Journal of the American Chemical Society
Article
BDE values are those recommended in Luo’s compilation.31 For
substrates where Luo provides more than one value without a
recommendation, the BDE was taken as the average of the
tabulated values. Luo’s compilation does not contain BDE
values for the tertiary C−H bond of 2,2,3-trimethylbutane (4) or
the C−H bonds α to the OH, NH2, or NH groups of 1,3-
propanediol (21), cis-4-tert-butylcyclohexanol (25), trans-4-tert-
butylcyclohexanol (26), isobutylamine (35) dibenzylamine
(38), N-tert-butylpyrrolidine (42), N-Boc-pyrrolidine (48),
and N-Boc-proline (49), and these substrates were omitted
from the plot. The C−H BDE values for HMPA (51) and
DMSO (54), 94.4 and 102.1 kcal mol−1, respectively, were taken
from our recent work in which we discussed the large
discrepancy between Luo’s tabulated value for DMSO of 94.0
kcal mol−1 and our computed value.47 The BDEs for the α-C−H
bonds of hexylamine (31) and octylamine (32) are assumed to
be identical to the tabulated value for pentylamine (30).31 The
BDE for the α-C−H bonds of dipropylamine (33) is assumed to
be identical to the tabulated value for tripropylamine (34).31 All
the data employed for the log kH′ vs C−H BDE plot displayed in
Figure 1 are collected in Table S1 in the SI. For comparison, the
O−H BDE of 2-phenylpropan-2-ol (cumyl alcohol) is given by
Luo as 104.7 0.2 kcal mol−1,31 essentially at the right axis of the
plots.
The overall view of Figure 1a suggests that there is not a
simple relationship between log kH′ and C−H BDE. However,
grouping the kinetic data based on substrate type, i.e., benzylic/
allylic hydrocarbons (black circles), saturated hydrocarbons
(white circles), alcohols and ethers (red circles), amines (green
circles), and other substrates (yellow circles), reveals two broad
relationships (Figure 1b) that are similar to those observed by
Tanko and co-workers for HAT reactions involving tBuO• with
their smaller set of substrates (SI, Figure S2).15 However, unlike
the results of Tanko, the plot displayed in Figure 1b shows no
leveling off of log kH′ at around 6.6.
high-level W1BD approach (mean absolute error, MAE = 0.26
kcal mol−1). Additional analysis of the calculated BDEs is
provided in the SI.
Compared to Luo’s set of compiled data,31 the computed
BDE for the benzylic C−H bonds of benzylamine (37) is
significantly lower (79.8 vs 88.0 kcal mol−1). The computed
value is also similar to that obtained for the corresponding C−H
bonds of dibenzylamine (38). The computed BDEs for the
benzylic C−H bonds of benzyl alcohol (23) and dibenzyl ether
(24) are now very similar, in line with expectations (83.2 and
83.7 kcal mol−1, respectively). The computations also produce
BDEs for the secondary and tertiary C−H bonds of adamantane
(9) that are very similar, viz., 100.5 and 100.1 kcal mol−1,
respectively.
Interestingly, with 1,2-ethanediol (20) and 1,3-propanediol
(21) calculations predict an intramolecular hydrogen-bonded
structure that in acetonitrile is more stable than the non-
hydrogen-bonded one by 2.5 and 3.5 kcal mol−1, respectively
(SI, Figure S28). The BDEs for the C−H bonds that are α to the
hydrogen bond donor (HBD) and hydrogen bond acceptor
(HBA) OH groups are 95.1 and 97.4 kcal mol−1 and 94.0 and
96.0 kcal mol−1, for 20 and 21, respectively. The corresponding
BDFEs are 86.8 and 88.4 kcal mol−1 and 85.7 and 87.2 kcal
mol−1. Based on these findings, it can be reasonably assumed
that with both substrates HAT to CumO• predominantly occurs
from the weaker and more electron-rich C−H bonds of a single
methylene unit, and accordingly, for these two substrates the kH′
values displayed in Table 2 have been obtained considering n =
2.
