Structural correlations for nucleophilic addition to the C=O group: The solvation angle

Article abstract of DOI:10.1002/hlca.200390105

We report some new observations in Buergi-Dunitz territory - structure-structure correlations for the addition of N nucleophiles to the C=O group. The amino aldehyde 1 exists predominantly in polar solvents, and exclusively in the crystal, as the carbonyl

Full text of DOI:10.1002/hlca.200390105

1222
Helvetica Chimica Acta Vol. 86 (2003)
Structural Correlations for Nucleophilic Addition to the CˆO Group:
The Solvation Angle
by John E. Daviesa), Anthony J. Kirby*a), and Igor V. Komarovb)
a) University Chemical Laboratory, Cambridge CB2 1EW, UK
(phone: ‡44 1223 336370 (direct), fax: ‡44 1223 336362 (central), e-mail: ajk1@cam.ac.uk)
b) Chemistry Department, Taras Shevchenko University, Kiev 252033, Ukraine
Dedicated to Professor Jack D. Dunitz on the occasion of his 80th birthday
We report some new observations in B¸rgiÀDunitz territory structureÀstructure correlations for the
addition of N nucleophiles to the CˆO group. The amino aldehyde 1 exists predominantly in polar solvents, and
exclusively in the crystal, as the carbonyl-addition structure 2, but appears to demand hydrogen-bonding
solvation as part of the process. In terms of CÀO bond lengthening, the new system lies somewhere nearly half-
way between an amino aldehyde and a stable quaternary ammonium aminal: we present the structure 2 ¥ Me of
‡
the N ÀCÀOMe derivative of 2 as an example. We have investigated the effects on the cyclisation equilibrium
in this system of hydration, which is important, and of the degree of methylation of the azaadamantane skeleton,
‡
which unexpectedly is not. In terms of the reverse process, i.e., breaking the CÀN bond: in each case, the
bond is lengthened by (n(O)Às*(CÀN)) electron-pair donation from the O-atom (the generalised anomeric effect,
as in 2 ¥ Me $ 3, below) [4]. This effect is strongest when the O-atom is negatively charged, reduced by
hydrogen-bonding, and minimised by alkylation.
Introduction. The work of the Dunitz school has had a major and lasting impact on
the way we think about and teach some very fundamental chemistry. The idea that
crystal structures to a first approximation static representations of stable molecules
can provide important insights into reactivity is by now so familiar, and supported by so
much evidence, that it no longer seems counterintuitive. We accept, for example, that
the B¸rgiÀDunitz angle defines the preferred trajectory of approach of a nucleophile
as it adds to a CˆO group, and, thus, contains fundamental information about the
geometry of the orbital interactions involved in bond formation [1].
We have learned also that crystal structure correlations can illustrate, or even
reveal, quite subtle conformational and basic stereoelectronic preferences in suitable
systems [2]. If the relevant data set is large enough, it is generally a reasonable
assumption that intramolecular and packing forces specific to individual structures will
−average out×, so that small molecules at least adopt the same conformation in the
crystal and in solution. For example, the conformational preferences determined by the
anomeric effect are observed in detail in series of crystal structures [2]; making the
leaving group XOÀ better drives the conformations of benzylic systems ArCH(Me)-
ÀOX towards the conformation required for optimal pÀs* overlap and subsequent
cleavage [3], and the directionality of H-bonding to a CˆO group can be related to the
shape of the electron density in the nonbonding orbitals of sp2-hybridised O-atom [4].
Bond lengths are the geometric parameters least susceptible to perturbation [5],
and, so, the most likely to yield sensible correlations from a small data set. Thus, the

Helvetica Chimica Acta Vol. 86 (2003)
1223
simple, linear correlation between the length of a bond CÀOX and the rate at which it
is cleaved, first observed for acetals containing the structural unit ROÀCÀOX,
appears to be general for bonds ZÀOX; where Z can be C, N or P (at least where
conformations, and, thus, through-bond stereoelectronic interactions, do not change
significantly across the series) [2]. We have called these relationships −real×
structureÀreactivity correlations [6]: in contrast to the structureÀreactivity correla-
tions familiar to the physical organic chemist, which are typically linear free energy
relationships, and correlate rate and equilibrium constants. We report here some new
observations in this general area, involving like the original observations of B¸rgi and
Dunitz structureÀstructure correlations relevant to the addition of N nucleophiles to
the CˆO group.
We reported recently that the amino aldehyde 1 exists predominantly in polar
solvents, and exclusively in the crystal, as the CˆO addition structure 2 [7]. The
thermodynamically stable adamantane unit provides an energy sink deep enough to
accommodate the high-energy tetrahedral zwitterion. The crystal structure is unique in
that it illustrates the addition of tertiary amine N-atom to an aldehyde group. (The
structures studied in the classic investigations of B¸rgi and Dunitz were all derived
from ketones, the second CÀCˆO bond providing the additional constraint necessary
to hold the two groups together productively.) We report here further details of our
investigation, define more clearly the factors driving cyclisation, and show that
compound 2, and its methylated derivative 2 ¥ Me, fit a structure correlation defined by
the B¸rgiÀDunitz series of amino ketones.
Results and Discussion. Crystal-Structure Correlations. We first compare the
structure of our amineÀaldehyde adduct 2 with the series of amino ketones studied in
the classic work of B¸rgi and Dunitz [1]. We are not unmindful of an important caveat
that we have articulated as follows: −for most purposes evidence based on a single
structure is basically anecdotal× [2]. Not only is 2 a single compound, it is observed in the
crystal −under one particular set of conditions, in the crystal in the presence of one
molecule of water of crystallization× [7]. Calculations suggest that it would exist in the
ring-opened form in the gas phase, and NMR evidence is consistent with increasing
amounts of NÀC bond-formation in more polar solvents [7]. So the detailed structure
is expected to depend on the environment in solution on the polarity of the solvent,
and in the crystal on the presence or absence of water of crystallisation. This is what is
observed.
The most convenient correlation for comparison, because our new system lacks the
second CÀCˆO bond, is between the CÀN and CÀO bond lengths at C(2). Fig. 1
shows the smooth curve, which correlates these bond lengths for the original data set of
B¸rgi and Dunitz (six partially or fully open circles) [1]. The point for compound 2
(filled circle) falls nicely on the same curve, and effectively fills a gap between amino

