Transition state for β-elimination of hydrogen from alkoxy groups on metal surfaces

Article abstract of DOI:10.1021/jp993474x

Experimental investigations of β-hydrogen elimination from alkoxy and alkyl groups bound to a Cu(111) surface have been coupled with computational studies of gas-phase analogues to provide insight into the transition state for catalytic hydrogenation and dehydrogenation on metal surfaces. Previous studies have shown that fluorination increases the activation barrier (ΔE_act) to β-hydrogen elimination in ethoxy groups (RCH₂O_(ad) → RCH=O_(ad) + H_(ad), where R = CH₃, CFH₂, CHF₂, CF₃) and propyl groups (RCH₂CH_2,(ad) → RCH=CH_2,(ad)+ H_(ad), where R = CH₃, CF₃) on the Cu(111) surface. The increase in barrier height with increasing fluorination was attributed to the inductive influence of fluorine, which energetically destabilizes a hydride-like transition state of the form [RC^δ+...H^δ-]^?. In this paper, deuterium kinetic isotope effects (DKIE) show that fluorination does not alter the mechanism for β-hydrogen elimination from ethoxy groups. Furthermore, the DKIE measurements confirm that the effects of fluorine on the kinetics of β-hydrogen elimination do not result from the change in mass when hydrogen is substituted by fluorine. A systematic study of fluorine substitution of surface-bound isopropoxy groups reveals combined steric and electronic effects. An excellent correlation is found between the ΔE_act for β-hydrogen elimination in adsorbed alkoxy groups and the calculated reaction energetics (ΔH_rxn) for gas-phase dehydrogenation of fluorinated alcohols in trans antiperiplanar conformations (e.g., RCH₂OH_(g) → RCH=O_(g) + H_2,(g), where the hydroxyl hydrogen is antiperiplanar to a carbon and the oxygen is antiperiplanar to a fluorine). Hammett plots for β-hydrogen elimination give a reaction parameter of ρ = -26. These correlations both suggest that the transition state for β-hydrogen elimination develops a greater partial positive charge on the carbinol carbon than is found in the adsorbed reactant. Furthermore, the transition state is energetically late in the reaction coordinate for β-hydrogen elimination.

Full text of DOI:10.1021/jp993474x

2476
J. Phys. Chem. A 2000, 104, 2476-2485
Transition State for â-Elimination of Hydrogen from Alkoxy Groups on Metal Surfaces
Andrew J. Gellman,* Mark T. Buelow,† and Shane C. Street‡
Department of Chemical Engineering, Carnegie Mellon UniVersity, Pittsburgh, PennsylVania 15213
Thomas Hellman Morton
Department of Chemistry, UniVersity of California, RiVerside, California 92521-0403
ReceiVed: September 28, 1999; In Final Form: January 11, 2000
Experimental investigations of â-hydrogen elimination from alkoxy and alkyl groups bound to a Cu(111)
surface have been coupled with computational studies of gas-phase analogues to provide insight into the
transition state for catalytic hydrogenation and dehydrogenation on metal surfaces. Previous studies have
shown that fluorination increases the activation barrier (∆Eact) to â-hydrogen elimination in ethoxy groups
(RCH2O(ad) f RCHdO(ad) + H(ad), where R ) CH3, CFH2, CHF2, CF3) and propyl groups (RCH2CH2,(ad)
f
RCHdCH2,(ad) + H(ad), where R ) CH3, CF3) on the Cu(111) surface. The increase in barrier height with
increasing fluorination was attributed to the inductive influence of fluorine, which energetically destabilizes
a hydride-like transition state of the form [RCδ+‚‚‚Hδ-]‡. In this paper, deuterium kinetic isotope effects
(DKIE) show that fluorination does not alter the mechanism for â-hydrogen elimination from ethoxy groups.
Furthermore, the DKIE measurements confirm that the effects of fluorine on the kinetics of â-hydrogen
elimination do not result from the change in mass when hydrogen is substituted by fluorine. A systematic
study of fluorine substitution of surface-bound isopropoxy groups reveals combined steric and electronic
effects. An excellent correlation is found between the ∆Eact for â-hydrogen elimination in adsorbed alkoxy
groups and the calculated reaction energetics (∆Hrxn) for gas-phase dehydrogenation of fluorinated alcohols
in trans antiperiplanar conformations (e.g., RCH2OH(g) f RCHdO(g) + H2,(g), where the hydroxyl hydrogen
is antiperiplanar to a carbon and the oxygen is antiperiplanar to a fluorine). Hammett plots for â-hydrogen
elimination give a reaction parameter of F ) -26. These correlations both suggest that the transition state for
â-hydrogen elimination develops a greater partial positive charge on the carbinol carbon than is found in the
adsorbed reactant. Furthermore, the transition state is energetically late in the reaction coordinate for â-hydrogen
elimination.
Introduction
the Cu(111) surface, eqs 1 and 2.1,2 These experiments used
temperature-programmed reaction spectra (TPRS) to measure
The transition states for elementary steps of surface-catalyzed
reactions ultimately dictate the overall reaction kinetics of
complex catalytic processes. â-Hydrogen elimination has served
as an ideal prototype reaction for exploring the use of substituent
effects in the study of transition states on surfaces.1,2 Moreover,
â-hydrogen elimination is an important reaction in its own right.
It is observed in the decomposition of alkoxy groups and alkyl
groups on many metal surfaces.3,4 The microscopic reverse of
â-hydrogen elimination also constitutes the initial step in
catalytic hydrogenation.5 Ultimately, a deeper knowledge of the
nature of the transition state for â-hydrogen elimination offers
an opportunity to understand the influence of various catalytic
surfaces on this very common elementary reaction. Here, we
present results that clarify several points in our current
understanding of the barrier to â-hydrogen elimination and
further refine the picture of the transition state.
activation barriers (∆Eact) for â-hydrogen elimination in these
ethoxy and n-propyl groups as a function of fluorine substitution.
RCH2O-Cu f RCHdO(ad) + H-Cu
R ) CF3, CHF2, CFH2, CH3 (1)
RCH2CH2-Cu f RCHdCH2,(ad) + H-Cu
R ) CF3, CH3 (2)
In the case of the ethoxy species, fluorination of the methyl
group systematically increases the ∆Eact by 55 kJ mol-1 from
121 ( 4 kJ mol-1 to 176 ( 6 kJ mol-1. In the case of n-propyl
groups, perfluorination of the terminal methyl group increases
the activation barrier by ∆∆Eact ) 35 ( 3 kJ mol-1. The
experiments on ethoxy groups gave very similar results on
Cu(100), Cu(110), and Ag(110) surfaces, revealing that the
fluorine substituent effect does not display sensitivity to the
structure of the surface or the nature of the metal. However, it
is important to note that although the magnitude of the
substituent effect (∆∆Eact) is similar on Cu and Ag surfaces,
the absolute values of the ∆Eact are quite different.
Previous work using substituent effects to probe the nature
of the transition state for â-hydrogen elimination examined the
reaction in substituted ethoxy and n-propyl groups bound to
* Author to whom correspondence should be addressed.
† Current address: Department of Materials Science, University of
Delaware, Newark, DE 19716.
The fluorine substituent effect on the ∆Eact for â-hydrogen
elimination has been rationalized in terms of a hydride-like
‡ Current address: Department of Chemistry, University of Alabama,
Tuscaloosa, AL 35487.
