Temperature dependence of regioselectivity in nucleophilic photosubstitution of 4-nitroanisole. The activation energy criterion for regioselectivity

Article abstract of DOI:10.1021/jo1015882

Photosubstitution of the nitro group vs the methoxy group of triplet 4-nitroanisole by hydroxide ion in water leads to product yields of about 80% 4-methoxyphenol and 20% 4-nitrophenol. The ratio depends slightly on temperature from 3 to 73 °C. The slight temperature variation in the yield ratio is reproduced almost perfectly with a simple Arrhenius model for a mechanism involving bonding of hydroxide ion with the triplet state of 4-nitroanisole. The competing transition states have activation energies of 2.2 and 2.6 kcal/mol, respectively. Correct prediction of regioselectivity can be done for this case by quantum chemical calculation of the competing triplet transition-state energies, or those of the corresponding triplet σ-complexes. Other models for aromatic photosubstitution regioselectivity in mechanisms of the S _N2Ar* type, such as those based on calculated electron densities, HOMO/LUMO coefficients, or energy gap sizes, are discussed and shown to be inferior to the relative activation energies model. The photoreaction in alcohol solvents, claimed by others to generate the same products as in water and to have an exceedingly large variation of the product ratio with temperature, may reflect chemical changes other than those reported.

Full text of DOI:10.1021/jo1015882

pubs.acs.org/joc
Temperature Dependence of Regioselectivity in Nucleophilic
Photosubstitution of 4-Nitroanisole. The Activation Energy Criterion
for Regioselectivity
Gene G. Wubbels,* Hanan Danial, and Danielle Policarpio
Department of Chemistry, University of Nebraska at Kearney, Kearney, Nebraska 68847, United States
wubbelsg@unk.edu
Received August 12, 2010
Photosubstitution of the nitro group vs the methoxy group of triplet 4-nitroanisole by hydroxide ion
in water leads to product yields of about 80% 4-methoxyphenol and 20% 4-nitrophenol. The ratio
depends slightly on temperature from 3 to 73 ꢀC. The slight temperature variation in the yield ratio is
reproduced almost perfectly with a simple Arrhenius model for a mechanism involving bonding of
hydroxide ion with the triplet state of 4-nitroanisole. The competing transition states have activation
energies of 2.2 and 2.6 kcal/mol, respectively. Correct prediction of regioselectivity can be done for
this case by quantum chemical calculation of the competing triplet transition-state energies, or those
of the corresponding triplet σ-complexes. Other models for aromatic photosubstitution regioselec-
tivity in mechanisms of the SN2Ar* type, such as those based on calculated electron densities,
HOMO/LUMO coefficients, or energy gap sizes, are discussed and shown to be inferior to the
relative activation energies model. The photoreaction in alcohol solvents, claimed by others to
generate the same products as in water and to have an exceedingly large variation of the product ratio
with temperature, may reflect chemical changes other than those reported.
Introduction
reaction mechanisms.2-5 It seemed odd that, despite the
long history of attempts to explain the regioselectivity of
nucleophilic aromatic photosubstitution reactions,6-13 no
one to our knowledge had examined the utility of relative
When it became clear that photochemical reactions in
condensed phases occurred typically from thermally relaxed,
electronically excited molecules, pioneering photochemists
suggested that the temperature dependence of reaction rate
constants from excited states would be governed by the
Arrhenius equation.1 Many studies have shown the useful-
ness of this model in the investigation of photochemical
(5) Gonzales, C.; Pincock, J. Photochem. Photobiol. 2006, 82, 301.
(6) Zimmerman, H. E.; Somasekhara, S. J. Am. Chem. Soc. 1963, 85, 922.
(7) Cornelisse, J.; Havinga, E. Chem. Rev. 1975, 75, 353.
(8) Epiotis, N. D.; Shaik, S. J. Am. Chem. Soc. 1978, 100, 29.
(9) van Reil, H. C. H. A.; Lodder, G.; Havinga, E. J. Am. Chem. Soc.
1981, 103, 7257.
(10) Mutai, K.; Nakagaki, R.; Tukada, H. Bull. Chem. Soc. Jpn. 1985, 58,
2066.
(1) Franck, J.; Rabinowitsch, E. Trans. Faraday Soc. 1934, 30, 120.
(2) Fischer, E. J. Am. Chem. Soc. 1960, 82, 3249.
~
ꢀ
(11) Cantos, A.; Marquet, J.; Moreno-Manas, M; Gonzalez-Lafont, A.;
ꢀ
(3) Encina, M. V.; Lissi, E. A.; Lemp, E.; Zanocco, A.; Scaiano, J. C.
J. Am. Chem. Soc. 1983, 105, 1856.
(4) Brozic, J. A.; Crosby, G. A. Coord. Chem. Rev. 2005, 249, 1310.
Lluch, J. M.; Bertran, J. J. Org. Chem. 1990, 55, 3303.
(12) Manickam, M. C. D.; Pitchumani, K.; Srinivasan, C. Proc. Indian
Acad. Sci.(Chem. Sci.) 2003, 115, 273.
