Monitoring mechanistic details in the synthesis of pyrimidines via real-time, ultrafast multidimensional NMR spectroscopy

Article abstract of DOI:10.1021/ja210154g

Recent years have witnessed unprecedented advances in the development of fast multidimensional NMR acquisition techniques. This progress could open valuable new opportunities for the elucidation of chemical and biochemical processes. This study demonstrates one such capability, with the first real-time Two-dimensional (2D) dynamic analysis of a complex organic reaction relying on unlabeled substrates. Implementing such measurements required the development of new ultrafast 2D methods, capable of monitoring multiple spectral regions of interest as the reaction progressed. The alternate application of these acquisitions in an interleaved, excitation-optimized fashion, allowed us to extract new structural and dynamic insight concerning the reaction between aliphatic ketones and triflic anhydride in the presence of nitriles to yield alkylpyrimidines. Up to 2500 2D NMR data sets were thus collected over the course of this nearly 100 min long reaction, in an approach resembling that used in functional magnetic resonance imaging. With the aid of these new frequency-selective low-gradient strength experiments, supplemented by chemical shift calculations of the spectral coordinates observed in the 2D heteronuclear correlations, previously postulated intermediates involved in the alkylpyrimidine formation process could be confirmed, and hitherto undetected ones were revealed. The potential and limitations of the resulting methods are discussed.

Full text of DOI:10.1021/ja210154g

Article
pubs.acs.org/JACS
Monitoring Mechanistic Details in the Synthesis of Pyrimidines via
Real-Time, Ultrafast Multidimensional NMR Spectroscopy
Zulay D. Pardo,‡ Gregory L. Olsen,† María Encarnacion Fernandez-Valle,§ Lucio Frydman,*,†
́
,‡,§
́
‡
́
Roberto Martínez-Alvarez, and Antonio Herrera*
†Department of Chemical Physics, Weizmann Institute of Science, 76100 Rehovot, Israel
‡Departamento de Química Organica I and §CAI de RMN y RSE, Facultad de Ciencias Químicas, Universidad Complutense, 28040
́
Madrid, Spain
S
* Supporting Information
ABSTRACT: Recent years have witnessed unprecedented advances
in the development of fast multidimensional NMR acquisition
techniques. This progress could open valuable new opportunities for
the elucidation of chemical and biochemical processes. This study
demonstrates one such capability, with the first real-time Two-
dimensional (2D) dynamic analysis of a complex organic reaction
relying on unlabeled substrates. Implementing such measurements
required the development of new ultrafast 2D methods, capable of
monitoring multiple spectral regions of interest as the reaction
progressed. The alternate application of these acquisitions in an
interleaved, excitation-optimized fashion, allowed us to extract new structural and dynamic insight concerning the reaction
between aliphatic ketones and triflic anhydride in the presence of nitriles to yield alkylpyrimidines. Up to 2500 2D NMR data
sets were thus collected over the course of this nearly 100 min long reaction, in an approach resembling that used in functional
magnetic resonance imaging. With the aid of these new frequency-selective low-gradient strength experiments, supplemented by
chemical shift calculations of the spectral coordinates observed in the 2D heteronuclear correlations, previously postulated
intermediates involved in the alkylpyrimidine formation process could be confirmed, and hitherto undetected ones were revealed.
The potential and limitations of the resulting methods are discussed.
reaction as well as to quantify their overall kinetics of
formation. Doing so, however, requires completing the 2D
NMR measurement in time scales compatible with the chemical
changes occurring following a sudden triggering of the reaction.