10.1021/jp993474x CCC: $19.00 © 2000 American Chemical Society
Published on Web 02/23/2000

Transition State for â-Elimination of Hydrogen
J. Phys. Chem. A, Vol. 104, No. 11, 2000 2477
transition state [RCδ+‚‚‚Hδ-]‡, in which the carbon atom is
electron deficient. The inductive effect of the fluorine in the
methyl groups serves to energetically destabilize such a transi-
tion state. The implication of this interpretation was that the
electron deficient nature of the carbon atom resulted from charge
compensation with the hydrogen leaving group. Within that
model for the transition state it was not clear why the
perfluorination of methyl has a greater effect in the case of an
ethoxy group (∆∆Eact ) 55 kJ mol-1) than in the case of
n-propyl (∆∆Eact ) 35 kJ mol-1). The observations made in
the work reported here rationalize this effect.
phase alcohol dehydrogenation becomes apparent only when
suitable rotamer structures are chosen for the fluorinated
alcohols. The favored geometry of gaseous trifluoroethanol, for
instance, brings the O-H group close to one of the fluorine
atoms, with a hydrogen-fluorine distance of 2.56 Å.9 The
analogous structure does not seem realistic for a trifluoroethoxy
group adsorbed on a surface. Imposition of appropriate con-
formational constraints on the gas-phase ethanols reveals a
correlation between the ∆Hrxn for gas-phase dehydrogenation
and the ∆Eact for â-hydrogen elimination in ethoxy groups. That
insight provides a deeper understanding of the nature of the
transition state for this extremely important elementary surface
reaction.
One intriguing aspect of the measured values of ∆Eact for
â-hydrogen elimination is that they are more sensitive to fluorine
substituent effects than the thermochemical energies (∆Hrxn) of
ethanol and propane dehydrogenation, eqs 3 and 4. Unfortu-
nately, experimental ∆Hrxn for dehydrogenation are not available
for all the fluorine-substituted ethanols and propanes that need
be considered if one wishes to make direct comparison of ∆Hrxn
to the ∆Eact measured for â-hydrogen elimination in the
corresponding ethoxy and propyl groups. Some relevant num-
bers are available. If the published theoretical prediction of the
heat of formation (at 298 K) of gaseous trifluoroacetaldehyde6
is combined with the experimental ∆H0f for trifluoroethanol,7
the estimated heat of reaction for the dehydrogenation of
trifluoroethanol to trifluoroacetaldehyde is ∆Hrxn ) 114 kJ
mol-1. This is substantially greater than the experimental value
for the heat of ethanol dehydrogenation to acetaldehyde, ∆Hrxn
) 69 kJ mol-1.7 Those data lead to a ratio of ∆∆Eact/∆∆Hrxn
) 1.2 for the dehydrogenation of ethanols. For the case of
trifluoropropane dehydrogenation (eq 4), one can obtain the
∆Hrxn ) 143 kJ mol-1 from an estimated ∆H0f of 1,1,1-
trifluoropropane (-757 kJ mol-1)8 and the experimental ∆H0f
for 1,1,1-trifluoropropene.7 The experimental value for the
Complementing the investigations that probe the electronic
structure of the transition state for â-hydrogen elimination,
published studies also shed light on the structure of the adsorbed
reactant states. The structure of the ethoxy reactant has been
determined using Fourier transform infrared reflection-absorp-
tion spectroscopy (FT-IRAS).10 On the Cu(111) surface, ethoxy
is oriented with its C-C bond roughly parallel to the plane of
the surface and the O-C-C plane of the molecule tilted by
approximately 20° with respect to the surface normal. This
structure looks like that proposed for ethoxy on the Ni(111)
surface.11 More importantly, it bears a great similarity to that
proposed for trifluoroethoxy on the Cu(111) surface.12 The fact
that adsorbed ethoxy and trifluoroethoxy both have nearly the
same orientations suggests that differences in reactant geometry
are not responsible for the effects of fluorination on the ∆Eact
to â-hydrogen elimination. Furthermore, the observation that
the methylene stretch modes have similar frequencies in both
ethoxy and trifluoroethoxy indicates that the C-H bond strength
is not influenced significantly by fluorination.10,12
Steric effects on the transition state for â-hydrogen elimina-
tion have been probed by a study of the kinetics in a series of
cyclic alkyl groups on the Cu(100) surface.13 The ∆Eact to
â-hydrogen elimination in cyclohexyl groups is observed to be
23 kJ mol-1 higher than in either cyclopentyl or cycloheptyl
groups. A number of arguments have been advanced to attribute
this primarily to the differences in the energies needed to achieve
a planar configuration of the Cu-C-C-H bonds in the
adsorbed alkyl. Both cyclopentyl and cycloheptyl groups have
stable structures, which are very close in energy to this
configuration. In the cyclohexyl group, which can adopt highly
nonplanar “twist-boat” or “chair” conformations, the energy
needed to achieve planarity is approximately 23 kJ/mol and
would account entirely for the differences observed in the
barriers to â-hydrogen elimination. The planar structure of the
transition state is also predicted by computational models.14
dehydrogenation of propane is ∆Hrxn ) 124 kJ mol-1 7
.
Comparing these to the measured ∆Eact for â-hydrogen elimina-
tion in propyl and trifluoropropyl groups on the Cu(111) surface
2
leads to a ratio of ∆∆Eact/∆∆Hrxn ) 1.8. Since experiment
has shown that fluorination of the methyl does not greatly
perturb the binding energies of ethoxy groups to metal surfaces,1
the fact that these ∆∆Eact/∆∆Hrxn ratios are g1 cannot be
attributed to reactant state stabilization. In other words, the
transition state energies for â-hydrogen elimination appear to
be more greatly influenced by fluorine substitution than the net
∆Hrxn of eqs 3 and 4. Since the ∆∆Eact/∆∆Hrxn ratios are based
on numbers obtained from widely disparate estimation proce-
dures, it is not easy to evaluate their quantitative significance.
Consequently, this paper explores the relationship between
experimental ∆Eact and estimates of ∆Hrxn based on a consistent
set of DFT calculations. We find that ∆Eact varies roughly
linearly with ∆Hrxn as fluorine is added to ethoxy groups, with
nearly the same proportionality constant as for propyl groups
(once the appropriate molecular conformations are taken into
account). The ∆∆Eact/∆∆Hrxn ratios are much closer to one
another than the above estimates would suggest, which in turn
implies that the apparent differences between â-hydrogen
elimination in alkoxy and alkyl groups mask an underlying
parallelism in substituent effects.
This paper provides additional experimental results that
address both the mechanism and the nature of the transition
state for â-hydrogen elimination. Studies of deuterium isotope
effects reported here confirm that the mechanism does not
change in going from ethoxy to trifluoroethoxy and that the
effects of fluorination on the barrier to â-hydrogen elimination
do not exhibit significant contributions that can be attributed to
the mass or steric bulk of fluorine.
Additional measurements enlarge the set of alkoxy groups
for which the ∆Eact has been measured. â-Hydrogen elimination
from adsorbed isopropoxy groups to form ketones (eq 5)
increases the variety of substituents directly attached to the
carbinol carbon. With this expanded set of compounds, it has
been possible to correlate ∆Eact with Hammett substituent
constants (both σF and σp)15 as well as with gas phase ∆Hrxn
for alcohol dehydrogenation. This approach demonstrates an
RCH2OH f RCHdO + H2
R ) CF3, CHF2, CFH2, CH3 (3)
RCH2CH3 f RCHdCH2 + H2 R ) CF3, CH3 (4)
The correspondence between fluorine substituent effects on
∆Eact for â-hydrogen elimination in alcohols and ∆Hrxn for gas-

2478 J. Phys. Chem. A, Vol. 104, No. 11, 2000
Gellman et al.
empirical means of quantifying the effects of substituents that
in the future can be applied to a wider variety of surface
reactions.
alkoxy to aldehyde. The kinetics of the â-hydrogen elimination
reaction on the Cu(111) surface are relatively insensitive to
alkoxy coverage making minor variations unimportant.1
The quadrupole mass spectrometer used for desorption
measurements was shrouded by a stainless steel tube with a
stainless steel cone as an aperture to the ionizer. The diameter
of the aperture was slightly smaller than the diameter of the
Cu(111) sample (∼1 cm). Desorption measurements were made
by positioning the sample 2-3 mm from the front of the aperture
and then heating the sample while monitoring ion fragments,
usually the parent aldehyde, with the mass spectrometer. Heating
was controlled by computer, using a proportional/differential
feedback routine to maintain a constant heating rate.
