Reactions of charged substrates. 4. The gas-phase dissociation of (4-substituted benzyl) dimethylsulfoniums and -pyridiniums

Article abstract of DOI:10.1021/jo951257w

The relative rates for the gas-phase dissociation RX^- → R^- + X° of five (4-Y-substituted benzyl)-dimethysulfoniums (Y = MeO, Me, H, Cl, and NO₂) and 24 (4-Y-substituted benzyl)-3′-Z-pyridiniums (complete series for Z = CN, Cl, CONH₂, and H, and 4-methoxy- and 4-nitrobenzyls for Z = F and CH₃CO) were measured using liquid secondary ion mass spectrometry. The Hammett plot (vs δΔG° or σ^-) is linear for the sulfoniums, but plots for the four pyridinium series have a drastic break between the 4-Cl and 4-NO₂ substrates. Bronsted-like plots for the pyridiniums show a strong leaving group effect only for 4-nitrobenzyls. An analysis of these linear free energy relations with supporting evidence from semiempirical computations suggests that collisionally activated pyridinium substrates dissociate by two pathways, direct dissociation and through an ion-neutral complex intermediate. Comparison of these results with results for the solution reactions of some of these compounds shows that the mechanism is different in the gas and solution phases. Sufficient experimental data are not available to assign a mechanism for dissociation to the sulfonium series, but computational results show characteristics of a direct dissociative mechanism.

Full text of DOI:10.1021/jo951257w

J . Org. Chem. 1996, 61, 2753-2762
2753
Rea ction s of Ch a r ged Su bstr a tes. 4. Th e Ga s-P h a se Dissocia tion
of (4-Su bstitu ted ben zyl)d im eth ylsu lfon iu m s a n d -p yr id in iu m s
Neil Buckley,* David Maltby, Alma L. Burlingame, and Norman J . Oppenheimer
The Department of Pharmaceutical Chemistry, School of Pharmacy, S-926, Box 0446, The University of
California, San Francisco, California 94143-0446
Received J uly 13, 1995 (Revised Manuscript Received J anuary 25, 1996X)
The relative rates for the gas-phase dissociation RX+ f R+ + X° of five (4-Y-substituted benzyl)-
dimethysulfoniums (Y ) MeO, Me, H, Cl, and NO2) and 24 (4-Y-substituted benzyl)-3′-Z-pyridiniums
(complete series for Z ) CN, Cl, CONH2, and H, and 4-methoxy- and 4-nitrobenzyls for Z ) F and
CH3CO) were measured using liquid secondary ion mass spectrometry. The Hammett plot (vs
δ∆G° or σ+) is linear for the sulfoniums, but plots for the four pyridinium series have a drastic
break between the 4-Cl and 4-NO2 substrates. Brønsted-like plots for the pyridiniums show a
strong leaving group effect only for 4-nitrobenzyls. An analysis of these linear free energy relations
with supporting evidence from semiempirical computations suggests that collisionally activated
pyridinium substrates dissociate by two pathways, direct dissociation and through an ion-neutral
complex intermediate. Comparison of these results with results for the solution reactions of some
of these compounds shows that the mechanism is different in the gas and solution phases. Sufficient
experimental data are not available to assign a mechanism for dissociation to the sulfonium series,
but computational results show characteristics of a direct dissociative mechanism.
Ch a r t 1
In tr od u ction
As part of continuing studies on the cleavage of the
nicotinamide-ribosyl bond in â-nicotinamide adenosine
dinucleotide (NAD+),1 we recently reported the results
of a study of the gas-phase dissociation of a series of 2′-
substituted â-nicotinamide arabinosides.2 The relative
rates of dissociation, measured by tandem positive-ion
liquid secondary ion mass spectrometry (LSIMS), follow
the Taft equation and correlate with the relative rates
for the dissociation in water.3 An analysis of the product
ratios and the AM1 energy profiles for dissociation of the
nicotinamide-ribosyl bond shows that an ion-neutral
complex (INC, used to designate the gas-phase interme-
diate) is involved. Rearrangements within the INC that
occur after the rate-limiting step are readily explained
by the relative energies of the possible products. The
system is complex, but easily understood.
A large part of our recent efforts to understand NAD+
hydrolysis in detail has centered on an extensive study
of the mechanisms of substitution of charged benzyl
substrates.4-6 The hydrolysis and nucleophilic substitu-
tion reactions of benzyl dimethylsulfoniums (1) (Chart
1) have been interpreted in terms of an ion-dipole
complex (IDC, used to designate the solution intermedi-
ate) mechanism.7 The kinetics of the nucleophilic sub-
stitution reactions of (4-methoxybenzyl)dimethylsulfo-
nium (1a ) with a large number of nucleophiles and added
salts have been measured.4 1a reacted only with nu-
cleophiles of intermediate hardness and not with hard
or soft nucleophiles. The hydrolysis reaction is compli-
cated by the establishment of an equilibrium among
starting material and products,5 but it is clear from the
kinetics and the selectivities that the reaction is mixed
SN1/SN2 under constant ionic strength. Kevill and his
colleagues8 reached the same conclusion independently.
Hammett plots for the azide reaction and hydrolysis show
distinct breaks but are not V-shaped.
X Abstract published in Advance ACS Abstracts, March 15, 1996.
(1) (a) J ohnson, R. W.; Marshner, T. M.; Oppenheimer, N. J . J . Am.
Chem. Soc. 1988, 110, 2257-2263. (b) Handlon, A. L.; Xu, C.; Mu¨ller-
Steffnar, H.; Schuber, F.; Oppenheimer, N. J . J . Am. Chem. Soc. 1994,
116, 12087-12088.
(2) Buckley, N.; Handlon, A. L.; Maltby, D.; Burlingame, A. L.;
Oppenheimer, N. J . J . Org. Chem. 1994, 59, 3609-3615.
(3) Handlon, A. L.; Oppenheimer, N. J . J . Org. Chem. 1991, 56,
5009-5010.
(4) Buckley, N.; Oppenheimer, N. J . Submitted. Some results have
been presented in ref 5.
The kinetics of the reaction between azide and the
benzylpyridinium substrates 2-4a -e in water have been
measured.6 The reactions are strictly SN2 with no
evidence for an IDC intermediate, despite the fact that
the values of âLG are quite large (-1.4 to -1.6). The rates
for the hydrolysis reactions of 2a -e are quite slow
compared with the rates for the azide reaction, but the
rates of the two reactions are comparable for 3a -e.
Hammett plots for hydrolysis of 2-3a -e are linear but
(5) Buckley, N.; Oppenheimer, N. J . J . Org. Chem. 1994, 59, 5717-
5723.
(6) Buckley, N.; Oppenheimer, N. J . Submitted.
(7) For reactions of 1a ,c: Sneen, R. A.; Felt, G. R.; Dickason, W. C.
J . Am. Chem. Soc. 1973, 95,638-639. For the ion-pair mechanism for
neutral substrates: Sneen, R. A. Acc. Chem. Res. 1973, 6, 46-53.
(8) Kevill, D. N.; Ismail, N. H. J .; D’Souza, M. J . J . Org. Chem. 1994,
59, 6303-6312.