Much progress has taken place during recent years in the area
of accelerating multidimensional NMRboth in the areas of
organic and biomolecular acquisitions. Efforts that have made
2D spectroscopy compatible with the tracking of rapid
transformations include Hadamard-encoded versions of hetero-
nuclear (HSQC) and homonuclear (TOCSY) NMR exploiting
a priori information,8 compressed sensing acquisitions where
the usual Nyquist demands for the FT-based sampling of a
NMR domain can be alleviated,9 and so-called “ultrafast” (UF)
acquisition schemes whereby arbitrary 2D NMR correlations
can be established in a single-scan fashion.10 The latter
approach in particular replaces the indirect-domain parametric
time encoding of 2D NMR with a spatiotemporal manipulation,
effectively imparting all the increments that would be involved
in a conventional experimentin their entiretywithin one
single scan.11 The spatial patterns imparted by indirect domain
evolution frequencies encoded in such a manner can be
subsequently read out with the aid of suitable magnetic field
1. INTRODUCTION
Heterocycles’ significance is central in modern chemistry.1
Among nitrogen-containing heterocycles, pyrimidines play
unique roles as integral constituents in many pharmaceutical,
functional, and natural materials.2 Understanding and control-
ling laboratory preparations of pyrimidinic derivatives have
therefore become an important field of synthetic chemistry.3
Recently we focused our attention on one-pot syntheses of
various heterocyclic derivatives and developed alternatives for
an easier preparation of a wide range of azaheterocyclic
compounds.4 Other elegant procedures have been also
described for the synthesis of pyrimidine derivatives5 and
oxazoles.6 Several of these studies revealed details of the
mechanism of pyrimidine formation when reacting carbonyl
compounds with strong electrophiles, such as trifluorometha-
nesulfonic acid anhydride (triflic anhydride, Tf2O).7 Still,
despite the importance of these reactions, large gaps in
knowledge remain about the variety of intermediates and
mechanistic routes involved in their formation, and key details
about the structural information associated to the kinetics of
these processes remain to be elucidated.
Real-time NMR measurements could open unique oppor-
tunities in the analysis of complex chemical problems such as
this one. Two-dimensional (2D) NMR, in particular, could
enable one to monitor and assign transient intermediates in the
Received: October 28, 2011
Published: January 4, 2012
© 2012 American Chemical Society
2706
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
gradients. Using such rapid acquisition 2D NMR methods, a
number of fast-reacting systems have been recently monitored
in real time. These have included biomolecular single- and
multiple-quantum heteronuclear correlations (HSQC and
HMQC) on proteins and nucleic acids12 as well as TOCSY
and multiple-bond correlation experiments on reacting organic
systems.13a Included among the latter were studies on the
synthesis of alkyl- and alkylaryl-pyrimidines from aliphatic and
aromatic ketones, whereby series of complete 2D NMR data
sets were collected on 100% abundant isotopes to observe the
changes in chemical species occurring as the reactions
progressed.
signal-to-noise ratio (SNR) is in turn also governed by the
γaGaL frequency span, which becomes the filter bandwidth that
needs to be used for this kind of measurement. Noise in this
kind of experiment thereby behaves proportionally to
√(SW1·SW2/Δv1). For the kind of HSQC and HMBC
experiments that would be needed to characterize a priori
unknown intermediates that can fall anywhere in the aliphatic/
olefinic/aromatic range, as is the case for the kind of reaction
depicted in Scheme 3, this relation would impose high strains
on the gradient strengths that would be necessary (≈80 G/cm
for covering an indirect domain spanning ≈150 ppm, a direct
domain of ≈4 kHz, and the typical 1.8 cm sample length). Even
worse, the associated noise penalties would become untenable
for the ensuing kind of filter bandwidths.
We have recently studied these reactions for the specific case
of the addition of 3-pentanone 1 to acetonitrile-d3 in the
presence of triflic anhydride (Scheme 1). In these studies 2D
One of the aims of this work consisted in developing a
methodology which would allow us to monitor the necessary
spectral widths while bypassing such stringent demands.
Strategies have been discussed in the recent literature capable
of alleviating the gradient strengths needed to decode a given
spectral bandwidth in single-scan 2D NMR, including the
folding of peaks along the direct and indirect domain as well as
shifting of peaks into a single, congruent band of detection.14 In
the present instance, however, we decided to explore an
alternative route that, while still alleviating the gradient’s
demands, could still furnish a direct reading of the chemical
shift information about the transient reaction intermediates. To
do so a new family of experiments was devised incorporating
(i) the selective excitation of different spectral bands of interest
(aliphatic, olefinic, aromatic) and (ii) the interleaving of real-
time ultrafast acquisitions targeting these spectral bands
through consecutive scans. With the aid of such strategies it
was possible to interrogate the nature of the chemical dynamics
in Scheme 3 using unlabeled compounds, recording literally
thousands of 2D heteronuclear correlation spectra within the
course of these hour-long processes. With the aid of such data,
of ancillary 1D NMR measurements, and of computer-based
estimators of chemical shift values, new transient intermediates
were detected, and their kinetics of generation/consumption
characterized. All this led us to propose a new two-pathway
mechanism for the triflate-mediated addition of acetonitrile-d3
to arylketones.