The alcohols used in this work were CH3CH2OH, CH3CD2-
OH, CF3CH2OH, CF3CD2OH, CH3CHDOH, (CF3)2CHOH, CF3-
CH(OH)CH3, and (CH3)2CHOH. The CH3CHDOH was syn-
thesized by reducing acetaldehyde with LiAlD4 in ether,
removing the solvent, quenching the reaction with excess
dodecanol, and distilling ethanol-d1 to afford a product that was
g99 at. % D. Other materials were purchased from commercial
sources. All compounds were degassed by several cycles of
freezing, pumping, and thawing before introduction into the
vacuum chamber, where their purity was checked by mass
spectrometry.
R′RCHO-Cu f R′RCHdO(ad) + Cu-H
R,R′ ) CF3, CH3 (5)
Experimental Section
The experiments described in this paper were performed in
a stainless steel ultrahigh vacuum (UHV) chamber. A base
pressure of <10-10 Torr is achieved by means of a cryopump
and a titanium sublimation pump. The chamber is equipped with
an Ar+ sputter gun, an electron gun, a retarding field analyzer
for Auger electron spectroscopy, and a quadrupole mass
spectrometer for both background gas analysis and temperature-
programmed reaction spectroscopy (TPRS). The chamber has
leak valves for backfilling with gas or vapor and a capillary
array gas doser for line-of-sight exposure of the Cu(111) surface
to vapor. Exposures are reported in Langmuirs (1 L ) 1 × 10-6
Torr s) and are not corrected for differences in ion gauge
sensitivity. It should be noted that for the gases used in this
work the multiplication factor when using the capillary array is
∼20.
The Cu(111) single crystal was obtained from Monocrystals
Inc. It was mounted on a sample holder by spot-welding between
two tantalum wires. The sample holder was attached to a
manipulator allowing precise x-, y-, and z-axis travel and 360°
rotation of the crystal inside the chamber. Once mounted on
the manipulator the sample was in mechanical contact with a
liquid nitrogen reservoir. The crystal could be cooled to T <
100 K and resistively heated to T > 1000 K. Sample temperature
measurement was made using a chromel-alumel thermocouple
spot-welded to the edge of the crystal. During heating the sample
temperature was controlled by a computer.
Cleaning of the Cu(111) surface followed standard proce-
dures. The surface was cleaned by cycles of heating to 1000 K
while Ar+ ion sputtering and by exposure to large doses (∼100
L) of oxygen at 800 K to remove carbon. After each experiment
involving the formation and reaction of adsorbed alkoxy the
crystal was annealed to 1000 K to dissolve any carbon left by
alkoxy decomposition into the bulk of the crystal. The crystal
surface was judged to be clean if no impurities were detected
by AES.
Computational
Experimental thermodynamic values for dehydrogenation of
fluorinated alkanes and alcohols have not been reported, so they
were computed using G3 and DFT methods. The gas-phase heats
of formation for unfluorinated alcohols and aldehydes agree well
with G3 values.16 We also note that ∆∆Hrxn for dehydrogenation
of trifluoropropane (whose heat of formation has not been
determined experimentally and must be estimated by group
equivalents) relative to propane is not badly fit by ab initio
calculations.16
In the course of these computations, it was necessary to
ascertain the most stable conformations of the alcohols and their
corresponding aldehydes. Calculation of the geometry reveals
that ethanol prefers an anti geometry, in agreement with
experiment. The favored conformations of the fluoroethanols
all have oxygen synclinal with a fluorine, again in conformity
with experiment, with the O-H hydrogen directed toward a
fluorine.17-19
Adsorbed alkoxy groups were formed by exposing the clean
Cu(111) surface to high background exposures of oxygen with
the crystal at 600 K followed by exposure to the corresponding
alcohol through the capillary array with the crystal at 250 K.
Holding the crystal at 250 K during exposure to the alcohol
allowed the reaction of the alcohol with the adsorbed oxygen
to form the alkoxy while the coproduct, water, desorbed.
Saturation of the surface with a given alkoxy species was
achieved when the aldehyde yield in the subsequent TPRS
measurement was maximized. The alcohol exposure in each case
was greater than that required to react with the available
adsorbed oxygen (usually 0.1 L of the alcohol when dosing
through the capillary array) so that the oxygen exposure was
the factor limiting the amount of alkoxy that was formed on
the surface. The maximum yields of the aldehydes occurred for
oxygen exposures in the range 30-70 L (uncorrected back-
ground). The amount of oxygen required to routinely achieve
the maximum yield of aldehyde was dependent on the initial
cleanliness of the surface and bulk of the crystal. Some of the
oxygen scavenged adsorbed carbon which was left behind as a
minor decomposition product by prior reaction of adsorbed
Results
Isotope Effects on â-Hydrogen Elimination. Previous work
to determine the nature of the transition state for â-hydrogen
elimination has compared the values of ∆Eact in a set of
fluorinated and unfluorinated ethoxy groups on Cu and Ag
surfaces.1 This comparison is meaningful only if all of the
ethoxy groups react by the same mechanism and via the same
transition state. Thus, this investigation has examined deuterium
kinetic isotope effects (DKIE) in order to test whether the TPR
spectra do in fact measure the kinetics of the same rate-limiting
step for both fluorinated and unfluorinated ethoxy groups. The
observation of the same primary DKIE in the TPR spectra of
â-hydrogen elimination in both ethoxy and trifluoroethoxy
provides strong evidence that both react via the same mecha-
nism.
To observe the DKIE for â-hydrogen elimination on the
Cu(111) surface, the TPR spectra shown in Figure 1 were
obtained for CH3CH2O(ad), CH3CD2O(ad), CF3CH2O(ad), and
CF3CD2O(ad). These spectra were obtained by starting with
saturated coverages of the ethoxy groups, as described previ-

Transition State for â-Elimination of Hydrogen
J. Phys. Chem. A, Vol. 104, No. 11, 2000 2479
Figure 1. TPRS of CH3CH2O(ad), CH3CD2O(ad), CF3CH2O(ad), and
CF3CD2O(ad) on the Cu(111) surface. The spectra were obtained by
monitoring the parent ion of the acetaldehyde products generated by
â-hydrogen elimination in the ethoxides. The peak desorption temper-
atures increase by 19 K on deuteration of the C1 carbon atom in both
ethoxy and trifluoroethoxy. The fact that a deuterium kinetic isotope
effect is observed in both cases indicates that â-hydrogen elimination
is rate limiting in both cases. Heating rate â ) 2 K/s.
Figure 2. TPRS for CH3CHDO(ad) on the Cu(111) surface. The spectra
were obtained by monitoring the parent ion of acetaldehyde products.
The ratio of the spectra of CH3CDdO to CH3CHdO is a measure of
the deuterium kinetic isotope effect. The ratio of CH3DdO to
CH3CHdO gives kHD/kDH ) 3.6 at 395 K. Heating rate â ) 2 K/s.
the mathematical relation ∆∆Eact ) RT ln(kH,H/kD,D), provided
that we also assume that the preexponential factors are
unchanged by isotopic labeling. This gives an estimate of ∆∆Eact
) 4.0 ( 0.7 kJ mol-1 for the ethoxy groups. That value is
consistent with an isotope effect, which results from differences
in the zero-point energy levels of the νs(CD2) and νs(CH2)
vibrational modes of the adsorbed ethoxy groups. The frequen-
cies of these modes can be obtained from the literature10 and
are νs(CH2) ) 2857 cm-1 and νs(CD2) ) 2085 cm-1. These
frequencies yield a predicted difference in the zero-point
energies of ∆∆Eact ) 4.8 kJ mol-1, in reasonable agreement
with the value determined experimentally.
