Antihydrophobic cosolvent effects for alkylation reactions in water solution, particularly oxygen versus carbon alkylations of phenoxide ions

Article abstract of DOI:10.1021/ja012293h

Antihydrophobic cosolvents such as ethanol increase the solubility of hydrophobic molecules in water, and they also affect the rates of reactions involving hydrophobic surfaces. In simple reactions of hydrocarbons, such as the Diels-Alder dimerization of 1,3-cyclopentadiene, the rate and solubility data directly reflect the geometry of the transition state, in which some hydrophobic surface becomes hidden. In reactions involving polar groups, such as alkylations of phenoxide ions or S_N1 ionizations of alkyl halides, cosolvents in water can have other effects as well. However, solvation of hydrophobic surfaces is still important. By the use of structure-reactivity relationships, and comparing the effects of ethanol and DMSO as solvents, it has been possible to sort out these effects. The conclusions are reinforced by an ab initio computer model for hydrophobic solvation. The result is a sensible transition state for phenoxide ion as a nucleophile, using its oxygen n electrons to avoid loss of conjugation. The geometry of alkylation of aniline is very different, involving packing (stacking) of the aniline ring onto the phenyl ring of a benzyl group in the benzylation reaction. The alkylation of phenoxide ions by benzylic chlorides can occur both at the phenoxide oxygen and on ortho and para positions of the ring. Carbon alkylation occurs in water, but not in nonpolar organic solvents, and it is observed only when the phenoxide has at least one methyl substituent ortho, meta, or para. The effects of phenol substituents and of antihydrophobic cosolvents on the rates of the competing alkylation processes indicate that in water the carbon alkylation involves a transition state with hydrophobic packing of the benzyl group onto the phenol ring. The results also support our conclusion that oxygen alkylation uses the n electrons of the phenoxide oxygen as the nucleophile and does not have hydrophobic overlap in the transition state. The mechanisms and explanations for competing oxygen and carbon alkylations differ from previous proposals by others.

Full text of DOI:10.1021/ja012293h

Published on Web 03/13/2002
Antihydrophobic Cosolvent Effects for Alkylation Reactions in
Water Solution, Particularly Oxygen versus Carbon
Alkylations of Phenoxide Ions†
Ronald Breslow,* Kevin Groves, and M. Uljana Mayer
Contribution from the Department of Chemistry, Columbia UniVersity,
New York, New York 10027
Received October 3, 2001
Abstract: Antihydrophobic cosolvents such as ethanol increase the solubility of hydrophobic molecules in
water, and they also affect the rates of reactions involving hydrophobic surfaces. In simple reactions of
hydrocarbons, such as the Diels-Alder dimerization of 1,3-cyclopentadiene, the rate and solubility data
directly reflect the geometry of the transition state, in which some hydrophobic surface becomes hidden.
In reactions involving polar groups, such as alkylations of phenoxide ions or SN1 ionizations of alkyl halides,
cosolvents in water can have other effects as well. However, solvation of hydrophobic surfaces is still
important. By the use of structure-reactivity relationships, and comparing the effects of ethanol and DMSO
as solvents, it has been possible to sort out these effects. The conclusions are reinforced by an ab initio
computer model for hydrophobic solvation. The result is a sensible transition state for phenoxide ion as a
nucleophile, using its oxygen n electrons to avoid loss of conjugation. The geometry of alkylation of aniline
is very different, involving packing (stacking) of the aniline ring onto the phenyl ring of a benzyl group in
the benzylation reaction. The alkylation of phenoxide ions by benzylic chlorides can occur both at the
phenoxide oxygen and on ortho and para positions of the ring. Carbon alkylation occurs in water, but not
in nonpolar organic solvents, and it is observed only when the phenoxide has at least one methyl substituent
ortho, meta, or para. The effects of phenol substituents and of antihydrophobic cosolvents on the rates of
the competing alkylation processes indicate that in water the carbon alkylation involves a transition state
with hydrophobic packing of the benzyl group onto the phenol ring. The results also support our conclusion
that oxygen alkylation uses the n electrons of the phenoxide oxygen as the nucleophile and does not have
hydrophobic overlap in the transition state. The mechanisms and explanations for competing oxygen and
carbon alkylations differ from previous proposals by others.
Introduction
reactions. It can be diagnosed by seeing the effect of added
substances that increase or decrease the hydrophobic effect;
Hydrophobic Effects on Solubilities and Rates. When
organic reactions are performed in water they can be influenced
by the hydrophobic effect,1 which reflects the high energy of a
water interface with a nonpolar molecule or region. As we have
reported previously, Diels-Alder reactions2-7 and the benzoin
condensation8-10 show such effects. The hydrophobic effect
itself can lead to higher rates for these reactions in water than
in other solvents, and higher endo selectivity for Diels-Alder
prohydrophobic additives are generally simple salts such as
sodium or lithium chloride, while antihydrophobic additives
include salts of large ions such as guanidinium perchlorate.11
We have shown that such antihydrophobic ions, which are often
used as protein denaturants, function by bridging between the
nonpolar surface and the water solvent.6
Cosolvents such as ethanol also serve as bridging species in
water, diminishing the hydrophobic effect. This is most easily
diagnosed from the increased solubility of nonpolar substances
in such mixed solvents. Because solubility is an equilibrium
constant, its log reflects a free energy of solution, and the
increased solubility produced from an antihydrophobic additive
to water can be described in terms of a change of the free energy
of solution. This free energy change is directly proportional to
the log of the ratio of the solubilities with and without the
cosolvent.
† Taken in part from the Ph.D. theses of M. Uljana Mayer (1999) and
Kevin Groves (2001), Columbia University.
* To whom correspondence should be addressed. E-mail: rb33@
columbia.edu.
(1) Tanford, C. The Hydrophobic Effect: Formation of Micelles and Biological
Membranes, 2nd ed.; John Wiley & Sons: New York, 1980.
(2) Breslow, R.; Rideout, D. C. J. Am. Chem. Soc. 1980, 102, 7817.
(3) Breslow, R.; Maitra, U. Tetrahedron Lett. 1984, 25, 1239-1240.
(4) Breslow, R. Acc. Chem. Res. 1991, 24, 159-164.
(5) Breslow, R. Hydrophobic and Antihydrophobic Effects on Organic Reac-
tions in Aqueous Solution; Cramer, C. J., Truhlar, D. G., Eds.; American
Chemical Society: Washington, DC, 1994; pp 291-302.
(6) Breslow, R.; Guo, T. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 167.
(7) Breslow, R.; Zhu, Z. J. Am. Chem. Soc. 1995, 117, 9923-9924.
(8) Kool, E. T.; Breslow, R. J. Am. Chem. Soc. 1988, 110, 1596-1597.
(9) Breslow, R.; Connors, R. V. J. Am. Chem. Soc. 1995, 117, 6601-6602.
(10) Breslow, R.; Connors, R. J. Am. Chem. Soc. 1996, 118, 6323-6324.
Antihydrophobic agents such as organic cosolvents have been
very useful as a method of experimentally measuring the
(11) Breslow, R.; Rizzo, C. J. J. Am. Chem. Soc. 1991, 113, 4340-4341.
9
3622
J. AM. CHEM. SOC. 2002, 124, 3622-3635
10.1021/ja012293h CCC: $22.00 © 2002 American Chemical Society

Cosolvent Effects for Alkylation Reactions
A R T I C L E S
Table 1. Solubilities in Water and Cosolvent Solubility
Perturbations by 20% Ethanol and 20% DMSO (Error e 5%)a
solvation properties of transition states for reactions in water,
particularly the amount of exposed hydrophobic surface area
relative to that of the starting materials.4,5,7,9,10,12,13 The use of
small amounts of cosolvents allows us to observe a change in
solvation of hydrophobic surfaces while preserving the es-
sentially aqueous environment. In choosing the concentration
of the cosolvent, we want to maximize the magnitude of
experimentally measured hydrophobic solubility perturbation
without changing the properties of the solution so drastically
that the solvent structure is no longer waterlike.
S
20% EtOH
S
20% DMSO
Swater (mM)
Swater
Swater
benzene
21 ( 1
1.7
1.8
1.8
3.9
2.3
2.2
1.8
N-methylaniline
benzaldehydeb
benzyl phenyl ether
p-nitrobenzyl chloride
p-nitrobenzyl bromide
0.080 ( 0.003
0.7
0.2
5.3
3.4
3.6
a The data except for those of footnote b are results from this work.
Thus, the table could be in the Results and Discussion section, but is placed
here for intellectual reasons. b Reference 9.
Grunwald has proposed two additive terms to describe the
partial molar thermodynamic properties of water-rich aqueous
alcohol solutions.14a The first term, called the “isodelphic” term,
describes effects on a solute where the bulk aqueous solvent
network theoretically remains unchanged upon addition of small
amounts of an alcohol cosolvent. The second term, called the
“lyodelphic” term, describes effects due to changes in the solvent
network resulting from the cosolvent. For our studies, we wish
to maximize the isodelphic effects, which include the solvation
of hydrophobic surfaces, while minimizing the lyodelphic
effects, which involve a significant change from pure water
structure. We can do this by keeping cosolvent concentrations
low enough as to remain under isodelphic conditions with
negligible lyodelphic effects.
(5% or better, we consider 20% ethanol to be sufficient for
measuring changes in the amount of hydrophobic surface of a
reaction as it reaches its transition state.
The aqueous solubility perturbations by 20% v/v ethanol on
benzaldehyde and N-methylaniline are essentially the same as
that of benzene (Table 1), indicating that interactions of the
cosolvent with the substituent aldehyde and amino groups are
small. We conclude that the dominant effect of 20% ethanol
on neutral hydrophobic molecules is to solvate hydrophobic
surfaces and that the bulk structure of the solvent network
remains similar to that of pure water.
Haak and Engberts14b have described a critical hydrophobic
interaction concentration, or “chic”, for organic cosolvents in
water and related it to Grunwald’s isodelphic/lyodelphic analy-
sis. In solutions with alcoholic cosolvent concentrations above
the chic, there are not sufficient water molecules for the
formation of complete hydration shells of the alcohol, and
lyodelphic effects become important. Under these conditions,
the alcohol molecules will form clusters, which create micro-
heterogeneity in the solution. Cosolvent effects on solutes will
then reflect interactions of the solute with the clusters, often
showing very complex solvation effects and large temperature
dependencies.14b Below the chic, however, lyodelphic effects
do not come into play, and the solutions remain isodelphic in
nature.
We find that DMSO (dimethyl sulfoxide) is even more
effective at solvating the hydrophobic surfaces than is ethanol
(Table 1), even though DMSO is a more polar molecule with
respect to dielectric constant. The contrast between ethanol and
DMSO cosolvents in water has proven to be a valuable tool in
exploring mechanisms.
We see that the free energy changes induced by ethanol
cosolvent are proportional to the extent of the nonpolar surface
in the solute; for a given concentration of ethanol, for instance,
compounds with two fully exposed phenyl groups, such as (E)-
diphenyloxirane 1, have twice as large a δ∆G° as those with
one phenyl group.10 Compounds with two partially overlapping
phenyl groups, such as (Z)-diphenyloxirane 2, have δ∆G°s
midway between 100% and 200% of those for monophenyl
compounds, reflecting the fact that partial packing of the phenyls
on each other diminishes the amount of nonpolar surface
exposed to the water solvent. Thus, we proposed that antihy-
drophobic cosolvent effects on reaction rates in water solution
could be quantitatively interpreted in terms of the amount of
hydrophobic surface that becomes shielded from the solvent in
the transition state (TS) for a reaction (Figure 1).
For the experiments in this work we have predominantly used
20% v/v ethanol (which corresponds to 7.2 mol % or 13 water
molecules per ethanol molecule) to probe organic displacement
reactions for changes in exposed hydrophobic surface area and
other effects. We have performed solubility measurements of
benzene in pure water and water with 5%, 9%, 13%, and 20%
v/v ethanol. The natural log of the ratio of the solubility with
cosolvent to that in pure water, which is proportional to the
change in free energy of solvation by the cosolvent, was plotted
against the ethanol concentration.15 We saw that the aqueous
free energy solubility perturbations of benzene were completely
linear with respect to ethanol concentration all the way up to
20% v/v. Thus ethanol-ethanol interactions in solutions at 20%
ethanol are minimal, and we have not drastically altered the
aqueous structure of the solvent. Because the 70% increase in
aqueous solubility of benzene by 20% ethanol is well outside
the experimental error for measuring rate constants, which is
(12) Breslow, R.; Connors, R.; Zhu, Z. Pure Appl. Chem. 1996, 68, 1527-
1533.
(13) Breslow, R.; Groves, K.; Mayer, M. U. Pure Appl. Chem. 1998, 70, 1933.
(14) (a) Grunwald, E. J. Am. Chem. Soc. 1984, 106, 5414-5420. (b) Haak, J.
R.; Engberts, J. B. F. N. J. Am. Chem. Soc. 1986, 108, 1705-1706.
(15) See ref 36, p 121.
9
J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002 3623