Scheme 1
ultrafast TOCSY homonuclear acquisitions revealed the
presence of intermediates 2−4, and the generation of the
final reaction products 5 and 6 was revealed (Scheme 2).13b
Two-dimensional ultrafast heteronuclear multiple bond corre-
lation spectroscopy (HMBC) experiments on specifically
labeled precursors were also implemented to elucidate the
chemical nature and evolution of reactive quaternary centers
involved in this reaction. Upon monitoring a similar
condensation that starting from 13C-carbonyl-labeled acetophe-
none 7 lead to alkylaryl-pyrimidine 9 as main product (Scheme
3), such experiments permitted the detection of the
intermediate (trifluorormethylsulfonyloxy)carbenium ion 8.13c
Unfortunately, label-based experiments such as these have a
narrow focus that may obscure a full and unambiguous
identification of all transient intermediates generated during
the course of the reaction. To extract such important
mechanistic information, it would be necessary to implement
more general 2D NMR correlations. Particularly useful could be
site-identifying heteronuclear correlations covering the full
possible range of chemical shifts that might arise, so that both
permanent as well as transient species can be characterized.
Running ultrafast HSQC experiments at natural abundance in
non-labeled samples, however, would not be a trivial endeavor.
Indeed in single scan sequences, the spectral widths SW2 and
SW1 along the directly and indirectly sampled dimensions are
related to one another. Moreover these parameters as well as
the effective spectral resolution Δν1 desired along the indirect
domain define the gradient strength Ga required to read out the
indirect domain spectral width according to (SW1SW2)/(Δv1)
∼ γaGaL, where L is the encoded sample length. The per scan
2. MATERIALS AND METHODS
To address the spectral-width/spectral-resolution issues just men-
tioned, we decided to attempt the observation of the reaction in
Scheme 3 at natural abundance by modifying a general, broadband
heteronuclear correlation experiment, into a multiple-window
observation mode. In other words, instead of implementing ultrafast
sequences covering in excess of 100 ppm in the 13C domain, the
targeted direct- and indirect-domain windows were split into regions
expected to possess the peaks of interest. Areas that were a priori,
known to be devoid of peaks, were consequently ignored. Moreover,
these multiple spectrally selective versions of single-bond hetero-
nuclear correlation ultrafast sequences were run in an interleaved
fashion so as to employ the relaxation recovery delay of one spectral
Scheme 2
2707
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
Scheme 3
Figure 1. Multiwindowed ultrafast HSQC pulse sequence involving: a selective excitation of protons in the targeted spectral window; an INEPT
magnetization transfer to carbon; four chirped pulses applied in the presence of suitable gradients and (π)H decoupling pulses to give a constant-time
spatial encoding of the carbon evolution; a selective transfer of this encoded information to protons; and low-strength Ga gradient pairs acting as
read-out of the resulting encoded signals. After shifting the irradiation frequencies to a second spectral window, the sequence is then immediately
repeated to acquire a second FID addressing a different spectral region. Selective excitation and spatial encoding used 1.42 ms sinc π/2 pulses and
2.5 ms π chirp pulses, respectively. Encoding gradient strengths were 25 G/cm. Read out of encoded signals typically used N2 =40 cycles and a Ga of
6.35 or 10.6 G/cm. The narrow independent windows targeting specific spectral regions of interest permits use of these low-strength gradients,
providing corresponding increases in SNR and enabling the detection of nonlabeled samples.
region to collect the data pertaining to another spectral regionvery
much like multislice imaging experiments perform their optimized
acquisitions in frecuency-shifted 3D MRI.15 To do so in the ultrafast
cycling of both the initial (π/2)C hard pulse and the receiver phase, for
ensuring a more complete suppression of the natural abundance
background. Acquisitions used gradient strengths of Ga = 6.35 or 10.6
G/cm and filter bandwidths of either 25 or 42 kHz (as noted in the
text). Further acquisition parameters included N2 = 40 cycles of
gradient pairs, a physical sampling dwell time of 2 μs, and gradient
durations Ta = 0.60 ms each. Square (π)C pulses were applied in
between all acquisition gradients for heteronuclear decoupling
purposes. Purge gradient pulses preceding acquisition were 1.60 ms
in duration, at 6.35 G/cm (10.6 G/cm for the experiment using
labeled starting material (7) to probe the olefinic region).
1
instances, care had to be made to either avoid the excitation of H
signals outside the targeted region or, alternatively, to make sure that
unencoded regions were addressed by an even number of (π)H
rotations so as to return their magnetizations to a relaxation-optimized
longitudinal state. (13C manipulations being less critical for all cases
1
given that H initiated, indirectly detected sequences were assayed.)
