Experimental and theoretical characterization of FSi(OCH3)2(OCH2)-: A gas phase fluoride-siloxirane adduct

Article abstract of DOI:10.1021/ja963382a

The structural characteristics and reactivity of the gas-phase FSi(OCH₃)₂(OCH₂)^- ion were investigated by a combination of ab initio calculations and FT-ICR techniques. The theoretical calculations for different possible structures reveal that carbanion and alkoxide ion type structures lead to ring closure upon geometry optimization to yield two different cyclic fluoride-siloxirane structures. The FSi(OCH₃)₂(cyc-OCH₂)^- ions containing the elusive siloxirane ring are estimated to be extremely stable with respect to F^- (69 kcal mol^-1) dissociation in agreement with earlier calculations on simpler systems. Experimentally, this ion is formed as a minor product (7%) in the gas-phase ion/molecule reaction of F^- with Si(OMe)₄ and is observed to undergo readily fluoride transfer to the parent neutral. This strongly suggests an ion with a structure corresponding to a fluoride adduct of a siloxirane species. Reaction of FSi(OCH₃)₂(OCH₂)^- with BF₃, hexafluorobenzene, and gas-phase acids more acidic than ethanol further suggests that this ion is capable of reacting as an alkoxide type nucleophile or base. This latter behavior has been associated with the possibility of ring opening of the siloxirane in the collision complex that mediates this ion/molecule reaction.

Full text of DOI:10.1021/ja963382a

1708
J. Am. Chem. Soc. 1997, 119, 1708-1716
Experimental and Theoretical Characterization of
FSi(OCH3)2(OCH2)-: A Gas Phase Fluoride-Siloxirane Adduct
Nelson H. Morgon,† Andre B. Argenton, Maria L. P. da Silva, and
Jose´ M. Riveros*
Contribution from the Institute of Chemistry, UniVersity of Sa˜o Paulo, Caixa Postal 26077,
Sa˜o Paulo, Brazil, CEP 05599-970
ReceiVed September 26, 1996X
Abstract: The structural characteristics and reactivity of the gas-phase FSi(OCH3)2(OCH2)- ion were investigated
by a combination of ab initio calculations and FT-ICR techniques. The theoretical calculations for different possible
structures reveal that carbanion and alkoxide ion type structures lead to ring closure upon geometry optimization to
yield two different cyclic fluoride-siloxirane structures. The FSi(OCH3)2(cyc-OCH2)- ions containing the elusive
siloxirane ring are estimated to be extremely stable with respect to F- (69 kcal mol-1) dissociation in agreement
with earlier calculations on simpler systems. Experimentally, this ion is formed as a minor product (7%) in the
gas-phase ion/molecule reaction of F- with Si(OMe)4 and is observed to undergo readily fluoride transfer to the
parent neutral. This strongly suggests an ion with a structure corresponding to a fluoride adduct of a siloxirane
species. Reaction of FSi(OCH3)2(OCH2)- with BF3, hexafluorobenzene, and gas-phase acids more acidic than ethanol
further suggests that this ion is capable of reacting as an alkoxide type nucleophile or base. This latter behavior has
been associated with the possibility of ring opening of the siloxirane in the collision complex that mediates this
ion/molecule reaction.
Siloxiranes (I) are among the simplest cyclic organosilane
systems which, unlike their well-known carbon analogs, have
eluded systematic chemical and structural characterization.
The thermodynamic stability of three-membered cyclic silanes
and of heterocyclopropanes in general is a subject of much
interest because of the presumed role of strain energy on the
enthalpy content.3 High-quality ab initio calculations predict
silacyclopropane to have a much higher strain energy than
cyclopropane, while negligible differences are computed for
higher cyclic organosilanes and cycloalkanes.4 On the other
hand, the successful spectroscopic identification of such exotic
species as silacyclopropyne,5 for which the strain energy is
estimated to be in the range of 100 kcal mol-1, suggests that
this effect alone may not be the only source of the apparent
difficulty in isolating siloxiranes.
(I)
Results from recent theoretical and experimental studies have
contributed to the revival of the discussion regarding the
potential stability of these three-membered moieties. For
example, the thermal decomposition of gaseous Si(OMe)4 has
been claimed to proceed initially by elimination of MeOH and
(MeO)2Si(cyc-OCH2).1 While this latter product defied isolation
in the high-temperature reaction, this fact was attributed to a
rearrangement of the siloxirane species to MeOSi(O)OEt.
The idea that siloxiranes may rearrange to more stable forms
finds support in molecular orbital calculations of model systems.
The extensive set of ab initio calculations carried out for simple
Si-O-H-C-containing molecules at the MP4(SDTQ)/6-31G**/
/HF/6-31G** level with empirical bond additivity parameter
corrections is particularly useful in deriving conclusions about
the thermochemistry of these compounds.1,2 This theoretical
approach has been shown to yield heats of formation in excellent
agreement with the few experimentally known values for Si-
containing compounds and provide a guideline for bond energies
in these systems. Calculations for siloxiranes such as (HO)2Si-
(cyc-OCH2) at this level do suggest that this species could be
prone to undergo interconversion to more stable structures as
claimed in ref 1.
Siloxirane type intermediates have long been recognized to
mediate molecular rearrangements leading to silicon migration
from carbon to oxygen,6 and Vice Versa.7
The mechanism involving ring closure for the transient anionic
intermediate is expected to be favored by the ability of Si to
form hypervalent species of the kind typically associated with
(3) (a) For a qualitative discussion on heterocyclopropanes, see: Driess,
M.; Gru¨tzmacher, H. Angew. Chem., Int. Ed. Engl. 1996, 35, 828. (b) For
thermochemical aspects of strain energy, see: Benson, S. W. Thermo-
chemical Kinetics, 2nd ed.; John Wiley & Sons, Inc.: New York, 1976; pp
60-63.
(4) Boatz, J. A.; Gordon, M. S.; Hildebrandt, R. L. J. Am. Chem. Soc.
1988, 110, 352.
(5) (a) Maier, G.; Reisenauer, H. P.; Pacl, H. Angew. Chem., Int. Ed.