1224
Helvetica Chimica Acta Vol. 86 (2003)
ketones with close N ¥¥¥ CˆO approach distances (d(CÀN) greater than ca. 2 ä) and the
two compounds with single NÀC bonds slightly lengthened by the generalised
anomeric effect. Points A and B of Fig. 1 refer, respectively, to retusamine, with a fully
‡
protonated N ÀCÀOH group (partial structure A; c.f. Fig. 3) [8], and the quaternary
‡
ammonium aminal group N ÀCÀOMe of the mitomycin system [9]. Compound 2 is
present in our crystal structure as the monohydrate, with the developing alkoxide O-
atom stabilised by H-bonding to H2O (part structure 2a) [7]. For a further comparison,
we have obtained the crystal structure (Fig. 2) of the quaternary ammonium aminal 2 ¥
Me (obtained by alkylation of 2 but only by using Meerwein×s reagent [7]).
Fig. 1. Correlation between CÀN and CÀO bond lengths at the former carbonyl centre for tetrahedral addition
compounds derived from amino ketones (circles, data from [1]; see the text) and the amino aldehyde (compound
2, filled circle). The filled diamond shows the point for compound 2 ¥ Me.
Closer inspection shows that the trend in the pattern of bond lengths represented by
the smooth curve in Fig. 1 is accompanied by increasing amounts of solvation of the
developing alkoxide O-atom by H2O. Only the three systems with d(CÀN) > 2.5 ä have
no H2O of crystallization. Clivorine (point C) in the crystal has a single H-bond from
the developing alkoxide O-atom to a molecule of H2O [10]: 2 has two, to two separate
H2O molecules [7]; while retusamine [8], as discussed, has a H-bonded OH group as
shown in Fig. 3 (partial structure A).
The systems represented by points lying at the bottom right hand corner of Fig. 1
are best considered as quaternary ammonium aminals, with N acting as a potential
‡
leaving group. In each case, the CÀN bond is lengthened by electron-pair donation
(n(O)Às*(CÀN)) from the O-atom (the generalised anomeric effect, as in 2 ¥ Me $ 5;
below) [4]. This effect is strongest when the O-atom is negatively charged, reduced by

Helvetica Chimica Acta Vol. 86 (2003)
1225
Fig. 2. Molecular structure of 2 ¥ Me (tetrafluoroborate salt; ORTEP: ellipsoids are drawn at the 30%
probability level and H-atoms represented as spheres of radius 0.12 ä. For details see the Exper. Part)
H-bonding, and minimised by alkylation. (The bond lengths of interest (Fig. 1, filled
diamond) are almost identical to those of the same system in the mitomycin system
(point B).)
Protonation (point A) is slightly less effective at reducing the electron-donor
properties of the O-atom because the OÀH bond is itself involved as a H-bond
acceptor (partial structure A; Fig. 3) [8].
At first sight, it seems remarkable that the data points for 2 ¥ Me (Fig. 1, filled
diamond) and that for the mitomycin system (point B), representing the much −tighter×
‡
ROÀCÀN system, fall on the same curve as the zwitterionic species. In the case of 2 ¥
Me, for example, the ring-opened system is the high-energy species 3 (not observed!).
The explanation is clearly that the trend represented by the original smooth curve
accommodates both the fundamental stereoelectronic effects involved in the addition
of the amine N-atom to CˆO and the (necessary) stabilization of the developing
alkoxide anion by increasingly strong bonding interactions. In the absence of these
increasingly strong bonding interactions, the curve might look quite different.
However, such speculation is of limited interest for the experimentalist because of
the difficulty in obtaining crystals containing naked oxyanions. This solvation problem
is discussed in more detail below.