A R T I C L E S
Breslow et al.
Figure 3. In the benzoin condensation, the rate-determining addition of
malononitrile anion to benzaldehyde occurs with packing - but only partial
packing - of the phenyl rings in the transition state, because of the
stereoelectronic requirements for addition to the carbonyl group.
We also saw that in the cyanide-catalyzed benzoin condensation
the phenyl ring of mandelonitrile anion overlaps, but only
partially, with the phenyl ring of benzaldehyde at the TS (Figure
3), again consistent with a sensible TS.9 The nucleophile
approaches from the rear of the carbonyl group, leading to about
40% overlap of the two phenyls at the TS.
These conclusions involved the assumption that the only
significant rate-modifying effect of the alcohol cosolvents is to
solvate the hydrophobic surfaces, and this is probably true for
the Diels-Alder reactions. For the benzoin condensation, we
are dealing with ions whose solvation might well be affected
by nonpolar cosolvents in water. We have also examined the
application of our procedure to other reactions such as alkyla-
tions, in which solvation of ions is surely important.
Figure 1. Cosolvents may lower the free energy of the reactants, products,
and transition state for a reaction performed in water. The lowering of the
free energies of reactants and products can be determined from the effects
of the cosolvents on their solubilities, while the lowering of the energy of
the transition state can be determined from the change in the reaction rate
if the cosolvents have no other effect.
Alkylation Reactions. There is much previous work studying
solvent polarity effects on reaction rates in alkylation reactions.
For example, Grunwald and Winstein16 studied the solvolysis
of tert-butyl chloride in various solvents and proposed a linear
free energy relationship in which a Y value for a given solvent
reflected the “ionizing power” of the solvent, its ability to solvate
the developing tert-butyl cation and chloride ion in the TS.
Bentley removed the possibility of nucleophilic participation
of solvent at the developing tert-butyl cation by examining
1-adamantyl chloride solvolyses,17 while Kevill and Anderson
removed the chloride solvation by examining adamantyl di-
methylsulfonium ion solvolysis.18-20
For SN2 displacements, Hughes and Ingold described four
charge types, and the dependence of their rates on solvent
polarity, depending on the charge on the nucleophile and the
electrophile.21 When a neutral nucleophile and electrophile
produce charged products in the TS, polar solvents speed the
reaction. When a charged nucleophile and charged electrophile
produce neutral products, polar solvents slow the reaction. When
there is no net change in charge in the reaction, but one
component is charged, there can be some slowing by polar
solvents if they solvate the TS less well than they solvate the
charged reactant.
Figure 2. The Diels-Alder dimerization of 1,3-cyclopentadiene in water
has a transition state in which one face of each ring is hidden from the
solvent. The Diels-Alder addition of N-methylmaleimide to anthacene-
carbinol has a transition state in which about 50% of an anthracene face is
covered, but in the product only 10% is covered. These results from our
experimental work are consistent with theoretical models.
As Figure 1 shows, if the transition state (TS) of a reaction
has the same amount of similar exposed hydrophobic surface
as occurs in the starting materials, and if the cosolvent really
has no effect other than to help solvate hydrophobic surfaces,
then the rate of the reaction will be unaffected by adding the
cosolvent to the water solution. That is, the ∆G° of the TS will
be lowered to the same extent as that of the starting materials,
so the activation energy and the rate will be unchanged. If the
TS is less hydrophobic than the starting materials - for example,
if some hydrophobic surface is hidden from solvent in the TS,
or if the hydrophobicity of a solvent-exposed reactant decreases
in the TS - the cosolvent should slow the reaction by an amount
corresponding to the difference in δ∆G° for starting materials
and TS. If the TS is more hydrophobic than the starting materials
- as happens when a delocalized phenoxide ion is partially
neutralized in the TS (see below) - the reaction will be faster
with the added cosolvent. The detailed mathematical treatment
of the effects has been published elsewhere.10,12
We confirmed this idea in the Diels-Alder dimerization of
1,3-cyclopentadiene.7 The reaction, forming the endo product,
was 20% slower in 15% v/v ethanol than in pure water, and
the solubility of cyclopentadiene was 25% greater in that solvent.
Translating into free energies, this indicates that 46% of each
cyclopentadiene ring is concealed from solvent in the TS,
completely reasonable for two rings face-to-face and consistent
with our quantum mechanical calculations of the TS (Figure
2). With this technique we also concluded that N-methylmale-
imide covers about 25% of the total surface of anthracene in
the TS of their Diels-Alder reaction, but that in the product
only 10% of it is covered, both reasonable results (Figure 2).
Specific solvation of the nucleophile and the leaving group
can also have great impact on SN2 reactions, especially in protic
solvents.22 Hydrogen bonding to the nucleophile and leaving
group is particularly important with smaller ions.
These previous studies on solvent effects have focused on
the solvation of polar species. There has been little attention
paid to the change in solvation of the nonpolar species, those
that in water exhibit a hydrophobic effect. For example, the
(16) Grunwald, E.; Winstein, S. J. Am. Chem. Soc. 1948, 70, 846.
(17) Bentley, T. W.; Carter, G. E. J. Am. Chem. Soc. 1982, 104, 5741-5747.
(18) Kevill, D. N.; Anderson, S. J. Am. Chem. Soc. 1986, 108, 1579-1585.
(19) Kevill, D. N.; Kamil, W. A.; Anderson, S. W. Tetrahedron Lett. 1982, 23,
4635-4638.
(20) Kevill, D. N.; Anderson, S. W.; Fujimoto, E. K. AdV. Chem. Ser. 1987,
269-283.
(21) Hughes, E. D.; Ingold, C. K. J. Chem. Soc. 1935, 244.
(22) Parker, A. J. Chem. ReV. 1969, 69, 1.
9
3624 J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002

Cosolvent Effects for Alkylation Reactions
A R T I C L E S
Table 2. Rate Constants in Water and with Added Ethanol and DMSO for the Reactions of 4-Carboxybenzyl Chloride (CBC) and
Haloacetamides with Various Nucleophiles under Initial Rate Conditions at 25 °C in M-1 min-1 (Error e 5%)
nucleophile
electrophile
kwater
k20% ethanol
k20% DMSO
k20% EtOH/kwater
k20% DMSO/kwater
N-methylaniline
phenoxide ion
2,6-dimethylphenoxide ion
2,6-dimethylphenoxide ion
4-nitrophenoxide ion
cyanide
methoxylamine
N-methylaniline
N-methylaniline
phenoxide ion
CBC 4
CBC 4
CBC 4 (O-alk)
CBC 4 (para-alk)
CBC 4
0.554
0.053
0.018
0.032
0.0082
0.0055
0.006
0.033
0.0534
0.022
0.009
0.013
0.0011
0.418
0.051
0.02
0.33
0.75
0.97
1.06
0.65
0.67
1.02
0.78
0.63
0.7
0.95
1
1.25
0.74
0.79
1.5
0.60
1
1.25
0.56
0.55
1.21
0.053
0.023
0.018
0.0045
0.0066
0.02
0.0055
0.0056
0.0047
0.021
0.037
0.021
0.009
0.017
0.0008
CBC 4
CBC 4
BrCH2CONH2
ICH2CONH2
BrCH2CONH2
ICH2CONH2
BrCH2CONH2 (O-alk)
BrCH2CONH2
BrCH2CONH2
methyl mesylate
methyl iodide
methyl mesylate
methyl iodide
0.0482
0.014
0.9
phenoxide ion
1.6
2,6-dimethylphenoxide ion
4-nitrophenoxide ion
methoxylamine
hydroquinone dianion
hydroquinone dianion
phenoxide ion
0.0011
1.02
0.109
0.0149
0.022
0.001
0.164
0.0228
0.036
1.7
1.6
2.1
phenoxide ion
0.0021
ionization of tert-butyl chloride to the tert-butyl cation and
chloride anion converts a hydrophobic tert-butyl group to a
hydrophilic one. Thus, part of the decreased rate in water when
ethanol is added comes from the better solvation of the substrate
by ethanol than of the ionic transition state. Our calculation,
using the techniques described below for computation of
hydrophobic interactions, indicates that this effect on hydro-
phobic character is significant but cannot account for more than
one-half of the entire ethanol rate effect; the various ionic
solvation interactions are still important.
ion attacks with its oxygen n electrons, and there is no overlap
of the phenyl rings in the TS. Phenoxide has this choice, which
aniline does not, and the proposed transition state avoids loss
of conjugation of the oxygen electrons with the benzene ring.
However, here too it is important to see what other aspects of
the TS the cosolvent can affect, besides the energy change that
would have resulted if there had been any overlap of the
hydrophobic reactants.
We found that the alkylation of N-methylaniline (3) by
p-carboxybenzyl chloride (4) in water is 25% slowed by 20%
v/v added ethanol (Table 2).23 A sensible TS (5) for this process
involves overlap of the two phenyl rings of the substrates, as
we proposed. The nitrogen electron pair is conjugated with the
π system, displacement at the benzylic carbon should come from
behind the C-Cl bond, and developing partial cationic character
on the benzylic carbon should orient the C-Cl bond perpen-
dicular to the benzene plane. This could lead to the rate decrease
with added ethanol. However, two neutral species are forming
two ions in this alkylation, and solvent effects on the energy of
developing ions in the TS could be important. For example,
we see that displacement by methoxylamine on 2-bromoacet-
amide, for which there is no hydrophobic contribution, is 20%
slower in the 20% v/v ethanol/water solvent than that in pure
water, reflecting solvation of the polar TS.
If charge is lost in the TS, as in the alkylation of a phenoxide
ion, the less polar mixed solvent can accelerate the reaction, as
Hughes and Ingold pointed out.21 The reactions of methyl
mesylate and methyl iodide with phenoxide ion (Table 1)
showed large rate increases - 60% and 110%, respectively -
upon addition of 20 volume % ethanol.23 (The difference reflects
different solvation demand by the leaving groups, with the larger
developing iodide ion requiring less solvent stabilization.)
Cyanide ion, with a basicity similar to that of phenoxide ion,
showed rate increases with added ethanol of 20% in the reaction
with methyl mesylate, and 70% with methyl iodide.23 We
conclude that a part of the rate increase upon addition of ethanol
is caused by decreased hydrogen bonding solvation of the
nucleophile lone pairs, but the larger rate effect in the reaction
of phenoxide involves an additional factor. In phenoxide ion
the negative charge is delocalized into the benzene ring, making
the ring less hydrophobic. As the TS is approached the charge
is partly lost, the ring becomes more hydrophobic, and this factor
should add to the acceleration by ethanol. Our computer
modeling, discussed below, supports this interpretation.
We also examined the alkylation of phenoxide ion by
p-carboxybenzyl chloride, and saw that in this case there was
essentially no rate effect of added ethanol (Table 2).23 This
makes sense for the TS 6 we proposed in which the phenoxide
O versus C Alkylation of Phenoxide Ions. The alkylation
of a phenoxide nucleophile usually results in the exclusive
formation of the ether product.24 However, in certain solvents,
(23) Breslow, R.; Groves, K.; Mayer, M. U. Org. Lett. 1999, 1, 117-120.
9
J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002 3625