Engl. 1994, 33, 1248. (b) Maier, G.; Pacl, H.; Reisenauer, H. P.; Meudt,
A.; Janoschek, R. J. Am. Chem. Soc. 1995, 117, 12712. (c) For a theoretical
description, see: Sherril, C. D.; Brandow, C. G.; Allen, W. D.; Schaefer,
J. F., III J. Am. Chem. Soc. 1996, 118, 7158.
(6) For reviews on these rearrangements and pertinent literature, see:
(a) Brook, A. G. Acc. Chem. Res. 1974, 7, 77. (b) Colvin, E. W. Silicon in
Organic Synthesis; Butterworths: London, 1981; Chapter 5.
(7) West, R.; Lowe, R.; Stewart, H. F.; Wright, A. J. Am. Chem. Soc.
1971, 93, 282.
† Present address: Institute of Chemistry, University of Campinas,
Campinas, SP, Brazil.
X Abstract published in AdVance ACS Abstracts, February 1, 1997.
(1) Chu, J. C. S.; Soller, R.; Lin, M. C.; Melius, C. F. J. Phys. Chem.
1995, 99, 663.
(2) (a) Ho, P.; Melius, C. F. J. Phys. Chem. 1995, 99, 2166. (b) For the
procedure used to develop empirical bond additivity corrections, see:
Allendorf, M. D.; Melius, C. F. J. Phys. Chem. 1992, 96, 428.
S0002-7863(96)03382-3 CCC: $14.00 © 1997 American Chemical Society

Characterization of FSi(OCH3)2(OCH2)-
J. Am. Chem. Soc., Vol. 119, No. 7, 1997 1709
nucleophilic substitution at a Si center.8 Theoretical calculations
confirm the notion that neutral siloxiranes should be strongly
stabilized by addition of an anion like fluoride to the silicon
center.9 This is actually a common feature of cyclic silanes
for which attachment to F- is proposed to lead to significant
stabilization.10 Gas-phase experimental measurements have
indeed corroborated this prediction, and a fluoride affinity of
40 ( 5 kcal mol-1 has been established for 1,1-dimethylsila-
cyclobutane.11
investigation of gas-phase ion/molecule reactions of the
FSi(OMe)2(OCH2)- ion unveils an unusually rich chemistry.
Computational Procedure
The first approach toward the structural elucidation of the product
ion of reaction 1b was to consider possible isomeric forms of
FSi(OMe)2(OCH2)- which would be consistent with mechanistic
considerations of reaction 1.15 Four different species, II-V, were
considered as likely candidates to arise from FSi(OMe)4- by elimination
of MeOH. These proposed structures and the outcome of the
The investigation of intrinsic properties of covalently bound
ionic intermediates and of molecules stabilized by binding to
an anion has been one of the most important domains of gas-
phase ion chemistry. For simple silicon-containing species, gas-
phase studies have already provided a unique view of the
reactivity of simple organosilanes and a wealth of important
thermochemical data.12 Thus, it is conceivable that ionic
siloxiranes intermediates, such as those responsible for Brook
type rearrangements,6 may be identified in the gas phase by a
judicious choice of experimental conditions. It is already known
that three-, four-, and five-membered species are likely to
mediate the reactions of silyl anions, and R-silyl carbanions,
with CO2 and N2O in the gas phase.13
calculations played an important role in the design of ion/molecule
reactions which could provide experimental insight into the structure
of the anion and a test of the theoretical predictions.
High-level ab initio calculations for Si species are usually limited
to small systems which can be handled by G2 methodology17 or
extensive correlation-consistent polarized valence basis sets.18 For more
complex species, one approach has been to mimic these molecules by
model systems with some of the substituents replaced by H atoms.
Another approach has been to use more modest basis sets which allow
calculations to be carried out to the MP4 level but for which empirical
corrections are essential to obtain the correct thermochemistry.1,2 For
our case, it was deemed more appropriate (a) to consider species which
exhibited different atom connectivities but which could correspond to
the actual experimentally observed ion and (b) to develop an appropriate
basis set for the anionic species at hand.
Ab initio calculations for ions II-V and some of the related neutral
systems were carried out using the GAMESS19 suite of programs. All
calculations were performed in the cluster of work stations located at
the CENAPAD-SP computing facilities of the University of Campinas.
The size of the systems represented by structures II-V precluded the
possibility of using very large basis sets or extending the calculations
to the highest levels of theory with our present computational facilities.
Yet, the study of systems II-V can be carried out with the help of
recent methods which are capable of generating good-quality basis sets
for anionic systems while circumventing some of the computational
limitations.20
The gas-phase reaction of F- with Si(OMe)4 was the subject
of an early study in a flowing afterglow tube and shown to
proceed exclusively by addition of F- with the subsequent
- 14
formation of a pentacoordinated adduct, FSi(OMe)4
.
This
same reaction studied in our laboratories at lower pressures and
under ion cyclotron resonance (ICR) conditions reveals a much
richer chemistry which presumably arises from unimolecular
elimination processes of energy rich pentacoordinated adduct
ions.15 The reported product distribution for this case is shown
in reaction 1. The rather unique reaction leading to the
FSi(OMe)2(OCH2)- ion (of unknown structure), and its possible
connection with siloxirane chemistry, prompted us to explore
the reactivity and structure of this ion by a combination of
experimental and theoretical techniques. It should be empha-
sized that this behavior is very specific of the reactions of F-
since products similar to that of reaction 1b are not observed in
reactions with much stronger gas-phase bases.15,16
This paper initially describes the results of ab initio calcula-
tions carried out for the FSi(OMe)2(OCH2)- ion. These
calculations predict the cyclic siloxirane-fluoride structure to
be the most stable form. In the second part, the experimental
The basic methodology used to obtain high-quality theoretical results
for these systems was to adapt the double ú valence set (DZV) of
primitive basis functions21 with the help of the generator coordinate
method (GCM).22 The GCM is a valuable tool in molecular orbital
calculations since it actually introduces a quality test for the basis set
used in a given atomic or molecular environment. In this method,
monoelectronic functions, ψ(1), are considered to be integral transforms
of the type
ψ(1) ) f(R) æ(1,R) dR
∫
where f(R) constitutes the so-called weight function, æ(1,R) represents
the generator function of Gaussian type orbitals (GTO), and integration
is performed over the space of the generator coordinate R, or exponent
(8) Corriu, R. Mechanisms of Nucleophilic Substitution at Silicon. In
Journal of Organometallic Chemistry Library 9, Organometallic Chemistry
ReViews; Seyferth, D., Ed.; Elsevier: Amsterdam, 1980; pp 357-373.