1226
Helvetica Chimica Acta Vol. 86 (2003)
Factors Driving the Cyclisation Reaction. We identified previously [7][11] two
factors (in addition to the stability of the azaadamantane) that appear to be important
in controlling the extent of cyclisation of 1 > 2: the array of ring-Me groups on the
adamantane skeleton, and H-bonding hydration of the alkoxide O-atom of 2.
We have spent considerable time and effort on attempts to obtain anhydrous 2. We
appear to have succeeded (see the Exper. Part) in three successive sublimations to give
a sample with microanalysis consistent with the parent molecular formula. A second,
pure sample gave a microanalysis consistent with a hemihydrate. However, the only
crystals we could obtain suitable for X-ray structure determination, by sublimation,
turned out to be the monohydrate of 2. This system crystallises as a tight dimer, with
the alkoxide O-atoms H-bonded to two H2O molecules shown in Fig. 3. [7]. This
species appears to retain its H2O even on sublimation under vacuum, and we are all but
convinced that the monohydrate actually sublimes as the dimer, without dissociating
[7].
Fig. 3. Successive stages of hydration of the developing alkoxide anion formed by the addition of amine N-atom
to CˆO. Partial structures of retusamine [8] and the dimer of 2 [7], A and 2a, respectively.
We know, qualitatively at least, that cyclisation is favored in polar solvents [7],
which will stabilise selectively the charged, zwitterionic species 2; and we may assume
that H-bond-donor solvents will be particularly effective, since the positive charge of 2
is shielded in the quaternary ammonium system. This, and the behaviour of the various
samples we have obtained, suggests that the cyclisation of 1, driven initially by the
stability of the tricyclic structure, generates the −near-alkoxide× 2, which is itself so
powerfully stabilised by H-bonding solvation that it behaves, in effect, as a dehydrating
agent, readily absorbing moisture from the atmosphere or, when dissolved, from the
solvent. We observed the same effect previously when trying to find a structur-
eÀreactivity correlation for the dianions of phosphate monoesters. Always the ester
found H2O from somewhere before crystallising [12].
This analysis suggested the following experiments. It is possible to observe the
cyclisation equilibrium by NMR, and we have followed the reaction as a function of
temperature under two sets of conditions in −dry× CD2Cl2. The samples of 1 > 2 were
our best −anhydrous× sample and the well-characterised hemihydrate, which adds a
controlled amount of H2O. Although the solvent was carefully dried, it is well-known
that it can be difficult to remove the last traces of moisture. So, in a third series of
measurements, we used CD2Cl2 saturated with H2O (ca. 6 equiv. with respect to solute
1 > 2). The results are summarised in Table 1.
At ambient temperatures the proton HÀC(2) is in fast exchange, and the
equilibrium constant can be calculated from the average chemical shift. (The shifts of
the aldehyde and aminal H-atoms are based on their separate signals, observed at

Helvetica Chimica Acta Vol. 86 (2003)
1227
Table 1. Effect of H2O on the Cyclisation Equilibrium 1 > 2
Anhydrous compounda)
Hemi-hydrate (0.5 H2O)b)
In the presence of an excess of H2O
(ca. 6 equiv.)c)
T/8 dobs/ppm
CHO/HÀC(2)
averaged
%
Kobs
dobs/ppm
CHO/HÀC(2)
averaged
peak
%
Kobs
dobs/ppm
CHO/HÀC(2)
averaged
peak
%
Kobs
cyclic
form
2
cyclic
form
2
cyclic
form
2
peak
30
25 7.12
8
6.85
6.52
6.22
5.89
61
68
75
82
1.56
2.13
3.00
4.56
6.13
5.47
5.12
4.96
68
81
88
91
2.13
4.26
7.33
10.1
44
0.78
0 6.75459.5
1.4
À 13
À 20 6.1
À 34
71
81
2.44
4.26
À 40 5.72
À 55
5.61
5.44
89
92
8.09
11.5
4.83
4.69
94
96
15.7
24.00
À 77
DH8 À15.6 kJ/mol DS8 À52 J/K/mol DH8 À9.1 kJ/mol DS8 À26 J/K/mol DH8 À10.6 kJ/mol DS8 À (20 28) J/K/mol
a) Chemical shifts used for the CHO and aminal HÀC(2) protons: 9.1 and 5.1; b) 9.6 and 5.1; c) 9.6 and 4.5 ppm.
temperatures below ca. À608). The data show that cyclisation is more favorable at
lower temperatures: only under the driest conditions is there less than 50% cyclisation,
and then only at the highest temperature (258) used. The extent of cyclisation also
increases, as expected, in the presence of increasing amounts of H2O. However, the
relatively small difference between the −anhydrous× sample and the hemihydrate
suggests that even the dried solvent contains trace amounts of H2O.
The variable-temperature data in Table 1 allow calculation of the thermodynamic
parameters for the cyclisation of 1 to 2. The results (Table 1) show that the observed
variation in equilibrium constant with H2O content represents a fine balance between
enthalpy and entropy control. That the enthalpy change becomes less favorable as the
H2O content of the medium increases is unexpected, and we can offer no simple
explanation beyond noting that the differences are small, and such parameters,
especially the entropy change, are often difficult to interpret when solvent reorganisa-
tion is a major part of the reaction. The entropy change for the cyclisation itself is
expected to be close to zero.
Our earlier observations suggested that the three ring Me-groups should make a
significant contribution to the high thermodynamic stability of the cyclic forms of
systems like 1, with aldehyde or carboxylic acid-derived groups held in close proximity
to the tertiary amino group. Thus, we could not make the twisted amide 4 (R ˆ H), by
sublimation of the unsubstituted aminoacid 5 (R ˆ H), which works well for the
trimethyl system 5 (R ˆ Me): similarly the MeO-exchange reaction of the correspond-
ing methyl esters, catalyzed by the MeN group [11] is fastest for the trimethyl
compound [13].
This all seems intuitively reasonable: single alkyl group substitution and especially
the geminal-dialkyl effect are well-known to favor ring formation in a wide range of
systems [14]. To test this proposition more quantitatively in our unusual, conforma-