A R T I C L E S
Breslow et al.
most notably water, considerable amounts of ortho and para
We have now examined various displacement reactions of
phenoxides in aqueous solution, and have found strong evidence
that the generally accepted set of factors governing the position
of phenoxide alkylation is incomplete. As in our previous work,
we have used cosolvents to observe how small changes in the
solvent properties affect the rates of reactions and product
selectivity. We have combined our interpretations of cosolvent
effects with careful control reactions and some computational
data to create a revised picture of the alkylation of phenoxides,
incorporating effects stemming from both the nucleophile and
the electrophile and their mutual interaction in aqueous solution.
The result is the discovery that hydrophobic interactions
dominate the choice between oxygen and carbon alkylation of
phenoxide ions. Another result is to confirm our picture of
oxygen alkylation by benzyl chloride involving an unpacked
geometry, with attack by the n electrons of the oxygen.
carbon alkylation can occur with some electrophiles.24-26
A
number of possible factors affecting the position of alkylation
of phenoxide ions have been discussed in the literature, including
the effects of solvent dielectric,27 reaction heterogeneity,26,28
SN1-SN2 character of the transition state,29 steric effects,30
selective solvation,24 and hydrogen bonding.25
Kornblum and Lurie have reported that the alkylation of
phenoxide salts that are tightly ion paired in nonpolar solvents
often leads to carbon alkylation that is not observed in polar
solvents.28 Curtin et al.26 also reported that the nature of the
cationic counterion influences the amount of carbon alkylation
in nonpolar solvents, with smaller cations such as lithium
favoring ortho alkylation, while the larger potassium ion favors
oxygen alkylation. In nonpolar solvents the cation is very closely
associated with the oxygen of the phenoxide anion. In our work
in media that are mostly aqueous, no tight ion pairing between
a phenoxide anion and a sodium cation is expected.
Results and Discussion
Alkylations. As further evidence for the increased hydro-
phobicity of a phenyl ring in the TS for phenoxide alkylation,
we have examined the reaction of hydroquinone dianion 7 with
methyl iodide and methyl mesylate (Table 2). The ethanol
cosolvent effects seen were only as large as those for cyanide
ion nucleophile, not the larger effects seen with phenoxide ion
nucleophile. The hydroquinone dianion never becomes hydro-
phobic during monoalkylation, since only part of one of the
two delocalized charges is neutralized in the TSs.
The SN1-SN2 character of a transition state affects the hard-
soft character of the electrophile and can influence the course
of phenoxide alkylation.28,31Steric hindrance can have a powerful
effect on the outcome of a phenoxide alkylation. Kornblum has
found that 2,6-di-tert-butylphenoxide gives mostly oxygen
alkylation with methyl iodide, but as the electrophile becomes
more bulky, as with ethyl iodide and isopropyl iodide, the
reaction is pushed entirely to the less hindered para position.30
It has been reported that solvents such as water and trifluo-
roethanol favor carbon alkylation of phenoxides, while only
oxygen alkylation is observed in most other polar solvents.24,25
The data should be taken with some caution, however, as the
reactions in water and trifluoroethanol were run as heteroge-
neous two-phase reactions because the allyl halide electrophiles
are not soluble in the solvents. To explain these observations,
Kornblum24 states that in these solvents “the oxygen of the
phenoxide is so intensely solvated that the availability of the
oxygen for nucleophilic displacement is greatly decreased; as
a consequence, displacements employing the otherwise unfa-
vored ortho and para carbon atoms compete successfully.”
Another factor, not mentioned by previous workers, actually
dominates the choice between O and C alkylation of phenoxide
ions in water. We have seen that carbon alkylation in water is
observed only with certain electrophiles, and we found it to be
accelerated in water over other solvents.23 It is not simply seen
by default once oxygen alkylation is suppressed by a protic
solvent, as others had suggested. In light of our past work
demonstrating the powerful role that hydrophobic packing at a
transition state can have in influencing the course of Diels-
Alder and other reactions, we postulated that hydrophobic effects
might also play a role in determining the regiochemistry of
phenoxide alkylation.
Simple phenoxide nucleophile becomes more hydrophobic
in the TS for reaction with a benzyl chloride as anionic
delocalized charge is neutralized. At the same time, the benzyl
chloride reactant develops some dispersed positiVe charge in
the TS, even though the reaction shows strictly second-order
kinetics, and this charge will decrease its hydrophobicity in the
TS. One component becomes more hydrophobic; the other one
becomes less hydrophobic. We have performed a number of
studies to sort out the various factors involved in such reactions
of phenoxide ions.
As one approach, we substituted DMSO for ethanol. As we
have described, DMSO at 20% v/v in water is about 50% more
effective than is ethanol in lowering the energy of phenyl
derivatives by solvation (Table 1), but the aqueous DMSO
solution is much more polar than is the ethanol solution.23
(However, hydrogen-bond donation could be more effective with
ethanol than with DMSO.) The dielectric constant of water is
78.5, that of 20% v/v aqueous DMSO is 77.1, while that of
20% v/v ethanol is 72.1.32,33 If the cosolvent effect is mainly to
decrease the dielectric constant of the solvent, and thus its ability
(24) Kornblum, N.; Berrigan, P. J.; LeNoble, W. J. J. Am. Chem. Soc. 1960,
82, 1257-1258.
(25) Kornblum, N.; Berrigan, P. J.; LeNoble, W. J. J. Am. Chem. Soc. 1963,
85, 1141.
(26) Curtin, D. Y.; Crawford, R. J.; Wilhelm, M. J. Am. Chem. Soc. 1958, 80,
1391-1397.
(27) Kornblum, N.; Seltzer, R.; Haberfield, P. J. Am. Chem. Soc. 1963, 85, 1148.
(28) Kornblum, N.; Lurie, A. P. J. Am. Chem. Soc. 1959, 81, 2705.
(29) Kornblum, N.; Smiley, R. A.; Blackwood, R. K.; Iffland, D. C. J. Am.
Chem. Soc. 1955, 77, 6269.
(30) Kornblum, N.; Seltzer, R. J. Am. Chem. Soc. 1961, 83, 3668-3671.
(31) March, J. Chapter 10; John Wiley & Sons: New York, 1992; pp 365-
368.
9
3626 J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002