The DKIE has also been determined for â-hydrogen elimina-
tion in CF3CH2O(ad) and CF3CD2O(ad). The ratio of rate constants
is kH,H/kD,D ) 2.9 ( 0.6 at 493 K, which yields a value of ∆∆Eact
) 4.4 ( 0.8 kJ mol-1. As with the nonfluorinated ethoxy group,
this is consistent with the difference in zero-point energy levels
of the νS(CH2) and νS(CD2) vibrational modes. The frequencies
of the vibrational modes reported in the literature12 are νS(CH2)
) 2937 cm-1 and νS(CD2) ) 2087 cm-1, which yield an
estimate of the difference in the zero point-energies of ∆∆Eact
) 5.3 kJ mol-1. The presence of this kinetic isotope effect in
both ethoxy and trifluoroethoxy and the approximately equal
value of the zero-point energy differences clearly identifies
â-hydrogen elimination as the rate-limiting step in the reactions
of both. This is an important point since it justifies the
comparison of the ∆Eact for â-hydrogen elimination in hydro-
carbon and fluorocarbon alkoxy groups as a means of probing
the transition state for this reaction.
ously, and then monitoring acetaldehyde or trifluoroacetaldehyde
desorption during heating of the surface at a rate of 2 K/s.
Because the rate constant for desorption of acetaldehyde is much
higher than the rate constant for â-hydrogen elimination, these
TPR spectra provide a measure of the â-hydrogen elimination
reaction kinetics.20 It is immediately evident that deuteration
of the carbon atom at the â-position with respect to the surface
increases the temperature at which the reaction occurs. This
temperature shift is approximately 19 K for both ethoxy and
trifluoroethoxy. The isotope effect identifies â-hydrogen elimi-
nation as a kinetically significant step in the aldehyde produc-
tion.
The DKIE can be quantified by determining the ratio of the
rate constants for â-hydrogen elimination in the deuterium
labeled and in the unlabeled ethoxy groups. By comparing the
rates of acetaldehyde desorption from the TPR spectra of
CH3CH2O(ad) and CH3CD2O(ad), one finds that kH,H/kD,D ) 3.4
(0.6 at 390 K. In this notation, the first subscript on the rate
constant refers to the atom that is removed during â-hydrogen
elimination and the second subscript refers to the atom remaining
on the â-carbon atom. The temperature at which the reaction
rate reaches its maximum is 390K for CH3CH2O(ad). This value
of kH,H/kD,D was determined by a technique referred to as the
“direct method” by Madix et al.21 Briefly, kH,H/kD,D ) (rH/θH)-
(θD/rD), where rH is the rate of â-H elimination and rD is the
rate of â-D elimination. At a given temperature, these are
proportional to the amplitudes of the TPR spectra. The quantities
θH and θD are the coverages of the corresponding alkoxy groups
and are determined by integrating the TPR spectra. The
advantage of this “direct method” is that absolute values of the
activation energies and pre-exponential factors for the reaction
need not be determined. The DKIE can be determined anywhere
in the temperature range over which the two TPR spectra
overlap. Assuming that the ∆Eact do not depend on coverage,
the difference in barrier heights can be calculated by means of
In an independent set of experiments, the DKIE was
determined by measuring the relative rate constants for â-H and
â-D elimination in CH3CHDO(ad) on the Cu(111) surface. The
TPR spectra showing the desorption of CH3CDdO and
CH3CHdO during reaction of a saturation coverage of
CH3CHDO(ad) are shown in Figure 2. The ratio of the yields of
CH3CDdO and CH3CHdO produced by this reaction is a direct
measure of the primary kinetic isotope effect. The first fact to

2480 J. Phys. Chem. A, Vol. 104, No. 11, 2000
Gellman et al.
TABLE 1: Kinetic Parameters and Substituent Constants,
σF, for â-Hydride Elimination on the Cu(111) Surfacea
adsorbed
reactant
substit-
uents
log10
ν
)
∆Eact
σp
∑σp
σF ∑σF (s-1
(kJ mol-1
)
CH3O
H
H
H
CH3
H
CFH2
H
CF2H
H
CF3
CH3
CH3
0.0
0.0
0.0 0.0 0.0
0.0
16.2
136
121
138
146
176
114
174
264
CH3CH2O
CFH2CH2O
CF2HCH2O
CF3CH2O
0.0 -0.17 0.0 0.0
-0.17 0.0
15.2
0.0
0.11
0.0
0.29
0.0
0.54
0.11 0.0 0.22 16.0
0.22
0.29 0.0 0.36 16.2
0.36
0.54 0.0 0.44 17.9
0.44
(CH3)2CHO
-0.17 -0.34 0.0 0.0
-0.17 0.0
-0.17 0.37 0.0 0.44 17.8
0.54 0.44
0.54 1.08 0.44 0.88 23.3
0.54 0.44
14.3
(CF3)CH(CH3)O CH3
CF3
(CF3)2CHO
CF3
CF3
a The kinetic parameters were determined by fitting variable heating
rate TPRS data to the model; rate ) ν exp(-∆Eact/RT)θ, where θ is
the alkoxy coverage. Addition of fluorine to the methyl substituent
increases the ∆Eact
.
Figure 3. TPRS for CH3CH2O(ad), CD3CH2O(ad), CD3CH2O(ad), and
CD3CD2O(ad) on the Cu(111) surface. The spectra were obtained by
monitoring the parent ion of the acetaldehyde products. There is no
observable effect of deuteration in the methyl group indicating that
there is no secondary isotope effect. Heating rate â ) 2 K/s.
note is that more CH3CDdO desorbs than CH3CHdO, thus
the rate of C-H bond breaking is greater than the rate of C-D
bond breaking. These TPRS curves both reach a maximum at
a temperature of 395 K for a heating rate of 2 K/s. By this
method, kH,D/kD,H ) 3.6 at 395 K and implies that ∆∆Eact
)
4.2 ( 0.8 kJ mol-1, which is in close agreement with the values
calculated above.
During â-elimination in CH3CHDO(ad) the isotope (H or D)
that is left on the carbon atom depends on which isotope is
eliminated (D or H). When there are isotopic differences
between atoms bound to the reaction center (other than the atom
that is involved in the reaction), these differences can give rise
to secondary isotope effects. The only way in which the use of
CH3CHDO(ad) and the comparison between CH3CH2O(ad) and
CH3CD2O(ad) can yield equal primary isotope effects is if these
secondary isotope effects are negligible, which is clearly the
case. A previous study of secondary isotope effects on â-hy-
drogen elimination in methoxy groups on the Cu(110) surface
also showed that secondary effects are very small.22
Secondary isotope effects resulting from deuterium substitu-
tion in the methyl group of ethoxy can be identified by
comparing the kinetics of â-hydrogen elimination between
CH3CH2O(ad) and CD3CH2O(ad) and between CH3CD2O(ad) and
CD3CD2O(ad). Figure 3 shows TPR spectra of the desorption of
acetaldehyde produced from â-hydrogen elimination of these
ethoxy groups. These spectra were collected at saturation
coverage of the ethoxy and at a heating rate of 2 K/s. The
absence of any change in peak temperature due to methyl group
deuteration indicates clearly that the secondary isotope effects
from substitution vicinal to the reacting center are immeasurable.
This is relevant to our comparison of â-hydrogen elimination
in substituted alkoxy groups, since it indicates that effects purely
due to atomic mass (mF ) 19 versus mH ) 1) can be neglected.
Electronic Substituent Effects on â-Hydrogen Elimination.