1228
Helvetica Chimica Acta Vol. 86 (2003)
Table 2. Effect of CÀMe Groups on the Cyclisation Equilibria 1 > 2 and 6 > 7 (R ˆ Me and R ˆ H).
T/8C
Compound 2
d/ppm
Compound 7 (R ˆ Me)
Compound 7 (R ˆ H)
Kobs d/ppm
CHO HÀC(2)
%
cyclised
Kobs d/ppm
%
%
cyclised
Kobs
cyclised
CHO HÀC(2)
CHO HÀC(2)
25
7.12
6.5459.5
6.1
5.72
5.42
44
1.4
71
81
0.78
5.79
2.44
4.26
6.74
72.5
46
0.85
6.29
4.26
5.66
7.07
55
1.22
4.8
0
À 20
À 40
À 60
À 70
7
2.63
73
2.7
5.49
5.22
5.07
4.99
81
85
5.76
5.69
5.19
5.13
83
5.31
7.50
À 75 8.39
À 80 8.46
À 90 8.58
À 100 8.68
À 110
5.20
5.18
5.11
5.1
86
88
88
6.14 7.91
7.33 8.07
7.33 8.22
8.31
4.95
4.9
4.87
4.87
76
79
80
79
3.16
5.10
5.08
5.04
5.01
3.76 9.23
4.00 9.34
3.76 9.45
83
78
4.8
3.5
DH8 À15.6 kJ/mol
DS8 À52 J/K/mol
DH8 À16.5 kJ/mol
DH8 À19.1 kJ/mol
DS8 À57 J/K/mol
DS8 À62 J/K/mol
tionally rigid system, we have prepared two homologues of 2: 6 > 7 (R ˆ Me and
R ˆ H), with one and no ring-Me groups. Data for the equilibrium 6 > 7 (R ˆ Me and
R ˆ H) are compared in Table 2 with those from Table 1 for the same process for 1 > 2,
using anhydrous samples and solvent CD2Cl2.
Here, too, the results are unexpected. The equilibrium constant at 08 scarcely
changes across the series, from 1.45 for the fully methylated system 1 > 2 to 1.33 for the
system lacking all three Me groups. It is clear that our earlier observations were
misleading: the compounds with fewer Me groups are much more volatile and reactive
than the original system 2, but evidently no less ready, thermodynamically, to cyclise.
The full set of variable-temperature data do not fit the Arrhenius equation (Keq
levels out at lower temperatures (possibly because H2O freezes out) so we can say little
with confidence about the thermodynamics of cyclisation. However, the data points in
the fast-exchange region show a reasonably linear dependence on 1/T and have been
used to calculate the enthalpy and entropy changes shown in Table 2. Here, too, the free
energy changes on cyclisation represent small differences between larger (but not
large) values of DH8 and TDS8; and, here too, the trend in DH8 is unexpected, the
enthalpy change becoming more favorable for the systems with fewer MeÀC groups.

Helvetica Chimica Acta Vol. 86 (2003)
1229
But the change is very small, and offset by (still smaller) changes in the entropy term.
The main conclusion from Table 2 is that the number of MeÀC groups has very little
effect on the cyclisation equilibrium. Such an effect would be expected to operate
primarily on the enthalpy term, but neither enthalpy nor entropy term changes
significantly with the number of Me groups. (This is at least consistent with structural
changes that are similar in detail for the three equilibria, which presumably are also set
up with similar degrees of hydration).
Experimental Part
General. Reagents were purified and dried as necessary by standard techniques. Melting points were
determined with a B¸chi 510 melting-point apparatus and are uncorrected. Anal. TLC was carried out on pre-
coated 0.25-mm thick Merck 60 F254 silica-gel plates, with visualization accomplished using I2 vapour. IR Spectra
were recorded on a Perkin-Elmer 1310 spectrometer; the samples were prepared as solns. in the indicated
solvents. One-dimensional NMR spectra and experiments (APT, NOE, low-temperature NMR measurements)
were performed with the following instruments. Bruker WM-400 (400 MHz for 1H), WP-100 (100 MHz for 1H)
or Bruker WM-250 (250 MHz for 1H). Two-dimensional experiments were performed on a Bruker WM-500
(500 MHz for 1H). The deuterated solvents indicated were used as reference or internal deuterium lock. Mass
spectra were recorded in the University Chemical Laboratory, Cambridge. Elemental analyses were carried out
by the staff of the Laboratory×s Microanalytical Department.
Determination of Thermodynamic Parameters. 1H-NMR Spectra were recorded for compounds 2 and 7
(R ˆ Me and R ˆ H) at different temperatures in CD2Cl2 (ca. 0.2m solns.). The percentage of the tricyclic forms
in each case was calculated either from the position of the CHO/HÀC(2) averaged peak (equilibration between
the aldehydes and zwitterionic form fast on the NMR time scale), or by integrating the separate CHO and
ÀC(2) peaks at lower temp. The data are summarized in Tables 1 and 2, and were used to calculate the
thermodynamic parameters DH0 and DS0.
Synthesis. We have now generated the tetrahedral structure 2, which we can prepare in at least three forms,
by three different routes (Scheme 1). Two are detailed below.
Scheme 1
i) LiAlH4 in Et2O, reflux 24h. ii) Jones reagent, sublimation. iii) LiAlH4 in Et2O, reflux. iv) Jones reagent.
v) MeI in benzene. vi) Na2CO3, sublimation.
In each case, compound 2 produced by oxidation of 9, reduction of 10 or removing the proton from 13, was
purified by sublimation. Careful sublimation of the compound as first produced gave a solid that appeared to be
almost if not completely anhydrous. But all crystals obtained by sublimation that were suitable for X-ray
structure determination turned out to be the monohydrate. Even three successive sublimations of the product
from 13 gave only the hemihydrate.