7726 J. Org. Chem. 2010, 75, 7726–7733
Published on Web 10/18/2010
DOI: 10.1021/jo1015882
r
2010 American Chemical Society

Wubbels et al.
JOCArticle
activation energies. Although it was not aimed directly at this
question, a recent study14 of the temperature dependence of
regioselectivity of photosubstitution on 4-nitroanisole by
hydroxide ion, which claimed a large dependence on tem-
perature of relative product yields, aroused our interest. A
subsequent report15 added a claim of a large temperature
dependence of photosubstitution regiochemistry for the
same photoreaction.
would be expected to generate photoproducts. We now
report that the temperature dependence of photosubstitution
regioselectivity in pure water is modest and that alkoxide ion
products are major photoproducts in alcohol solutions.
Results
Photolyses of 4-nitroanisole (2.0 ꢀ 10-4 M) in air-
saturated 5% acetonitrile/water containing 0.010 M NaOH
were carried out initially at 313 nm (Scheme 1). We found by
UV-vis monitoring of the reactions that the absorption of the
4-nitrophenoxide ion (PNPh) product (λmax 400 nm, ε 18,300)
was easily identifiable, but that of the 4-methoxyphenoxide
ion (PMPh) product (λmax 306 nm, ε 2920) was obscured by
absorptions of unknown origin centered at 350 nm.18 This
result had been noticed by the original investigators of this
reaction,17 who attributed the unknown absorptions to
degradation of the PMPh product. They found that the
photoreaction gave the two products shown in Scheme 1
quite cleanly if air was excluded. We confirmed this result,
and we found that use of a uranium glass filter (λ > 330 nm)
further improved the cleanness of the reaction.
Overlaid UV-vis spectra (225-500 nm) for the photolysis
at ∼25 ꢀC of 4-nitroanisole (5.4 ꢀ 10-5 M) in argon-purged
pure water solution with 0.10 M NaOH through the uranium
filter are shown in Figure 1. The isosbestic points at 365 and
262 nm indicate a single-stage, simple reaction, after which
the absorptions of the 4-nitrophenoxide ion at 400 nm and
the 4-methoxyphenoxide ion at 306 nm are clearly visible.
These absorptions allowed quantitative monitoring of the
two products. The isosbestic point at 365 nm was sharp
under all temperature conditions. The isosbestic point at
262 nm was sharp at 3 ꢀC, but slightly ragged at 37 ꢀC and at
73 ꢀC. When samples were irradiated exhaustively at 3 ꢀC,
the UV-vis absorptions of the two products in three runs
Our investigation16 of the element effect of halogen leav-
ing groups in nucleophilic aromatic photosubstitution of
2-halo-4-nitroanisoles had shown normal, fast, elementary
rate constants for the photosubstitutions of the nitro and
methoxy groups of triplet 4-nitroanisole by hydroxide ion in
aqueous solution. These findings for 4-nitroanisole extended
and agreed with those of the original investigation of these
photoreactions.17 That study found that hydroxide ion in
aqueous media displaced the nitro group and the methoxy
group of triplet 4-nitroanisole in a ratio of about 4:1 at 26 and
3 ꢀC. That photolysis of 4-nitroanisole in solutions of sodium
hydroxide in alcohol solvents was reported14 to give product
ratios varying from 99:1 at -20 ꢀC to 0.6:1 at 196 ꢀC was very
surprising. That the photoreactions were reported to give rise
exclusively to hydroxide ion photodisplacement products
was also unexpected. Prototropy would cause such solutions
to contain little hydroxide ion and much alkoxide ion that
SCHEME 1. Photosubstitution Reactions of Hydroxide Ion
with 4-Nitroanisole
FIGURE 1. Electronic spectra for photolysis of 4-nitroanisole (5.4 ꢀ 10-5 M) in aqueous 0.10 M NaOH after irradiation for 0, 2, 4, 8, 16, and
24 min at 25 ꢀC.
J. Org. Chem. Vol. 75, No. 22, 2010 7727

Wubbels et al.
JOCArticle
TABLE 1. Photolyses of 4-Nitroanisole (5.4 ꢀ 10-5 M) in Water Containing 0.020 M NaOH
trial
T, ꢀC
A¥, 400 nm
PNPh, M ( ꢀ 105)
% PNPh
mean % PNPh
mean % PMPh
1a
1b
2a
2b
3a
3b
3
0.190
0.189
0.203
0.204
0.214
0.215
1.04
1.03
1.11
1.12
1.17
1.18
19.1
18.9
20.4
20.6
22.1
21.5
19.0
81.0
3
37
37
73
73
20.5
21.8
79.5
78.2
TABLE 2. Arrhenius Theory Calculations for Nitro vs Methoxy Sub-
stitution by Hydroxide Ion in Triplet 4-Nitroanisole with Activation
Energies of 2.17 and 2.58 kcal/mol, Respectively
temperature, ꢀC knitro ꢀ 10-7 0.5kmethoxy ꢀ 10-7 % PNPh % PMPh
accounted for 99.6, 99.4, and 97.2% of the moles of 4-
nitroanisole. This indicates a quantitatively clean reaction giving
two products. At the higher temperatures, the two product
absorptions accounted for about 95% of the reactant, owing
to some slight thermal or photochemical degradation of the
4-methoxyphenoxide product, as reported previously.17 At
all temperatures, therefore, we used the clearly quantifiable
amount of 4-nitrophenoxide as the measure of methoxide ion
displacement, and the difference from the total molarity of
4-nitroanisole as the quantitative measure of nitrite ion
displacement, giving 4-methoxyphenoxide ion. The results
of exhaustive irradiations at three temperatures for the two
product yields determined in this way are shown in Table 1.