(9) Schlegel, H. B.; Skancke, P. N. J. Am. Chem. Soc. 1993, 115, 10916.
(10) Skancke, P. N. J. Phys. Chem. 1994, 98, 3154.
(11) Sullivan, S. A.; DePuy, C. H.; Damrauer, R. J. Am. Chem. Soc.
1981, 103, 480.
(12) For comprehensive reviews, see: (a) Schwarz, H. In The Chemistry
of Organic Silicon Compounds, Part 1; Patai, S., Rappoport, Z., Eds.; John
Wiley and Sons: Chichester, 1989; pp 445-510. (b) Damrauer, R.; Hankin,
J. A. Chem. ReV. 1995, 95, 1137.
(13) DePuy, C. H.; Damrauer, R.; Bowie, J. H.; Sheldon, J. C. Acc. Chem.
Res. 1987, 20, 127.
(14) DePuy, C. H.; Bierbaum, V. M.; Flippin, L. A.; Grabowski, J. J.;
King, G. K.; Schmitt; R. J.; Sullivan, S. A. J. Am. Chem. Soc. 1980, 102,
5012.
(15) Silva, M. L. P.; Riveros, J. M. J. Mass Spectrom. 1995, 30, 733.
(16) (a) Damrauer, R.; Krempp, M. Organometallics 1990, 9, 999. (b)
Krempp, M.; Damrauer, R. Organometallics 1995, 14, 170.
(17) Schlegel, H. B.; Darling, C. L. J. Phys. Chem. 1993, 97, 8207.
(18) Nicholas, J. B.; Feyereisen, M. J. Chem. Phys. 1995, 103, 8031.
(19) GAMESS: Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert,
S. T.; Gordon, M. S.; Jensen, J. J.; Koseki, S.; Matsunaga, N.; Nguyen, K.
A.; Su, S.; Windus, T. L.; Dupuis, M.; Montgomery, J. A. J. Comput. Chem.
1993, 14, 1347.
(20) (a) Morgon, N. H.; Custodio, R.; Riveros, J. M. Chem. Phys. Lett.
1995, 235, 436. (b) Morgon, N. H.; Linnert, H. V.; Riveros, J. M. J. Phys.
Chem. 1995, 99, 11667. (c) Morgon, N. H. J. Phys. Chem. 1995, 99, 17832.
(21) Dunning, T. H., Jr.; Hay, P. J. In Methods of Electronic Structure
Theory; Shaefer, H. F., III, Ed., Plenum Press: New York, 1977; Chapter
1, pp 1-27.
(22) (a) Custodio, R.; Goddard, J. D.; Giordan, M.; Morgon, N. H. Can.
J. Chem. 1992, 70, 580. (b) Custodio, R.; Giordan, M.; Morgon, N. H.;
Goddard, J. D. Int. J. Quantum Chem. 1992, 42, 411.

1710 J. Am. Chem. Soc., Vol. 119, No. 7, 1997
Morgon et al.
FA ) E[c-H2Si(CH2)2] + E(F-) - E[c-H2FSi(CH2)2]-
of the GTOs. The existence of weight functions (amounting to the
graphical representation of the linear combination of atomic orbital
functions) is an essential condition for using the GCM. An explicit
analysis of the behavior of the weight functions by the GCM is then
used as a criterion for adjusting the set of atomic functions necessary
to obtain a proper description of the electronic cloud.
The actual calculations required the internal electrons of all atoms
other than H to be represented by an effective core potential (ECP).
Under these conditions, the combination of the generator coordinate
method with an effective core potential is used to define the cutoff
point for the pseudopotential. For the valence electrons, the GCM
actually yields the choice of primitive diffuse functions to be added to
the original basis set for a correct description of anions.
A value of 42.6 kcal mol-1 was obtained after correction for
zero-point vibrational energies with vibrational frequencies
scaled by 0.89. Our result is comparable to values of 43.8 kcal
mol-1 obtained at the MP2//HF/6-31++G* level and 44.8 kcal
mol-1 obtained at the MP2//HF/MC-311++G* level.10 Like-
wise, the calculated Si-C, Si-H, C-C, and C-H bond
distances for c-H2Si(CH2)2 and [c-H2FSi(CH2)2]- differ on the
average by 0.03 Å with those obtained for the calculations at
the MP2//HF/6-31++G* level and by 0.02 Å with those
obtained at the MP2//HF/MC-311++G* level of ref 10.
Furthermore, our calculated H-C-Si, H-C-C, H-Si-C and
F-Si-C angles were found to differ on the average by 0.2°
and 0.3° with the two levels of calculation previously dis-
cussed.10
The internal electrons of Si were represented by the effective core
potential of Stevens, Basch, and Krauss (SBK),23 while the valence
region was corrected by addition of diffuse functions following the
procedure described before20 and outlined above. The resulting basis
set for Si was then ECP+7s5p1d (effective core potential for the inner
electrons and a 7s, 5p, and 1d primitive basis function set for the outer
electrons). The weight functions of the DZV primitive 1s, 2s, and 2p
atomic orbitals of Si, including all of the electrons (ae), display a normal
asymptotic behavior with decreasing values of R as shown and need
no further correction. However, this condition is not obeyed for small
values of R by the valence orbitals 3s and 3p. Since anionic systems
require corrections to account for the diffuse nature of the electronic
cloud, the GCM in conjunction with the effective core potential is used
to define the cutoff point at which the internal electrons are represented
by a pseudopotential. Addition of extra diffuse functions then leads
to a correct description of the valence orbitals, and the result of this
procedure was then verified in atomic and molecular environments.
A similar approach was used for other atoms which may bear anionic
character in structures II-V. Thus, the descriptions of the outer
electrons were corrected with the addition of diffuse functions for C,
Similar agreement was found between calculations using our
- 26
basis set and those previously reported for the case of SiF3
,
and SiOH.28
The good agreement with experiment and with previous ab
initio calculations on molecular Si-containing species illustrates
the fact that the basis sets developed here for larger Si
environments fare well when adapted to much simpler molecular
species.