Also included in Table 3 are the computed gas-phase C−H
BDFEs for substrates 1−56, calculated using the (RO)CBS-
QB3 approach. The BDFEs are on average 8.6 0.7 kcal mol−1
lower than the corresponding BDEs (uncertainty is 1σ). This
difference is primarily due to the entropy of H•(g), TΔS°(H•)(g)
= 8.17 kcal mol−1.50 The agreement between these values
indicates that the entropies of R−H and R• are close to the
same.51 Because there is close to a constant shift between BDE
and BDFE, the plots in this report look very similar using either
parameter, with just a change in the horizontal axis (see below).
For comparison, the computed O−H BDE and BDFE for
cumyl alcohol (2-phenylpropan-2-ol) are 106.6 and 98.2 kcal
mol−1.
Analysis of the data points displayed in Figure 1b shows that
benzyl alcohol (23) and dibenzyl ether (24) do not follow the
same general trend established by the alcohol and ether
substrates, nor does triallylamine (39) fit the general trend
associated with the amine substrates. Given their very weak C−
H bonds, these substrates would have been expected to react
with much higher rate constants. HAT from 23, 24, and 39 to
CumO• occurs selectively from the benzylic and allylic C−H
bonds, and the corresponding log kH′ values appear to fit fairly
well to the benzylic/allylic correlation (black circles).
Computations of a Consistent Set of BDEs and BDFEs.
One would reasonably expect very similar BDEs for the benzylic
C−H bonds of benzyl alcohol (23) and dibenzyl ether (24), but
quite surprisingly the tabulated values differ by 6.8 kcal mol−1,
viz., 79.0 and 85.8 kcal mol−1,31 respectively. From this
perspective, the tabulated BDE of 88.0 kcal mol−1 for the
benzylic C−H bonds of benzylamine (37) seems to be too
high.31 These apparent discrepancies in the BDEs, along with
the absence of BDE values for the C−H bonds of some of the
substrates listed in Tables 1 and 2, prompted us to use
computational methods to generate a consistent set of gas-phase
C−H BDEs for substrates 1−56. We calculated the relevant C−
H BDEs using the (RO)CBS-QB3 approach and present these
data in Table 3 (column 4). For comparison, Luo’s tabulated
values are shown in column 3 of Table 3. According to the
benchmarking data we present in the SI, the (RO)CBS-QB3
approach predicts BDEs that are in excellent agreement with the
BDEs we computed for 22 out of the 56 substrates using the
Rate Constant−Bond Strength Correlations. In Figure
2, we plot the measured log kH′ values for HAT from substrates
1−56 to CumO• taken from Table 1 and Table 2 against the
calculated C−H BDEs from Table 3.
Figure 2 reveals trends in the relationship between log kH′ and
C−H BDE that are not apparent in Figure 1a, clarifying the
trends that are roughly present in Figure 1b. Specifically, the data
associated with the benzylic and allylic hydrocarbons (black
circles), i.e., the unsaturated group, show a relatively good
correlation, in particular with the inclusion of the data associated
with benzyl alcohol (23), dibenzyl ether (24), benzylamine
(37), dibenzylamine (38), and triallylamine (39). Figure 2 also
shows that the points for saturated hydrocarbons, alcohols,
ethers, diols, amines, and carbamates (saturated group) tend to
cluster around a different line with a slope that is steeper than
that associated with the benzylic/allylic set, albeit with a lower
correlation coefficient. Collectively, Figure 2 demonstrates that,
depending on the nature of the substrate, there are two distinct
EP relationships, and these results provide strong support for the
bimodal behavior observed previously in purely theoretical
studies of reactions promoted by other HAT reagents.16,19
11767
J. Am. Chem. Soc. 2021, 143, 11759−11776