0022-3263/96/1961-2753$12.00/0 © 1996 American Chemical Society

2754 J . Org. Chem., Vol. 61, No. 8, 1996
Buckley et al.
Ta ble 1. Log [R+/(R+ + M+)] for th e Ga s-P h a se
have different slopes. While 4a ,e react very slowly with
azide, 4b-d did not react with either water or 1.7 M
azide after 6 months at 96 °C.6
Dissocia tion of (4-Y-ben zyl) d im eth ylsu lfon iu m s a n d
3-Z-p yr id in iu m s
In contrast, Katritzky and his colleagues9 found that
the reactions between 4-substituted benzyl substrates
with highly arylated pyridine leaving groups (LGs) and
neutral amine nucleophiles in solvent chlorobenzene
exhibited borderline kinetic behavior for the 4-methoxy-
benzyl compounds but not for substrates with less
electron-donating substituents. These results were in-
terpreted in terms of an IDC mechanism, which was
supported by the results of ion cyclotron resonance (ICR)
experiments10,11 and a computational study.12
LG/Y
SMe2
3-CN-Py
3-Cl-Py
nicotinamide
3-H-Py
MeO
Me
H
Cl
NO2
-0.78
-0.95
-0.95
-1.00
-0.99
-1.06
-1.05
-1.07
-1.11
-1.25
-1.23
-1.14
-1.15
-1.21
-1.29
-1.14
-1.16
-1.21
-1.20
-1.31
-1.71
-1.83
-1.89
-2.05
-2.30
RAM) with Releases 3.0 and 4.0 of the Hyperchem software
using methods described elsewhere.2
To construct the energy profiles for 5a -e, the benzyl-C7-
pyridine bond length was increased in steps from the initial
length of 1.49-1.52 Å using the restraint function to a final
restraint force constant of 105. Each structure was minimized
completely with no restraint other than the reaction coordi-
nate. Transition state structures were refined until a single
negative (imaginary) frequency14 was obtained using the
Vibrational Analysis subroutine. Energies of the carbenium
ions were calculated after manual removal of the LG from the
INC. Stabilities of the INCs were assessed by removing the
reaction coordinate constraint from the structure correspond-
ing the minimum in the energy profile and reminimizing.
For the norcaradienylcarbenium ions (9), structures were
formed from the corresponding benzylcarbenium ions (8) and
minimized in MM+ before being fully minimized in AM1.
These wildly different results prompted us to examine
the collisionally activated gas-phase dissociation of 1a -e
and four series of benzylpyridiniums (2-5a -e) with
tandem LSIMS. Hammett and Brønsted-like plots based
on relative rates for gas-phase and solution dissociation
of 1a -e and 2-5a -e show differences among and
between the two series in both phases. These results
were used to analyze the kinetics of the gas-phase
reaction. It appears that 1a -e may react by direct
dissociation and that 2-5a -d react through direct
dissociation and INC mechanisms, while 2-5e react only
through direct dissociation.
Resu lts
Exp er im en ta l Section
Exp er im en ta l. LSIMS Resu lts. For all compounds,
LSIMS spectra obtained in MSII had two peaks that
corresponded to the parent molecular ion M+ and the
respective benzylcarbenium ion R+, consistent with het-
erolytic cleavage. The relative rates for dissociation were
obtained as the ratio of the relative abundances [R+/(R+
+ M+)], with M+ t 100. These values are summarized
in Table 1. Because the average error in the measure-
ments is <5%, in plots of these data the error bars are
within the symbols.
Spectra for the 4-nitropyridinium compounds 2-5e
had additional peaks, not found in any spectrum for the
2-5a -d compounds, that corresponded to the respective
pyridine radical-cation and its decomposition products,
consistent with homolytic cleavage. The relative abun-
dances of these species were less than or equal to those
for the carbenium ion (1-2% of M+). These small yields
have insignificant effects on the relative rates.
Gen er a l. All chemicals and solvents were obtained from
Aldrich and used without further purification. NMR spectra
were recorded in D2O at 300 MHz.
Syn th esis. All substrates were prepared by mixing 1 equiv
of the appropriate benzyl chloride with 1.1 equiv of either SMe2
or the appropriate pyridine in chloroform at ambient temper-
ature. Reactions were followed by TLC (silica, neat chloro-
form) until all chloride had disappeared. The chloroform layer
was extracted with water, and the aqueous layer was back-
extracted extensively with chloroform and then with ether to
give water-white solutions. Rotary evaporation at reduced
pressure with repeated flashing with ethanol to remove traces
of water gave products that were >95% pure by proton NMR.
For several of the 3-cyanopyridine substrates, the alkylation
was performed by adding the chloride to a melt of the pyridine,
which was held in a drying oven at ca. 110 °C for 10-20 min.
After cooling, the (usually dark brown) solid was dissolved in
water and carried through the extraction procedure. Heating
the water layer with Norit A and filtering through a pad of
Filtre-Aid, followed by evaporation, gave pure compound. All
compounds are known with various counterions and were
characterized only by proton NMR and LSIMS.
Each sulfonium spectrum had a very small peak at
m/ z 61 that may correspond to the sulfonium product
formed in a four-centered hydride reduction reaction,
LSIMS Sp ectr a . Tandem mass spectra were recorded
under identical conditions in the positive ion mode on a four-
sector Kratos Concept II HH mass spectrometer fitted with
an optically coupled 4% diode array detector on MS II.
Substrates were sputtered from a glycerol matrix with Cs+ (18
keV), and the molecular ion was sorted into MSII where
fragmentation was induced by collisional activation with
helium (13 eV). Seven to 10 determinations were collected and
averaged in the computer (<5% error).
ArCH2SMe2 f ArCH3 + [CH2dSMe]+. No pyridinium
+
spectrum had a peak corresponding to 3-Z-PyH+.
Lin ea r F r ee En er gy Rela tion s (LF ERs). Because
it is advisable to use a scale derived from gas-phase data,
Hammett plots of log [R+/(R+ + M+)] for 1a -e (Figure 1)
and 2-5a -e (Figure 2) were constructed vs δ∆G°t-cumyl
for the gas-phase reaction 4-Y-C6H4C(Me)dCH2 + H+
h
4-Y-C6H4CMe2+; values have been summarized by Taft
and Topsom.15 (The δ∆G°t-cumyl and σ+ scales correlate
well; for substituents a -c,e, r ) 0.9999 with a slope of
14.) The plots are clearly different, with good linearity
for the sulfoniums and a distinct break for the pyridini-
ums. The slopes “γ+” are -0.042 (r ) 0.999 99 with the
point for 1d excluded, r ) 0.9964 with the point for 1d
included) for 1a -e, and -0.020 (r ) 0.983), -0.022 (r )
Com p u ta tion a l Meth od s. Semiempirical computations in
MNDO, AM1, and PM313 were performed on a 486/66 (16 MB
(9) Reviewed in: Katritzky, A. R.; Brycki, B. E. Chem. Soc. Rev.
1990, 19, 83-105 and elsewhere.
(10) Katritzky, A. R.; Watson, C. H.; Dega-Szafran, Z.; Eyler, J . R.
J . Am. Chem. Soc. 1990, 112, 2471-2478.
(11) Katritzky, A. R.; Malhotra, N.; Dega-Szafran, Z.; Savage, G.