1230
Helvetica Chimica Acta Vol. 86 (2003)
1,3,5,7-Tetramethyl-1-azoniatricyclo[3.3.1.13,7]decan-2-olate (2). Procedure A. The amino alcohol 9 [11]
(0.759 g, 3.6 mmol) was dissolved in acetone (30 ml) and Jones× reagent (prepared from 0.901 g of CrO3, 0.85 ml
of H2SO4 and 1.7 ml of H2O) was added. The mixture was stirred for 2 h at r.t., then H2O (75 ml) was added,
followed by i-PrOH (20 ml) to destroy excess oxidant. After stirring for a further 0.5 h, a sat. soln. of Na2CO3
was added (care!) to bring the pH to ca. 9. The resulting mixture was evaporated, and the product was sublimed
(808, 0.05 mm) from the remained solid, then resublimed to give crystalline 2 (0.585 g, 77.9%).
Procedure B. To a suspension of the amino acid 10 [11] (0.1 mmol), LiAlH4 (0.6 mmol) was added carefully,
and the mixture was refluxed under Ar for 2 h. H2O was added to quench the excess reducing agent, the product
was extracted with CH2Cl2, the extract was dried over 3 g of K2CO3, evaporated without filtration, and the dry
product was sublimed to give crystalline 2, as before; 0.585 g (77 % yield).
Procedure C (not described), starting from the amino alcohol 11 [11], gave an identical product, M.p. 638.
Data of 2. IR: 2956, 2200, 1459, 1216. 1H-NMR (400 MHz, CDCl3) d: 6.61 (s, HÀC(2)); 2.74( d, J ˆ 10.83,
HÀC(8), HÀC(8a)); 2.23 (s, MeN); 2.08 (dd, J ˆ 10.83, 2.45, HeqÀC(8), HeqÀC(9)); 1.86 (dd, J ˆ 12.6, 2.45,
HeqÀC(4), HeqÀC(10)); 1.17 (dt, J ˆ 11.8, 2.45, HeqÀC(6)); 1.095 (dt, J ˆ 11.8, 2.45, HaxÀC(6)); 1.025 (dd, J ˆ
12.6, 2.45, HaxÀC(4), HaxÀC(10)); 0.86 (s, MeÀC(5), MeÀC(7)); 0.85 (s, MeÀC(3) (assigned from COSY,
HMQC, HMBC). 13C-NMR (100.59 MHz, CDCl3): 130.0; 63.58; 47.46; 46.41; 44.29; 39.89; 31.07; 26.57; 26.32.
‡
‡
‡
‡
EI-MS: 209 (4%, M ) 208 (13%, [M À H] ), 181 (68%, [M À CO] ), 180 (51%, [M À H À CO] ), 58 (100,
‡
‡
‡
C3H5N ). FAB-MS (positive): 419 (18, [2M ‡ H] ), 210 (100, [M ‡ H] ), 93 (8), 58 (6). HR-FAB-MS:
210.18590 (C13H23NO ‡ H; calc. 210.18578).
1H-NMR Spectra of a triply-sublimed sample, taken in carefully dried solvents, showed additional peaks
assigned to H2O: a br. s (4.33 ppm) in CD2Cl2 and a s (1.56 ppm) in C6D6), which careful integration showed to
be equivalent to 0.5 H2O. Microanalysis was consistent with this conclusion: calc. for the hemihydrate:
C13H23NO ¥ 0.5 H2O: C 71.51, H 11.08, N 6.42; found: C 71.91, H 11.08, N 6.42. The compound prepared by
Procedure A had essentially the same spectral data, except that no H2O peak is visible in the 1H-NMR taken in
CD2Cl2, and the analysis is most consistent with an anhydrous product: calc. for C13H23NO: C 74.59, H 11.02,
N 6.69; found: C 73.55, H 11.07, N 6.64.
2-Methoxy-1,3,5,7-tetramethyl-1-azoniatricyclo[3.3.1.13,7]decane Tetrafluoroborate (2 ¥ Me). The methyla-
tion of the nominally alkoxide O-atom of 2 needed Meerwein×s reagent. The reaction was carried out in a flame-
dried flask under Ar. Compound 2 (50 mg, 0.24mmol) and trimethyloxonium tetrafluoroborate (42.4mg,
0.29 mmol), were placed in the reaction flask, and dry CH2Cl2 (2 ml) was added. The mixture was stirred at r.t.
for 1 h, then a new portion of the alkylating agent (10 mg, 0.07 mmol) was added. The mixture was stirred
overnight, then treated with MeOH (5 ml). All volatile products were evaporated, and the solid residue was
dissolved in CH2Cl2 (0.5 ml) and filtered. The product was precipitated from the filtrate soln. with Et2O, filtered,
and dried under vacuum to give crystals of 2 ¥ Me (59.9 mg, 80.6%). M.p. 161 1628. 1H-NMR (400 MHz,
CDCl3): 4.47 (s, HÀC(2)); 3.75 (s, MeO); 3.50 (d, J ˆ 12.3, HeqÀC(9)); 3.13 (d, J ˆ 12.3, HaxÀC(9)); 3.12 (d, J ˆ
12.3, HeqÀC(8)); 2.99 (s, MeN); 2.80 (d, J ˆ 12.3, HaxÀC(8)); 1.78 (d, J ˆ 13.4, HeqÀC(6)); 1.64( d, J ˆ 13.09,
HeqÀC(10)); 1.49 (d, J ˆ 13.4, HeqÀC(4)); 1.26 (d ‡ d, J ˆ 13.4, HaxÀC(4), HaxÀC(6)); 1.01 (d, J ˆ 13.09,
HaxÀC(10)); 0.99 (s, MeÀC(3)); 0.94( s, Me); 0.93 (s, Me) (assigned from COSY, HMBC, HMQC). 13C-NMR
(100.59 MHz, CDCl3): 101.46 (C(2)); 68.4 (C(9)); 63.5 (MeO); 61.9 (C(8)); 49.1 (MeN); 46.35 (C(6)); 45.70
(C(4)); 40.32 (C(10)); 36.45 (C(3)); 30.82 (C(5), C(7)); 25.55 (Me); 24.79 (Me); 23.73 (Me) (assigned from
APT, HMBC, HMQC).
The preparation of the homologues of 2 lacking two or all three ring MeÀC groups, required a different
approach [15]. These systems (6 > 7, R ˆ Me and H) were synthesized as outlined in Scheme 2. (There is a
possibility of racemization of 14 during its hydrolysis to 16. None was observed; the tricyclic product 17 was
formed in high yield.)
Ethyl 3-Benzoyl-7-methyl-3-azabicyclo[3.3.1]nonane-7-carboxylate (15). A soln. of BuLi (17.6 ml of 1.6m
(ca. 15%) in hexanes) was added to a soln. of i-Pr2NH (3.67 ml in 200 ml of Et2O) at 08 (ice-water bath). After
stirring at 08 for ca. 1 h, a soln. of 14 [14] (3.9 g in 100 ml of Et2O) was added dropwise, the ice bath was
removed, and the soln. was stirred for a further 2 h at r.t. Freshly distilled Me2SO4 (2.65 ml) was carefully added
to the stirred mixture. The reaction was exothermic, and a white precipitate formed immediately. Stirring was
continued overnight, then the precipitate was filtered off. The filtrate was washed with H2O, 5% NH3 and 1n
HCl, brine, then dried (MgSO4) and evaporated. The colourless viscous oil obtained (3.4g, 85% yield) was
subjected to acid hydrolysis (aq. HBr) without further purification. The product could be purified by column
chromatography (SiO2; AcOEt/hexane 1:1): crystals. Rf 0.34(SiO 2; AcOEt/hexane, 1:1). IR (CDCl3): 2912,
2857 (CH); 1710 (CˆO); 1621 (CˆO); 1453 (Ph); 1416. 1H-NMR (CDCl3, 500 MHz): 7.48 (dd, J ˆ 7.0, 1.7,
2 Ho); 7.36 (m, 2 Hm, 1 Hp); 4.47 (d, J ˆ 13.7); 4.23 (dq, J ˆ 7.07, 3.42, 1 H, MeCH2); 4.13 (dq, J ˆ 7.07, 3.42, 1 H,