The initial search of the best geometries for species II-V
had to contend with the likelihood of several local minima due
to conformations arising from rotation of the groups. For
species V, near-axial or near-equatorial arrangements could also
be envisioned as a result of distorted trigonal bipyramid
geometries. For simple hypervalent Si species like SiH4X- and
SiH3XY-, calculations reveal that the preferred conformation
for F and OMe groups cannot be established a priori in multiply
substituted siliconates.29 These considerations led us to carry
out an initial analysis with the most probable structures chosen
by using the genetic algorithm of SPARTAN.30 These structures
were then optimized at the ab initio level.
-
in the CH2 moiety of II, and O and F in FSi(OMe)2CH2O- (III),
(CH2F)Si(OMe)2O- (IV), and [FSi(OMe)2(cyc-CH2O)]- (V) and rep-
resented by an ECP+8s7p1d basis set. For the other atoms, the original
double ú basis functions augmented with polarization functions (p on
H and d on other atoms) were used and are here identified as the DZV**
basis set.
Computational Results
The initial structures assumed for II and III were found to
lead to ring closure upon geometry optimization but to yield
different isomeric structures of V. Thus, at this level of
calculation II and III are not predicted to be local minima in
the energy surface. The carbanion structure II yields a
siloxirane-containing siliconate ion which is identified in Figure
1 as structure V(II). By comparison, the alkoxide structure III
results in the siloxirane-containing siliconate anion labeled V-
(III). Given the complexity of the systems under consideration,
no attempt was made to find the transition state connecting
structures V(II) and V(III). Figure 1 shows the optimized
geometric parameters obtained for all ionic species, while Figure
2 shows the optimized geometries obtained for different neutral
systems used in comparing thermochemical data. These include
Si(OMe)4 (usually known as TMOS), FSi(OMe)3, FSi(OMe)2-
(cyc-CH2O), (MeO)2Si(cyc-CH2O), and FSi(OMe)2(CH2OH).
The fact that structures II and III could not be established
as local minima is not altogether surprising. The earlier
calculations9 for the much simpler FH2SiCH2O- and
The generator coordinate method is known to be capable of
generating basis sets of extremely good quality. For simple
systems, calculations carried out to high levels of theory (in
terms of corrections for correlation energy) and with consider-
ably more refined basis sets by the GCM than those used in
the present case yield gas-phase acidities comparable to or
better24 than those obtained with the G2 method which is often
used as the benchmark in this field.25
The quality of the basis set developed for this work was first
tested on simple Si-containing systems relevant to the problem
at hand. For example, the proton affinity of SiH3- is estimated
to be 377 kcal mol-1 with our basis sets in calculations carried
out for geometries optimized at the Hartree-Fock level, energies
calculated at the MP2 level, and correcting for zero-point
vibrational energies. This calculation compares well with values
of 378 kcal mol-1, obtained using MP2(FULL)//HF/6-31++G-
(d,p), and 375 kcal mol-1 at the MP4SDTQ//HF/6-311++G-
(2df,p) level.26 In all cases, the calculations are above the
-
experimental ∆H°acid(SiH4) ) 372.8 ( 2 kcal mol-1 27
.
FH2SiOCH2 systems established the alkoxide and carbanion
structures to be shallow minima (14.5 and 23.7 kcal mol-1
,
The fluoride affinity (FA) of silacyclopropane, c-H2Si(CH2)2,
was also calculated with the present basis set according to the
formal definition,
respectively, above the corresponding siloxirane anions) display-
ing very low energy barriers for ring closure (2.0 and 1.9 kcal
(23) Stevens, W. J.; Basch, H.; Krauss, M. J. Chem. Phys. 1984, 81,
6086.
(24) Morgon, N. H. Manuscript in preparation.
(25) Smith, B. J.; Radom, L. J. J. Phys. Chem. 1991, 95, 10549.
(26) Rodrigues, C. F.; Hopkinson, A. C. Can. J. Chem. 1992, 70, 2234.
(27) Wetzel, D. M.; Salomon, K. E.; Berger, S.; Brauman, J. I. J. Am.
Chem. Soc. 1989, 111, 3835.
(28) Allendorf, M. K.; Melius, C. F.; Ho, P.; Zachariah, M. R. J. Phys.
Chem. 1995, 99, 15285.
(29) Gordon, M. S.; Davis, L. P.; Burgraff, L. W. In AdVances in Gas
Phase Ion Chemistry; Adams, N. G., Babcock, L. M., Eds.; JAI Press Inc.:
Greenwich, CT, 1992; pp 203-223.
(30) SPARTAN version 4.0, Wavefunction, Inc., 18401 Von Karman
Ave., #370, Irvine, CA 92715.

Characterization of FSi(OCH3)2(OCH2)-
J. Am. Chem. Soc., Vol. 119, No. 7, 1997 1711
Figure 1. Optimized geometries for (MeO)2Si(CH2F)O- (IV) and for the two isomeric forms of FSi(OMe)2(cyc-CH2O)- , V(II) and V(III).
mol-1, respectively). For those systems, ring closure was also
claimed to yield two different conformational isomers of [FH2-
Si(cyc-CH2O)]- with significant differences in Si-C and Si-O
bond distances as in our case. The transition state for inter-
conversion between these two species was found in ref 9 to
involve a mechanism resembling a pseudorotation.
promoting either fluoride or methoxide transfer to other
substrates. The dissociation energy for V(II) to yield either
fluoride or methoxide ion is calculated to be
The geometric parameters obtained for the siloxirane moiety
in our calculations bear strong similarities to those obtained in
ref 9. A much shorter Si-O bond (1.634 Å) is calculated for
the anion originating from the carbanion, V(II), than for V-
(III) (1.748 Å). By comparison, the FH2Si(cyc-CH2O)- orig-
inating by ring closure of the carbanion was calculated to have
a Si-O bond distance of 1.660 Å as opposed to a bond distance
of 1.775 Å in the siloxirane anion obtained by ring closure of
the alkoxide structure.9 Likewise, our calculated Si-cycC bond
distance of 1.891 Å is much longer in V(II) than in V(III),
where Si-cyc-C ) 1.827 Å. The equivalent anions FH2Si-
(cyc-CH2O)- were calculated in ref 9 to have Si-cyc-C bond
distances of 1.927 and 1.833 Å, respectively. Finally, the
calculated O-Si-C angles in the siloxirane moiety are similar
within 1° of those previously calculated.