A control experiment showed that PNPh, a possible thermal
hydrolysis product, was not formed detectably by UV-vis
monitoring when 4-nitroanisole was heated in the dark in
aqueous 0.020 M NaOH at 80 ꢀC for a time period equal to
that of the photochemical reaction.
3
37
73
5.74
8.85
12.8
1.36
2.27
3.51
19.1
20.4
21.6
80.9
79.6
78.4
effects, nucleofugality is related to acidities of the conjugate
acids,20 which are very similar.21 The actual rate constant for
attack at the methoxy-bearing carbon, therefore, should be a
factor of 2 greater than that inferred from the yield of PNPh.
That is, the total rate constant for loss of starting material
(9.6 ꢀ 107 M-1 s-1) is equal to knitro þ 0.5kmethoxy. We also
know from the observed product yields of 80% and 20% that
knitro/0.5kmethoxy is equal to 4. The solution of the simulta-
neous equations yields the nucleophilic attack rate constants
at 25 ꢀC of 7.68 ꢀ 107 M-1 s-1 for attack at the nitro-bearing
carbon (knitro), and 3.84 ꢀ 107 M-1 s-1 for attack at the
methoxy-bearing carbon (kmethoxy). We assume that the A
factor in the Arrhenius equation (ln k = -(Ea/RT) þ ln A) is
the fastest intermolecular reaction in the medium at 25 ꢀC,
which is approximated by the diffusion rate constant. For
reactants in water with no ionic attraction or repulsion, that
It is clear that temperature increases cause a modest
increase in the yield of the methoxy displacement product,
PNPh, and a corresponding decrease in the amount of the
nitro displacement product, PMPh.
If the two reactions represent a simple competition be-
tween two pathways from the triplet having different activa-
tion energies, it should be possible to reproduce the changes
in product yields as a function of temperature with Arrhenius
theory. The rate constants were obtained from results we
reported for triplet 4-nitroanisole and hydroxide ion at
approximately 25 ꢀC in 33% acetonitrile/water.16 The sum
of the two reaction rate constants from the triplet state of
4-nitroanisole is 9.6 ꢀ 107 M-1 s-1 for displacement of the
nitro group and displacement of the methoxy group.16 The
total rate constant was obtained by Stern-Volmer kinetics,
which depend on reactant disappearance as the photoprod-
ucts are formed. We judged16 the intermediate σ-complex
for nitro displacement to partition fully to product because
nitrite ion is far superior to hydroxide ion as a nucleofuge.19
The σ-complex for methoxy displacement is expected to
partition between product and reactant in a ratio of about
1:1 because the methoxide and hydroxide nucleofuges are
comparable. In such cases in the absence of complicating
rate constant is 3 ꢀ 109 M-1 s-1 22
. On this basis we calculated
the activation energies for the two operative reactions. The
values are 2.17 kcal/mol for attack at the nitro-bearing car-
bon and 2.58 kcal/mol for attack at the methoxy-bearing
carbon. We then calculated the rate constants for each dis-
placement reaction at the various temperatures. The results
and the consequent product ratios are given in Table 2. The
agreement of the product yields at various temperatures cal-
culated by using the Arrhenius equation with those of ex-
periment is excellent. This supports the view that the nucleo-
phile attacks the triplet at two sites that are differentiated by
slight differences of activation energy.
The large temperature effects on photosubstitution ratios
in 4-nitroanisole by hydroxide ion in alcohol solvents were
attributed14 to nitro displacement occurring with a very low
activation energy and methoxy displacement involving a
considerable activation energy. The claimed changes from
a product ratio (nitro:methoxy) of 99:1 at -20 ꢀC, to 9:1 at
25 ꢀC, to 0.6:1 at 196 ꢀC cannot be rationalized, however,
with an Arrhenius model. With the activation energies
needed to give a 9:1 ratio at 25 ꢀC, as reported,14 the change
to the ratio of 99:1 by Arrhenius theory would require a tem-
perature below -200 ꢀC, and a temperature of 500 ꢀC would
predict a product ratio of only 1.3:1. In fact, a temperature
nearing infinity could cause the product ratio only to
ꢀ
ꢀ
(13) Pinter, B.; De Proft, F.; Veszpremi, T.; Geerlings, P. J. Org. Chem.
2008, 73, 1243.
ꢀ
ꢁ ꢁ
ꢀ
(14) Klan, P.; Ruzicka, R.; Heger, D.; Literak, J.; Kulhanek, P.; Loupy,
ꢀ
A. Photochem. Photobiol. Sci. 2002, 1, 1012.
ꢀ
€
(15) Muller, P.; Loupy, A.; Klan, P. J. Photochem. Photobiol., A 2005,
172, 146.