Cosolvent Effects for Alkylation Reactions
A R T I C L E S
Table 3. Rate Constants and Cosolvent Effects for the Reactions of p-Carboxybenzyl Chloride 4 and p-Carboxybenzyl Dimethylsulfonium
Chloride 8 with Various Phenoxide Ions in Watera
4
8
kwater
k20% EtOH
kwater
k20% DMSO
kwater
kwater
(M-1 min-1
)
k20% EtOH
kwater
k20% DMSO
kwater
(M-1 min-1
)
phenoxide
5.3 × 10-2
1.8 × 10-2
3.2 × 10-2
8.2 × 10-3
5.5 × 10-1
0.97
1.06
0.65
0.67
0.75
1.00
1.25
0.56
0.55
0.60
6.6 × 10-5
3.9 × 10-5
4.4 × 10-5
1.1 × 10-5
2.1 × 10-3
1.50
1.68
0.97
0.84
0.89
1.40
1.62
0.72
0.69
0.70
2,6-dimethylphenoxide (oxygen)
2,6-dimethylphenoxide (p-carbon)
p-nitrophenoxide
N-methylaniline
a Data are initial rate at 25 °C (error e 5%).
to support charges, ethanol should slow the reactions more than
does DMSO. If the cosolvent effect is mainly to solvate nonpolar
sections of the reactants and TS - slowing reactions in which
either the hydrophobicity or the amount of exposed nonpolar
surface decreases in the TS, and speeding others in which they
increase - the effect of DMSO should be greater than that of
ethanol.
As another approach to sorting the factors involved, we have
replaced the p-carboxybenzyl chloride 4 by a p-carboxyben-
zylsulfonium ion 8, so the leaving group would not develop
negative charge. This lets us learn the extent to which altered
solvation of a leaving chloride ion plays a role in the cosolvent
effects on rates. We prepared compound 8 by reaction of
p-carboxybenzyl chloride with dimethyl sulfide and examined
it in reactions with N-methylaniline and with phenoxide and
p-nitrophenoxide ions. As Table 3 shows, the rate decrease
induced by ethanol in reaction of 8 with N-methylaniline was
significantly less than that for reaction of the benzylic chloride
4, indicating that some of the chloride rate decrease with added
cosolvents results from poorer solvation of the developing
chloride ion. Very similar changes were also seen with nucleo-
philic p-nitrophenoxide ion, again indicating the importance of
chloride ion solvation.
The data for attack by simple phenoxide ion are most striking
(Table 3). While in the benzylic chloride case 4 there was almost
no cosolvent effect, with the sulfonium salt 8 there was a 50%
rate increase with added ethanol, and 40% with added DMSO.
Apparently with the benzylic chloride 4 the rate-increasing
ethanol cosolvent effect in somewhat desolvating the nucleo-
philic oxyanion of phenoxide ion - and in solvating the
increasingly hydrophobic phenyl ring of the phenoxide - was
compensated by the rate-decreasing cosolvent effect on the
solvation of the leaving chloride ion, while no such compensa-
tion occurred with the sulfonium substrate 8. Ionic solvation
dominates the cosolvent effects, as the greater effectiveness of
ethanol than of DMSO indicates. The net hydrophobic effects
are not large; the increased hydrophobicity of phenoxide ion in
the TS is compensated by the decreased hydrophobicity of the
partially cationic benzyl group in the TS. There is no reason to
invoke hydrophobic packing. Support for the proposal that there
is no hydrophobic packing in the TS for phenoxide attack on
benzylic chlorides also comes from our studies on cases in which
phenoxide ion is alkylated competitively on both oxygen and
carbon, as described below.
The results are shown in Table 2. The reaction of N-
methylaniline with 2-iodoacetamide and of p-nitrophenoxide ion
with 2-bromoacetamide both show rate decreases of about 30%
with added 20% v/v ethanol as compared with pure water, but
both reactions are almost unaffected by added DMSO. Thus,
the cosolvent effects for these reactions reflect the dielectric
constant of the medium, not any hydrophobic effects. Both the
aniline and the nitrophenoxide put only little negative charge
into their phenyl rings. The amino group of aniline is much
less basic than is the oxyanion of phenoxide, so there will be
much less charge distributed into the aniline ring than for
phenoxide ion. In p-nitrophenoxide the negative charge is
distributed between the oxyanion and the already polar nitro
group, with little on the benzene carbons. Thus, in alkylations
of aniline and of p-nitrophenoxide ion by simple electrophiles
there is no large change in the hydrophobic effect, in contrast
to the increased hydrophobicity at the TS in such alkylations
of unsubstituted phenoxide ion.
However, the reaction rate of N-methylaniline with p-
carboxybenzyl chloride is decreased by 25% with added ethanol,
and 40% with added DMSO (Table 2). The reaction of this
benzyl chloride with p-nitrophenoxide also shows more slowing
with DMSO than with ethanol. Thus, in these alkylations there
is a decrease in the hydrophobic character of the TS relative to
the starting materials. This decrease reflects either (1) some
packing of the phenyl rings of electrophile and nucleophile in
the TS, or (2) a decrease in the hydrophobicity of the benzyl
system as the partial positive benzylic charge in the TS is
distributed throughout the phenyl ring. Because the effect is
seen with both the aniline and the nitrophenoxide ion nucleo-
philes, the latter of which surely has no packing interaction with
the benzyl chloride, partial delocalized cationic charge in the
electrophile must be involved. Our computational models
described below indicate that a benzyl cation will be much less
hydrophobic than a neutral benzene, because of the charge
distribution into the phenyl ring. Possible hydrogen bonding to
the leaving group could also be altered to a different extent by
the two cosolvents.
We have also used p-nitrobenzyl chloride (9), in which there
is surely less delocalized positive charge induced in the TS for
substitution than with p-carboxybenzyl chloride. In fact, the rate-
accelerating effects of a p-nitro group in SN2 reactions of benzyl
halides almost certainly reflect some negatiVe charge into the
benzyl group in the TS, probably by three-center bonding of
the nucleophile to both the benzyl and the ipso carbons.
However, that charge goes into the nitro group (which even
without the charge is polar, not hydrophobic), not onto benzene
ring carbons.
(32) Timmermans, J. Physicochemical Constants of Binary Systems; Interscience
Publishers: New York, 1960; Vol. 4.
(33) Kaatze, U.; Pottel, R.; Schafer, M. J. Phys. Chem. 1989, 93, 5623.
We see that 20% v/v ethanol increases the rate of reaction of
9 with phenoxide ion by 20%, while 20% v/v DMSO increases
9
J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002 3627

A R T I C L E S
Breslow et al.
Table 4. Minimized Interaction Energies of Ethane with Aromatic
Substrates in Water, As Calculated by the ab initio Jaguar
Program, and Comparison with Some Experimental Values
of the ethanol stabilization of the hydrophobicity of phenol is
likely to be lost when it ionizes to the anion. Conversely, when
a phenoxide ion is the nucleophile there will be a significant
increase in hydrophobicity, and in its stabilization by ethanol
or DMSO, as it loses some of its charge in the TS. Our Bro¨nsted
plots of the alkylation of various para substituted phenoxides
with p-carboxybenzyl chloride 4 indicate that 29% of the charge
on the phenoxide ion is neutralized at the TS in water, and 34%
in 20% v/v ethanol.34
calculated ethane
interaction energy,
kcal/mol
experimental 20%
ethanolic solubility
perturbation, kcal/mol
delocalized
charge?
benzene
benzaldehyde
no
no
no
yes
no
yes
no
no
-1.40
-1.45
-1.44
-0.46b
-1.31b
-0.16b
-1.21b
-1.19b
(0)b,c
-0.35
-0.37a
-0.35
N-methylaniline
phenoxide ion
phenoxy radical
thiophenoxide ion
thiophenoxy radical
anilinium cation
N-methylanilinium
radical cation
When an aniline is protonated, there is a small decrease in
its interaction energy with ethane in water. Thus, in a TS in
which the aniline nitrogen is the nucleophile we may see a small
decrease in rate caused by nonpolar cosolvents such as ethanol
or DMSO, in contrast to the rate increase induced in phenoxide
ion as nucleophile. However, when the positive charge is
delocalized into the ring, in the cation radical of N-methylaniline,
all hydrophobicity disappears. This is the result of delocalization
of charge, not of the radical, since phenoxy radical is almost as
hydrophobic as is benzene. Thus, we conclude that a full benzyl
cation would also be minimally hydrophobic and that some
hydrophobicity is probably lost in the partially cationic TS for
SN2 displacement on 4 and 8, but not on 9.
yes
a Reference 9. b Average interaction energy with the face of the benzene
ring. c A slightly positive interaction energy was calculated as a result of
forcing the two molecules to occupy the same solvent cavity. We consider
this to be zero interaction energy as the real molecules would have the
freedom to occupy separate cavities.
the aqueous rate by 30%.34 Apparently with 9 we are seeing
the increased hydrophobicity of phenoxide ion as it loses some
charge in the TS without seeing the compensating decrease in
hydrophobicity of a benzyl group as it develops some delocal-
ized charge in the TS.
O versus C Alkylation of Phenoxide Ions. In examining
the aqueous alkylation of phenoxides we found that in water
2,6-dimethylphenoxide 10 undergoes 64% para alkylation with
p-carboxybenzyl chloride 4, as well as 36% oxygen alkylation
(Figure 4). We saw no para alkylation in common organic
solvents (ethanol, DMSO, THF). No para alkylation was seen
in water (or other solvents) with unmethylated phenoxide 11
as the nucleophile.23 No para alkylation in water was observed
for the reactions of 10 with methyl iodide or 2-bromoacetamide
either, clearly demonstrating that the factors that control the
position of alkylation are not limited to the nucleophile.
A Computational Model. To address the contribution of
delocalized charge in aromatic rings to hydrophobic effects on
reactions, we developed an ab initio computational model using
Jaguar 3.035 to probe the relative hydrophobicity of aromatic
surfaces.13 In particular, we were interested in modeling how
delocalized charge affects the hydrophobicity of a benzene ring
in aqueous solution, and how cosolvents might interact with
substrates having, or lacking, delocalized charge.
In the model, a hydrophobic probe molecule, ethane, is
brought to the surface of various aromatic substrates in simulated
water solvent. The position of the ethane is optimized by energy
minimization, with the 6-31G** basis set and continuum solvent
SCRF, to determine the magnitude of interaction with the surface
at several local minima. The ethane molecule and the substrates
are first energy minimized individually, and all intramolecular
coordinates are constrained during minimization of the inter-
molecular interactions.
The calculated values are listed in Table 4. The obserVed
effects of 20% v/v ethanol in water on the solubility of the
compounds, in kcal/mol, are also in that table. These numbers
are smaller, since fully coordinated ethane in the computer
overestimates the effect of only occasionally coordinated ethanol
in the experiments. However, it is gratifying that the observed
solubility δ∆G°s for benzene, benzaldehyde, and N-methyl-
aniline run parallel to the calculated values for ethane coordina-
tion, each observed value being ca. 25% of that calculated for
full coordination of an ethane molecule to the substrate in water.
This finding gave us an important way to isolate the factors
involved in phenoxide ion benzylations. The starting materials
and the leaving groups are the same for oxygen and carbon
alkylation, so solvent and cosolvent effects reflect differences
in the transition states for the two competing processes.
When we ran the reactions with cosolvents, we observed a
difference in the rate perturbation by 20% ethanol and 20%
DMSO cosolvents on the para alkylation, which is slowed by
both cosolvents, from that of the oxygen alkylation, which is
unaffected by 20% ethanol and shows a modest rate increase
with 20% DMSO (Table 3). Our results in Table 3 suggest that
there is a concealed hydrophobic surface area at the TS for the
para alkylation reaction of 1 with 2, but no concealed surface
in the TS leading to the oxygen alkylation product. Hydrophobic
packing is driving the reaction toward para alkylation in water.
The calculations indicate that a phenoxide ion is only one-
third as hydrophobic as is a phenoxy radical (and phenol itself).
Because the poorly hydrophobic phenoxide ion will also
coordinate ethanol on its phenyl ring to a lesser extent than
does neutral phenol, these calculations suggest that almost all
(34) Mayer, M. U. Ph.D. Thesis, Columbia University, 2000.
(35) Jaguar 3.0; Schrodinger, Inc.: Portland, Oregon, 1997.
9
3628 J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002