The geometric parameters obtained for TMOS reveal Si-O
and C-O bond distances of 1.612 and 1.403 Å, respectively.
These values should be compared with bond distances of 1.623
and 1.404 Å obtained in ref 2.
Table 1 lists the relative energies at the MP2 level for the
different ion structures including zero-point vibrational energy
and thermal corrections for enthalpy calculations at 298 K. It
is clear that structures V(II) and V(III) are more stable than
the isomeric structure IV, and that structure V(II) is the most
stable form predicted at the MP2//HF level.
The high value calculated for the first process predicts a very
high fluoride affinity for (MeO)2Si(cyc-CH2O). This number
is consistent with the previously calculated value of 61.8 kcal
mol-1 for H2Si(cyc-CH2O) at the MP4(SDTQ)/6-31++G**//
RHF/6-31++G** level including zero-point energy corrections.9
Such a high value for the fluoride affinity of a siloxirane system
was claimed to reflect the decrease in strain energy in going
from a tetrahedral-like environment in the neutral to a distorted
trigonal bipyramid in the anionic species.
An analysis of the charge distribution for the different anions
can provide some information regarding the expected reactivity
for these systems. While several approaches can be used for
this purpose, the simple Mulliken charges provide a useful
qualitative guide. A high positive Mulliken charge (ranging
from 1.5 to 1.78) is calculated on Si for both neutrals and anions
considered here. For anion IV, the siloxide oxygen is calculated
to have a Mulliken charge of -1.0, as might be expected for a
more localized anion. For structures V(II) and V(III), the extra
Mulliken negative charge is distributed among F, the cyclic O,
the O-(Me), and to a lesser extent the cyclic C.31
Finally, the question of relative proton affinities for the
different species deserves some consideration. The occurrence
of reaction 1b is indicative of a mechanism in which a nascent
Other predicted thermochemical properties deserve special
attention for comparison with the experimental results. The first
aspect refers to the possibility of FSi(OMe)2(cyc-CH2O)-
(31) The Mulliken charges on all atoms for the ionic species and neutrals
are available from the authors upon request.

1712 J. Am. Chem. Soc., Vol. 119, No. 7, 1997
Morgon et al.
Figure 2. Optimized geometries for Si(OMe)4, FSi(OMe)3, FSi(OMe)2(CH2OH), FSi(OMe)2(cyc-CH2O), and (MeO)2Si(cyc-CH2O).
Table 1. Relative Energies Calculated at the MP2 Level for
FSi(OMe)2(CH2O)- Ions Including ZPE and Thermal Corrections
for 298 K
Experimental Section
The general experimental procedure was similar to that used in our
earlier investigation of the ion/molecule reactions of Si(OMe)4.15 Ion/
molecule reactions were investigated in a homemade FT-ICR spec-
trometer with a magnetic field set at either 1.0 or 1.25 T and interfaced
to an Omega IonSpec Fourier transform data system. The cell was a
standard cubic cell with sides 2.5 cm long.
ionic species
∆H(298 K)/kcal mol-1
IV
V(II)
V(III)
17.2
(0.0)
7.6
Fluoride ions were generated by electron impact from NF3 (Pennwalt
Ozark Mahoning) at pressures around 8 × 10-8 Torr (ion gauge reading)
and at electron energies of 2 eV. Tetramethoxysilane, Si(OMe)4
(Aldrich, 99+%), was thoroughly degassed, dried, and purified by
distilling off in situ any volatile methanol produced by hydrolysis of
the sample. Typical partial pressures used in this work were in the
range of 1-2 × 10-8 Torr. Under these conditions, fluoride ions react
completely after 400-800 ms. The subsequent ion chemistry of a given
product ion generated by the F-/Si(OMe)4 reaction, reaction 1, was
studied by selective isolation of that ion after 600 ms of reaction time.
All unwanted ions were ejected with a combination of short (2 ms)
on-resonance radio frequency (rf) bursts and long rf sweeps (10-15
ms) over a wide mass range. In the case of FSi(OMe)2(OCH2)-, most
experiments were carried out by reisolating the ions after 1200 ms to
improve thermal equilibration of the ions.
methoxide ion is capable of abstracting a proton from the initial
pentacoordinated adduct, [FSi(OMe)4]-. The proton affinity of
(CH2F)Si(OMe)2O-, IV, is calculated to be 350 kcal mol-1
,
which is similar to an estimated proton affinity of 355 kcal
mol-1 for (MeO)2(Me)SiO-, (MeO)3SiO-, F(Me)2SiO-, and
(MeO)2(H)SiO- from bracketing experiments.32 The proton
affinities of the putative II and III structures are more difficult
to establish since no stable geometries could be found in the
energy surface.
The theoretical calculations were then considered an ap-
propriate background for the experimental results described
below which provide a critical assessment of the theoretical
predictions.
Reagents used to test the reactivity of FSi(OMe)2(OCH2)- were
introduced in the vacuum chamber via a pulsed valve (General Valve,
(32) Damrauer, R; Simon, R.; Krempp, M. J. Am. Chem. Soc. 1991, 113,
4431.

Characterization of FSi(OCH3)2(OCH2)-
J. Am. Chem. Soc., Vol. 119, No. 7, 1997 1713
known.33 Yet, reaction 2a strongly argues in favor of a
siloxirane structure for the [FSi(OMe)2(OCH2)]- ion. An ion
with a structure similar to IV (or to structures II or III) would
not be expected to react by F- transfer.
The appearance of FSi(OMe)2O- in reaction 2b suggests that
reaction 2a is sufficiently exothermic to promote partial
displacement of a MeO- in the pentacoordinated adduct,
FSi(OMe)4-. A nascent MeO- ion in a [FSi(OMe)3‚‚‚-OMe]
type collision complex could then promote an internal nucleo-
philic displacement similar to that claimed for reaction 1c. While
the thermochemistry of the fluoride transfer reaction is not
known, the product distribution reported earlier for reaction 1
reveals that MeO- is easily displaced from the pentacoordinated
-
FSi(OMe)4 ion. Nevertheless, these experiments cannot rule
out the possibility that the nature of the neutral products may
be different from those shown in reaction 2b. However, it
should be emphasized that reasonable neutral products are
difficult to suggest if the reactant ion in reaction 2b is assumed
to have structure IV (or II and III).