The ∆Eact for â-hydrogen elimination in several alkoxy and
fluorinated alkoxy groups were measured in order to probe the
nature of the transition state. The use of highly electronegative
Figure 4. TPRS of (CH3)2CHO(ad)
, (CF3)(CH3)CHO(ad), and
(CF3)2CHO(ad) on the Cu(111) surface. The spectra were obtained by
monitoring the parent ion of the acetone products generated by
â-hydrogen elimination in the reactant alkoxides. Fluorination sub-
stantially increases the peak desorption temperature indicating an
increase in the barrier to â-hydrogen elimination. Heating rate â ) 2
K/s.
fluorine as a substituent can have a dramatic effect on the ∆Eact
if there is a difference in the charge density at the reaction center
on going from the reactant to the transition state. The work
reported here expands the set of alkoxy groups for which the
∆Eact to â-hydrogen elimination has been measured on the Cu-
(111) surface. The previously measured values of ∆Eact for
methoxy and ethoxy groups are listed in Table 1. In this work
the ∆Eact for â-hydrogen elimination in isopropoxy groups, eq
5, have also been measured. The TPR spectra of acetone and
fluorinated acetones produced by â-hydrogen elimination in
these isopropoxy groups are shown in Figure 4. These were
collected at saturation coverage of the isopropoxy and using a
heating rate of 2 K/s. Each of the trifluoromethyl groups

Transition State for â-Elimination of Hydrogen
J. Phys. Chem. A, Vol. 104, No. 11, 2000 2481
son with the effects of fluorine substitution on the ∆Eact for
â-hydrogen elimination in propyl groups on the Cu(111)
surface.2
RCH2O-Li f RCHdO + Li-H
(6)
The ∆Hrxn for dehydrogenation of alcohols and alkanes given
in Table 2 were estimated using G3 and DFT calculations and
compared to values obtained from thermochemical tables.7 G3
theory is reported to give excellent values for the ∆H0f,298 of
ethanol, 2-propanol, acetaldehyde, and acetone,16 but we find
that the ∆H0f,298 for trifluoroethanol predicted by G3 theory
(859 kJ mol-1) does not agree well with the reported experi-
mental value (888 kJ mol-1).7 Nor does the G3 prediction of
∆H0f,298 for trifluoroacetaldehyde agree well with the theoreti-
cal value calculated using BAC-MP4 theory.23 Nevertheless the
G3 value of ∆Hrxn ) 112 kJ mol-1 for trifluoroethanol
dehydrogenation is in good agreement with the value of ∆Hrxn
) 114 kJ mol-1 predicted from the experimental ∆H0f,298 for
trifluoroethanol and the theoretical BAC-MP4 ∆H0f,298 for
trifluoroacetaldehyde. We need only the relative values, ∆Hrxn
,
for heats of dehydrogenation of the fluorinated alcohols. Hence,
although the ∆H0f,298 for the fluorinated alcohols are not as
close to the experimental values as one might like, the ∆Hrxn
predicted by G3 theory are deemed to be sufficiently accurate
for the purposes of this work.
2
Figure 5. Plots of ln(Tp /â) versus 1/Tp for the TPR spectra of
isopropoxy groups on the Cu(111) surface (Tp, peak desorption
temperature; â, heating rate). The heating rate was varied from â )
0.2 to 10 K/s. The slopes give the activation barriers for â-hydrogen
elimination.
The most important observation of the calculations is that
∆Hrxn for alcohol dehydrogenation increases systematically with
the degree of fluorination of the methyl groups. This agrees
with the observed effect of fluorination on the experimental
∆Eact for â-hydrogen elimination in alkoxy groups on the
Cu(111) surface. For purposes of establishing a correlation
between the ∆Hrxn and the ∆Eact, we make use of electronic
energy differences (∆Eel) between reactants and products of
alcohol dehydrogenation calculated using DFT. We neglect
contributions from zero-point energy changes because, as Table
2 summarizes, they are nearly the same for all the reactions
under consideration. For similar reasons, we do not include
corrections for basis set superposition error. While the absolute
values of ∆Eel predicted by DFT are substantially higher than
the G3 values for ∆Hrxn, a plot of one versus the other gives a
straight line (slope ) 1.05, r2 ) 0.98). Thus, we believe that
estimating ∆∆Hrxn by using the DFT values for ∆∆Eel is just
as suitable as using values predicted by G3 theory for exploring
the relationship between thermodynamics of alcohol dehydro-
genation and the ∆Eact to â-hydrogen elimination in alkoxy
groups.
increases the temperature of the product desorption peak in the
TPR spectra by roughly 100 K. Clearly the fluorination of the
methyl groups of isopropoxy causes a substantial increase in
the ∆Eact for â-hydrogen elimination.
To estimate the ∆Eact for â-hydrogen elimination, TPR spectra
have been obtained for each of the isopropoxy groups at heating
rates in the range 0.2-10 K/s. These variable heating rate spectra
can be analyzed to give the ∆Eact without the need to assume
a value for the pre-exponential factor in the rate constant for
2
â-hydrogen elimination. Figure 5 shows plots of ln(Tp /â) versus
1/Tp for the three isopropoxy groups, where â symbolizes the
heating rate. The values of ∆Eact are determined from the slopes
of these plots and are listed in Table 1 with the pre-exponential
factors. The effects of fluorine substitution on the ∆Eact to
â-hydrogen elimination of the secondary alkoxy groups agree
qualitatively with those observed in the ethoxy species, increas-
ing systematically with the degree of fluorine substitution of
the methyl groups.
Electronic Substituent Effects on Alcohol Dehydrogena-
tion. To understand the effects of fluorine substitution on the
energetics of the transition state for â-hydrogen elimination, it
has been instructive to compare this reaction (eq 1) to
comparable reactions in the gas phase. â-Hydrogen elimination
(eq 1) and alcohol dehydrogenation (eq 3) are analogous in the
sense that both produce the same unsaturated products, but in
one case the hydrogen atom forms Cu-H while in the other
case it forms H2. Since very few experimental values have been
reported for the dehydrogenation energies (∆Hrxn) of the
fluorine-substituted alcohols in eq 3 they have been calculated
here for a variety of reactant conformations. In addition
electronic energy changes (∆Eel) have been calculated for the
gas-phase elimination of lithium hydride from lithium ethoxides,
eq 6. Values of ∆Hrxn for dehydrogenation of propane and
trifluoropropane have also been calculated, to make a compari-
The values of ∆Hrxn for alcohol dehydrogenation depend on
the conformations of the reactants and the products. It is quite
possible that the most stable gas-phase conformations of the
fluorinated alcohols are not analogous to the conformations
adopted by the fluorinated alkoxy groups on the Cu(111) surface.
When the most stable conformers of the substituted ethanols
are used as reactants a plot of experimental ∆Eact for â-hydrogen
elimination versus calculated values of ∆Eel for alcohol dehy-
drogenation gives a poor correlation, regardless of whether G3
or DFT values are used for the thermochemistry. Likewise, ∆Eact
does not correlate well with ∆Eel calculated for gas-phase

2482 J. Phys. Chem. A, Vol. 104, No. 11, 2000
Gellman et al.