Three spectral windows were selected for these observations, usually
1
involving ≈1.5 ppm along the H and ≈10−20 ppm along the 13C
dimension. These windows encompassed: (i) an ‘aliphatic’ region
centered at 2.2 ppm (1H) and 28.7 ppm (13C); (ii) an ‘olefinic’ region
at 5.8 ppm (1H) and 110.9 ppm (13C) set to follow transient
intermediates; and (iii) an aromatic ‘pyrimidinic’ region centered at 8.2
ppm (1H) and 117.4 ppm (13C) to track formation of the final reaction
product 2,4-dimethyl-6-phenylpyrimidine-d6 (9). A typical scheme of
these pulse sequences is illustrated in Figure 1. This consists of a series
of back-to-back heteronuclear correlations involving two continuous
constant-time spatial encoding processes for monitoring the indirect
domain 13C evolution. As each one of these additive processes
involved a (π)H decoupling pulse in their center, no overall rotation of
the proton magnetizations resulted from the t1 period. The two
INEPT transfers in each of these sequences incorporated sinc(π/2)H
excitations and two hard (π)H pulses, thereby fulfilling the band-
selectivity requirements of these sequences. These multiwindowed
constant-time ultrafast 2D NMR data sets were collected on a Varian
VNMRS 600 MHz NMR spectrometer using a triply tuned HCN
single-gradient probe at 298 K. The constant-time spatial encoding
adiabatic (π)C pulses needed for the spatial encoding were generated
using the Varian Pbox software package, with these chirped pulses
sweeping a bandwidth [Oi−Of] = 49 519 Hz in 2.5 ms, applied in
conjunction with 25 G/cm encoding gradients. All protons were
selectively excited in their appropriate bandwidths with 1.42 ms long, 1
kHz bandwidth (π/2)H pulses. All experiments used 180° phase
The reaction described in Scheme 3 was initially monitored using
1
spectral windows spanning 800.8 Hz in H and 2500 Hz in 13C. To
permit an enhanced sensitivity, the 13C window sizes were further
reduced to 1500 Hz in subsequent tests. Gradients and bandwidths
were adjusted accordingly, as noted below. Each 0.5 mL
acetophenone-acetonitrile solution was pretuned and preshimmed
prior to injection of a small aliquot (≈56 μL) of Tf2O. A mixing device
was used for this injection, including a syringe feeding directly into the
NMR tube inside the magnet (see Supporting Information). Data
collection was initiated prior to injection of Tf2O. Preliminary tests
monitoring 13C-labeled acetophenone (7) used two scans per 2D
spectrum. Reactions monitoring unlabeled acetophenone (7) were
acquired using 12 scans per FID and a four-step phase cycle, again with
180° phase cycling of both 13C π/2 pulses and of the receiver phase.
Each double-window HSQC data set thus required 4.16 s per scan
when using the 2 s recycle delay. Total times per 2D spectrum were
8.32 and 49.9 s for monitoring of labeled and nonlabeled reactants,
respectively. This back-to-back application of separate ultrafast HSQC
experiments, targeting different windows of interest, permitted the
parallel observation of multiple spectral windows in the same real-time
acquisition batch. While as many as four distinct spectral regions were
monitored in certain runs, only two windows per experiment (Figure
1) are presented in this work. The resulting concatenated time-domain
signals were separated into FIDs corresponding to the individual
2708
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
Figure 2. Real-time 1D 1H NMR series recorded as a function of time (only a small subset of the resulting spectra hereby shown). Colored arrows
show the positions of specific signals from products and intermediates participating in the reaction between acetophenone (7) (0.183 mmol) and
triflic anhydride (0.275 mmol) in acetonitrile-d3 (as both a coreactant and solvent) at 298 K. The red arrow at (2.60 ppm, 0.0 min) indicates the
signal of methyl group from starting acetophenone. Light-blue and purple arrows show signals from final products, aromatic proton H-5 (8.32 ppm,
2.1 min) in pyrimidine (9), and the methylidene protons (5.85 and 5.49 ppm, 51.2 min) in vinyl triflate (10), respectively. Gray (3.28 ppm), blue
(2.11 ppm), green (2.04 ppm), and magenta (1.87 ppm) arrows represent the position of methyl group in different aliphatic intermediates; gold
(5.90/5.57 ppm) arrows present the positions of methylidene protons from olefinic intermediates as well.