Cosolvent Effects for Alkylation Reactions
A R T I C L E S
Figure 4. Alkylations of substituted phenoxide ions in water at oxygen, para, and ortho positions.
Our results are contrary to what we would expect only on
the basis of selective solvation of the phenoxide oxygen, which
was considered to be one of the dominant factors controlling
phenoxide alkylation.24,25,31 If oxygen solvation were the only
factor in aqueous solution, we would expect little or no effect
of cosolvents on the para alkylation and rate increases for the
oxygen alkylation as the cosolvents interfere with water
solvation of phenoxide ion. Yet we observe exactly the opposite;
the large effect of cosolvents is to slow the carbon alkylation,
while having a lesser effect on the oxygen alkylation.
Cation association has been implicated in driving phenoxide
alkylations toward carbon in nonpolar solvents,26,28 but consid-
ering that the reactions that we have examined were run in at
least 80% water and with the reactants completely in solution,
the effects of close association of the sodium cation with the
phenoxide oxygen do not apply to our results. We would expect
greater association of the cation with the oxygen in the solutions
with the nonpolar cosolvents, if anything, and hence a greater
amount of carbon alkylation with cosolvent; instead, we observe
that cosolvents lead to less carbon alkylation.
nucleophilic substitutions of benzyl halides in aqueous acetone
and benzyl sulfonium ions in water are less than first order in
nucleophile, and concluded that a borderline SN1-SN2 ion-pair
mechanism was operative. Thus, we determined whether we
were dealing with a pure SN2 reaction, an SN1 reaction, or a
borderline ion-pair mechanism.
We performed kinetic order experiments on the aqueous
reactions of phenoxide 11, p-nitrophenoxide 12, and 2,6-
dimethylphenoxide 10 (both oxygen and para carbon alkylations)
with 4. For each of the nucleophiles, formation of products and
formation of the competing hydrolysis product were monitored
over a range of concentrations. In each case, the rate of
alkylation of the phenoxides showed a linear relationship with
the concentration of phenoxide ion (first order), and, more
importantly, the rate of hydrolysis of the benzylic halide was
unaffected by the concentration of phenoxide ion.36 This last
result indicates that there is no competition between the
nucleophile and the water solvent for a free benzyl cation or a
borderline ion-pair reactive intermediate under our conditions,
25 °C in pure water, and that the reaction is truly SN2.
The effect of the SN1-SN2 character of our reactions was of
some concern, as changes in the mechanism can affect the
position of alkylation.29,31 If the degree of SN1-SN2 character
were to change in water relative to other solvents, a highly polar
solvent such as water would likely favor a more SN1-like
transition state and a subsequent preference to alkylate the
harder, more electronegative oxygen of the phenoxide ion. This
is the opposite of our observation that 10 prefers to alkylate at
the softer para position in water. However, Kornblum and Lurie
have reported that highly reactive benzyl cations show a lack
of selectivity toward the position of alkylation on phenoxide
ions, sometimes resulting in carbon alkylation.28 Additionally,
Sneen et al.37,38 have reported that the kinetics for some
Steric hindrance was a serious consideration for the inter-
pretation of our initial results. We found that ortho substitution
promoted para alkylation of phenoxide with a relatively large
electrophile such as 4, but not with small electrophiles such as
methyl iodide, a comparable result to that of Kornblum
mentioned previously.30 However, like selective solvation, steric
effects should allow carbon alkylation to compete only by
slowing oxygen alkylation, not speeding carbon alkylation. We
find that para alkylation is actually accelerated in water; with
the two methyl groups in 10, the overall rate of alkylation by 4
in water decreases by a factor of only 3 as compared with the
rate for simple phenoxide 11, but the proportion of that rate
that leads to carbon alkylation increases by at least 30-fold. Thus,
we conclude that there must be forces that actiVate para
alkylation in aqueous solution, not just suppress oxygen
alkylation. The only escape from this conclusion would be a
(36) Groves, K. Ph.D. Thesis, Columbia University, 2001.
(37) Sneen, R. A.; Larsen, J. W. J. Am. Chem. Soc. 1969, 91, 6031-6035.
(38) Sneen, R. A.; Felt, G. R.; Dickason, W. C. J. Am. Chem. Soc. 1973, 95,
638-639.
9
J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002 3629

A R T I C L E S
Breslow et al.
Table 6. Second-Order Rate Constants (M-1 min-1 × 10-3) and
20% Ethanolic Rate Perturbations of the Reactions of Various
Phenoxides with 4 in Watera
oxygen alkylation
ortho alkylation
para alkylation
kwater k20%EtOH
(M-1 min-1
kwater k20%EtOH
kwater k20%EtOH
(M-1 min-1
)
kwater
(M-1min-1
)
kwater
)
kwater
11
10
13
14
15
53
18
70
28
37
0.97
1.08
0.91
0.99
0.96
32
28
39
10
0.66
0.64
0.61
0.67
Figure 5. Oblique overlap in the para coupling of (a) 10 with 4 and (b) 13
with 4.
55
23
4.3
0.71
0.78
0.84
Table 5. Product Distributions for the Reactions of Alkyl
Substituted Phenoxides with 4 in Water at 25 °C
a Data are initial rate (error e 4%).
oxygen %
ortho %
para %
also show that something else is dominating the selectivity in
aqueous media.
11
10
13
14
15
100
36
46
30
75
64
18
45
16
36
25
9
We compared the aqueous reaction rates and cosolvent effects
of 20% ethanol for the alkylation of a series of alkyl substituted
phenoxides by 4. In addition to simple phenoxide 11 and 2,6-
dimethylphenoxide 10, 3,5-dimethylphenoxide 13 was chosen
to control any steric effects the 2,6-disubstitution might have
at the oxygen, and 2,6-diethylphenoxide 14 and 2,6-diisopro-
pylphenoxide 15 were chosen to examine the effect of the size
of the alkyl substituents on the rates of both carbon and oxygen
alkylations. All phenoxides gave some oxygen alkylation, while
10 also gave para alkylation, and 13, 14, and 15 also gave both
ortho and para alkylation, with the ortho alkylation of the 2,6-
disubstituted phenoxides resulting in cyclohexadienone products
(Figure 4). Authentic products for the reactions were synthesized
in water, as synthesis in any other solvent gave exclusively
oxygen alkylation, and characterized by 1H NMR. Table 5 shows
product distributions for each of the alkylation reactions in pure
water. Aqueous rate data and 20% ethanol rate perturbations
for the phenoxide series are shown in Table 6.
mechanism in which the two reactants form a complex in the
rate-determining step, and this complex then partitions between
oxygen and carbon alkylation subsequently. Our further evi-
dence, described below, shows that this is not the explanation
for an increased rate of carbon alkylation in water solution.
The nature of the phenoxide substituents was found to be
crucial for determining the amount of carbon alkylated products
in water. For instance, of five para monosubstituted phenoxides
that we examined, only p-cresol, with a methyl substituent, gave
any carbon alkylation with 4 in water, in this case 16% ortho
alkylation. The other para substituents - methoxyl, chloro,
cyano, and nitro - gave exclusively oxygen alkylation of the
phenols.
We have also found that hydrophobic methyl groups at any
position on the phenoxide ion induced carbon alkylation with
4 in water. We proposed that a stacked transition state leading
to carbon alkylation would allow methyl substituents to make
contact with the hydrophobic surface of 4, concealing hydro-
phobic surface area and resulting in the observed rate decrease
by 20% ethanol and a preference for carbon alkylation in water
(Figure 5).23 The oxygen will prefer to attack with its n non-π
electrons (not conjugated with the benzene ring) such that the
two phenyl rings of the reactants are perpendicular to one
another, as we have proposed above and previously.12 For carbon
alkylation, however, the two phenyl rings must be parallel,
which allows packing to occur. Such a transition state would
also explain why only hydrophobic alkyl substituents result in
carbon-alkylated products. We propose (Figure 5) that the
packing in the para alkylation is oblique, allowing overlap of a
methyl group of the phenol while minimizing overlap of the
phenoxide oxyanion.
It has long been believed that para alkylation of 2,6-
disubstituted phenoxide ions reflected the steric hindrance of
the substituents for attack at the oxygen atom and that para
alkylation occurred then by default. The clearest rebuttal of this
idea - at least for our benzylation reactions in aqueous media
- comes from our results with 3,5-dimethylphenoxide ion 13,
as shown in Table 5. In 13 there is no steric hindrance at oxygen;
if anything, there is steric hindrance at the ortho and para
carbons, and yet benzylation occurs at these carbons, while it
does not occur in unsubstituted phenoxide ion. Of course our
other findings on substituent effects, and on cosolvent effects,
The results of the kinetic studies for the alkylation of the
phenoxides by 4 revealed, not unexpectedly, that steric effects
of 2,6 substitution as well as basicity of the phenoxide affect
the rate of the reaction at oxygen. The rate of the oxygen
alkylation of 3,5-dimethylphenoxide 13 is about 30% faster than
that of simple phenoxide 11, reflecting an increase in nucleo-
philicity due to increased basicity (13 had an identical rate to
that of p-methoxyphenoxide, which has the same pKa). The rate
of oxygen alkylation of 10 is 66% slower than that of 11, even
though it is more basic, reflecting the steric effects of the ortho
methyl substituents. Oxygen alkylation of 14 and 15 was
somewhat faster than for 10, again reflecting increased basicity
and possibly decreased oxygen solvation caused by the alkyl
substituents, which apparently outweighs any slowing from the
increase in steric bulk around the oxygen relative to 10. All
oxygen alkylations showed almost no effect on the rate by 20%
v/v ethanol, just as we had seen for a simple phenoxide ion
previously.
The rates for alkylation at the para position of phenoxides
are more interesting, however. No para alkylation at all was
observed with simple phenoxide 11, but para alkylation was
observed for all of the alkyl substituted phenoxides. Figure 6
provides a graphical representation of the second-order rate
constants for the oxygen and para alkylations of each of the
phenoxides by 4. The stacked transition state allows for
favorable hydrophobic interaction of the benzyl group with the
phenoxide phenyl ring and alkyl substituents, accelerating the
reaction in water. The position of the alkyl groups does not
9
3630 J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002