P.; Eyler, J . R.; Watson, C. H. Org. Mass Spectrom. 1992, 27, 1317-
1321.
(12) Katritzky, A. R.; Malhotra, N.; Ford, G. P.; Anders, E.; Tropsch,
J . G. J . Org. Chem. 1991, 56, 5039-5044.
(13) AM1:Dewar, M. J . S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J .
J . P. J . Am. Chem. Soc. 1985, 107, 3902-3909. PM3: Stewart, J . J . P.
J . Comput. Chem. 1989, 10, 209-220, 221-264.
(14) McIver, J . W., J r.; Komornicki, A. Chem. Phys. Lett. 1971, 10,
303-306.
(15) Taft, R. W.; Topsom, R. D. Prog. Phys. Org. Chem. 1987, 16,
2-83.

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 8, 1996 2755
F igu r e 3. Plots of the LSIMS data for 2-7a (b), 2-5b (9),
2-5c (2), 2-5d ([), and 2-7e (1) vs the relative free energy
for gas-phase protonation of the respective pyridine LGs.
Slopes for the a -d compounds are 0.01-0.04 and for the e
compounds is 0.12.
F igu r e 1. Plot of the LSIMS data for 1a -e vs the relative
free energy for gas-phase formation of tert-cumylcarbenium
ions. The point off the line is for 1d ; the regression line is for
the other points with a slope of “γ+” ) -0.042 (r ) 0.9999).
With the point for 1d included, r ) 0.996. Left to right the
points are for 1a , 1b, 1d , 1c, and 1e.
F igu r e 4. AM1 energy profiles for the gas-phase dissociation
of 5a (b), 5b (9), 5c (2), 5d (3, shown open for clarity), and
5e ([). The open symbols are values for ∆HR for complete
dissociation of the carbenium ion and LG. The bond length at
the transition state is relatively constant at ca. 2.2 Å.
F igu r e 2. Plots of the LSIMS data for 2a -e (b), 3a -e (9),
4a -e (1), and 5a ([) vs the relative free energy for gas-phase
formation of tert-cumylcarbenium ions. Left to right the points
are for a -e. The slopes of “γ+” are as follows: 2a -d ) 0.020
(r ) 0.98), 3a -d ) 0.022 (r ) 0.96), 4a -d ) 0.020 (r ) 0.996),
and 5a -d ) 0.030 (r ) 0.97).
Sch em e 1
0.960), -0.020 (r ) 0.996), and -0.030 (r ) 0.970) for
2a -d , 3a -d , 4a -d , and 5a -d , respectively. Thus, as
expected, the rates decrease with the basicity of the
pyridine. (Hammett plots for 1a -e and 2a -d against
σ+ are shown in Figures 6 and 7 with the data for
hydrolysis.)
For Brønsted plots of the experimental data for the four
complete series of pyridiniums, with additional points for
6a ,e and 7a ,e, log [R+/(R+ + M+)] was plotted vs values
of δ∆G°Py for the gas-phase protonation of pyridines15
(Figure 3). The value for nicotinamide is not available,
but was estimated by interpolation against solution pKa
values, which assumes that the effects are constant
between the gas phase and solution. For 2-7a and
2-5b-d , plots are grouped together and have small
slopes; for 2-7e, however, there is a definite slope, which
corresponds to the pattern seen for the spread of points
for 2-5e in Figure 2.
Com p u ta tion a l. P yr id in iu m s. Because there is
little effect of the LG on the experimental data, complete
energy profiles (Scheme 1) were calculated only for 4a -
e, which is the simplest pyridinium computational sys-
tem. Our results agree with those reported by Katritzky
et al.12 that the PM3 energy profiles for pyridiniums have
no distinct transition states (not shown; the profile for
4c has been reported2). In AM1, profiles for 4a -d had
distinct transition states, but the profile for 4e was
consistent with dissociation without an INC (Figure 4).
We calculated several profiles for various pyridiniums in

2756 J . Org. Chem., Vol. 61, No. 8, 1996
Buckley et al.
Ta ble 2. AM1 En th a lp ies of F or m a tion for th e
Ga s-P h a se Dissocia tion of (4-Y-ben zyl)p yr id in iu m s
(k ca l/m ol)a
Ta ble 4. AM1 ∆Hf a n d ∆HR, fr om th e P yr id in iu m s, of
(4-Y-su bstitu ted ben zyl)ca r ben iu m (4-Y-su bstiu ted
n or ca r a d ien yl) a n d 4-Y-Su bstitu ted Tr op yliu m
Ca r ben iu m Ion s (k ca l/m ol)
starting
transition
state
ion-dipole
carbenium
ion
structure
complex
MeO
Me
H
Cl
NO2
170.3
201.0
209.8
204.5
221.8
201.0
234.8
245.1
239.9
197.6
233.4
244.6
239.1
173.0
209.6
221.9
216.1
243.9
a
∆Hf pyridine ) 31.9 kcal/mol.
Ta ble 3. AM1 Activa tion En th a lp ies a n d En th a lp ies of
Rea ction for th e Ga s-P h a se Dissocia tion of
(4-Y-ben zyl)p yr id in iu m s (k ca l/m ol)a
∆Hq k1
∆Hq k-1
∆Hq k2
∆HR
MeO
Me
H
Cl
NO2
30.7
33.8
35.5
35.4
3.4
1.4
0.8
0.5
7.3
8.2
9.2
8.9
34.6
40.5
44.0
43.5
54
a
Designations of rate constants refer to Scheme 1.
Sch em e 2
a
Reverts to the benzylcarbenium ion upon minimization.
carbenium ions are not good (Figure 1 of ref 11), using
the published values for the APs and computed ∆HR, we
found that the correlation is measurably better for the
benzyl energies (r ) 0.90) than for the tropylium energies
(r ) 0.82) in plots of AP vs ∆HR (not shown). If the points
for 4-NO2 and 4-H are ignored, the correlation coefficients
are 0.94 and 0.95 for benzyls and tropyliums, respec-
tively, which reflects the effects of the substituents on
the stabilities of the carbenium ions, not on the ability
of the benzyls to rearrange. By themselves, these fits of
data and computed energies show that the benzylcarbe-
nium ions are stable within the INC for both groups of
pyridine LGs.
Dewar and his colleagues have calculated the barriers
to isomerization of benzyl-17 and substituted benzylcar-
benium18 ions into the corresponding tropylium carbe-
nium ions through the sequence shown in Scheme 2. In
MINDO/3, the barrier for 8 f 11 is ca. 33 kcal/mol. While
we did not determine the activation barriers, we calcu-
lated ∆HR in AM1 (Table 4) for the conversion 8 f 9.
The (4-methoxynorcaradienyl)carbenium ion (9a ) reverts
to the (4-methoxybenzyl)carbenium ion upon minimiza-
tion; the other compounds have values of ∆HR 4- to 6-fold
greater than ∆Hq for dissociation of the INC into the
carbenium ion and the LG. The sequence 9 f 10 f 11
would require more energy; ∆Hf for the tropyliums in
PM3 are 10-20 kcal greater than the corresponding
MINDO/3 values,18a which suggests that the total energy
of conversion would be 40-50 kcal/mol, assumings
crudelyssimple additivity between the methods. If ben-
zylcarbenium ions other than the 4-MeO- isomerized to
tropylium ions in our systems, a break between the
energies for 1-5a and the other compounds would be
expected; there is none. Note that the break at the 4-NO2
compound for the pyridiniums 2-5 is not the result of
an isomerization. First, if this were the case the line in
MNDO and found that the energies were several kcal/
mol lower than found for AM1; the MNDO profile for 4c
has been reported.2 Values for the various AM1 energies
are listed in Table 2; the activation values obtained from
these energies are summarized in Table 3.