(16) Wubbels, G. G.; Brown, T. R.; Babcock, T. A.; Johnson, K. M.
J. Org. Chem. 2008, 73, 1925.
(17) Letsinger, R. L.; Ramsay, O. B.; McCain, J. H. J. Am. Chem. Soc.
1965, 87, 2945.
(18) The phenolic products, PMP and PMP, are ionized completely by the
alkaline medium to PMPh and PNPh after they are formed.
(19) Broxton, T. J.; Muir, D. M.; Parker, A. J. J. Org. Chem. 1975, 40,
3230.
(20) Stirling, C. J. M. Acc. Chem. Res. 1979, 12, 198.
(21) Ballenger, P.; Long, F. A. J. Am. Chem. Soc. 1960, 82, 795.
(22) (a) Eigen, M. Angew. Chem., Int. Ed. Engl. 1964, 3, 1. (b) Weller, A.
Z. Physik. Chem. 1958, 17, 224.
7728 J. Org. Chem. Vol. 75, No. 22, 2010

Wubbels et al.
JOCArticle
SCHEME 2. Photoreaction Species and Sequence for 4-Nitroanisole and Hydroxide Ion
approach 1:1. This suggested that products other than those
claimed were being measured. That the reaction system
could give other products could be inferred from a proto-
tropic equilibrium effect that was not noted in the report.
Ethanol has acidity comparable to that of water. With
acidity constants of 1.3 and 1.8 ꢀ 10-16 M-1 for ethanol and
water, respectively,21 the prototropic equilibrium in the
reported medium of pure ethanol (or the other primary
alcohols used) containing 0.2 M NaOH can be calculated
quickly to have about 2% of the base in the form of hydrox-
ide ion, with most of it in the form of alkoxide ion. We
verified only that the products of photolysis of 4-nitroanisole
in 0.2 M NaOH in absolute ethanol were indeed complex. In
our hands, GC/MS analysis of the photolysis products
indicated 4-ethoxyanisole (41%), 1-ethoxy-4-nitrobenzene
(7%), 4-methoxyphenol (38%), and 4-nitrophenol (9%). We
also detected minor amounts of photoreduction products
such as 4-methoxyaniline and 4,40-dimethoxyazoxybenzene,
a process with precedent for other nitroaromatics in alco-
holic alkali.23 As in the earlier work,14 we did not dehydrate
our absolute ethanol. Adventitious water may be the cause of
the substantial yields of photosubstitution products of hy-
droxide ion.
Scheme 2, and the calculated structures and Gibbs free
energies at 298 K are shown in Table 3. The Gibbs energies
are calculated from the total energy, after calculation of the
vibrational frequencies, by correcting for the zero-point
vibrational energies, and the entropy and enthalpy changes
from zero to 298 K. We also calculated most of the species at
higher levels of theory including Hartree-Fock 6-31G*,
6-31þG*, and DFT-B3LYP/6-31G*. These more costly
calculations did not produce significantly different or more
reasonable energy increments or geometries. We could not
apply solvation models to open-shell species, but we could
for triplet transition states. The energy differences of
solvated vs nonsolvated transition states were comparable,
implying that calculation errors and solvation effects can-
celed each other out for comparable species. The vibrational
frequencies for the transition states included the expected
imaginary frequency for stretching the nucleophile-carbon
bond, with the exception of entry 6. This calculation con-
verged to an energy maximum saddle point but did not give
an imaginary frequency. Of several encounter complexes
found, we have shown as entry 1 only the one of highest
energy.
Because we were interested in a model for regioselectivity
with predictive power, we carried out quantum chemical
calculations on the reactants, transition states, and intermediates
of the probable mechanism to see if the regiochemical find-
ings could be discovered by calculation. The Spartan ’08
quantum chemical software at the Hartree-Fock/3-21G(*)
level of theory efficiently found the molecular species
postulated. They are identified in the reaction sequence in
Discussion
It has been shown by photosensitization24 and transient
spectroscopy16 that the excited state of 4-nitroanisole popu-
lated rapidly and efficiently by photoexcitation in water is
the triplet (Et ≈ 60 kcal/mol).25 Kinetic17 and quenching26
evidence suggested that nucleophiles react directly with
(24) Letsinger, R. L.; Steller, K. E. Tetrahedron Lett. 1969, 1401.
(25) Brinen, J. S.; Singh, B. J. Am. Chem. Soc. 1971, 93, 6623.
(26) Bonhila, J. B. S.; Tedesco, A. C.; Nogueira, L. C.; Diamantino, M. T.
R. S.; Carreiro, J. C. Tetrahedron 1993, 49, 3053.
(23) Frolov, A. N.; Kuznetsova, N. A.; El’tsov, A. V.; Rtishchev, N. I. Zh.
Organ. Khim. 1973, 9, 963.
J. Org. Chem. Vol. 75, No. 22, 2010 7729

Wubbels et al.