TABLE 2: Calculated Gas-Phase Endothermicities for Dehydrogenation of Alcohols and of Loss of LiH from Lithium
Alkoxides (All Energies in kJ mol-1
)
density functional theory
(B3LYP/6-311G**)
G3 theorya
298
el
Y)Li
el
reaction
∆H
∆E
∆E
BSSEb
∆ZPEc
Y)H
Y)H
CH3OY f CH2O
CH3CH2OY f CH3CHO
FCH2CH2OY f FCH2CHO
88
252
232
276
111
83
8
8
8
-38
-38
-40
67a
87
106
(251 anti)d
292
(96 anti)d
122
F2CHCH2OY f F2CHCHO
CF3CH2OY f CF3CHO
103
112
55a
8
8
-38
-38
(280 anti)d
299
(112 anti)d
129
(299 anti)d
211
(119 anti)d
68
(CH3)2CHOY f (CH3)2CO
CF3(CH3)CHOY f CF3(CH3)2CO
7
11
-38
-37
273
106
(273 anti)d
336
(96 anti)d,e
135
(CF3)2CHOY f (CF3)2CO
CH3CH2CH3 f CH3CHdCH2
CF3CH2CH3 f CF3CHdCH2
11
5
7
-36
-36
-35
124
146
150
169
a Reference 16. b Basis set superposition error estimated by counterpoise. c Unscaled. d Refers to a geometry in which the O-Y bond is antiperiplanar
to a fluorinated methyl group and a C-F bond is antiperiplanar to the oxygen. e A 10.52 kJ/mol electronic energy difference between synclinal and
antiperiplanar conformations, as compared to a 10.95 kJ/mol electronic energy difference calculated for geometries optimized at B3LYP/6-31+G*
(Schaal, H.; Haeber, T.; Suhm, M. A. J. Phys. Chem. A 2000, 104, 265-274).
elimination of Li-H from substituted lithium alkoxides if the
most stable conformation of the lithium alkoxides are used, eq
6. However, both the fluorinated alcohols and the fluorinated
lithium alkoxides exhibit attractive intramolecular interactions
between fluorine and the substituent attached to oxygen (H in
the case of the alcohols and Li in the case of the lithium
alkoxides). If we use DFT to calculate ∆Eel for alcohol
dehydrogenation using the less stable conformational isomers
in which this interaction is absent, the correlation with the ∆Eact
remains poor for the lithium alkoxides but improves for the
alcohols. Similarly we find poor correlations of ∆Eact with ∆Eel
for the fluorinated 2-propanols and the lithium isopropanoxides
in their most favorable conformations, but ∆Eact displays an
excellent correlation with ∆Eel for dehydrogenation of the
2-propanols when reactant conformations have the O-H bond
antiperiplanar to a fluorinated methyl group. The two linear plots
are drawn in Figure 6. Note that the electronic energy difference,
∆Eel, for methanol dehydrogenation does not fall on either of
these lines.
Discussion
Isotope Effects and the â-Hydrogen Elimination Mecha-
nism. This work provides a deeper understanding of the nature
of the transition state for â-hydrogen elimination on metal
surfaces. The approach has been to use fluorine substitution in
alkoxy and alkyl groups on the Cu(111) surface as a means of
perturbing the reaction energetics. The effects of fluorine
substitution are then used to probe the nature of the transition
Figure 6. Correlation between ∆Eact for â-hydrogen elimination in
state. The measurements of deuterium isotope effects presented
in this paper serve to justify some of the assumptions implicit
alkoxides on Cu(111) to ∆Eel for alcohol dehydrogenation in the gas
phase. The alcohols have conformations in which the O-H bond is
in the use of substituent effects to study reaction energetics.
The isotopic ∆∆Eact values determined from intermolecular
DKIE measurements are 4.0 ( 0.7 kJ mol-1 for ethoxy and 4.4
( 0.8 kJ mol-1 for trifluoroethoxy. These do not differ
significantly from one another, supporting the conclusion that
fluorination of the methyl group in ethoxy does not alter the
reaction mechanism and that the â-hydrogen elimination step
is rate limiting in both species. Furthermore, in agreement with
previous measurements of â-hydrogen elimination in methoxy
species on the Cu(110) surface,22 we find no evidence for
secondary isotope effects. The net implication of these observa-
tions is to support our discussion below that describes the effects
antiperiplanar to a fluorinated methyl. The separate lines for ethoxy
groups and isopropoxy groups indicate that substitution at the carbinol
carbon affects the transition state of â-hydrogen elimination on the
surface but that within a given series the transition state is more sensitive
to fluorination than are the overall energetics. The data point corre-
sponding to methanol does not lie on either line.
of fluorination on the ∆Eact to â-hydrogen elimination in terms
of electronic effects which energetically destabilize the transition
state with respect to the reactant alkoxy species.
The Transition State for â-Hydrogen Elimination. The
nature of the transition state for â-hydrogen elimination proposed

Transition State for â-Elimination of Hydrogen
J. Phys. Chem. A, Vol. 104, No. 11, 2000 2483
Figure 7. One-dimensional potential energy diagram showing the
effects of fluorination on the ∆Eact to â-hydrogen elimination in
ethoxides on the Cu(111) surface. The left-hand side of the diagram
illustrates the energetics for formation of ethoxy groups from gas-phase
alcohols and the subsequent reaction via the transition state to
â-hydrogen elimination. At the right are the energetics for the gas-
phase dehydrogenation of the ethanols in their trans antiperiplanar
conformations to acetaldehydes. The effects of fluorination map ∆∆Eact
proportionately onto ∆∆Eel.
Figure 8. Linear free energy relationship correlating the barriers to
â-hydrogen elimination in adsorbed alkoxides with the Hammett σp
constants of the substituents attached to the carbinol carbon (cf. Table
1). The steep slope gives a Hammett reaction constant of F ) -26 (
3 that is consistent with substantial development of positive charge at
the C1 carbon in the transition state.
The comparison of the ∆Eact for a surface reaction with
theoretical estimates of ∆Eel for the analogous gas-phase
dehydrogenation reaction is only one approach to quantifying
the effects of substituents on surface reaction energetics. A
second approach that can be used to understand the effects of
substituents in a wide range of reactions is to examine linear
free energy relationships by means of empirically derived
substituent constants. Such substituent constants have been
tabulated through the study of a wide range of solution-phase
and gas-phase organic reactions.15 They reflect empirical
measures of a number of different properties of the substituents.
Table 1 lists the traditional Hammett substituent constants (σp
and σF) of the methyl and fluorinated methyl groups used as
substituents on the alkoxy groups studied in this work. These
have been used as measures of the properties of the substituents
and can be correlated with the ∆Eact, as illustrated in Figure 8.
The use of linear free energy relationships requires that
entropy changes scale with energy changes. Plots of the
logarithm of the frequency factor ν (log10 ν) versus ∆Eact give
good linear correlations (r2 > 0.99), so that criterion is met.
Figure 8 plots the ∆Eact for â-hydrogen elimination against the
sum of the substituent constants (∑σF) in a given alkoxy species
and reveals a strong linear correlation between the two. There
are a number of substituent constants to choose from. The field
substituent constant, σF, attempts to isolate the electrostatic field
effect of a substituent on nearby charge. The venerable Hammett
substituent constant, σp, is an empirical measure of combined
effects of resonance and electrostatic field of the substituent.
Both give equally good fits to the set of ∆Eact measured in this
work. Even better fits can be obtained by plotting linear free
on the basis of previous results was one in which the â-carbon
atom loses electron density due to the elimination of a hydride-
like species [RCδ+‚‚‚Hδ-]‡.1,2 In this type of transition state,
the â-carbon is electron deficient with respect to the reactant
ethoxide, and fluorination of the methyl group (R) increases
the ∆Eact by energetically destabilizing the charge distribution
in the transition state.1 The results presented in this paper provide
a deeper insight into the nature of the transition state. The
â-hydrogen elimination energetics are illustrated in the potential
energy diagram of Figure 7. The left-hand portion of the diagram
shows the energetics for the reaction of substituted ethanols to
form ethoxy groups followed by â-hydrogen elimination. The
right side of the diagram shows the energetics for the gas-phase
dehydrogenation of the ethanols to form acetaldehydes. Clearly,
the effects of fluorine on the energy of the dehydrogenation
reaction map onto the energies of the transition states for
â-hydrogen elimination. This is illustrated in a more quantitative
fashion by the plot in Figure 6 of ∆Eact versus calculated values
of ∆Eel.