Su lfon iu m s. No distinct transition states were found
in the PM3 or AM1 reaction profiles for 1a -e (not
shown).
Tr op yliu m s. Benzylcarbenium ions (8) may rear-
range to tropylium carbenium ions (11) in the gas phase
(Scheme 2).16 Values of ∆Hq calculated for heterolytic
cleavage of the benzyl methylene-pyridine bond would
not be affected by rearrangements that take place in the
INC after bond cleavage, but rates for the recombination
reaction would be affected because the affinity of
tropyllium and benzylcarbenium ions for nucleophiles are
different. Thus, the stabilities of the carbenium ions is
of some importance.
Katritzky10,11 measured the appearance potentials
(APs) for carbenium ions formed by collisional activation
of laser-desorbed (4-Y-substituted benzyl)pyridiniums.
Because of a correlation between the APs and AM1
values of ∆HR for benzylcarbenium, but not for tropylium
carbenium, ions formed from simple, unhindered pyri-
diniums, Katritzky10 argued that no benzyl-to-tropylium
rearrangement took place within the INC (although the
4-Br and 4-Cl compounds may have undergone this
rearrangement). For highly hindered 4-Y-substituted
benzyl substrates (Y ) MeO, Me, H, F, Cl, NO2) with
2,4,6-triphenylpyridinium LGs, however, he11 argued that
the measured APs and the calculated AM1 ∆HR were
consistent with benzyl-to-tropylium rearrangements within
the INC. While the correlations between AP and the
AM1 ∆HR for dissociation to either tropylium or benzyl-
(17) Cone, C.; Dewar, M. S. J .; Landman, D. J . Am. Chem. Soc. 1977,
99, 372-376.
(18) (a) Dewar, M. J . S.; Landman, D. J . Am. Chem. Soc. 1977, 99,
4633-4639. (b) Dewar, M. J . S.; Landman, D. J . Am. Chem. Soc. 1977,
99, 7439-7445.
(16) (a) Kuck, D. Mass Spectrom. Rev. 1990, 9, 187-233. (b) For an
extensive review of earlier work, see: Bursey, J . T.; Bursey, M. M.;
Kingston, D. G. I. Chem. Rev. 1973, 73, 191-234.

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 8, 1996 2757
Figure 2 should break up, not down. Second, it has been
estimated that the energies for the isomerization of the
(4-methoxybenzyl) and (4-nitrobenzyl)carbenium ions are
the same.16a Thus, we are confident that the benzylcar-
benium ions do not rearrange in the sulfonium and
pyridinium INCs under our experimental conditions.
isomerize to a more stable species (through a 1,2 hydride
shift, MeCH2CH2 f Me2HC+)19 and/or there may be a
+
proton transfer from the cation to the base (such as the
commonly observed â-elimination Me2HC+ + X f
CH2dCHMe + XH+).19,26 Proton transfer is less endo-
thermic than dissociation.26 Both processes occur and
provide evidence for the existence of an INC intermedi-
ate.19,21,26 The conclusions are bolstered by isotope label-
ing and stereochemical studies on gas-phase reactions,
including Morton’s work collecting and analyzing neutral
hydrocarbon products.19 In the absence of this daunting
task, however, it is often sufficient to show that proton
transfer takes place, signaled by the presence of XH+
among the products.22,26
Discu ssion
In the gas phase, a molecule with sufficient internal
energy may vibrate so violently that the weakest bond
is broken, producing two fragments by either homolytic
or heterolytic cleavage.19-23 Heterolytic cleavage of posi-
tively charged molecules produces a cation and a neutral,
with the charge borne by the less polarizable fragment20
;
In our earlier study of the collisionally activated, gas-
phase dissociation of 2′-substituted â-nicotinamide ara-
binosides,2 protonated nicotinamide was a prominent
product formed by abstraction of a proton from the 5′-
OH in all compounds and from 2′-â-NH2 and 2′-â-OH
groups in two others, which established that the reaction
proceeded through an INC. With one exception, this
diagnostic reaction is not available in the benzylsulfo-
nium or -pyridinium systems. It is possible that a methyl
proton could be abstracted from the 4-Me carbenium ion
to form the quinoid p-xylylene, but no protonated pyri-
dine or sulfonium is detected in any series. As noted in
the Results, it is doubtful that any of the benzylcarbe-
nium ions rearrange to tropylium carbenium ions under
our reaction conditions. In the absence of this classic
evidence we must deduce from the available data whether
or not an INC can be identified in either the sulfonium
or pyridinium series.
For the sulfoniums, it is not possible to choose between
direct dissociation and an INC mechanism. The fact that
the plot of log krel vs δ∆G° (or σ+ ) is linear (Figure 1)
suggests only that the mechanism is the same within the
series and that there is considerable carbenium ion
character, which would be characteristic of either a direct
dissociation or INC mechanism.
For the pyridiniums, however, it is possible to speculate
about mechanism based on the shape of the gas-phase
Hammett (Figure 2) and Brønsted-like plots (Figure 3).
The Hammett plot has a break down, which is usually
interpreted as the signature for a change in rate-limiting
step, not in mechanism.27 By itself, this break implies
that there are at least two kinetically significant steps
in the reaction.28 The Brønsted plots show that there is
no LG effect for the 4-MeO substrates, a slight effect for
the 4-Me, 4-H, and 4-Cl substrates (although the values
upon which they are based are within or near the 5%
error), and a marked effect for the 4-NO2 substrates that
follows pyridine basicity (bond strength) and is consistent
with rate-limiting cleavage of the R-X+ bond.
the positive charge will switch fragments upon dissocia-
tion, RX+ f R+ + X°, as found for all compounds studied
here.24 Molecules with sufficient energy may fling the
fragments sufficiently far apart (>10 Å) that mutual
attraction is overcome and they do not recombine.
Completely dissociated products fall into potential wells,
separated from RX+ by the Longevin transition state.22
If there is not sufficient energy to cause complete dis-
sociation, an intermediate INC [R+ X°] may form.20,22
The general criterion that must be met for [R+ X°] to
be an intermediate is that both fragments must be able
to rotate about an axis orthogonal to the interfragment
axis, a maneuver prevented in the parent by the R-X
bond.22,23 The fate of the INC depends on a number of
factors related to the free rotationsor lack of itsof the
fragments in either of two “critical configurations” de-
termined by the “density of states,” kinetic-statistical
quantities from RRKM theory.25 In general, however,
there are three possibilities. One is dissociation into R+
and X°, which can occur if the energy associated with
the internal degrees of freedom is converted into suf-
ficient motion along the reaction coordinate that the
attractive forces are overcome. The second is recombina-
tion to RX+. In order for this to occur, a “locked rotor”
critical configuration must be achieved in which the
HOMO of X and LUMO of R+ are aligned properly to form
a covalent bond.22 The locked rotor critical configuration
does not correspond to a saddle point on a potential
energy surface. Because of free translation of the ion and
neutral, the fragments may exist in what Morton has
called an “orbiting critical configuration.” Formation of
the locked rotor occurs with loss of translational entropy,
and there is an entropic “bottleneck” to recombination;
attractive forces create an enthalpic barrier to complete
dissociation of the fragments. If [R+ X°] can exist with a
nonzero lifetime between the two critical configurations,
it is an INC. This intermediate may, but does not have
to, correspond to a minimum on a potential energy
surface.23
(26) Bowen, R. D. Acc. Chem. Res. 1991, 24, 364-371 and references
cited therein for the work of the Cambridge group.