JOCArticle
TABLE 3. 4-Nitroanisole-Hydroxide Ion Reaction Species Calculated
and Kronenberg.27 This rationale lost relevance when it
became clear that the nitroaromatic nucleophilic photosub-
stitutions proceeded from the triplet state.24 A reexamina-
tion of the charge criterion using the triplet-state charge
densities, however, concluded that the charge criterion was
still unsatisfactory.9 Application to the photosubstitutions
of 4-nitroanisole was noted as a particular failing of the
approach. That calculated charge at nuclear positions is
unrelated to predominant regioselectivity of substitution
could be inferred also from the absence of regiochemical
effects of charge of nucleophiles. In the case of triplet
4-nitroanisole, for example, the negatively charged nucleo-
phile, hydroxide ion, displaces mainly at the positive charge
favored nitro-bearing carbon, but the neutral nucleo-
philes, ammonia and pyridine, displace exclusively at that
position.7,17
by Hartree-Fock/3-21G(*)
Epiotis and Shaik8 suggested on the basis of frontier
molecular orbital theory (FMO) that the regioselectivity of
nucleophilic aromatic photosubstitution could be predicted
by the ring carbon atom (bearing a displaceable group)
calculated to have the largest orbital coefficient (or electron
density) in the HOMO of the ground state singlet.28 The
model was extended11,12,29 to photoreactions involving elec-
tron transfer from the nucleophile as the initial step
(SN(ET)Ar*),30 wherein the regioselectivity was claimed to
be related to the maximum LUMO coefficient of the ground-
state reactant.
The simple FMO model, or more sophisticated versions
such as those based on spin-polarized Fukui functions,13
predicts the major products of many nucleophilic aromatic
photosubstitutions of nitrophenyl ethers, but it fails in
several cases. It seldom accounts for the ordering of minor
products,16 and it predicts incorrectly for the cases of 4-nitro-
and 2-nitroanisole.13 Allowing for the absence of a nucleo-
fuge at the meta position, it predicts, as observed, that the
favored photodisplacement site for 4-nitroanisole and hy-
droxide ion is the nitro-bearing carbon. It does not predict
correct regioselectivity for the photoreaction of 4-nitroani-
sole with cyanate or cyanide ion.7,16 It should be noted that
these regioselectivity predictions for the nitrophenyl ethers
are not risky, and yet they display a high level of error. The
simple generalization that SN2Ar* reactions occur meta to
the nitro group and SN(ET)Ar* reactions occur para to the
nitro group does as well as the FMO model or its extensions
to predict regioselectivity. It is possible that the HOMO and
LUMO criteria amount only to a cumbersome method of
identifying the positions meta and para to the nitro group.
We question the FMO model for regioselectivity first
because the relationship of electron density in the ground
state singlet HOMO or LUMO to reaction from an excited
state triplet should be faint. The orbitals of the ground state
upon which the predictions rest do not even exist in the
reacting, excited molecule. Moreover, the bond formation
with the nucleophile in SN2Ar3 is electron-paired and prob-
ably has nothing to do with the singly occupied, excited state
triplet nitrophenyl ethers. The halogen element effect evi-
dence16 indicated that the initial steps of nucleophilic photo-
substitutions proceeded within the triplet manifold through
triplet σ-complexes. That is, reaction of the triplet with the
nucleophile appeared to be adiabatic, leading rapidly to a σ-
complex of the same multiplicity. Nonadiabatic reaction was
unlikely because there is no electronic triplet state below the
ground state triplet, and intersystem crossing was expected
to be too slow to account for the nucleophilic reaction rates.
The reaction rates approach the diffusion rate, and spin-
orbit coupling effects of halogens on the nitroaromatics
could not be discerned in the rates of nucleophilic attack.16
The photosubstitutions of triplet nitrophenyl ethers and
many other nitroaromatics with nucleophiles have been the
subject of many attempts to explain the reactivity and
regioselectivity.6-13 The unusual orientation effect of sub-
stituents in nucleophilic reactions, (and other reactions) of
excited aromatic molecules was recognized early by the
name, the “meta effect”.6 The effect was rationalized6 by
suggesting the occurrence of meta-bridged valence bond
structures for zwitterionic excited-state reactants. This ra-
tionale, however, was mostly symbolic since it could not
accommodate the open-shell nature of the excited states and
the triplet multiplicity.
(27) Havinga, E.; Kronenberg, M. E. Pure Appl. Chem. 1968, 16, 137.
(28) This was originally stated as the maximum difference of the squares
of the HOMO and LUMO coefficients. The authors, however, use the square
of the HOMO coefficient (HOMO electron density) as a sufficient measure.
(29) Bunce, N. J.; Cater, S. R.; Scaiano, J. C.; Johnston, L. J. J. Org.
Chem. 1987, 52, 4214.
The suggestion that charge density in the excited singlet
state could explain regiochemistry in electrophilic and nu-
cleophilic photosubstitutions was put forward by Havinga
(30) Mutai, K.; Kobayashi, K.; Yokoyama, K. Tetrahedron 1984, 40,
1755.
7730 J. Org. Chem. Vol. 75, No. 22, 2010

Wubbels et al.