A question arises as to the importance of screening by the
metal as a mechanism for influencing the magnitude of
substituent effects. Electrostatic calculations have been made
of the influence of conducting metal surfaces on substituent
effects for â-hydrogen elimination from ethoxy groups (where
the substituent is relatively far from the surface, >3 Å).24 These
indicate that the substituent effect on ∆Eact is not significantly
influenced by the conducting nature of the metal surface.

2484 J. Phys. Chem. A, Vol. 104, No. 11, 2000
Gellman et al.
energy relationships independently for the ∆Eact measured for
the ethoxy and isopropoxy groups. This might be expected from
the correlations of Figure 6 which revealed some differences
between the ethoxy and isopropoxy groups. When the data for
ethoxy and isopropoxy are plotted independently the correlation
using the σp constants is better than that obtained using just the
σF constants. The interpretation of Figure 8 is that the fluoro-
methyl groups interact by electrostatic field effects with a
â-carbon which is electron deficient in the transition state. The
slope of the linear free energy relationship is 150 ( 19 kJ mol-1
which converts into a reaction constant of F ) -26 ( 3. For
comparison a large value of F (∼12) was recently measured
for a gas-phase hydride transfer equilibrium reaction.25 This use
of substituent parameters points to a simple approach for
accounting for substituent effects that can be used to correlate
effects on surfaces with those observed in other environments
such as the gas phase or solution.
literature.20 ∆H5 is estimated from the hydrogen bond strength27
and the desorption energy of hydrogen.28 It has also been
reported by B. E. Bent and co-workers.29 These energies provide
an estimate of the heat of reaction for â-H elimination in ethoxy
of ∆Hâ-H > 41 ( 18 kJ mol-1. It has also been possible to
estimate the energetics for another relevant â-hydrogen elimina-
tion reaction, the conversion of propyl groups to adsorbed
propene by â-hydrogen elimination on the Cu(100) surface.28
In that case the heat of reaction of â-hydrogen elimination in
propyl groups has been estimated at ∆Hrxn ) 27 ( 15 kJ mol-1
.
The reaction energetics are probably quite similar on the Cu(111)
surface. Within the context of a discussion based on Hammond’s
postulate, it should be noted that the endothermicity of the
â-hydrogen elimination reaction in both the alkoxy and the alkyl
groups is consistent with our proposal that the transition state
occurs late in the reaction coordinate. It should be noted that
the substituent effects observed for â-hydrogen elimination in
the ethoxy groups on the Cu(111) surface have also been
observed with the same magnitude on the Cu(100) surface and
on the Ag(110) surface.1 Thus, it is quite likely that in all these
cases â-hydrogen elimination is endothermic and occurs with
a late transition state.
Reaction Energetics for â-Hydrogen Elimination. On the
basis of the arguments above, we propose that the transition
state for â-hydrogen elimination occurs late in the reaction
coordinate. Hammond’s postulate suggests that the transition
states for exothermic reactions should be reactant-like, while
the transition states for endothermic reactions should be product-
like.26 The overall reaction energetics for ethoxy conversion to
adsorbed acetaldehyde on the Cu(111) surface are not known.
However, the implication of the proposed late transition state
is that the reaction is endothermic. We can estimate the
energetics on the Cu(111) surface using the following hypotheti-
cal reaction path:
â-Hydrogen Elimination in Alkoxy versus Alkyl. The
deeper insight offered by these results regarding the nature of
the transition state for â-hydrogen elimination also offers a
plausible explanation for the differences between the substituent
effects observed in alkoxy and in alkyl groups on the Cu(111)
surface. The ∆Eact for â-hydrogen elimination in CH3CH2O(ad)
increases by ∆∆Eact ) 55 kJ mol-1 as a result of perfluorination
of the methyl group. However, perfluorination of the methyl
group in CH3CH2CH2,(ad) increases the ∆Eact for â-hydrogen
elimination by only ∆∆Eact ) 35 kJ/mol. These differences are
also reflected in the heats of the gas-phase dehydrogenation
reactions for alcohols and alkanes. DFT calculations predict that
perfluorination of the methyl group results in a ∆∆Eel ) 36 kJ
mol-1 for ethanols in their trans conformations but only 19 kJ/
mol in going from propane to trifluoropropane (the experimental
heat of hydrogenation of trifluoropropene is not available). The
ratio ∆∆Eact/∆∆Eel () 1.8) in going from propane to 1,1,1-
trifluoropropane is close to that observed on going from ethanol
to trans-1,1,1-trifluoroethanol (∆∆Eact/∆∆Eel ) 1.5). In our
view, this warrants the conclusion that the same electronic
effects operate in dehydrogenation of surface-bound alkyl groups
as in alkoxy groups. In the alkoxy case, the reaction results in
the formation of a CdO double bond from what originally was
a C-O single bond in the ethoxy reactant. In the alkyl case,
the reaction results in the formation of a CdC double bond
from what was a C-C single bond in the propyl reactant. The
substituent effect reflects differences in the charge density on
the â-carbon on going from reactant to product. The fact that
these ratios are ∆∆Eact/∆∆Eel > 1 is interesting. Apparently, a
greater partial positive charge develops in the transition state
than in the final product.
CH3CH2OH(g) f CH3CH2O(g) + H(g)
∆H1 ) 437.7 ( 3.4 kJ mol-1 (1)
CH3CH2O(g) f CH3CH2O(ad)
∆H2 < -264 kJ mol-1
CH3CH2O(ad) f CH3CHdO(ad) + H(ad)
∆Hâ-H
CH3CHdO(ad) f CH3CHdO(g)
(2)
(3)
∆H4 ) 56 ( 8 kJ mol-1
∆H5 ) 230 ( 16 kJ mol-1
∆H6 ) -436 kJ mol-1
(4)
(5)
(6)
H
(ad) f H(g)
2H(g) f H2,(g)
CH3CH2OH(g) f CH3CHdO(g) + H2,(g)
∆Hrxn ) 64.6 ( 1.6 kJ mol-1
The point of decomposing the thermodynamic cycle for the
dehydrogenation of ethanol in this manner is to estimate the
heat of reaction for step 3 (∆Hâ-H), â-hydrogen elimination on
the Cu(111) surface. ∆H1 and ∆H6 were obtained from the
appropriate bond strengths.27 The desorption energy of the
alkoxy (∆H2) must be greater than the barrier to â-hydrogen
elimination of (CF3)2CHO(ad) (otherwise it would desorb rather
than undergoing â-hydrogen elimination); therefore, ∆H2 must
be less than -264 kJ mol-1. This assumes that the alkoxy bond
strength to the metal is unaffected by fluorination of the methyl
groups. This is justified on the basis of measurements of the
relative heats of formation of ethoxy and trifluoroethoxy on the
Cu(111) surface from their corresponding alcohols. The de-
sorption energy of acetaldehyde, ∆H4, was found in the
Substitution at the Carbinol Carbon. The ratio ∆∆Eact/
∆∆Eel in going from 2-propanol to 1,1,1-trifluoro-2-propanol
() 2.2) is greater than for the substituted ethanols () 1.5). Given
that the plot for the isopropoxy groups in Figure 6 has a steeper
slope than for the ethoxy groups, it appears that a combination
of effects operate for isopropoxy groups. On one hand, a greater
partial positive charge develops in the dehydrogenation of
secondary alkoxy groups than in primary alkoxy groups. This
is consistent with the qualitative electronic effect expected with
increasing substitution at the reacting center. On the other hand,
the ratio ∆∆Eact/∆∆Eel in going from ethanol to 2-propanol is

Transition State for â-Elimination of Hydrogen
J. Phys. Chem. A, Vol. 104, No. 11, 2000 2485
0.5 and has a value comparable to the ratio in going from
methanol to ethanol (∆∆Eact/∆∆Eel ) 0.5). Apparently steric
effects attenuate the electronic effects of increasing alkyl
substitution at the carbinol carbon.
than is the overall thermochemistry. Finally, good linear
correlations are obtained between â-hydrogen elimination rates
and Hammett substituent parameters (σp or σF), which take into
account both resonance and electric field effects.