While reactions 2a and 2b are reasonably conclusive regarding
the nature of the [FSi(OMe)2(OCH2)]- anion, further experi-
ments were considered necessary in order to establish the
strength of the fluoride binding in this anion and the stability
of the siloxirane ring.
(b) Reactions with BF3. Boron trifluoride, BF3, was
considered a particularly attractive substrate because of its high
fluoride affinity,33 estimated to be in the range 72 ( 5 kcal
mol-1, and its capability to undergo nucleophilic attack readily.34
A rapid series of fluoride-methoxide exchange reactions
(reaction 3) was observed to take place between [FSi-
(OMe)2(OCH2)]- and related anions with BF3.
Figure 3. The time evolution of reagent and products in the reaction
of FSi(OMe)2(CH2O)- with Si(OMe)4, reaction 2. The experimental
conditions are given in the Experimental Section.
Model 9) after isolation of the reagent ions, or used under static partial
pressures maintained in the neighborhood of 1-2 × 10-8 Torr. Typical
pulse widths of 2-5 ms were used for the pulsed valve, and the
maximum pressure in the ion gauge after the pulse never exceeded 5
× 10-7 Torr. A typical delay time of 5 s was used between successive
pulse sequences to allow for proper pumping of the system in these
experiments.
Acetylacetone, nitromethane, acetone, 2-propanol, acetic acid, etha-
nol, methanol and hexafluorobenzene were obtained commercially
(Aldrich) and subjected to several freeze-pump-thaw cycles before
sample introduction. Positive ion ICR spectra were recorded to test
the purity of the samples. BF3 (Matheson) was subjected to a triple
low-temperature distillation process in order to remove impurities
observed in its mass spectra.
FSi(OMe)2(OCH2)- + BF3 f
F2Si(OMe)(OCH2)- (m/z 127) + F2BOMe (3a)
Experimental Results
f F3Si(OCH2)- (m/z 115) + FB(OMe)2
(3b)
The reaction of F- with Si(OMe)4 in our FT-ICR spectrometer
has been shown to yield four different ionic products (see
reaction 1).15 While our main objective was to explore the
reactivity of the [FSi(OMe)2(CH2O)]- ion, it was necessary in
several cases to compare its behavior with that of the ions
The kinetic scheme is complicated by the fact that
F2Si(OMe)(OCH2)- (m/z 127) itself reacts rapidly with BF3.
F2Si(OMe)(OCH2)- + BF3 f F3Si(OCH2)- + FB(OMe)2
-
(4)
formed in reactions 1a and 1c since presumably FSi(OMe)4
corresponds to a pentacoordinated Si adduct, and FSi(OMe)2O-
to a siloxide type anion.
By comparison, fluoride transfer to BF3 from the different
-
ions occurs slowly and formation of BF4 is only observed to
(a) Reaction of [FSi(OMe)2(OCH2)]- with Si(OMe)4. Our
ion of interest reacts with the parent Si(OMe)4 by fluoride
transfer to yield FSi(OMe)4- and a neutral generically identified
as [Si(OMe)2(OCH2)], reaction 2a. Reaction 2b, which gener-
ates FSi(OMe)2O-, is observed to occur competitively as shown
in Figure 3. Two neutral products are assumed to be formed
in this case, Si(OMe)2(OCH2) and Me2O.
become important at long reaction times. All of the different
species apparently contribute to the formation of BF4-, but
F3Si(OCH2)- undergoes exclusively fluoride transfer with BF3,
-
and thus BF4 becomes the dominant ion at reaction times
longer than 5 s. Fluoride transfer, reaction 5, can then be
considered a slow competitive reaction in this system.
F1+nSi(OMe)2-n(OCH2)- + BF3 f
-
FnSi(OMe)2-n(OCH2) + BF4 (5)
n ) 0-2
No attempt was made to determine the absolute rate constants
due to the uncertainties associated with measuring the absolute
pressure of Si(OMe)4, but the reagent ion is observed to
disappear completely through a simple exponential decay.
Reaction 2a is a typical example of a fluoride transfer reaction,
suggesting that Si(OMe)4 has a higher fluoride affinity than [Si-
(OMe)2(OCH2)]. However, this observation cannot be used as
a quantitative thermochemical upper limit for the fluoride
affinity of Si(OMe)2(OCH2) since that of Si(OMe)4 is un-
Figure 4 shows the time dependence of the product distribu-
tion during the first 1200 ms after isolation of the FSi-
(OMe)2(OCH2)- ion. This time dependence is clearly consistent
with the set of reactions shown above, and with the fact that
the fluoride transfer type reactions (reaction 5) are slow under
these conditions.
(33) Larson, J. W.; McMahon, T. B. J. Am. Chem. Soc. 1985, 107, 766.
(34) For a recent example in the gas phase, see: Taylor, W. S.; Babcock,
L. M. J. Am. Chem. Soc. 1995, 117, 6497.

1714 J. Am. Chem. Soc., Vol. 119, No. 7, 1997
Morgon et al.
Figure 4. Product distribution as a function of time in the reaction of
FSi(OMe)2(CH2O)- with BF3. The time scale is characteristic of the
complete disappearance of the reagent ion. Notice that BF4 grows
with time and eventually becomes the most important species at long
reaction times (5 s). (See reactions 3-5.)
Figure 5. Product distribution as a function of time in the reaction of
-
FSi(OMe)4- with BF3. The time scale is characteristic of the complete
-
disappearance of the reagent ion. Notice that BF4 grows with time
and eventually becomes the most important species at long reaction
times (5 s). (See reactions 6-8.)
A very precise quantitative kinetic analysis of these reactions
in the presence of a stationary pressure of BF3 (as opposed to
experiments where BF3 is introduced through a pulsed valve)
is complicated by the fact that FSi(OMe)2(OCH2)- also
undergoes reaction 2 with Si(OMe)4, and both product ions of
reaction 2 are found to react further with BF3 as described below.
In order to understand the significance of reactions 3-5, the
-
reactivity patterns of FSi(OMe)4 and FSi(OMe)2O- in the
presence of BF3 were also studied independently. Since the
structures of these two anions are less likely to be controversial,
these experiments were considered to be a valuable frame of
reference for the reactivity of fluoride-silane adducts and
siloxide type anions with BF3.