JOCArticle
HOMO, or the LUMO, to say nothing of the ground state
HOMO or LUMO. More importantly, it suffers the power-
ful weakness of predicting favored reaction transition states
solely on the basis of properties of the reactant. This idea has
an inglorious history in the annals of reactivity and regio-
selectivity discussions. It is analogous to inferring the regio-
selectivity of electrophilic aromatic substitution reactions
from the resonance structures of reactants, a method that has
long been proscribed.31 Standard practice for use of the
resonance model is to consider the relative resonance stabi-
lization of cationic intermediates rather than the delocaliza-
tion or polarization of reactants to gain insight about the
relative energies of the preceding transition states. This logic
rests on the Bell-Evans-Polanyi principle,32 which holds
that activation energy and enthalpy change are correlated for
a set of comparable, elementary reaction steps.
energy than the favored meta triplet σ-complex. We
concluded16 that the energy gap interpretation of regioselec-
tivity of these reactions was not correct. Indeed, the evidence
pointed to nucleophilic aromatic photosubstitution of nitro-
phenyl ethers occurring within the triplet manifold, that is,
adiabatically. Evidence of the element effect on the elemen-
tary rate constants and the relative nucleophilicities of
different attacking reagents suggested, moreover, that the
bond formation process was electron-paired formation of a
σ-bond by the nucleophile with the triplet nitroaromatic
molecule.
A recent theoretical study13 of the orientation rules em-
phasized photohydrolysis of nitrophenyl ethers including
4-nitroanisole. It modeled the reactions with density func-
tional theory by calculating the geometries and energies of all
the likely encounter complexes, σ-complexes, and transition
states within the ground state singlet or triplet manifolds. It
found local energy minimum σ-complex intermediates for
the reactions that had oxygen nucleofugic groups. For triplet
4-nitroveratrol (3,4-dimethoxynitrobenzene) and hydroxide
ion, the triplet transition state leading to the favored meta
σ-complex was found by B3LYP calculation to be 6.1 kcal
lower in energy than the triplet transition state leading to the
para σ-complex. A Møller-Plesset calculation of the same
species, however, reversed the energy difference (to -5.0
kcal). Both methods of calculation corroborated at high
levels of theory the earlier finding9 that the meta triplet
σ-complex for 4-nitroveratrol and hydroxide ion was lower
in energy (by 9.5 kcal) than the para triplet σ-complex. By
confining its view to relative energies of successive species
within spin manifolds, however, it omitted consideration of
whether triplet σ-complexes might undergo intersystem
crossing before completing the substitution reaction. For the
case of 4-nitroanisole, the regiochemical discussion ignored
the predominant nitro displacement process. The regioselec-
tivity conclusions of the calculations of this study were
mixed, but it was the first to use quantum chemical methods
to estimate the relative energies of competing photosubstitu-
tion transition states of the same multiplicity.
Our experimental results for the temperature dependence
of the product ratio from 4-nitroanisole and those calculated
with a simple Arrhenius model agree very well. This validates
the model of competing transition states in the triplet mani-
fold, and it provides a sound experimental basis for assessing
regioselectivity by quantum chemical calculation of the
relative energies of transition states.
We found several energy minimum encounter complexes
of triplet 4-nitroanisole with hydroxide ion. The one calcu-
lated to have the highest energy, shown as entry 1 in Table 3,
placed hydroxide ion near the methoxy-bearing carbon of
triplet 4-nitroanisole. The transition states (entries 2 and 3)
lie 13.3 and 16.5 kcal above the energy of this encounter
complex, at the Hartree-Fock 3-21G(*) level. The correct
reference energy of the excited-state reactants appears from
experiment to be about 2 kcal below the energy of the first
transition state (entry 2), and we have shown energies
referenced to this value in column 5 of Table 3. Reaction
species discussed below are referenced to this excited reactant
energy level (-619.824577 au). The difficulty of finding the
correct reference energy of the reactants by calculation is not
unexpected because significant calculation errors occur between
species having different solvation, charge distributions, and
Dewar has emphasized33 that there is no theoretical
justification for reactivity or regioselectivity predictions in
alternate or nonalternate π-electron hydrocarbon systems
based on reactant properties at nuclear positions such as free
valence, self-polarization (charge density), and applications
in perturbation theory of HOMO coefficients of frontier
orbital theory. The successes that occur owe to the coin-
cidence of calculated positional properties and relative tran-
sition-state energies. The sorting of preferred pathways
occurs on the basis of energy. Energy of a molecular species
can be defined only for the whole and not for particular
nuclear positions. Estimating the contribution of certain
positional properties of reactants to prospective competing
transition states is clearly inferior to estimating directly the
energies of the competing transition states.