6. Conclusions
Acknowledgment. This paper is dedicated to Professor W.
A. Goddard III. This work was supported by NSF Grants CHE-
9701924 and CHE 9522604. M. T. Buelow was supported by
a fellowship from the Shell Foundation.
The â-hydrogen elimination reaction of fluorine-substituted
alkoxy groups on the Cu(111) surface yields the corresponding
aldehydes and ketones. Three relationships between kinetics and
thermochemistry have been explored: deuterium isotope effects,
fluorine substituent effects, and increasing alkyl substitution at
the carbinol center. Deuterium kinetic isotope effects are of
roughly the same magnitude for both ethoxy and trifluoroethoxy.
This indicates that the â-hydrogen elimination step is rate
limiting in both species. Experimental ∆∆Eact values scale
linearly with ∆∆Eel for gas-phase dehydrogenation for alcohols
in their trans antiperiplanar conformations, with a constant of
proportionality of >1. The same scaling is found when the distal
carbon of an n-propyl group is fluorinated, but experimental
∆Eact for unfluorinated alcohols do not give such a good linear
correlation with increasing methyl substitution at the reacting
center (i.e., methoxy f ethoxy f isopropoxy, where ∆∆Eact
scales with ∆∆Hrxn by a factor of e0.5).
References and Notes
(1) Gellman, A. J.; Dai, Q. J. Am. Chem. Soc. 1993, 115 (2), 714-
722.
(2) Forbes, J. G.; Gellman, A. J. J. Am. Chem. Soc. 1993, 115 (14),
6277-6283.
(3) Weldon, M. K.; Friend, C. M. Chem. ReV. 1996, 96 (4), 1391-
1412.
(4) Friend, C. M.; Xu, X. Annu. ReV. Phys. Chem. 1991, 42, 251-
278.
(5) Zaera, F. Chem. ReV. 1995, 95 (8), 2651-2693.
(6) Zachariah, M. R.; Westmoreland, P. R.; Burgess, D. R., Jr.; Tsang,
W.; Melius, C. F. J. Phys. Chem. 1996, 100, 8737-8747.
(7) (a) Domalski, E. A.; Hearing, E. D. J. Phys. Chem. Ref. Data 1993,
22, 805-1159. (b) Afeefy, H. Y.; Liebman, J. F.; Stein, S. E. Neutral
Thermochemical Data. In NIST Chemistry WebBook; Mallard, W. G.,
Linstrom, P. J., Eds.; NIST Standard Reference Database Number 69;
National Institute of Standards and Technology: Gaithersburg, MD,
November 1998 (http://webbook.nist.gov).
The correlations between kinetics and thermochemistry
exhibit a variety of trends. The observed kinetic isotope effects
indicate that dehydrogenation on the Cu(111) surface reaches
the top of the barrier at a point where the alkoxy C-H bond
has dissociated to a substantial extent (i.e., a late transition state).
The correlation between ∆Eact and calculated gas-phase ∆Eel
for dehydrogenation of fluorinated ethanols in their trans
antiperiplanar conformations contrasts with the poorer correla-
tions based on thermochemistry of their most stable conforma-
tions or on the thermochemistry calculated for elimination of
LiH from lithium ethoxides (regardless of conformation). This
suggests that on the surface the oxygen-metal bond is anti-
periplanar to a fluorinated carbon and that the oxygen is
antiperiplanar to one of the fluorines. This interpretation accords
with the experimental finding that fluorination does not greatly
perturb binding to the metal surface. The effect of fluorination
does not arise from stabilization of the adsorbed alkoxy reactant
but is instead a consequence of developing positive charge on
the â-carbon in the transition state [RCδ+‚‚‚H]‡. A similar effect
is observed for â-hydrogen elimination from adsorbed alkyl
groups, where the same type of charge buildup on carbon must
be taking place. The net result is that the ∆Eact are 1.5-2.2
times more sensitive to fluorine substitution than is the overall
∆Eel for the analogous gas-phase dehydrogenation reactions.
Alkyl substitution at the carbinol carbon has two opposing
effects. On one hand, it stabilizes the emerging positive charge
at the carbinol carbon atom in the transition state. On the other
hand, it creates steric hindrance at the reacting center. The net
result is that the ∆Eact are less sensitive to methyl substitution
(8) Joshi, R. M. J. Macromol. Sci. Chem. 1974, 8, 861-865.
(9) Xu, L.-H.; Fraser, G. T.; Lovas, F. J.; Suenram, R. D.; Gillies, G.
W.; Warner, H. E.; Gillies, J. Z. J. Chem. Phys. 1995, 103, 9541-9548.
(10) Street, S. C.; Gellman, A. J. J. Chem. Phys. 1996, 105 (16), 7158-
7170.
(11) Xu, J.; Zhang, X.; Zenobi, R.; Yoshinobu, J.; Xu, Z.; Yates, J. T.,
Jr. Surf. Sci. 1991, 256, 288-300.
(12) Street, S. C.; Gellman, A. J. J. Phys. Chem. 1996, 100 (20), 8338-
8348.
(13) Teplyakov, A. V.; Bent, B. E. J. Am. Chem. Soc. 1995, 117 (40),
10076-10087.
(14) Higashi, G. S.; Raghavachari, K.; Steigerwald, M. L. J. Vac. Sci.
Technol. B 1990, 8, 103-105.
(15) Hansch, C.; Leo, A.; Taft, R. W. Chem. ReV. 1991, 91 (2), 165-
195.
(16) Curtiss, L. A.; Raghavchari, K.; Redfern, P. C.; Rassolov, V.; Pople,
J. A. J. Chem. Phys. 1998, 109, 7764-7776.
(17) Hagen, K.; Hedberg, K. J. Am. Chem. Soc. 1973, 95, 8263-8266.
(18) Xu, L.-H.; Fraser, G. T.; Lovas, F. J.; Suenram, R. D.; Gillies, C.
W.; Warner, H. E.; Gillies, J. Z. J. Chem. Phys. 1995, 103, 9541-9548.
(19) Marstokk, K. M.; Møllendahl, H. Acta Chem. Scand. Ser. A 1980,
34, 765-770.
(20) Bowker, M.; Madix, R. J. Appl. Surf. Sci. 1981, 8, 299.
(21) Madix, R. J.; Telford, S. G. Surf. Sci. 1992, 277, 246-252.
(22) Madix, R. J.; Telford, S.G. Surf. Sci. 1995, 328, L576-L581.
(23) Zachariah, M. R.; Westmoreland, P. R.; Burgess, D. R., Jr.; Tsang,
W.; Melius, C. F. J. Phys. Chem. 1996, 100, 8737-8747.
(24) Buelow, M. T.; Gellman, A. J. J. Phys. Chem., in press.
(25) Tashiro, H.; Kikukawa, K.; Ikenaga, K.; Shimizu, N.; Mishima,
M. J. Chem. Soc., Perkin II 1998, 2435.
(26) Hammond, G. S. J. Am. Chem. Soc. 1955, 77, 334.
(27) Lide, D. R., Ed. CRC Handbook of Chemistry and Physics, 78th
ed.; CRC Press: Boca Raton, 1997.
(28) Anger, G.; Winkler, A.; Rendulic, K. D. Surf. Sci. 1989, 220, 1-17.
(29) Jenks, C. J.; Xi, M.; Yang, M. X.; Bent, B. E. J. Phys. Chem. 1994,
98 (8), 2152-2157.