Helvetica Chimica Acta Vol. 86 (2003)
1231
Scheme 2
i) LDA, Et2O, 08, 2 h, then Me2SO4, 24h. ii) 45% aq. HCl, reflux, 6 days. iii) LiAlH 4, Et2O, reflux, 2 h. iv) MeI,
benzene, 10 h. v) Na2CO3, sublimation.
MeCH2); 3.63 (d, J ˆ 12.4, HeqÀC(4)); 3.04 (dd, J ˆ 14.1, 1.3, HeqÀC(4)); 3.02 (dd, J ˆ 13.7, 4.1, HaxÀC(2)); 2.75
(dd, J ˆ 12.4, 3.3, HaxÀ(6)); 2.63 (dd, J ˆ 14.2, 1.3, HeqÀC(8)); 2.12 (s, HÀC(1)); 1.89 (s, HÀC(5)); 1.61 (m,
CH2(9)); 1.47 (dd, J ˆ 3.3, 14.7, HaxÀC(6)); 1.38 (dd, J ˆ 14.2, 4.1, HaxÀC(8)); 1.26 (t, J ˆ 7.08, MeCH2); 1.15 (s,
MeÀC(7)). (Assignments were made on the basis of a DQF-COSY spectrum.) 13C-NMR (CDCl3, 100 MHz):
171.7; 173.3; 137.4; 128.8; 128.2; 126.7; 61.1; 52.5; 46.9; 40.8; 39.2; 38.6; 32.9; 32.3; 28.3; 27.6; 14.0. 1-D NOE
Experiment (MeÀ(7)): HaxÀC(8) 11%; HeqÀC(8) 5%; no NOE between MeÀC(7) and CH2(9). EI-MS: 315
‡
(27, M , 286 (4), 242 (6), 211 (14), 210 (100), 182 (5), 136 (10), 105 (85). EI-HR-MS: 315.18333 (C 19H25NO3;
calc. 315.1834305).
7-Methyl-3-azabicyclo[3.3.1]nonane-7-carboxylic Acid (16; R ˆ Me). Conc. aq. HBr soln. (ca. 45%, 20 ml)
was added to 15 (3.5 g), and the mixture was stirred under reflux for 6 d. At the end of this period, all starting
material had dissolved. The PhCOOH formed was removed by steam distillation, and the remaining mixture
evaporated, the product extracted with H2O (3 Â 20 ml), the extract concentrated by evaporation to ca. 10 ml,
and pH adjusted to 7.45. H2O was removed under vacuum, MeOH (5 ml) was added and the soln. transferred to
a sublimation apparatus, evaporated to dryness, then sublimed (60 1208, oil pump). The sublimed product was
resublimed twice (608, oil pump) to obtain 3-methyl-1-azatricyclo[3.3.1.13,7]decan-2-one as white crystals
possessing a strong camphor-like smell (0.75 g, 41.7% yield). This compound was dissolved in H2O (20 ml), and
the soln. was evaporated in high vacuum. Crystals of 16 (R ˆ Me) obtained were dried in a desiccator over P2O5.
3-Methyl-1-azatricyclo[3.3.1.13,7]decan-2-ol (17; R ˆ Me) and 1-Azatricyclo[3.3.1.13,7]decan-2-ol (17; R ˆ
H) were prepared by LiAlH4 reduction of 16 (R ˆ Me and H, resp.), as described for 2 (Procedure B, above).
1
Data of 17 (R ˆ Me): H-NMR (CDCl3, 250 MHz): 7.54(br. s, OH); 4 .09 (s, HÀC(2)); 3.45 (d, J ˆ 12.5,
1 H, CH2N); 2.67 (d, J ˆ 12.5, 1 H, CH2N); 3.18 (d, J ˆ 13.4, 1 H, CH2N); 2.95 (d, J ˆ 13.4, 1 H, CH2N); 2.07 (d,
J ˆ 12, HeqÀC(6)); 1.50 1.85 (m, 6 H); 1.26 (d, J ˆ 12, HaxÀC(6)); 0.8 (s, MeÀC(3)).
Data of 17 (R ˆ H): 1H-NME (400 MHz, CDCl3): 6.27 (br., s, 1 H); 4.12 (d, 1 H); 3.09 (m, 2 H); 2.65 (m,
1 H); 2.35 (m, 1 H); 2.24( m, 1 H); 1.89 (dt, J ˆ 12.42, 2.06, 1 H); 1.33 (dd, J ˆ 12.42, 2.77, 2 H); 1.20 (d, J ˆ
12.42, 1 H); 1.15 (dt, J ˆ 12.42, 2.06, 2 H); 0.83 (dd, J ˆ 12.42, 2.77, 2 H).
2-Hydroxy-1,3-dimethyl-1-azoniatricyclo[3.3.1.13,7]decane iodide (18; R ˆ Me) and 2-hydroxy-1-methyl-1-
azoniatricyclo[3.3.1.13,7]decane iodide (18, R ˆ H) were synthesized from 17 (R ˆ Me and H, resp.) by refluxing
the starting compound (0.1 mmol) in benzene (20 ml) with an excess of MeI (1 mmol) for 10 h. After cooling the
mixtures to r.t., the crystals formed were filtered and used in the next step without further purification.
1
Data of 18 (R ˆ Me): H-NMR (CDCl3, 250 MHz): 6.62 (br. s, OH); 5.31 (s, HÀC(2)); 3.79 (d, J ˆ 12.5,
CH2N); 3.06 (d, J ˆ 12.5, CH2N); 2.91 (s, MeN); 1.15 2.20 (m, 8 H); 0.92 (s, MeÀC(3)).
Data of 18 (R ˆ H): 1H-NMR (CDCl3, 250 MHz): 5.95 (br. s, OH); 5.17 (s, HÀC(2)); 3.50 3.80 (m,
2 CH2N); 3.11 (s, MeN); 1.50 2.30 (m, 9 H).
1,3-Dimethyl-1-azoniatricyclo[3.3.1.13,7]deca-2-olate (7; R ˆ Me) and 1-methyl-1-azoniatricyclo
[3.3.1.13,7]deca-2-olate (7; R ˆ H) were obtained from 18 (R ˆ Me and H resp.). The starting salts (200 mg)
were dissolved in 15% aq. KOH (10 ml), and the products were extracted with CH2Cl2 (4 Â 30 ml). The extracts
were dried (K2CO3), evaporated, and compounds 7 were sublimed to give amorphous solids.