Another model for regioselectivity of SN2Ar* reactions
claimed that the energy gap law for radiationless electronic
transitions should govern the favored pathways.9 The energy
gap law in electronic spectroscopy states that internal con-
versions between states close in energy are faster than those
between states far apart in energy. Since singlet meta-to-nitro
σ-complexes or those for displacement of the nitro group
have little resonance stabilization, they are much higher in
energy than the delocalized ortho or para σ-complexes. Thus
the transition from the triplet encounter complex to the less
stable (meta) σ-complexes was thought to be faster than to
the more stable (ortho or para) σ-complexes.9
The energy gap law applies primarily to internal conver-
sion processes that are fast. The application to photosub-
stitution regioselectivity9 required intersystem crossing
during nucleophilic attack on the triplet nitroaromatic lead-
ing to the singlet σ-complex. With rate constants of nucleo-
philic attack approaching the rate of diffusion, the processes
appeared to be much too fast to permit intersystem cros-
sing.16 Moreover, spin-orbit coupling effects of attached
halogens that should have facilitated the intersystem cross-
ing correlated inversely with the observed rate constants.16
For reaction within the triplet manifold, the energy gap law
would predict the wrong regioselectivity since the para triplet
σ-complex was found by calculation9 to be much higher in
(31) March, J. Advanced Organic Chemistry: Reactions, Mechanisms, and
Structure; McGraw-Hill: New York, 1968; p 387.
(32) Bruckner, R. Advanced Organic Chemistry; Harcourt/Academic
Press: San Diego, 2002; p 10.
(33) Dewar, M. J. S. The Molecular Orbital Theory of Organic Chemistry;
McGraw-Hill: New York, 1969; p 362.
J. Org. Chem. Vol. 75, No. 22, 2010 7731

Wubbels et al.
JOCArticle
numbers and types of bonds. For closely comparable species,
however, such as two transition states or two σ-complexes,
the error in the calculated energy difference should be much
smaller because of cancellation of calculation errors. Indeed,
the HF 3-21G(*) energy difference for entry 3 minus entry 2 is
3.2 kcal, which may be compared to the difference of 0.4 kcal
that was found experimentally. The favored transition state
is the one calculated to have lower energy. At higher levels of
theory, the energy differences (in kcal) were as follows: HF/
3-21G(*) in water (SM8), 6.0; HF/6-31G*, 6.7; HF/6-31G* in
water (SM8), 11.4; HF/6-31þG* 5.5; HF/6-31þG* in water
(SM8), 10.7;34 B3LYP/6-31G*, -1.0; B3LYP/6-31G* in
water, 7.8. Only one of these methods, B3LYP (vacuum),
misses the favored transition state (by about 1 kcal). The
other seven methods are successful at identifying the favored
transition state.
As shown by the results discussed above, we prefer the HF/
3-21G(*) calculations for this system because of their lower
cost and adequate accuracy. Calculation of the species in
Table 3 at the HF/6-31G* level yielded geometries and
energy differences very similar to those of the HF/3-21G(*)
level. While the calculations, save one, identify the lower-
energy transition state, we do not know the reason that the
calculated energy differences of the competitive transition
states are typically larger than the experimental values, even
at higher levels of theory that provide large valence sets,
more polarization, diffuse functions for charges, electron
correlation, and solvation.
The favored regiochemistry could also be identified by
comparing the energies of the triplet σ-complexes. These
(entries 4 and 5) were found to lie 20.8 and 18.3 kcal below
the respective transition states (entries 2 and 3). In this case,
the energy difference at the HF/3-21G(*) level for entry 5
minus entry 4 is 5.7 kcal. These differences (in kcal) at higher
levels of theory were: HF/6-31G*, 6.5; HF/6-31þG*, 2.7;
and B3LYP/6-31G*, 5.0. The calculations all identify the
favored pathway, and the differences cluster more closely
than those for the transition states. The lower-energy path-
way is correctly identified by all methods, and they agree
with results for the transition states. The agreement supports
the view that the mechanism involves electron-paired bond
formation within the triplet manifold, since in that case, these
transition states and intermediates would be related by an
elementary reaction step. The correspondence of their rela-
tive energies is a manifestation of the Bell-Evans-Polanyi
principle. In view of the agreement of the σ-complex energy
differences, calculation of their relative energies may be a
more reliable indicator of regioselectivity than calculation of
relative transition-state energies.
that the results reported in ref 14 could be caused by an
analogous rearrangement mechanism. That suggestion now
appears unnecessary because the para substitution on 4-
nitroanisole by hydroxide ion appears to fit a straightfor-
ward SN2Ar* model.
The calculations of the triplet transition states for expul-
sion of nitrite ion or methoxide ion shown as entries 6 and 7,
respectively, reveal that they lie 14.4 and 18.6 kcal above the
respective triplet σ-complexes. These present large activation
energy barriers. They suggest that expulsion of leaving
groups from the triplet σ-complexes would be slow, giving
the triplet σ-complexes sufficient lifetime to undergo inter-
system crossing. The corresponding ground state singlet
σ-complex for nitro displacement (entry 8) lies 48.5 kcal
below the triplet σ-complex (entry 4), and that for methoxy
displacement (entry 9) lies 10.9 kcal below the corresponding
triplet σ-complex (entry 5). The former singlet (entry 8),
however, reached energy minimization with an impossibly
˚
long C-N bond of 3.92 A. This implies that an energy mini-
mum structure was not found with a normal C-N distance
and that the singlet σ-complex is dissociative.36 It implies
that nitrite ion loss from the singlet σ-complex would have no
activation energy and would follow intersystem crossing very
rapidly. The similarity of the energy gap (67.3 kcal) between
entry 8 and the excited reactants, and the spectroscopic
triplet energy of 4-nitroanisole (60 kcal)25 suggests that entry
8 represents the ground singlet state of the products, 4-meth-
oxyphenol and nitrite ion. This view was confirmed by direct
calculation of the energy of the singlet hydrogen-bonded
complex of 4-methoxyphenol and nitrite ion, which was 68.2
kcal below that of the excited reactants. The methoxide ion
expulsion is reasonably formulated as an activated reaction
that competes with hydroxide ion expulsion from the singlet
σ-complex that lies about 24 kcal below the excited reactants.