Figure 3. Normalized integrals as a function of time, taken from the real-time 1D 1H NMR series shown in Figure 2. The integrals were measured by
taking the characteristic signals for the initial ketone (7, red), the final pyrimidine (9, light blue), vinyl triflate (10, purple), and the intermediate (8,
gray) as well as for other new intermediates detected in these studies. The right-hand panel presents an expanded description of these compounds’
time dependencies.
2709
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
Figure 4. Representative selection of real-time 2D HSQC NMR spectra arising from the reaction of triflic anhydride, ketone (7) and acetonitrile-d3.
Spectra show species (7 and 11−13) present in the aliphatic window range (1.54−2.87 ppm for 1H, 23.7−33.7 ppm for 13C) at key time points as
the reaction progressed, depicted by arrows (red, blue, green, and magenta, respectively) (see text for further details).
windows, then processed into 2D spectra in the usual ultrafast fashion,
and characterized using custom-written Matlab scripts. To aid in the
elucidation of the reaction intermediate structures, H NMR chemical
shifts were estimated using Advanced Chemistry Development, Inc.
(ACD/Laboratories) Software V8.0, and 13C NMR shifts were
calculated within the GIAO approximation16 on the PCM-M06-2X/
6-31+G* (solvent = acetonitrile)17,18 with optimized geometries using
the Gaussian 09 suite of programs.19
show doublets produced by the diastereotopic protons from the
methylidene group. Figure 3 presents a more quantitative
description on the appearance and disappearance of these
different species, showing the time-dependent intensities of
products 7, 9, and 10 as well as of intermediate 8 (cf. Scheme
3). Also shown is the kinetic behavior of an additional number
of aliphatic and olefinic intermediates arising during the
reaction.
1
3.2. Ultrafast 2D Heteronuclear Correlations. The
identities of the intermediates involved cannot be determined
exclusively from 1D 1H NMR experiments. To obtain structural
insight into these peaks, real-time ultrafast NMR techniques
3. RESULTS AND DISCUSSION
3.1. Real-Time 1D Measurements. A variety of
1
preliminary 1D H NMR data were taken to delineate the
spectral windows to be used for the main 2D studies. Figure 2
illustrates a representative set of such tests. In addition to
signals corresponding to reagents and products (Scheme 3),
1
correlating data from H and 13C nuclei were recorded. Based
on the temporal series of 1H NMR experiments, three different
spectral windows were selected for examination by real-time 2D
NMR. One focused on an aliphatic region: 1.54−2.87 ppm for
1H with 23.7−33.7 ppm for 13C; another in an olefinic one:
1
these tests revealed over a dozen transient signals in the H
aliphatic and olefinic regions. These signalsarising on top of
the 1H NMR resonances belonging to reagents and products
suggest the presence of a number of candidate aliphatic and
olefinic intermediates that had not been previously noticed.
Aliphatic intermediates will be characterized by singlet signals
from methyl group present, whereas olefinic intermediates will
1
5.17−6.50 ppm for H with 106.0−116.0 ppm for 13C; and a
1
third in an aromatic/pyrimidinic one: 7.49−8.81 ppm for H
with 112.40−122.40 ppm for 13C. A series of UF-HSQC
spectra was taken for each window of interest; they are
2710
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
Figure 5. Representative selection of the real-time 2D HSQC NMR spectral series arising from the reaction of triflic anhydride, ketone (7), and
acetonitrile-d3. Spectra taken as the reaction progressed show a cross peak (blue arrow) in the aromatic/pyrimidinic window range (7.49−8.81 ppm
1
for H, 112.40−122.40 ppm for 13C) arising from the H5−C5 single bond of the pyrimidine (9, light-blue arrow) (see text for further details).
illustrated in Figures 4 and 5. Colored arrows in these plots
denote cross-peaks representing reagents, products, and
intermediates that participate in the reaction. Cross-peaks
assigned to starting ketone (7) and final products vinyl triflate
(10) and pyrimidine (9) were confirmed by 1D and 2D spectra
(see the Supporting Information). Values of chemical shifts of
highlighted (bold) nuclei are shown in Scheme 4.
cross-peaks at 2.02/28.10 ppm (green arrow) and 1.82/30.30
ppm (magenta arrow), respectively, were clearly detected. The
intensity of these cross-peaks changed with time, rising and
falling during the course of the reaction (see Figure 6).