The pentacoordinated adduct ion, FSi(OMe)4-, was observed
to react through multiple fluoride-methoxide exchange reac-
tions as shown in reaction 6. Unlike the reactions of
FSi(OMe)4- + BF3 f FSi(OMe)3 + [F3BOMe]- (m/z 99)
Figure 6. Product distribution as a function of time in the reaction of
(6a)
-
FSi(OMe)2O- with BF3 (reactions 9 and 10). Notice that BF4 is not
formed in this case.
f F2Si(OMe)2 + F2B(OMe)2- (m/z 111)
(6b)
On the other hand, FSi(OMe)2O- ions generated either by
2b or 1c react with BF3 exclusively through the fluoride-
methoxide exchange reactions 9 and 10. No BF4- is obserVed
FSi(OMe)2(OCH2)-, the anionic product species are boron-
containing ions. Isolation of either product ion of reaction 6
-
reveals that F2B(OMe)2 can react further with BF3 to yield
[F3BOMe]- (reaction 7).
FSi(OMe)2O- (m/z 125) + BF3 f
F2B(OMe)2- (m/z 111) + BF3 f
F2Si(OMe)O- (m/z 113) + F2BOMe (9)
F2BOMe + [F3BOMe]- (m/z 99) (7)
F2Si(OMe)O- (m/z 113) + BF3 f
F3SiO- (m/z 101) + F2BOMe (10)
-
Fluoride transfer from these different species to yield BF4
is observed to be slow but becomes important at long reaction
times. As in reaction 5, [F3BOMe]- can only undergo fluoride
transfer, reaction 8. The time evolution of the product ions is
in these reactions eVen at long reaction times. Figure 6 shows
the time evolution of the product ions in this case.
These reactions with BF3 allow further conclusions to be
-
F3B(OMe)- + BF3 f F2BOMe + BF4
(8)
derived regarding the nature of the FSi(OMe)2(OCH2)- ion. For
-
example, the reactions of FSi(OMe)4 are consistent with the
plotted in Figure 5. The ions F2B(OMe)2-, F3B(OMe)-, and
BF4- display a time dependent behavior similar to that observed
for the product ions of FSi(OMe)2(OCH2)- plotted in Figure 4.
Thus, considerable similarities are observed in these reactions
ease with which this ion can transfer MeO- to substrates of
higher methoxide affinity. Reactions 6a and 6b can then be
rationalized by a scheme in which exchange of fluoride between
the B and Si centers is promoted by the initial transfer of a
methoxide group from the Si to the B center, Scheme 1. In
-
between FSi(OMe)4 and FSi(OMe)2(OCH2)-.

Characterization of FSi(OCH3)2(OCH2)-
J. Am. Chem. Soc., Vol. 119, No. 7, 1997 1715
Scheme 1
Scheme 2
-
this scheme, long-lived collision complexes must be present to
account for the multiple fluoride-methoxide exchanges. In this
situation, the final partition between neutral and anionic moieties
would be determined by the relative stabilities of the corre-
sponding Si and B species.
By comparison, FSi(OMe)4 and FSi(OMe)2O- are also
observed to abstract a proton from the first two substrates but
not from the others. The behavior of FSi(OMe)4 can be
rationalized in terms of the ease with which methoxide ion can
be released to promote proton abstraction.
-
Reactions 9 and 10 are indicative of processes initiated by
nucleophilic attack of the siloxide ion on boron as depicted in
Scheme 2. Given the schemes proposed to account for the
FSi(OMe)4- + MeCOOH f
-
-
reactions of FSi(OMe)4 and FSi(OMe)2O- with BF3, two
FSi(OMe)3 + MeOH + MeCO2 (11)
possible alternatives can be proposed to account for reactions
3 and 4. The first alternative is to assume that reactions 3 and
4 reflect a siloxide (IV) or alkoxide (III) type behavior for the
[FSi(OMe)2(OCH2)]- ion, in which case these reactions could
be explained by a sequence similar to that shown in Scheme 2.
The second alternative is to consider the possibility that the ion/
neutral association of [FSi(OMe)2(cyc-OCH2)]- with BF3
activates the siloxirane ring and reaction then proceeds by ring
opening and nucleophilic attack of the nascent FSi(OMe)2CH2O-
ion. This reasoning then suggests that F2Si(OMe)(OCH2)- and
F3Si(OCH2)- ions also correspond to siloxirane species. The
fact that the different Fn+1Si(OMe)2-n(OCH2)- species can
promote F- transfer to BF3, albeit slowly, as in the case of the
ions derived from FSi(OMe)4-, strongly argues for this second
alternative and in favor of a siloxirane type structure for the
FSi(OMe)2CH2O- ion.
On the other hand, the reaction of FSi(OMe)2O- is consistent
with the gas-phase acidity trend for silanol and substituted
silanols.32,36
The base-like behavior of [FSi(OMe)2(OCH2)]- resembles
that of a substituted alkoxide ion. In fact, the gas-phase acidity
of an RCH2OH alcohol bearing a Si-containing substituent could
well fall in the range between that of 2-propanol and ethanol.
While these experiments are more consistent with an alkoxide
ion type structure, like that represented by structure III, the
results can also be accommodated by a mechanism in which
proton abstraction can activate ring opening of the siloxirane
structure V as proposed for the case of reaction with BF3.
Few examples are known of similar type reactions in gas-
phase ion chemistry. Ring opening has been observed in the
gas-phase reaction of neutral ethylene oxide37 with strong bases
like OH-, but these reactions are better described as addition-
elimination reactions.
(c) Reactions with Gas-Phase Acids. The fact that the
reaction of [FSi(OMe)2(OCH2)]- with BF3 can be construed as
a nucleophilic attack on BF3 mediated by activation of the
siloxirane ring led us to explore the possibility of this ion
reacting with gas-phase acids.
(d) Reaction with N2O and C6F6. A preliminary investiga-
tion15 showed [FSi(OMe)2(OCH2)]- to be unreactive toward
N2O. This observation has been used to rule out (1) a carbanion
structure for our ion of interest and (2) activation of the
siloxirane ring to yield a carbanion type behavior.