The third possibility is reactions within the INC and
between the ion and neutral in the INC. R+ may
(27) Leffler, J . E.; Grunwald, E. Rates and Equilibria of Organic
Reactions; Wiley; New York, 1963; pp 190-191 (Dover edition, 1989).
(28) The often-stated assumption that a break down in a Hammett
plot signals a change in rate-limiting step assumes a priori that there
is only a single process involved. In the majority of cases where this
has been documented for solution reactions, it is of course true. There
is no good theoretical or practical reason to suppose, however, that
this is generally true. For instance, a break up may signal either a
change in mechanism (see refs 5 and 31) or a change in the structure
of the activated complex (see ref 33). There is no reason to suppose
that the break down should be any different. All that is required for
any break in a Hammett plot are two kinetically distinguishable steps
irrespective of whether or not they repesent a single mechanism in
which kobsd is related to several individual rate constants (and their
relative rates) or represent two mechanisms with different rate
(19) Morton, T. H. Tetrahedron 1982, 38, 3195-3243.
(20) McAdoo, D. J .; Morton, T. H. Acc. Chem. Res. 1993, 26, 295-
302.
(21) Kondrat, R. W.; Morton, T. H. J . Org. Chem. 1991, 56, 952-
957.
(22) Morton, T. H. Org. Mass. Spectrom. 1992, 27, 353-368.
(23) Kondrat, R. W.; Morton, T. H. Org. Mass. Spectrom. 1991, 26,
410-415.
(24) That the (4-nitrobenzyl)pyridiniums also undergo homolytic
cleavage is not suprising, because of all possible carbenium ions that
can be produced, only the (4-nitrobenzyl)carbenium will not be
stabilized by resonance, while the 4-nitrobenzyl radical would be.
(25) Forst, W. Theory of Unimolecular Reactions; Academic Press;
New York, 1973.
constants that contribute to kobsd
.

2758 J . Org. Chem., Vol. 61, No. 8, 1996
Buckley et al.
Under the conditions of our experiment, in which the
internal energies of all RX+ are assumed to have a
Boltzmann distribution around the average energy of the
collision gas, 13 eV, there are populations of RX+ with
sufficient energy to dissociate directly or to form INCs;
there are also populations that do not have sufficient
internal energy to undergo bond cleavage. Kondrat and
Morton23 have shown that steady-state kinetics can be
used to analyze relative rates for the separate processes
under these conditions. Their approach is not directly
applicable to our system, however, because we obtain a
single detectable product, R+.
plotssan uncompensated decrease in kd that parallels
R-Py+ bond strength. The fact that the relative abun-
dance of the (4-nitrobenzyl)carbenium ion is so low
suggests that the INC mechanism is dominant for the
other substrates.
In his reviews on these matters,19,20,22 Morton has
pointed out that gas-phase dissociative reactions do not
require that the intermediates are energy minima on
potential energy surfaces, a condition that may be present
but is not required by RRKM theory.25 In addition to the
INC, three other possible intermediates may be associ-
ated with potential energy surfaces.22 One is a hydrogen-
bonded intermediate, the second is a π complex in
analogy with the electrophile-aromatic complexes that
are thought to precede the Wheland intermediate in
electrophilic aromatic substitution reactions, and the
third is a potential energy minimum associated with a
long bond extension.
Possible pathways are given in eq 1. Each channel will
be considered separately. Direct dissociation alone (kd
k1
k2
9
d
R+ + X°
9
8 R+ + X°
(1)
k-1
pathway) is not consistent with the Hammett plot.
Formation of the INC alone with partitioning between
dissociated products and starting material is a multistep
We have examined these factors by computing in AM1
the potential energy surfaces for reaction of 4a -e. (The
barriers for the various rate constants are given in
Scheme 1.) The results confirmsbut do not provesthe
assumptions made above concerning the relative barriers
for partitioning of the INC (Tables 3 and 4 and Figure
4). The energy profiles show distinct transition states
for all but the 4-NO2 substrate 4e, which has a slight
“hump” in the curve but no distinct minimum. For 4a -
d , values of ∆Hq are high and relatively constant for k1
and low and relatively constant for k2 but decrease
progressively for k-1 (Table 3), consistent with the
assumptions made above. Because of the absence of a
distinct minimum for the INC, these values are not
available for 4e. For 4a -d , the values of log K estimated
from the computed ∆∆Hq ) log k1/k-1 values give a linear
plot vs δ∆G° (r ) 0.998, not shown), consistent with the
proposed mechanism.
process; steady-state treatment gives kobsd ) k1k2/(k-1
+
k2). If k2 > k-1, kobsd ) k1, a pathway kinetically
indistinguishable from the kd pathway and inconsistent
with the Hammett plot. (As shown below, this assump-
tion does not fit the AM1 energy.) If k2 ∼ k-1, then kobsd
is either k1/2, which is not probable, or Kk2/2. The case
in which k-1 > k2 yields kobsd ) Kk2. This is the most
reasonable case based on the assumption that the barrier
to recombination is smaller than the barrier to dissocia-
tion of the INC. Both this and the Kk2/2 case, however,
are difficult to rationalize in terms of the Hammett and
Brønsted plots. Assuming that k2 would be relatively
constant across the series (see below), kobsd will vary
based on the value of K, which is mediated by obverse
(and compensating) effects of the stability of R+ on
The R-Py+ bond length is ca. 2.8 Å in the constrained
structures that give an energy minimum consistent with
an INC (Figure 4). If the bond length constraint is
removed in the INCs for 4a -d (4iii) and the resulting
structures are minimized, all remain stable species,
consistent with the presence of an enthalpic barrier to
collapse. The 4e “INC”, however, spontaneously collapses
back to the starting structure (4i) upon minimization.
Moreover, minimizing the π complex between the cation
and the pyridine (4ii), formed by rotating the latter 90°
about its center of mass, leads directly back to the INC
for 4a -d ; the 4e “INC” collapses back to the starting
structure (Scheme 3). Rotation of the pyridine by 180°
about its center of masssN lone pair pointing 180° away
from the carbenium ion with a benzyl-pyridine separa-
tion of ca. 3.6 Åsand minimizing in AM1 produces a
stable structure (4iv) that could contribute to an enthal-
pic bottleneck against collapse of the INC. Hydrogen-
bonded structures are not found in any INC. Thus,
within the limits of the semiempirical methodology, these
structures appear to be type 1 INCs.22
k1sdecreasing with
a
decrease in stabilitysand
k-1sincreasing with a decrease in stability. It seems
reasonable that K should decrease smoothly over the
series just as the equilibrium protonation of 2-(4-Y-
phenyl)propenes, the basis of the gas-phase δ∆G° scale
for tert-cumyl carbenium ions,15 changes smoothly and
continuously over a wide range of Y. It is difficult to
rationalize the discontinuity in the Hammett plot and
the dependence on the leaving group for the 4-NO2
substrates in terms of this continuous monotonic rate
expression. Thus, attempting to fit the results into a
mechanism involving only INC intermediates is unsuc-
cessful.