A series of representative 2D UF-HSQC spectra were
recorded in succession on the olefinic window (5.17 - 6.50
1
ppm, H; 106.0 - 116.0 ppm, 13C) and on the aromatic/
1
pyrimidinic region (7.49 - 8.81 ppm, H; 112.4 - 122,4 ppm,
13C). Throughout the series of 500 HSQC spectra recorded in
these windows, doublet cross-peaks from vinyl triflate (10) and
olefinic intermediates were not detected. Figure 5 illustrates the
2D heteronuclear correlation observed for the aromatic/
pyrimidine window. The spectra for this window are dominated
by a cross-peak (8.33/119.0 ppm, light blue arrow) which
appears with variable intensity in the aromatic region,
corresponding to the H5−C5 single-bond of the pyrimidine
(9). Its assignment to the pyrimidine (9) was confirmed
independently by 1D and 2D standard NMR analyses. Figure 6
summarizes the time-evolution of key cross-peaks observed in
the various windows studied. The aliphatic window (left) shows
the kinetic behavior for reagent acetophenone (7) and
intermediates (11), (12), (13). The aromatic window (right)
reflects the formation of final product pyrimidine (9). (A 3D
animation showing the complete time-series of spectra for the
aromatic window is available for download as Supporting
Information).
Scheme 4
Figure 4 illustrates a series of representative 2D UF-HSQC
spectra recorded for the aliphatic window at ca. 30 s delay. A
total of 500 HSQC spectra was taken in a typical kinetic
progression, where data acquisition was begun immediately
prior to the sudden addition of triflic anhydride to a solution of
ketone (7) in acetonitrile-d3 (Figure 4 highlights only 16 of
these, a full 3D animation made from the total set of HSQC
experiments is given as Supporting Information). Prior to the
Tf2O injection, the cross-peak of the methyl group of (7) is
clearly observed (0 min, red arrow). 1.66 min after initiation of
the reaction a new cross-peak at δ = 2.09/28.40 ppm (blue
arrow) is seen. Then, at ca. 7.48 and 14.97 min, two additional
To further examine the olefinic window, the experiment was
repeated with labeled 13C-carbonyl-acetophenone (7) as the
1
reagent. A spectral window targeting 5.17−6.50 ppm for H
and 102.6−119.2 ppm for 13C permitted the observation and
Figure 6. Integral values versus time-evolution of key cross-peaks as observed for the aliphatic and the aromatic windows studied.
2711
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
Figure 7. Representative selection of real-time 2D HSQC NMR spectra arising from the reaction of triflic anhydride, labeled 13C-methyl-
acetophenone (7), and acetonitrile-d3. Spectra show species (10) and (14/15) present for the olephinic window range (5.17−6.50 ppm 1H, 102.6−
119.2 ppm 13C) indicated by purple and dark-yellow arrows, respectively (top). The left bottom panel shows the time-dependent peak intensities
(integrals) of the vinyl triflate (10) and the newly found intermediates (14/15) which represent the variation in their respective concentrations
1
during the course of the reaction. The right bottom panel shows the olefinic peaks in a H NMR spectrum of isolated and purified vinyl triflate.
comparison of olefinic products and intermediates. A total of
1200 HSQC spectra taken under these conditions within the
course of the reaction revealed, in addition to the expected
doublet of cross-peaks from vinyl triflate (10, purple arrow), a
new doublet of cross-peaks (Figure 7). The rise and fall of these
cross-peaks (dark-yellow arrows) show its character as
intermediate. Figure 7 (bottom) also displays the kinetic
behavior of the new olefinic intermediates (14) and/or (15).
Additionally the presence of a time-increasing cross-peak at
6.20/106.5 was observed. As seen in Figure 7, the kinetics of
the vinyl triflate (10) are in good agreement with those seen
initially in the nonlabeled reaction. For a 3D animation see the
Supporting Information.
electrophilic attack of the S-atom of one of the SO2 groups of
triflic anhydride on the O-atom of the carbonyl group of
acetophenone (7) leading to the rapid formation of an
intermediate (8), whose presence we have detected in the
past by ultrafast 2D HMBC experiments,13c,20 but whose cross-
peaks fall outside the window targeted here. Evolution of (8) by
nucleophilic capture by acetonitrile-d3 leads to the complex
(11), which evolves to the nitrilium-salt intermediate (12).21,22
The reaction continues, and an additional nucleophilic capture
by the acetonitrile-d3 takes place forming intermediate (13). In
order to confirm these intermediates, measurements were
repeated using labeled 13C-methyl-acetophenone (7) and a
standard UF-HSQC sequence,23 which permits monitoring of
(examination) an expanded acquisition window (4.50−0.50
ppm, 1H; 63.60−3.60 ppm, 13C). These results from
monitoring of the methyl labeled reactant (7) confirmed the
formation of intermediate (8) and its evolution to inter-
mediates (11−13) (Figure 9).