1232
Helvetica Chimica Acta Vol. 86 (2003)
Data of 7 (R ˆ Me): 1H-NMR (CD2Cl2, 250 MHz, 258): 6.74( s, HÀC(2)); 2.98 (d, J ˆ 12, 1 H); 2.50 (two
overlapping d, 2 H); 2.17 (s, MeN); 1.96 (d, J ˆ 12, 2 H); 1.25 1.55 (m, 7 H); 0.8 (s, MeÀC(3)).
Data of 7 (R ˆ H): 1H-NMR (CD2Cl2, 250 MHz, 258): 7.07 (s, HÀC(2)); 2.45 3.35 (m, 7 H); 2.21 (s, MeN);
1.30 2.25 (m, 6 H).
Compounds 7 (R ˆ Me and H) are both thermally stable, but extremely volatile and basic compounds. Like
2, compound 7 (R ˆ Me) is the final product from the reduction of the corresponding amino acid 19 or from
oxidation of the amino alcohol 20 (Scheme 3). The oxidation proceeded in high yield, but formation of 7 (R ˆ
Me) by reduction was accompanied by overreduction to the amino alcohol 20 (Scheme 3).
Scheme 3
i) Jones reagent, 2 h, basic workup, sublimation. ii) LiAlH4, Et2O, reflux, 2 h.
Table 3. Crystal Data and Structure Refinement for 2 ¥ Me
Empirical formula
Formula weight
Temp.
C14H26BF4NO
311.17
150 K
Wavelength
0.71069 ä
Crystal system
Space group
Monoclinic
P21/c
Unit-cell dimensions
a ˆ 8.498(3) ä
b ˆ 12.406(3) ä
c ˆ 15.475(3) ä
1572.4(7) ä3
4
a ˆ 908
b ˆ 105.47(2)8
g ˆ 900
V
Z
Density (calculated)
Absorption coefficient
F(000)
1.314Mg/in 3
0.112 mmÀ1
664
Crystal size
q Range for data collection
Index ranges
0.30 Â 0.25 Â 0.15 mm
2.73 to 22.498
0 ꢀ h ꢀ 9, 0 ꢀ k ꢀ 13, À16 ꢀ l ꢀ 16
2214
Reflections collected
Independent reflections
Absorption correction
Refinement method
Data / restraints / parameters
Goodness-of-fit on F2
Final R indices [I > 2s(I)3]
R indices (all data)
Largest diff. peak and hole
2053 (Rint ˆ 0.1019)
None
Full-matrix least-squares on F2
2043 / 0 / 190
1.072
R1 ˆ 0.0716, wR2 ˆ 0.1597
R1 ˆ 0.1348, wR2 ˆ 0.2141
0.272 and À0.246 eäÀ3

