen and Y.-T. S. Liang, J. Am. Chem. Soc.,
1966, 88, 347.
1782
J. Chem. Soc., Perkin Trans. 2, 1997