4. DISCUSSION
The time-evolution of the peak intensities observed in the
various real-time 1D and 2D spectra illustrated in the preceding
section suggest a complex sequence of events underlying the
evolution of the starting acetophenone (7) to the final
pyrimidine (9). A mechanism capable of rationalizing all the
observed peaks is depicted in Figure 8. The HSQC series
Elimination of triflic anhydride from intermediate (8) would
lead to vinyl triflate (10), which should form as a side product.
Cross-peaks from (10) and/or from other possible olefinic
intermediates like (14), however, were not observed in the
1
recorded on the 1.54−2.87 ppm, H, and 23.7−33.7 ppm, 13C
1
aliphatic window, indicates that the reaction starts with an
HSQC of the olefinic window (5.17−6.50 ppm, H; 106.0−
2712
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
Figure 8. Proposed mechanism for the reaction of triflic anhydride, acetophenone (7), and acetonitrile-d3 in accordance with the structural and
kinetic insight gained from the ultrafast NMR experiments. The NMR results suggest two possible routes (an aliphatic and an olefinic) that can yield
the final product (9) of the reaction.
Figure 9. Representative series of real-time 2D HSQC NMR spectra recorded for the aliphatic window of a reaction between labeled 13C-methyl-
acetophenone (7, red arrow) with triflic anhydride and acetonitrile-d3. From the crosspeaks detected by the UF-HSQC sequence in this spectral
window, the formation of the intermediate (8, gray arrow) and its evolution to the intermediates (11, blue arrow), (12, green arrow), and (13,
magenta arrow) can be monitored (see text for further information).
116.0 ppm, 13C) starting from the unlabeled acetophenone (7).
Still, intermediate (14) must also participate in the reaction,7b
by elimination of triflic acid in intermediate (13). This would
lead to the formation of the crucial double bond present in
(14), in whose absence the cyclization step which leads to the
pyrimidine ring cannot take place. Suspecting that these
intermediates could not be seen due to a transient nature
leading to maximal concentrations below our sensitivity limits,
reactions were repeated using labeled 13C-methyl-acetophenone
(7) and the olefinic window revised with this aid (Figure 7).
The results show the expected double cross-peaks from vinyl
triflate (10) as well as a new pair of cross-peaks representing
diastereotopic protons of a methylidene group (purple and
dark-yellow arrows in Figure 7). We ascribe this new peak’s
origin to the triflate salt intermediates olefinic (14) and/or
(15) containing two and one molecules of the nucleophilic
acetonitrile-d3, respectively. Cross-peaks from intermediates
(14) and/or (15) are present since the first moments of the
reaction (Figure 7); the formation of intermediate (15) would
take place by the elimination of triflic acid from intermediate
(12) and prior to a second nucleophilic capture by the
acetonitrile-d3 which leads to (14). The kinetic behavior of
intermediates (14) and/or (15) (Figure 7) shows that these
species form quickly and then decay, coexisting until the last
moments of the reaction. Intermediate (13) (Figure 8) is the
first step of the so-called aliphatic route,13b whereas
intermediate (15) would be the first of the olefinic and (14)
the common final intermediate. No unambiguous explanation
for the cross-peak at 6.20/106.5 could be found; its presence
under laboratory conditions was not detected.14 The progress
of the reaction through the aliphatic and/or olefinic route and
the relative rate of both pathways will depend on the stability of
intermediate (12).
Overall, the findings arising from these HSQC correlations
reveal for the first time the presence of olefinic intermediates
(14) and/or (15) and confirm preliminary results previously
observed in ultrafast HMBC measurements where labeled
reactants were used.13c In order to get further confirmation on
2713
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
the nature of the detected intermediates, we exploited the
valuable insight arising from the chemical shifts observed along
and complete mechanism involving two different and
concurrent pathwaysone aliphatic and the other olefinic
for the progress of the reaction. The evolution from the starting
ketone to the final pyrimidine could be monitored and
rationalized in high detail. In summary, we believe that real-
time dynamic 2D NMR methodologies of the kind described in
this work could open valuable new ways in the kinetic
monitoring and characterization of complex dynamic chemical
systems and in the elucidation of the variety of mechanisms
underlying these transformations.