Cosolvent Effects for Alkylation Reactions
A R T I C L E S
The para alkylation of 15 can proceed with hydrophobic
packing only if one of the isopropyl groups is rotated into a
less favorable orientation such that a methyl group lies more
or less in the plane of the aromatic ring (14°-17° out of plane,
whereupon it does not interfere with the packing). The energy
cost of rotating an isopropyl substituent to thus flatten one
surface of 15 was estimated to be about 1.1-1.4 kcal/mol by a
gas-phase MM2* calculation on 2,6-diisopropylfluorobenzene
using MacroModel.39 This number is roughly the same amount
by which the transition state for the alkylation of 2,6-diisopro-
pylphenoxide 15 is destabilized relative to 10, 13, and 14.
For all para alkylations, 20% ethanolic rate perturbations were
the same, resulting in a rate decrease by about one-third (Table
6). Such a rate decrease in 20% ethanol is expected if the
transition state involves hydrophobic packing of the starting
materials. We judge from these results that the amount of
hydrophobic overlap is roughly the same in each case. The data
also indicate that hydrophobic packing of the benzyl group
against an alkyl substituent (of the right size) on the phenoxide
is essential to obtain para alkylation. In the absence of alkyl
substituents, or in organic solvents where there is no hydro-
phobic effect, or with electrophiles with no significant hydro-
phobic surface, only oxygen alkylation is observed.
Figure 6. The second-order rate constants of the oxygen and para carbon
alkylation reactions of various phenoxide ions with 4 in water.
The rates of alkylation by 4 at the ortho position of the
phenoxides show a less clear trend (Table 6). Alkylation at the
ortho position was not observed for 10 or 11, although the rate
of ortho alkylation was almost as fast as oxygen alkylation for
13 and 14. Relatively slow ortho alkylation was observed in
the reaction of 15 with 4. The observed ortho alkylations showed
20% ethanolic rate decreases ranging between 20% and 30%.
The cosolvent rate decreases of the ortho alkylations surely
reflect the undoubted hydrophobic packing in the transition
states. However, the variability in the rates of ortho alkylation
across the series of phenoxides, perhaps as a result of varying
steric constraints of the substituents, makes these data more
difficult to interpret than those of the para alkylations.
Table 3 shows several contributing factors affecting the
overall rates in water and with added cosolvents. (1) Alkylations
of phenoxide ions at oxygen lead to an increase in hydrophobic-
ity, and thus a rate increase with cosolvents, which we ascribe
to partial neutralization of the delocalized negative charge in
the TS. (2) This effect is suppressed with p-nitrophenoxide ion.
(3) Cosolvents diminish the rate by interfering with solvation
of the leaving chloride ion, so this effect is lost with a sulfonium
ion leaving group. (4) In water the alkylations of 2,6-dimeth-
ylphenoxide ion by p-carboxybenzyl chloride or sulfonium ion
are somewhat faster at the para position than at the oxygen,
but the para alkylation is slowed by added ethanol and even
more by added DMSO. The rate of oxygen alkylation is either
slightly or significantly increased by the cosolvents. (5) DMSO
is more effective than is ethanol at solvating hydrophobic
surfaces. Ethanol has a larger effect than does DMSO in
decreasing the dielectric constant of the water medium, so the
larger cosolvent effects of DMSO in the table indicate that
hydrophobic effects are more important than are changes in the
dielectric constant of the medium.
Figure 7. (a) Proposed stacked transition state for the para alkylation of
2,6-dimethylphenoxide 10 with benzyl chloride. (b) Theoretical extended
transition state of the reaction of 2,6-diisopropylphenoxide 15 with benzyl
chloride, illustrating how isopropyl groups protrude above the face of the
phenoxide and interfere with a stacked transition state. The phenoxide ion
oxygens are represented by fluorine atoms.
matter since the phenoxide can simply rotate to maximize
hydrophobic overlap.
As shown in Figure 6, the rate of para alkylation was
relatively unaffected by the position of the alkyl groups, as 10
and 13 showed approximately the same rate of para alkylation.
Para alkylation of 14 was just a little faster than that of 10.
Most interestingly, para alkylation of 15 was significantly slower
than for 10 or 14, 4 times slower than 2,6-diethylphenoxide
14. From this result we conclude that the attack at the para
position has a transition state that is more sensitive to the steric
effects of the ortho isopropyl groups of 15 than is attack at the
nearby phenolic oxygen.
The steric hindrance from isopropyl groups for para alkylation
of 4 in water indicates that the transition state does indeed
involve packing of the aromatic ring of the benzyl group over
the ring and alkyl group of the phenoxide, with which the bulky
isopropyl groups interfere. The ethyl groups of 14 have the
opportunity to turn the outer methyl group away from the face
of the phenoxide involved in stacking, providing essentially the
same amount of hydrophobic overlap as with 10. Compound
14 therefore shows no rate disadvantage over 10. The isopropyl
groups of 15, however, which energetically prefer to remain
with their methyl groups almost perpendicular to the aryl ring
of the phenoxide, effectively block favorable packing of the
benzyl group and slow the para reaction. Figure 7 shows possible
transition states of the reactions of benzyl chloride with 2,6-
dimethylphenoxide 10 and 2,6-diisopropylphenoxide 15 and
illustrates how the isopropyl groups might interfere with the
packed transition state proposed for 10.
For O-alkylation, the hydrophobicity increase for phenoxide
ion and the decrease for the benzyl group at the TS apparently
roughly cancel each other. These effects should still be present
(39) Macromodel 7.0; Schrodinger, Inc.: Portland, Oregon, 2000.
9
J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002 3631

A R T I C L E S
Breslow et al.
Experimental Section
for C-alkylation, but that process has an extra factor - the
packing of the benzyl group onto the phenoxide ring with its
added methyl groups. This packing is the only sensible
interpretation of the various effects of phenyl substituents
discussed earlier. Thus, we conclude that the two processes -
oxygen and carbon alkylation - do indeed have the transition
states shown in Figure 2. This supports our previous conclusion
that oxygen alkylation of phenoxides uses their n electrons and
has no hydrophobic packing in the TS.
Kinetics. p-Carboxybenzyl chloride 4 was recrystallized from
ethanol/ether. p-Carboxybenzyl dimethylsulfonium chloride 8 was
recrystallized from acetonitrile/ether. p-Nitrophenol was dissolved in
methylene chloride (with 5% ethyl acetate to aid solubility), dried over
sodium sulfate, filtered, and crystallized by slow addition of hexanes.
All of the above reagents were dried under vacuum after purification.
All other reagents were purchased with the highest purity possible and
used without further purification.
Kinetics studies were performed on an HP1090 Series II liquid
chromatograph with a DR5 pumping system and a temperature-
controlled autosampler. Rainin Microsorb C-18 reverse phase analytical
columns (4.6 mm diameter × 150 mm length) with 5 mm particle size
were used with Rainin C-18 guard columns. Analytes were detected
with a diode-array UV-vis detector. All HPLC solvents were HPLC
grade purchased from Fisher Scientific Co. HPLC water was deionized
through a U.S. Filter type I mixed bed deionizer and filtered through
Osmonics or Whatman nylon membrane filters with 0.2 mm pore size.
Buffers were prepared using an Orion 701 A pH meter with an Orion
8103 Ross combination electrode. The electrode was calibrated against
standard buffer solutions (pH 4.0, 7.0, 10.0) obtained from Aldrich
Chemical Co.
Conclusions
The hydrophobic effect can be involved in every reaction in
water, or in water with cosolvents. In nonpolar reactions - such
as the Diels-Alder dimerization of cyclopentadiene in water
- the effect of cosolvents such as ethanol is entirely due to a
modification of the hydrophobic effect as the ethanol helps
solvate the hydrocarbon substrate. The magnitude of the
cosolvent effects, in free energy terms, is completely consistent
with the transition state (TS) for the process and the extent to
which hydrophobic surfaces become hidden from solvent in the
TS.
All second-order kinetics were monitored as initial rates. With the
exception of N-methylaniline 3, the nucleophile was used in excess.
Stock solutions of the nucleophile and electrophile were prepared fresh
with water freshly distilled under argon, and kept in argon-flushed
conical flasks with a rubber septum and a needle delivering positive
pressure of argon. Necessary base (LiOH) was added to nucleophile
solutions. Reagent solutions were sonicated and vortexed to ensure full
dissolution of reagents and homogeneity of the solutions. Additionally,
stock solutions of pure water, 60% v/v 200 proof ethanol, and 60%
v/v DMSO were prepared.
In reactions with polar reagents and in which there are polarity
changes in the TS, cosolvents also modify the ability of the
solvent to stabilize polar groups. However, even in such polar
reactions as SN2 displacements on benzylic halides by anilines
or phenoxides there are hydrophobic effects on the rates. Some
have to do with physically shielding hydrophobic surfaces in
the TS, but others have to do with changes in the hydrophobicity
of the reactants at the TS. For example, as nucleophilic
phenoxide ion attacks an alkylating agent there is less delocal-
ized negative charge into the phenyl ring, and it becomes more
hydrophobic. As nucleophilic attack occurs on a benzylic halide
there is some positive charge delocalized into the benzene ring,
with a decrease in hydrophobicity at the TS.
For each reaction, a mixing vial was flushed with argon, and 0.500
mL of each of the two reactants and 0.500 mL of either water, 60%
ethanol, or 60% DMSO were added with a pipet. The mixing vial was
sealed and vortexed for 45 s during which time an amber HPLC vial
was flushed with argon. The reaction solution was transferred to the
HPLC vial via syringe and pushed through an Acrodisc 0.2 mm syringe
filter. The sample was then thermostated to 25 ( 0.5 °C in the HPLC
sample compartment. The HPLC sequence was then started which takes
6-10 injections at fixed intervals. An appropriate method (solvent
system/gradient) was devised with mixtures of the reactants and
authentic reaction products.
We have elucidated the factors involved in the rate effects
of antihydrophobic cosolvents such as ethanol by the contrasting
effects of ethanol and DMSO, by replacing a leaving chloride
with a dimethylsulfonium group, by a study with p-nitrobenzyl
chloride, and by a computer model of the hydrophobic effect.
The computer model reveals that charge delocalization is
important in hydrophobicity changes for aromatic reactants.
From these studies, we have concluded that phenoxide ion as a
nucleophile uses its n electrons, not π electrons, and that it does
not stack on the benzyl electrophile in the TS. This contrasts
with aniline nucleophiles, in which a very different geometry
of the TS results from the use of π-like electrons. Additional
evidence supporting these conclusions comes from the study
of competing oxygen and carbon alkylation in phenoxide
nucleophiles.
Peaks were detected by UV-vis at wavelengths appropriate to the
reactants and products. Reactant and product HPLC peak areas were
integrated automatically or manually, as appropriate. Rate of formation
of product was used for computation of all rate constants. Starting
material concentrations were also monitored to ensure that the reaction
was still in its initial rate (<4% complete).
Each HPLC injection generated a text file that contained the time
of injection, and a table of peak retention times and integrated peak
areas, as well as other information. A series of output text files was
read using a short script written with Perl38 (Practical Extraction and
Report Language) that reads the time of injection and peak area for a
specified retention time and generates a text file that contains the relative
time of injection and peak area in a format readable by the Kaleidegraph
plotting program. Linear regressions were performed by Kaleidegraph,
and slopes (R > 0.992) were calibrated on the basis of synthesized,
characterized, and purified authentic products. Second-order rate
In the case of phenoxide nucleophiles, decreases in solvent-
exposed hydrophobic area - resulting from packing of alkyl
substituents against the aromatic ring of a benzylic electrophile
- influence the regiochemistry of the reaction, selecting for
carbon alkylation over oxygen alkylation in water. The use of
two cosolvents with different properties, and comparison of
various alkylating reagents, have allowed us to distinguish
hydrophobic effects from other effects on these reactions, and
to confirm our suggested transition states for phenoxide alky-
lations at oxygen and at carbon.
constants (M-1 min-1) were obtained from the calculated rate (M min-1
)
divided by the initial concentrations of the reactants. Reactions were
run in at least triplicate for each solvent system used, and all rate
constants were reproducible to within at most 5%, usually 3-4%.
For all reactions monitored by HPLC, each kinetic method was
calibrated with an authentic sample of the product as well as each of
9
3632 J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002