Considering the entire mechanism in eq 1 and applying
a steady state treatment yields kobsd ) kd + k1k2/(k-1
+
k2). With the assumption that k-1 > k2 for the reasons
give above, kobsd ) kd + Kk2. Because the mass of the
parents does not vary greatly and because a majority of
the internal energy imparted to RX+ by collisional
activation is released in bond breaking,23 kd should
remain relatively constant over the series and changes
in kobsd should be dominated by Kk2 when Kk2 > kd. The
Brønsted plots in Figure 2 show no or little effect of the
LG for the MeO, Me, H, and Cl substrates, which
suggests that this condition is satisfied. Under the
assumption that k2 is relatively constant, as R+ becomes
less stable changes in the values of k1 and k-1, but
especially in k-1, cause the value of K to diminish to the
point that kd > Kk2 for the 4-NO2 substrate, which could
account for the break in the Hammett plot and the
leaving group dependence exhibited in the Brønsted
In a very recent study of the gas-phase dissociation of
protonated tert-butyl alkyl ethers, Morton and his col-
leagues29 found that [t-Bu+ HOR] did not correspond to
a minimum on the potential energy surface calculated
at the 3-21G//3-21G level of theory. Because the mech-
anisms leading to various products require the existence
of an INC, however, its existence was inferred and
(29) Audier, H. E.; Berthomieu, D.; Morton, T. H. J . Org. Chem.
1995, 60, 7198-7208.

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 8, 1996 2759
Sch em e 4
dine, but in the benzyls it is localized almost entirely on
the carbenium ion. Thus, with the exception of the
inability of the benzyl INCs to undergo proton transfer,
the rate-energy correlations and molecular orbital pic-
tures are virtually identical in the two systems. While
this strong correspondence does not constitute proof that
both systems react by the same mechanism, it strength-
ens the arguments in favor of the INC mechanism for
the benzylpyridiniums.
F igu r e 5. HOMO and LUMO for 2d displayed in the
ChemPlus option of Hyperchem for the AM1-minimized struc-
ture. The LUMO of the carbenium ion has the same structure
as the INC.
Sch em e 3
As noted above, there is not sufficient experimental
evidence available to speculate about the mechanism of
dissociation for the dimethylsulfonium substrates. The
Hammett plot is linear (Figure 1), which suggests that
the mechanism is the same across the series, but this
would fit any of the schemes considered above for the
pyridiniums. The value of F+, -0.6, is 2-fold greater than
the value of F+ for the pyridiniums (-0.24 to -0.35),
suggesting that there is much more “carbenium ion
character” in the activated complex for the sulfoniums
(later transition state in Hammond terms). Semiempiri-
cal results provide some insight, however. The energy
profiles for 1a -e calculated in PM3 or AM1 (not shown)
do not have distinct transition states, and there are no
stable structures corresponding to INCs on the potential
energy surface. In fact, removing the constraint on
+
structures with benzyl-SMe2 bond lengths as great as
5 Å and minimizing leads to spontaneous collapse to the
starting structure. This result is not suprising given the
softness of both the carbenium ion and SMe2. If the SMe2
is rotated 180° about its center of mass to give a structure
(1a i) in which the Me’s shield the S and the structure is
minimized in AM1, unlike the equivalent benzylpyridine
structure shown above it collapsed directly back to the
starting structure through the “locked rotor” structure
1a ii (Scheme 4]. Thus, for the sulfoniums, stable struc-
tures corresponding to neither entropic nor enthalpic
bottlenecks are identified computationally. In this con-
text, the lack of break in the Hammett plot, the higher
value of F+, and the semiempirical results are consistent
with but do not prove the existence of a direct dissociation
mechanism (kd in eq 1).
ascribed to an entropic rather than an enthalpic bottle-
neck. As noted above, our semiempirical results are
consistent with the existence of structures for both
bottlenecks on the potential energy surface.
The molecular orbital picture for these species is
consistent with an INC as well. In the four stable
structures for 4a -d , the HOMO is localized on the
pyridine, while the LUMO is localized on the carbenium
ion (Figure 5 for the 4d INC). The LUMO in the INC
has the same shape found in the respective carbenium
ions, although there is slight delocalization into the
pyridine, which is consistent with polarization of the
neutral by the carbenium ion.
Arguing by analogy can be dangerous, but in this
instance there are very clear correspondences between
the experimental and computational results for the gas-
phase dissociation of the 2′-substituted â-nicotinamide
arabinosides and 2-5a -d that appear to react through
an INC.2 First, the relative rates for both series follow
LFERs. Second, both show excellent correlations be-
tween the experimental rate constants and the computed
values for the various relevant computed ∆Hq and ∆HR
values (not shown). Third, INCs are stable upon mini-
mization. Fourth, both the arabinosides and the benz-
ylpyridiniums have the HOMO localized on the pyridine
moiety; in the arabinosides, the LUMO is localized on
the oxocarbenium ion moiety and partially on the pyri-
Dissocia tion in Wa ter . Unlike the corresponding
plot for the gas-phase results, the Hammett plot (vs σ+)
for the hydrolysis in pure water (no added salt) of a series
of 3- and 4-substituted benzyldimethylsulfonium sub-
strates5,30 shown in Figure 6 has the break between 1a
and 2a typically found for the solvolysis and substitution
reactions of neutral benzyl substrates.31 It was reported
recently5,8 that this break is the result of a change in
(30) Friedberger, M. P.; Thornton, E. R. J . Am. Chem. Soc. 1976,
98, 2861-2865.
(31) See, for example: Fujio, M.; Goto, M.; Susuki, T.; Akasaka, I.;
Mishima, M.; Tsuno, Y. Bull. Chem. Soc. J pn. 1990, 63, 1146-1153.
Reference 48 has an extensive review of studies on the substitution
reactions of benzyl derivatives.

2760 J . Org. Chem., Vol. 61, No. 8, 1996
Buckley et al.
from flat and linear to V-shaped suggests that the charge
on the substrate and the nucleophile, and the HSAB rank
of the nucleophile, has a significant influence on the
structure of the activated complex.) Despite the border-
line kinetic behavior, there was no evidence for an IDC
in any system; in fact, rate and selectivity data for 1a
show unequivocally that an IDC is not involved.4,5,8
These pronounced differences between the solution and
gas phases show that it is not always wise to assign a
mechanism to one phase based on a mechanism in
another.