Helvetica Chimica Acta Vol. 86 (2003)
1233
Crystal-Structure Determination1) (Table 3). Crystals of 2 ¥ Me were grown by pouring a CHCl3 solution
into Et2O. Fine crystals formed over 24h. Monoclinic, C 14H26BF4NO, Mr 311.17 g molÀ1, crystal size 0.30 Â
0.25  0.15 mm, space group P21/c, a ˆ 8.498(3) ä, b ˆ 12.406(3) ä, c ˆ 15.475(3) ä, b ˆ 105.47(2)8, Vˆ
1572.4(7) ä3, Z ˆ 4; Dx ˆ 1.314Mg m À3. Data were collected at 150(2) K with a Rigaku AFC7R diffractometer
equipped with an Oxford Cryosystems Cryostream cooling device. No. of reflections measured ˆ 2214,
observed ˆ 2053 (I > 2s(I)), qmax ˆ 22.58 (MoKa radiation). The structure was solved with SHELXS-97 [16] and
refined with SHELXL-97 [15]; No. of refined parameters ˆ 190, R factorˆ 0.072, wR ˆ 0.16, diff. densitymax
ˆ
0.272 e äÀ3, diff. densitymin ˆ À0.246 e äÀ3
.
We are grateful to the EPSRC for financial assistance towards the purchase of the Nonius CCD
diffractometer.
REFERENCES
[1] J. D. Dunitz, −X-Ray Analysis and the Structure of Organic molecules×, Cornell University Press, Ithaca and
London, 1979: especially pp. 366 372.
[2] A. J. Kirby, Adv. Phys. Org. Chem. 1994, 29, 87.
[3] M. R. Edwards, P. G. Jones, A. J. Kirby, J. Am. Chem. Soc. 1986, 108, 7067.
[4] A. J. Kirby, −The Anomeric Effect and Related Stereoelectronic Effects at Oxygen×s, Springer-Verlag,
Berlin and Heidelberg, 1983.
[5] T. M. Krygowski, Prog. Phys. Org. Chem. 1990, 17, 239.
[6] P. G. Jones, A. J. Kirby, J. Am. Chem. Soc. 1984, 106, 6207.
[7] A. J. Kirby, I. V. Komarov, V. A. Bilenko, J. E. Davies, J. M. Rawson, Chem. Commun 2002, 2106.
[8] J. A. Wunderlich, Acta. Crystallogr., Sect. B 1967, 23, 846.
[9] A. Tulinsky, J. H. van der Hende, J. Am. Chem. Soc. 1969, 89, 2907.
[10] K. B. Birnbaum, Acta. Cryst. 2825.
[11] A. J. Kirby, I. V. Komarov, N. Feeder, J. Chem. Soc., Perkin Trans. 2 2001, 522.
[12] P. G. Jones, G. M. Sheldrick, A. J. Kirby, K. W. Y. Abell, Acta Crystallogr., Sect. C 1984, 40, 54 7.
[13] I. V. Komarov, unpublished work.
[14] A. J. Kirby, Adv. Phys. Org. Chem. 1980, 17, 183.
[15] H. Stetter, W. Reinartz, Chem. Ber. 1972, 105, 2773.
[16] G. M. Sheldrick, SHELXS-97 / SHELXL-97, University of Gˆttingen, Germany, 1997.
Received February 5, 2003
1
)
Supplementary crystallographic data have been deposited with the Cambridge Crystallographic Data
Centre as deposition No. CCDC-101432. Copies of the data can be obtained, free of charge, on application
to the CCDC, 12 Union Road, Cambridge CB2 1EZ UK (fax: ‡44(1223)336033; e-mail: deposit@ccdc.-
cam.ac.uk).