Cosolvent Effects for Alkylation Reactions
A R T I C L E S
the reactants. For a calibration, three stock solutions of a compound
were separately prepared at a concentration comparable to the
concentrations in the kinetic runs. Each sample was then injected into
the HPLC three times and eluted with the same method used for the
kinetics. Obtained peak areas were divided by the concentration of stock
solutions, and all values were averaged to obtain a correction factor in
units of M/area.
different aromatic molecules. In this way, the initial orientations of
the ethane to all model aromatic solutes were identical. Several
geometries for interaction of ethane with the face of the benzene ring
as well as parallel and perpendicular orientations of the ethane on the
edge of the benzene ring were used. All other coordinates (i.e., all
intramolecular coordinates) were constrained for the calculations of
intermolecular interaction energies.
Solubilities. The solubilities of benzene and benzyl phenyl ether in
water, 20% ethanol, and 20% DMSO were measured by HPLC. Three
2.0 mL samples of each solvent mixture were equilibrated with 50 mL
(0.56 mmol) of benzene or 2.0 mg (0.011 mmol) of benzyl phenyl
ether at 25 ( 0.2 °C for 24 h with slow stirring. The samples were
then allowed to stand without stirring for an additional 24 h before
being sampled with a needle. For benzene, 40.0 mL of solvent was
drawn from underneath the benzene layer and diluted to 2.00 mL with
ethanol and injected onto the HPLC three times. For benzyl phenyl
ether, 0.500 mL of sample was drawn and injected directly into the
HPLC three times. In each case, the HPLC peak areas were calibrated
against three samples of the authentic compounds at known concentra-
tions comparable to injected concentrations in pure ethanol. Benzene
solubility in pure water was measured at 20.9 ( 1.0 mM as compared
to the IUPAC value of 20.7 mM, confirming the accuracy of the
measurements.
Computations. All ab initio calculations were performed with the
Jaguar 2.5 or Jaguar 3.0 software packages35 using the 6-31G** basis
set. For calculations with water solvent, continuum solvent SCRF was
used. Energies are calculated in hartrees, then converted to kcal/mol,
where 1 hartree ) 627.5 kcal/mol. Input structures were created with
Macromodel39 and minimized on the mm2* force field in the gas phase
prior to export to Jaguar. The Macromodel files were converted to xyz
and Jaguar z-matrix files with Babel.41Xmol42was used to view xyz input
and output files and to measure coordinates of input structures in xyz
format. Output files were searched and parsed with the aid of several
scripts written in Perl40 specifically for this series of calculations.
An ab initio computational method for estimating the effects of
delocalized charge on the hydrophobicity of various aromatic surfaces
was developed. For the calculations a molecule of ethane was used as
a hydrophobic probe molecule, and it was brought into van der Waals
contact with the surface of various aromatic substrates. Individual
molecules (ethane and aromatic substrates) were first constructed with
Macromodel and minimized on the MM2* force field in the gas phase.
The coordinates were then converted to Jaguar format and minimized
in water solution. The minimized coordinates for the aromatic substrates
were then converted to a z-matrix text format. The z-matrix coordinates
for ethane were then manually appended to the z-matrix coordinates
of the aromatic solute. Six new z-matrix coordinates were then input
to specify the relative orientation of the ethane molecule to the aromatic
ring. The first coordinate is the distance from one of the ethane carbons
to the carbon in the four position (arbitrarily chosen) of the monosub-
stituted benzene ring. The next two are the angle from the ethane carbon
to the four and three carbons and the torsional (dihedral) angle from
the ethane carbon to the four, three, and two carbons of the aromatic
ring. The next two coordinates specify the angle and torsional angle
from one of the hydrogens on the chosen ethane carbon through the
ethane carbon to the four and three carbons of the aromatic ring. Finally,
the last coordinate is the torsional angle from a second hydrogen on
the ethane carbon, to the first hydrogen then to the ethane carbon and
the four position of the aromatic ring.
Output files were searched for the geometry with the lowest solution
phase energy. Energy surfaces were generally rather flat, so geometric
convergence was rare. Energetic convergence was obtained with good
accuracy, however. To find the interaction energy, the sum of the
calculated solution phase energies of the separate model aromatic solute
and ethane was subtracted from the total solution phase energy
calculated for the ethane and the model aromatic solute placed together.
For the neutral aromatic molecules, including benzene, N-methylaniline,
anilinium cation, benzaldehyde, 1,3-dinitrobenzene, and thiophenoxy
and phenoxy radicals, the interactions on the face and edge were
equivalent. For these molecules, results of all optimizations (corre-
sponding to different input geometries) were averaged. For phenoxide
and thiophenoxide anions and N-methylanilinium radical cation,
however, the edge and face interactions were different and were
therefore treated separately. The results are listed in Table 4.
Syntheses. 2,6-Diethylphenol.43 Three grams (20 mmol) of 2,6-
diethylaniline was dissolved in 60 mL of water and 40 mL of
concentrated sulfuric acid in a 1 L round-bottom flask with a stir bar.
Twenty-five grams of ice was added, and the solution was immersed
in an ice bath. One and one-half grams (22 mmol) of sodium nitrite
was dissolved in 40 mL of water with 10 g of ice, and the solution
was transferred to an addition funnel attached to the 1 L reaction flask
with the 2,6-diethylaniline solution. The sodium nitrite solution was
added over about 15 s with vigorous stirring (note: the sodium nitrite
solution can be added quickly only if the solutions are fairly dilute, as
considerable heat is released and colored byproducts are formed if the
solutions are more concentrated). The solution turned light orange in
color, and bubbles of nitrogen were observed. The solution was
transferred to a 1 L flask with 50 mL of 50% sulfuric acid that was
heated to steaming (∼50 °C), and the combined solution was heated
to 90-100 °C for 15 min. The solution was then transferred to a 2 L
separatory funnel, and the product was extracted with three 150 mL
portions of methylene chloride. The methylene chloride was dried over
sodium sulfate and evaporated. The resulting oil was placed in a
sublimation apparatus with a 0° coldfinger at 0.25 mmHg and was
heated at 40-50 °C. The product was collected as a white, crystalline
solid on the coldfinger. A total of 1.6 g (53%) of product was collected
from several distillations in the sublimation apparatus. 1H NMR (CD2-
Cl2, 300 MHz): d 1.24 (t, J ) 7.6 Hz, 6H), 2.64 (q, J ) 7.6 Hz, 4H),
6.83 (t, J ) 7.5 Hz, 1H), 7.01 (d, J ) 7.5 Hz, 2H).
p-Carboxybenzyldimethylsulfonium Chloride 8. p-Carboxybenzyl
chloride 4 (1.7 g, 10 mmol) was dissolved in 20 mL (340 mmol) of
dimethyl sulfide and 4 mL of ethanol in a 100 mL round-bottom flask
with a stir bar and reflux condenser. The solution was stirred and
immersed in an oil bath at 55-60 °C and refluxed for 7-10 days during
which time a white precipitate formed. The solvent was evaporated,
and the resulting solid was dissolved in 50 mL of hot ethanol. The
ethanol was poured into a 500 mL beaker containing 250 mL of ethyl
acetate resulting in the formation of bright, white crystals. The crystals
were filtered and washed with three 25 mL portions of ethyl acetate
and two 20 mL portions of hexanes and dried under vacuum yielding
2.0 g (86%) of pure product. 1H NMR (DMSO-d6, 300 MHz): d 2.89
(s, 6H), 4.90 (s, 2H), 7.61 (d, J ) 8.3 Hz, 2H), 8.01 (d, J ) 8.3 Hz,
2H), 13.2 (bs, 1H).
The coordinates were obtained by maneuvering the molecules
together to the desired relative orientation with Macromodel, converting
the Macromodel file to xyz format and measuring the desired coordinates
with Xmol. Once a set of input geometries was created, the orientation
coordinates were cut and pasted into other input files for each of the
2,6-Dimethylphenoxide (10) Oxygen and Para Alkylation Prod-
ucts with 4. p-Carboxybenzyl chloride 4 (514 mg, 3.0 mmol) and 1.5
(40) Wall, L. Perl, 5th ed.; O’Reilly: Sebastopol, CA, 1994.
(41) Walters, P.; Stahl, M. Babel; University of Arizona: Tucson, Arizona, 1994.
(42) Xmol; Research Equipment, Inc., Minnesota Supercomputer Center, 1993.
(43) LeNoble, W. J. L.; Hayakawa, T.; Sen, A. K.; Tatsukami, Y. J. Org. Chem.
1971, 36, 193-196.
9
J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002 3633