Im p lica tion s for th e Solu tion Rea ction s of th e
Rela tive Sta bilities of th e Ribosyl Oxoca r ben iu m
a n d (4-Meth oxyben zyl)ca r ben iu m Ion s in th e Ga s
P h a se. All evidence, including high values of the Taft
FI,1,3 high, negative Brønsted âLG values,35 and positive
∆Sq values,36 points to a well-behaved dissociative reac-
tion for NAD+ and 2′-substituted â-nicotinamide arabi-
nosides and ribosides that may occur through an IDC in
water3,37 and an INC in the gas-phase.2 A large body of
historical evidence38-40 is consistent with an A-1 mech-
anism for the specific acid-catalyzed hydrolysis of methyl
and other alkyl glucosides40 and acetals and ketals
formed from “normal” aldehydes, ketones, and aliphatic
alcohols.41 There is similar evidence that glucosylpyri-
diniums react through a dissociative mechanism.42 In a
very recent study, Bennet and colleagues43 showed that
2-deoxy-â-D-glucosylpyridiniums react with nucleophiles
(s ∼ 0) through a dissociative mechanism, with nucleo-
phile trapping at the solvent-separated ion pair. These
results require that ribosyl- and glucosyloxocarbenium
ions are intermediates, either as solvent-equilibrated
species or as elements of IDCs.
F igu r e 6. Plots of the gas-phase relative rates (k ) R+/[R+
+
M+], O) for 1a -e and the hydrolysis rate constants kobsd at 70
°C for 1-c and the (3-methylbenzyl)- and (3-bromobenzyl)-
dimethylsulfonium compounds from ref 30 (b) vs σ+. For the
gas-phase data, F+ ) -0.57 (r ) 0.9999 with the point for Cl
excluded). For the hydrolysis data, F+ ) -7.7 for the regression
line between 1a and 1b and F+ ) -0.82 (r ) 0.993) for the
compounds with σ+ > -0.06.
Despite this historical evidence, for some years J encks
argued that the glycosyloxocarbenium ion was too un-
stable to exist as an intermediate in either solvolysis44
or acid-catalyzed hydrolysis45,46 of glucosyl substrates.
The initial Young-J encks45 estimate of the lifetime of
the glycosyl oxocarbenium ion, 10-15 s, has been revised
downward over the years to 10-12 s.46 On the basis of a
study of the nucleophilic substitution reactions of R-D-
glucopyranosyl fluoride, Banait and J encks47 concluded
that the the glucosyloxocarbenium ion has a finite
lifetime in solution but that it does not become solvent
equilibrated and has no significant lifetime in contact
with a strong nucleophile.
F igu r e 7. Plots of the gas-phase relative rates (k ) R+/[R+
+
M+], O) for 2a -e and the hydrolysis rate constants kobsd at 70
°C for 2a -e (9) and 3a -e (b) vs σ+. For the gas-phase data
for 2a -d , F+ ) -0.24 (r ) 0.998), which is one-half the value
for 1a -e (Figure 6). For the hydrolysis data, F+ ) -1.25 (r )
0.997) for 2a -e and 0 for 3a -e.
mechanism from SN1 for the 4-MeO compound (a “kC”
process) to SN2 (a “kS” process) for compounds with
4-substituents with σ+ > -0.31. Unlike the correspond-
ing plot for the gas-phase results, the Hammett plot
based on estimated rate constants for 2a -e is linear with
a slope (F+ ) 1.25) and for 3a -e the Hammett plot is
linear and flat (F+ ∼ 0, Figure 7), consistent with a single
kS mechanism. (Reasons for the difference in slopes for
the pyridiniums are discussed elsewhere6 in terms of
(34) While the reaction of 1e with azide was studied only cursorily,
the rate constants for reaction with 2 M NaN3 are much lower than
those for 1a , which shows that the Hammett plot is not V-shaped as
found for the benzylpyridiniums. Buckley, N. Unpublished results.
(35) Tarnus, C.; Schuber, F. Bioorg. Chem. 1987, 15, 31-42. Tarnus,
C.; Mu¨ller, H. M.; Schuber, F. Ibid. 1988, 16, 38-51.
(36) Handlon, A. L. Ph.D. Dissertation, The University of California,
San Francisco, 1991.
(37) Ta-Shma, R.; Oppenheimer, N. J . Unpublished results.
(38) Schaleger, L. L.; Long, F. A. Adv. Phys. Org. Chem. 1963, 1,
1-34.
(39) Bunnett, J . F. J . Am. Chem. Soc. 1961, 83, 4956-4968; 4968-
4973; 4973-4978; 4978-4983.
Pross-Shiak theory.32
) The fact that both plots are
linear shows that the structure of the activated complex
changes constantly and smoothly as a function of sub-
stituent, unlike the recent results of Richard and Yeary33
for the reaction of azide with benzyl bromides that is
marked by curvature resulting from nonsynchronous
resonance demand in the activated complex. (The fact
that the Hammett plots for the second-order rate con-
stants for the azide reactions of the benzyl dimethylsul-
foniums is a smooth curve34 (concave up) and that the
plots for the azide reactions of 2-4a -e change shape
(40) Capon, B. Chem. Rev. 1969, 69, 407-498.
(41) Cordes, E. H.; Bull, H. G. Chem. Rev. 1974, 74, 581-603.
(42) Hosie, L.; Marshall, P. J .; Sinnott, M. L. J . Chem. Soc., Perkin
Trans. 2, 1984, 1121-1131.
(43) Huang, X.; Surry, C.; Hiebert, T.; Bennet, A. J . J . Am. Chem.
Soc. 1995, 117, 10614-10621.
(44) Sinnott, M. L.; J encks, W. P. J . Am. Chem. Soc. 1980, 102,
2026-2032.
(45) Young, P. R.; J encks, W. P. J . Am. Chem. Soc. 1977, 99, 8238-
8248.
(32) Pross, A. Adv. Phys. Org. Chem. 1985, 21, 99-195.
(33) Richard, J . P.; Yeary, P. E. A Simple Explanation for Curved
Hammett Plots for Nucleophilic Substitution Reactions at Ring-
Substituted Benzyl Derivatives; presented at the 206th ACS National
Meeting, Chicago, IL, 22-27 Aug, 1993.
(46) Amyes, T. L.; J encks, W. P. J . Am. Chem. Soc. 1989, 111, 7888-
7900.
(47) Banait, N. S.; J encks, W. P. J . Am. Chem. Soc. 1991, 113, 7951-
7958.

Reactions of Charged Substrates
J . Org. Chem., Vol. 61, No. 8, 1996 2761
Ch a r t 2
1-3a all have the same potential intermediate, the (4-
methoxybenzyl)carbenium ion that is known to be a
solvent-equilibrated intermediate in highly aqueous sol-
vent systems.5,8,48 These substrates, however, exhibit the
full range of possible mechanisms for the reaction with
azide in water under identical conditions: 1a , mixed SN1/
SN2 with comparable rates for both pathways;5,8 2a , direct
displacement by both azide and water,6 with the rate for
the water reaction much slower than the azide reaction
(krel ) 4 × 103); 3a , direct displacement by both azide
and water that occur at comparable ratesstypical bor-
derline kinetic behavior. Despite the fact that the pKa’s
of the LGs for 1a and 2a differ by 7-8 units, the second-
order rate constants for the azide reaction at 80 °C under
identical conditions differ only by a factor of 2.5. ∆Hq is
4.6 kcal/mol lower for 2a than for 1a , but ∆Sq is 15 gibbs/
mol more positive for 1a , the result of different solvation
of the LGs in the activated complex.5 This large differ-
ence in ∆Sq is apparently the source of the inability of
2a to undergo an SN1 reaction.6 There was no evidence
for ion pairs under conditions of constant ionic strength
in any system. Thus, by merely changing the LG in
substrates with the same putative stable, potentially
solvent-equilibrated carbenium ion,5,8,48 both relative
rates and mechanism change. In the benzyl series at
least, solvation is much more important than the stability
of the putative intermediate. While it is of course true,
as J encks has contended for many years,49 that a
substrate that cannot generate a stable intermediate is
“forced” to undergo a concerted displacement reaction,
the mere presence of a potentially stable intermediate
in a substrate does not guarantee that a stepwise
mechanism will be followed in a solvolysis or nucleophilic
substitution reaction.50
solvent-separated IDC was shown to be the intermediate
from which product formed.