The behavior described above for the [FSi(OMe)2(OCH2)]-
ion suggests additional experiments with substrates which have
been effective in recognizing ambidentate reactivity. While a
number of reagents have been used in this respect,39 hexafluo-
Experiments were then carried out by either pulsing the gas-
phase acid into the cell or using a static pressure (1 × 10-8
Torr) of the acid. Proton abstraction was observed to occur
readily from acetylacetone (∆Gacid ) 336.6 kcal mol-1), acetic
acid (∆Gacid ) 341.6 kcal mol-1), nitromethane (∆Gacid ) 349.7
kcal mol-1), and acetone (∆Gacid ) 361.9 kcal mol-1), and more
slowly from 2-propanol (∆Gacid ) 368.8 kcal mol-1).35 No
proton abstraction was observed from ethanol (∆Gacid ) 370.8
kcal mol-1) or methanol (∆Gacid ) 374.0 kcal mol-1).
(35) NIST Standard Reference Database 19B, NIST NegatiVe Ion
Energetics Database, Version 3.0, compiled by J. E. Bartmess, 1993.
(36) Grimm, D. T.; Bartmess, J. E. J. Am. Chem. Soc. 1992, 114, 1227.
(37) Bierbaum, V. M.; DePuy, C. H.; Shapiro, R. H.; Stewart, J. H. J.
Am. Chem. Soc. 1976, 98, 4229.
(38) DePuy, C. H.; Damrauer, R. Organometallics 1984, 3, 362.
(39) See, for example: (a) Brickhouse, M. D.; Squires, R. R. J. Am.
Chem. Soc. 1988, 110, 2706. (b) Freriks, I. L.; de Koning, L. J.; Nibbering,
N. M. M. J. Am. Chem. Soc. 1991, 113, 9119. (c) Zhong, M.; Brauman, J.
I. J. Am. Chem. Soc. 1996, 118, 636.

1716 J. Am. Chem. Soc., Vol. 119, No. 7, 1997
Morgon et al.
robenzene was chosen in this case as an additional test for the
nature of the [FSi(OMe)2(OCH2)]- ion. Reaction 12 was
observed to occur readily and exclusively. This result confirms
that two structurally different FSi(OMe)2(OCH2)- ions are
generated in this system. Thus, the idea has been advanced
that the siloxirane ring can be activated by suitable substrates
resulting in reactions which are typically those of an alkoxide
ion. The reactions with BF3 suggest that this ring activation is
indeed more favorable than fluoride transfer, a fact that may
not be surprising because of the large binding energy of F- to
these Si species. Furthermore, this ease of ring activation is in
agreement with the experimental difficulty associated with
isolating siloxirane derivatives and the fact that such structures
become intermediates, and not stable species, in the Brook
rearrangement.5
[FSi(OMe)2(OCH2)]- + C6F6 f
C6F5O- + [FSi(OMe)2(CH2F)] (12)
that the [FSi(OMe)2(OCH2)]- ion can react very efficiently in
a manner similar to that of alkoxide ions, or through a reaction
where the siloxirane ring is activated to yield an alkoxide-like
ion in the collision complex.
Finally, a word of caution needs to be introduced in
comparing the experimental results for the proton abstraction
and hexafluorobenzene reactions with theoretical predictions.
Firstly, the actual structures of the neutral products are not
known. Since rearrangements have been invoked as a likely
possibility in neutral siloxirane chemistry, this possibility needs
to be taken into consideration. Secondly, the calculation of the
thermochemistry for ring opening involves nonisodesmic reac-
tions. If the enthalpies for Si-containing compounds are not
well described at this level of calculation unless empirical bond
additivity corrections are introduced,1,2,28 then the estimated
enthalpy changes for such reactions may not be very reliable.
Discussion
The theoretical calculations presented in the early part of the
paper as well as calculations reported earlier9 clearly establish
that the most stable structure for R2FSi(CH2O)- ions corre-
sponds to an anion containing a siloxirane ring. The fact that
two sets of calculations with different basis sets, but including
corrections for correlation energy, arrive at similar conclusions
provides substantial support for the theoretical contention.
While a more complete investigation was carried out in ref 9
for the energy surface of a simpler system, it is clear that a full
exploration of all possible interconverting paths for FSi-
(OMe)2(OCH2)- would be a formidable task. The fact that we
could not observe a stable alkoxide or carbanion structure for
this anion at our level of calculation is not altogether different
from the results of ref 9 where these minima for FH2Si(CH2O)-
were claimed to be uncertain if larger basis sets or higher levels
of theory were to be used.
Acknowledgment. The authors thank the Sa˜o Paulo Science
Foundation (FAPESP) for generous support of this work through
Grant 94/4404-6. The authors also acknowledge the support
of the Brazilian Research Council (CNPq) in the form of a
Young Investigator Fellowship (NHM), an Undergraduate
Science Fellowship (ABA), and a Senior Research Fellowship
(JMR). We also wish to thank the High Performance Computing
Center (CENAPAD-SP) and the Institute of Chemistry of the
University of Campinas for use of their computing facilities.
The experimental results clearly show that FSi(OMe)2(OCH2)-,
and other related ions containing the CH2O moiety attached to
Si, can undergo fluoride transfer to substrates of high fluoride
affinity. This behavior strongly supports the theoretical predic-
tion of a stable siloxirane structure for this ion. Likewise, the
fact that in all cases FSi(OMe)2(OCH2)- is observed to react
completely with the different substrates suggests that a mixture
of isomeric species is unlikely. Thus the symbiotic relationship
between theory and experiment proves to be a valuable guideline
in this type of problem.
Supporting Information Available: Table S1 of the
exponents of the primitive basis functions obtained by the GCM
for C, F, O, and Si atoms using the SBK pseudopotential, Table
S2 with Hartree-Fock and MP2 calculated energies for all ions
and neutrals, and Figures S1 and S2 with the weight functions
for the atomic orbitals of Si in atomic Si and in Si(OCH3)4 (4
pages). See any current masthead page for ordering information
and Internet access instructions.
Nevertheless, several of the reactions reported above reveal
that FSi(OMe)2(OCH2)- can also promote reactions akin to those
induced by anions bearing a negatively charged oxygen. The
kinetic behavior of our ion of interest suggests that it is unlikely
JA963382A