Journal of the American Chemical Society
Article
ketone (with a known solid-state quantum yield of Φ = 0.2),
which is completely converted to the corresponding photo-
product under the same experimental conditions. If our
assumption holds, the time constant for the decarbonylation
reaction (step 4, Scheme 1) must be at least 2 orders of
magnitude greater than that of the intersystem crossing step 5
in Scheme 1. Thus, a long-lived transient with a time constant
of 63 μs for the decay of the radical pair from 1d implies a
decarbonylation time constant of at least 6.3 ms. This lower-
bound value demonstrates that the crystalline solid state can
drastically lengthen the decarbonylation lifetime by more than
3 orders of magnitude.
The lifetime enhancement observed in the solid state for the
radical pairs featuring aliphatic acyl radicals (1b−d) is
consistent with reports of restricted radical pairs and biradicals
exhibiting longer lifetimes in the solid state than in
solution.39−43 The structural rigidity and least-motion
character of reactions in solids suggested by the topochemical
postulate44 suggest that it is reasonable to analyze the lifetime
enhancement in terms of structural parameters of the ketone
molecular and crystal structure.45,46 Thus, on the assumption
that the acyl radical resides in an sp2 orbital, the two singly
occupied orbitals would remain in the original orientation of
the σ bond after the α-cleavage step, coaligned and close to the
ketone α bond length distance of 1.54 Å. As mentioned before
and suggested in Figures 1 and 2, the proximity of the two
radical centers is expected to result in large zero-field splitting
and a large singlet−triplet energy gap. When we consider that
energy must be conserved when there is a change in spin
angular momentum, the large energy gap between singlet and
triplet states poses a challenge for the intersystem crossing
step. Furthermore, on consideration that the open-shell singlet
should be ideally poised for bond formation, it is reasonable to
expect that dephasing and equilibration between the T0 and S
radical pair states, potentially facilitated by the relatively
different g factors of the two radical centers, should be very
unlikely. An alternative mechanism based on coupled
electron−nuclear spin flips determined by hyperfine coupling
(hfc) interactions are known to be effective when the S-T gap
is relatively small, at relatively long inter-radical distances.
Spin-flipping mechanisms caused by the dynamic modulation
of magnetic fields through spin−lattice relaxation requires fast
molecular dynamics, such as spin-rotation interactions and
solvent motion,47,48 which are ineffective under the rigidity of
the crystal lattice. Finally, while it may be expected that the
most efficient ISC mechanism for a triplet radical pair in a
crystalline solid will be based on spin−orbit coupling (SOC), it
is anticipated that the collinear arrangement of the two singly
occupied orbitals will be unfavorable, as there is a limited
change in angular momentum mediated by electron exchange
and/or mixing of zwitterionic states.15−17 While there is much
work to be done to ascertain the initial polarization, spin
sublevel equilibration, and relative contributions of the
different ISC mechanisms, it is clear that the restrictions
imposed by the crystal lattice are able to maintain these spin
systems for much longer times in comparison to any other
medium. To the best of our knowledge, the long-lived
Figure 4. Transient absorption spectra of 1-((1R,3R,5R,7R)-
adamantan-2-yl)-2,2-diphenylethan-1-one (1c) in acetonitrile solution
(top) and in aqueous nanocrystalline suspension (bottom). The fitted
exponential decay curves in the insets were used to determine the
lifetimes of the transient absorption signals.
We were able to confirm that the short-lived component
observed with nanocrystalline suspensions does not originate
from a small amount of the sample in solution. The observed
transient and the decay kinetics were not affected by oxygen,
which would have trapped free radicals in solution.
Furthermore, a 1H NMR analysis of recovered samples showed
only the starting material. It is possible that the short-lived
component may arise from molecules in less rigid sites either at
the surface or at defect sites, but we cannot discount that
different triplet radical pair sublevels formed by state-selective
intersystem crossing in the starting ketone may have access to
different intersystem crossing mechanisms.33−38
The first key observation from the data shown in Table 1 is
the remarkable difference between the phenylacyl (benzoyl)
radical pair in 1a and the alkyl−acyl radical pairs in 1b−d. The
second most notable observation pertains to the unprece-
dented long lifetime for an acyl-containing radical pair state.
We will center our attention on the long-lived ca. 50−60 μs
components for ketones 1b−d, which account for the majority
of the transients formed, and compare them with solution and
gas-phase decarbonylation lifetimes drawn from the literature
for analogous acyl radicals26,27 (Table 1). An important
conclusion from ketone 1d is that the rate of decarbonylation
in the solid state is much slower than that in solution. This can
be shown by comparing the decarbonylation rate of 1.9 μs in
solution with a time constant for intersystem crossing of 63 μs
in the crystalline state. Given that no photoproducts are
observed by 1H NMR, we take an upper bound of 1%
decarbonylation to estimate a lower bound for the time
constant of this bond-cleavage reaction. This analysis assumes
3
component of the RP-1 state of ketones 1b−d (53−63 μs)
has the longest lifetimes for an acyl−alkyl triplet biradical or
radical pair reported to date, with previous lifetimes reaching
no more than ca. 5.5 μs.29,49−54 Further studies based on
magnetic resonance methods such as CIDEP will yield
3
that each molecule in the sample accesses the RP-1 state at
least once, which we believe to be a conservative estimate on
the basis of a comparison with an analogous sample of dicumyl
8889
J. Am. Chem. Soc. 2021, 143, 8886−8892