Thus, the combination of solvent and LG effects makes
comparison of solution data as a measure of the relative
stabilities of carbenium ions extremely difficult. The
relative intrinsic gas-phase stabilities of the ribosyl
oxocarbenium ion and the (4-methoxybenzyl)carbenium
ion derived from arabinosyl compounds and 3a can be
compared directly from our data (Chart 2). For instance,
R+/[R+ + M+] for 2-deoxyribosyl-â-nicotinamide (12),
which forms the most stable carbenium ion in Handlon’s
series,2,3 is 0.24, while that for 3a is 0.10. The AM1-
calculated ∆Hq for the carbon-nicotinamide cleavage2 is
20.1 kcal/mol for 12 and 30.1 kcal/mol for 3a . The
corresponding values for the 2′-F-arabinoside, which
forms the least-stable oxocarbenium ion in the arabinosyl
series, are 0.11 and 26.0 kcal/mol, respectively. The same
trend is found in ∆HR; for 3a the value is 34.6 kcal/mol,
while the values for 2′-H, 2′-NH2, 2′-OH, 2′-NAc, and 2′-F
arabinosyl â-nicotinamide compounds, listed in the order
of decreasing stability of the oxocarbenium ions, are 27.5,
29.8, 32.7, 33.3, and 37.9 kcal/mol, respectively. Thus,
with nicotinamide as the LG, in the gas phase the ribosyl
INCs and oxocarbenium ions form at faster rates than
the quite stable 4-MeO-benzyl INCs and carbenium ions.
By this measure, the intrinsic stabilities of the ribosyl
oxocarbenium ions within the INC, and the bare species
that remain after diffusion away of the LG, are greater
than the (4-methoxybenzyl)carbenium ion. Even by the
Young-J encks rate comparison criterion,45 the same
trend is found in solution: the krel for hydrolysis of 12
and 3a is 6.4 × 104 at 96 °C in favor of 12. This
comparison of solution rate constants is invalid, however,
because 12 appears to solvolyze through an IDC,36,37
while 3a solvolyses by direct solvent displacement.6
One consequence of this stability and the effect of
solvent can be seen readily in the reactions of substrates
in enzyme active sites. Shuber and his colleagues35
suggested that the difference between the selectivity
The ground states and activated complexes for the
benzyl and arabionsyl and ribosyl substrates will be
solvated to much different extents, which makes a direct
comparison of results among the series risky at best.
Nonetheless, certain patterns are apparent. Despite the
unfavorable entropy of solvation of the LG, the â-
nicotinamide ribosides and arabinosides appear to un-
dergo fully dissociative reactions,1,3,35-37,51 which suggests
in turn that in this instance the stability of the oxocar-
benium ion is sufficient to overcome the negative driving
force of solvation entropy. Handlon found that addition
of exogenous nicotinamide did not affect the rate of
hydrolysis of several of the substrates.36 The lack of a
common LG effect rules out a solvent-equilibrated oxo-
carbenium ion as an intermediate, but is consistent with
a solvent-separated IDC, which in turn is consistent with
structures seen by Schro¨der et al.,51 in a computational
study of the hydrolysis of â-nicotinamide riboside, and
with selectivity and rate patterns seen by Ta-Shma and
Oppenheimer37 in a study of the solvolysis and azide
substitution reactions of NAD+ in methanol-water mix-
tures. In Bennet's recent study43 of glucosyl pyridinium
hydrolysis and nucleophilic substitution reactions, a
k
MeOH/kHOH for the solvolysis (kMeOH/kHOH ) 2) and enzyme-
catalyzed reactions (kMeOH/kHOH ) 100) of NAD+ analogs
was the result of the much greater selectivity of a fully
formed, stable ribosyl oxocarbenium ion in the active site
of the enzyme. This contention was recently confirmed
by a structure-activity study in the same enzyme.1b
Richard and his colleagues52 have recently reached a
similar conclusion for â-galactosidase, an enzyme that
yields a galactosyl oxocarbenium ion in the active site
that gives kMeOH/kHOH ) 130. Consistent with the gas-
phase stabilities, these results show that both ribosyl-
and pyranosyloxocarbenium ions appear to be stable
intermediates in enzyme active sites.
(48) Amyes, T. L.; Richard, J . P. J . Am. Chem. Soc. 1990, 112, 9507-
9512.
(49) J encks, W. P. Acc. Chem. Res. 1980, 13, 161-169. J encks, W.
P. Chem. Soc. Rev. 1981, 10, 345-375.
(50) The last sentence of Knier, B. L.; J encks, W. P. J . Am. Chem.
Soc. 1980, 102, 6789-6798, states that “[A]ll solvolysis and substitution
reactions at saturated carbon that proceed through SN2 displacement
mechanisms do so simply because the intermediate in the alternative
SN1 mechanism is too unstable to exist [italics added].”
(51) Schro¨der, S.; Buckley, N.; Oppenheimer, N. J .; Kollman, P. A.
J . Am. Chem. Soc. 1992, 114, 8231-8238.
Thus, rate and structure-activity data are mediated
by a complex interaction of solvent (∆Hq and ∆Sq), leaving
(52) Richard, J . P.; Westerfeld, J . G.; Lin, S.; Beard, J . Biochemistry
1995, 34, 11713-11724.

2762 J . Org. Chem., Vol. 61, No. 8, 1996
Buckley et al.
group (primarily ∆Sq), and substrate charge (differences
in solvation of ground state and activated complex
depending on charge) effects that are often difficult to
sort out. In the absence of gas-phase results, attempts
to assess these data in terms of the stability of the
intermediate may lead to entirely reasonable interpreta-
tions; if the intrinsic stabilities are known from gas-phase
studies, however, solution data may be interpreted
considering the range of possible interactions and effects.
State of California (N.J .O., N.B.), and a Research Award
from the UCSF Graduate Division (N.B.). Funds for the
purchase and support of the tandem spectrometer in the
UCSF Mass Spectrometry Facility (A. L. Burlingame,
Director) were provided by a grant from the Division of
Research Resources of the NIH (RR 01614) and a grant
from the NSF (DIR 8700766). We thank Randy Rad-
mer, J im Caldwell, and Peter Kollman (UCSF) for
helpful discussions concerning semiempirical calcula-
tions and the two reviewers for their insightful com-
ments on the manuscript.
Ack n ow led gm en t. Supported in part by NIH Grant
GM-22982 (N.J .O.), a Biotechnology Grant from the
J O951257W