We conclude that the required intersystem crossing from
triplet reactants to singlet products occurs at the σ-complex
stage, and that leaving group expulsion occurs from the
singlet rather than the triplet σ-complex.
The mechanistic conclusions afforded by the experimental
and quantum chemical results of this study agree with those
based on element effects16 for nucleophilic reactions of trip-
let nitrophenyl ethers. The main features are the following:
(1) nucleophilic bond formation occurs adiabatically within
the triplet manifold; (2) competitive nucleophilic reactions
occur through thermally competitive transition states; and
(3) the elementary reactions produce triplet σ-complexes that
undergo intersystem crossing to singlet σ-complexes, which
dissociate anionic leaving groups to make the products. The
relative transition-state energy predictor of regioselectivity
in these reactions appears to have a sound basis. We intend to
test its application to regioselectivity by quantum chemical
methods for other reactions of the SN2Ar* type that have
been reported.7
Several studies11,12 have suggested that nucleophilic
photosubstitution para to the nitro group may occur by a
mechanism different from SN2Ar*, that being the electron
transfer-radical coupling (SN(ET)Ar*) mechanism. We re-
ported evidence that the para photosubstitutions by amines
on 4-nitroanisole may actually involve bond formation by
the amine at the meta position followed by a fast sigmatropic
rearrangement to the para σ-complex.35 We also suggested
Experimental Section
Except as noted below, chemicals and solvents were commer-
cial materials of high purity. Ethanol was anhydrous, and
attempts to dehydrate it further were not made. Acetonitrile
(34) The calculation by HF/6-31þG* in water for entry 2 converged only
to twice the normal gradient tolerance, which did not permit calculation of
the vibrational frequencies. The energy difference is that for the total
energies.
(36) The calculation seeks geometry optimization by stretching the C-N
bond that increasingly produces nitrite ion in the gas phase. As its energy goes
up, the energy minimum is realized.
(35) Wubbels, G. G.; Johnson, K. M. Org. Lett. 2006, 8, 1451.
7732 J. Org. Chem. Vol. 75, No. 22, 2010

Wubbels et al.
JOCArticle
was chromatographic grade. 1-Ethoxy-4-nitrobenzene and 4-
ethoxyanisole were made by alkylating the corresponding phe-
nol with ethyl iodide in 2-butanone containing K2CO3. These
substances gave satisfactory 1H NMR spectra and literature
melting points.
UV-vis spectra were obtained with a diode-array spectro-
photometer with samples in quartz cuvettes. Gc/ms was carried
out with a 25 m silicone oil capillary column and an ion trap
mass spectrometer.
Photolyses for preparative purposes in ethanolic NaOH were
carried out in Pyrex tubes in a Rayonet RPR-208 reactor fitted
with eight 300 nm broadband lamps. Workups for chromato-
graphic product analyses were carried out by diluting irradiated
samples 5-fold with water, acidifying with H2SO4, extracting
with CH2Cl2 or toluene, concentrating, and diluting to volume
with CH2Cl2. Products were identified by coinjection with
authentic samples as well as by the mass spectra.
Temperature-controlled photolyses were carried out in the
Rayonet reactor with seven 350-nm broadband lamps in quartz
septum-capped cuvettes suspended in water or water/ice at
controlled temperature in a uranium glass filter sleeve within a
vacuum-jacketed quartz well. Reaction solutions were purged
with argon through syringe needles for 10 min. UV-vis analyses
were done directly in the irradiation cuvettes. The extinction
coefficients for the quantitative analyses were: 4-nitroanisole:
317 nm (9700); 4-nitrophenoxide ion: 400 nm (18,300), 306 nm
(1050); 4-methoxyphenoxide ion: 306 nm (2920).
The quantum calculations were carried out with Spartan ’08
software running on a desktop computer with the Windows
operating system. The structures were found by convergence to
equilibrium geometry at an energy minimum or a saddle point with
a gradient tolerance of 4.5 ꢀ 10-4 hartree bohr-1 and a distance
tolerance of 1.8 ꢀ 10-3 A. We calculated the vibrational frequencies
˚
for the transition states, and they showed with one exception the
imaginary frequency for the transition-state vibration.
Acknowledgment. This paper is dedicated to Professor
Robert L. Letsinger on the occasion of his 89th birthday.
Acknowledgment is made to the donors of the Petroleum
Research Fund, administered by the American Chemical
Society, for support of this work.
Supporting Information Available: Cartesian coordinates of
the structures reported in Table 3, the total energies, the
imaginary frequencies of transition states, and chromatogram
of photoproducts in ethanolic NaOH. This material is available
free of charge via the Internet at http://pubs.acs.org.
J. Org. Chem. Vol. 75, No. 22, 2010 7733