A R T I C L E S
Breslow et al.
g (12 mmol) of 2,6-dimethylphenol were dissolved in 20 mL of absolute
ethanol. Lithium hydroxide monohydrate (650 mg, 15 mmol) was
dissolved in 80 mL of water freshly distilled under argon. The solutions
were combined in a 250 mL round-bottom flask under argon and stirred
at room temperature for 24 h. The solution was neutralized with 100
mL of 150 mM pH 7.2 phosphate buffer and extracted with three 100
mL portions of ethyl acetate. The ethyl acetate layers were combined,
dried over sodium sulfate, and evaporated to an oily residue. Fifty
milliliters of hexanes was added, and a white precipitate formed. The
precipitate was filtered, washed with hexanes, and dried under vacuum
to give 340 mg (44%) of a mixture of the oxygen and para alkylation
products. The original aqueous layer was acidified to pH 3 with 1 M
HCl and was again extracted with ethyl acetate as above yielding more
product as well as starting material and hydrolysis product. The first
extract containing only the two products was purified by reverse phase
column chromatography eluting with 75:25:1 methanol:water:acetic
acid.
Oxygen Alkylation Product 16. mp ) 153-155 °C. 1H NMR
(DMSO-d6, 300 MHz): δ 2.23 (s, 6H), 4.87 (s, 2H), 6.95 (t, J ) 7.4
Hz, 1H), 7.05 (d, J ) 7.4 Hz, 2H), 7.61 (d, J ) 8.2 Hz, 2H), 7.98 (d,
J ) 8.2 Hz, 2H), 12.95 (bs, 1H).
Para Carbon Alkylation Product 17. mp ) 189-191 °C. 1H NMR
(DMSO-d6, 300 MHz): δ 2.09 (s, 6H), 3.81 (s, 2H), 6.75 (s, 2H), 7.29
(d, J ) 8.3 Hz, 2H), 7.83 (d, J ) 8.3 Hz, 2H), 8.02 (s, 1H), 12.71 (bs,
1H).
(13 mmol) of 2,6-diethylphenol, and 0.56 g (13 mmol) of lithium
hydroxide monohydrate were dissolved in 150 mL of water freshly
distilled under argon in a 250 mL round-bottom flask with a stir bar.
The solution was warmed to 90 °C for 2 h. The solution was extracted
with two 100 mL portions of methylene chloride while still basic to
remove unreacted phenol. The aqueous solution was then neutralized
with 150 mL of 140 mM pH 7.2 phosphate buffer and extracted with
three 100 mL portions of ethyl acetate. The ethyl acetate layers were
combined, dried over sodium sulfate, and evaporated to a residue
containing a mixture of the three products and some phenol. The residue
was dissolved in 25 mL of methylene chloride and poured into 100
mL of hexanes upon which a white precipitate formed. The precipitated
was filtered and washed with hexanes to yield 275 mg (32%) of a
mixture of the three products (the rest of the products remained with
the starting material and some hydrolysis product in the aqueous layer).
The 275 mg (0.97 mmol) of product mixture was dissolved in 40 mL
of freshly distilled methylene chloride in an oven-dried 250 mL round-
bottom flask under argon. One gram (9.0 mmol) of N-hydroxysuccin-
imide and 575 mg (3 mmol) of 1-(3-dimethylaminopropyl)-3-
ethylcarbodiimide hydrochloride (EDC) and the previous product were
stirred at room temperature for 4 h. The resulting hydroxysuccinimide
esters were purified by silica gel chromatography eluting with methylene
chloride with 2% ethyl acetate. The pure fractions were evaporated
and then dissolved in 25 mL of 1 M sodium hydroxide, acidified with
concentrated HCl, extracted with ethyl acetate, dried over sodium
sulfate, and evaporated. The ether product and the para product were
obtained pure by this procedure, but the ortho cyclohexadienone product
had some impurity, presumably a decomposition product due the
photosensitivity of cyclohexadienones. Final purification of the dienone
was achieved by reverse phase column chromatography eluting with a
gradient from 60:40 to 75:25 methanol:water with 1% acetic acid. The
ortho product must be kept in the dark.
3,5-Dimethylphenoxide (13) Oxygen, Ortho and Para Alkylation
Products with 4. p-Carboxybenzyl chloride 4 (1.7 g, 10 mmol) and
6.25 g (50 mmol) of 3,5-dimethylphenol were dissolved in 10 mL of
absolute ethanol in a 250 mL round-bottom flask with a stir bar under
argon. Lithium hydroxide monohydrate (2.6 g, 62 mmol) was dissolved
in 100 mL of water freshly distilled under argon and added to the flask.
The solution was stirred at room temperature for 24 h. About 3 mL of
concentrated HCl was added to bring the solution to a pH of about
8-9. One-hundred milliliters of 140 mM pH 7.2 phosphate buffer was
then added, and the solution was extracted with one 100 mL and three
50 mL portions of methylene chloride followed by one 100 mL and
three 50 mL portions of ethyl acetate. The aqueous layer was then
acidified with concentrated HCl to pH 2 and extracted again with ethyl
acetate as above. Each set of extracts was combined, dried over sodium
sulfate, and evaporated. The residue from the methylene chloride
extracts, containing unreacted 3,5-dimethylphenol and the oxygen
alkylation product, was triturated in 60 mL of hot hexanes, cooled,
and the solid filtered resulting in about 400 mg (16%) of pure oxygen
alkylation product. The residue from the first (neutral) ethyl acetate
extract contained a mixture of all three products. Crude yield was high
(>75%); however, separation of the ortho and para alkylation products
was difficult, but small amounts of the pure products were finally
obtained by reverse phase (RP-18) preparative HPLC (Dynamax 21.4
mm i.d., 250 mm L, 100 Å pore size, 5 mm particle size) eluting at 6
mL/min with a continuous gradient from 50:50 to 90:10 methanol/pH
3 25 mM phosphate buffer. From 150 mg of the mixture, about 35 mg
(23%) of the ortho product and 16 mg (11%) of the para product were
obtained pure off of the column.
1
Oxygen Alkylation Product 21. H NMR (CDCl3, 300 MHz): δ
1.25 (t, J ) 7.6 Hz, 6H), 2.70 (q, J ) 7.6 Hz, 4H), 4.92 (s, 2H), 7.10
(m, 3H), 7.61 (d, J ) 8.3 Hz, 2H), 8.17 (d, J ) 8.3 Hz, 2H).
Para Carbon Alkylation Product 22. 1H NMR (CDCl3, 300
MHz): δ 1.22 (t, J ) 7.6 Hz, 6H), 2.59 (q, J ) 7.6 Hz, 4H), 3.95 (s,
2H), 6.81 (s, 2H), 7.26 (d, J ) 8.2 Hz, 2H), 8.02 (d, J ) 8.2 Hz, 2H).
Ortho Carbon Alkylation Product 23. 1H NMR (CDCl3, 300
MHz): δ 0.68 (t, J ) 7.5 Hz, 3H), 0.94 (t, J ) 7.5 Hz, 3H), 1.58 (dq,
J ) 5.6 Hz, 7.5 Hz, 1H), 2.09 (dq, J ) 5.6 Hz, 7.5 Hz, 1H), 2.20 (q,
J ) 7.5 Hz, 2H), 2.75 (d, J ) 12.7 Hz, 1H), 3.25 (d, J ) 12.7 Hz,
1H), 6.20 (m, 2H), 6.56 (m, 1H), 7.10 (d, J ) 8.3 Hz, 2H), 7.85 (d,
J ) 8.3 Hz, 2H).
2,6-Diisopropylphenoxide (15) Oxygen, Ortho and Para Alky-
lation Products with 4. p-Carboxybenzyl chloride (614 mg, 3.6 mmol),
2.0 mL (13 mmol) of 2,6-diisopropylphenol, and 606 mg (15 mmol)
of lithium hydroxide monohydrate were dissolved in 150 mL of water
freshly distilled under argon in a 250 mL round-bottom flask with a
stir bar. The solution was warmed to 90 °C for 8 h. The solution was
extracted with two 50 mL portions of ethyl acetate while still basic to
remove unreacted 2,6-diisopropylphenol. The aqueous solution was then
neutralized with 150 mL of 140 mM pH 7.2 phosphate buffer and
extracted with three 50 mL portions of ethyl acetate. The neutral ethyl
acetate extracts were combined, dried over sodium sulfate, and
evaporated to give about 300 mg (27%) of a yellowish solid containing
a mixture of the three products. The 275 mg (0.97 mmol) of product
mixture was dissolved in 40 mL of freshly distilled methylene chloride
in an oven-dried 250 mL round-bottom flask under argon. One gram
(9.0 mmol) of N-hydroxysuccinimide and 500 mg (2.6 mmol) of 1-(3-
dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (EDC) were
added, and the solution was stirred at room temperature for 4 h. The
resulting hydroxysuccinimide esters were purified by silica gel chro-
matography eluting with methylene chloride. The pure fractions were
evaporated and then dissolved in 25 mL of 1 M sodium hydroxide,
1
Oxygen Alkylation Product 18. H NMR (DMSO-d6, 400 MHz):
δ 2.21 (s, 6H), 5.13 (s, 2H), 6.58 (s, 1H), 6.63 (s, 2H), 7.53 (d, J )
8.3, 2H), 7.94 (d, J ) 8.3 Hz, 2H), 12.93 (bs, 1H).
1
Para Carbon Alkylation Product 19. H NMR (DMSO-d6, 400
MHz): δ 2.07 (s, 6H), 3.94 (s, 2H), 6.46 (s, 2H), 7.08 (d, J ) 8.3 Hz,
2H), 7.81 (d, J ) 8.3 Hz, 2H), 9.05 (s, 1H), 12.7 (bs, 1H).
1
Ortho Carbon Alkylation Product 20. H NMR (DMSO-d6, 400
MHz): δ 2.07 (s, 3H), 2.15 (s, 3H), 3.94 (s, 2H), 6.44 (s, 1H), 6.52 (s,
2H), 7.20 (d, J ) 8.1 Hz, 2H), 7.79 (d, J ) 8.1 Hz, 2H), 9.22 (s, 1H),
12.70 (bs, 1H).
2,6-Diethylphenoxide (14) Oxygen, Ortho and Para Alkylation
Products with 4. p-Carboxybenzyl chloride 4 (0.50 g, 3 mmol), 2.0 g
9
3634 J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002

Cosolvent Effects for Alkylation Reactions
A R T I C L E S
acidified with concentrated HCl, extracted with ethyl acetate, dried over
sodium sulfate, and evaporated. Ninety-five milligrams (32%) of the
ether product and 37 mg (12%) of the para product slightly impure
with the ether were obtained. The dienone was not isolable by this
method, but could be positively identified by its crude 1H NMR
spectrum.
Ortho Carbon Alkylation Product 26. 1H NMR (CD3CN, 300
MHz): δ 0.69 (d, J ) 6.8 Hz, 3H), 0.75 (d, J ) 6.8 Hz, 3H), 0.93 (d,
J ) 6.8 Hz, 3H), 1.04 (d, J ) 6.8 Hz, 3H), 2.20 (m, 1H), 2.75 (m,
1H), 3.02 (d, J ) 12.4 Hz, 1H), 3.15 (d, J ) 12.4 Hz, 1H), 6.25 (dd,
J ) 6.2 Hz, 9.6 Hz, 1H), 6.40 (d, J ) 9.6 Hz, 1H), 6.49 (d, J ) 6.2
Hz, 1H), 7.05 (d, J ) 8.1 Hz, 2H), 7.72 (d, J ) 8.1 Hz, 2H).
1
Oxygen Alkylation Product 24. H NMR (DMSO-d6, 300 MHz):
δ 1.16 (s, 6H), 1.18 (s, 6H), 3.27 (m, 2H), 4.85 (s, 2H), 7.13 (m, 3H),
7.60 (d, J ) 8.3 Hz, 2H), 8.00 (d, J ) 8.3 Hz, 2H), 12.9 (bs, 1H).
Para Carbon Alkylation Product 25. 1H NMR (CD3CN, 300
MHz): δ 1.15 (d, J ) 6.8 Hz, 12H), 3.17 (m, 2H), 3.91 (s, 2H), 5.85
(bs, 1H), 6.91 (s, 2H), 7.32 (d, J ) 8.1 Hz, 2H), 7.89 (d, J ) 8.1 Hz,
2H).
Acknowledgment. We thank Professor Richard Friesner for
making his computer programs and expertise available. Support
of this work by the NIH, the NSF, and the EPA is gratefully
acknowledged.
JA012293H
9
J. AM. CHEM. SOC. VOL. 124, NO. 14, 2002 3635