1
the correlated H and 13C dimensions. Table 1 shows the
Table 1. Observed and Calculated Chemical Shifts of the
Different Species Involved in the Reaction
observed
a
b
d
(CD3CN)
1H
calculated (CDCl3)
calculated Gaussian 09
13C
13C
1H
f
7
8
2.64
26.33
196.90
23.71
2.60
24,70
f
e
193.05 (216.00)
ASSOCIATED CONTENT
* Supporting Information
■
e
3.26
28.75
S
e
195.2
202.40
Further details regarding experimental procedures, 1D and 2D
spectra of products, and 3D animations. Complete ref 19. This
material is free of charge via the Internet at http://pubs.acs.org.
ac
,
e
9
8.33
5.85
5.49
2.09
2.01
1.84
5.89
5.54
5.89
5.54
119.01
105.76
6.85
5.94
4.50
123.02
f
e
10
103.20 (118.60 )
e
11
12
13
14
28.40
28.05
30.26
114.50
31.21
AUTHOR INFORMATION
Corresponding Author
lucio.frydman@weizmann.ac.il; aherrera@quim.ucm.es
e
■
1.64
2.04
5.08
4.97
5.15
33.60
e
33.30
e
110.70
ACKNOWLEDGMENTS
e
■
15
114.50
146.60
The authors acknowledge financial support from MICINN
(grant CTQ2010-14936), the Israel Science Foundation (ISF
447/09), ERC Advanced grant no. 246754, EU’S BioNMR
grant no. 261863, a Helen and Kimmel Award for Innovative
Investigation, and the generosity of the Perlman Family
Foundation. G.O. acknowledges postdoctoral support from
the Feinberg Graduate School (Dean Fellowship) and from the
Fulbright Foundation. Z.P. acknowledges Universidad Com-
plutense de Madrid for a predoctoral fellowship. We gratefully
4.99
a
b
See the Supporting Information. Averaged values of chemical shifts
c
ACD/Laboratories (release 8.00). Differences between observed and
d
calculated chemical shifts are due to the nature of solvent. Calculated
with TMS (13C = TMS value Gaussian 09 − value 13C compound
e
f
Gaussian 09). Obtained with M06-2X approximation. Obtained with
B3LYP approximation.
experimental chemical shifts arising from the ultrafast series and
compares them against expectations stemming from the various
structures that are postulated to participate in the reaction
́
́ ́
acknowledge Dr. Israel Fernandez and Angel Sanchez (UCM)
for theoretical calculations and predictions of 13C chemical
shifts.
1
scheme proposed in Figure 8. The H NMR chemical shift
estimations were performed for the nitrilium-salt intermediates
(12−15), using the commercial chemical shift calculator
software ACD/Laboratories (release 8.00). 13C NMR shifts
were calculated within the GIAO approximation on the PCM-
M06-2X/6-31+G* (solvent = acetonitrile) with optimized
geometries using the Gaussian 09 suite of programs. In all cases
we find very good agreement between the observed values and
the values predicted/calculated for the postulated structures;
these calculations permit to distinguish between (14) and (15)
and assign (14) to the triflic olefinic nitrilium-salt intermediate
formed.
REFERENCES
■
(1) (a) Pozharskii, A. F.; Soldatenkov, A. T.; Katritzky, A. R. In
Heterocycles in Life and Society; Wiley-VCH: Weinheim, Germany,
1997. (b) Eicher, T.; Hauptmann, S. In The Chemistry of Heterocycles,
2nd ed.; Wiley-VCH: Weinheim, Germany, 2003. (c) Katritzky, A. R.
In Advances in Heterocyclic Chemistry; Academic Press: Oxford, U.K.,
2002; Vol.82.
(2) Rewcastle, G. W. In Comprehensive Heterocyclic Chemistry III;
Aitken, A., Ed.; Elsevier Science: Oxford, U.K., 2008; Vol, 8, pp117.
Ahmad, O. K.; Hill, M. D.; Movassaghi, M. J. Org. Chem. 2009, 74,
8460−8463.
(3) (a) von Angerer, S. In Science of Synthesis; Yamamoto, Y., Ed.;
Thieme: Stuttgart, Germany, 2003; Vol. 16, pp 379. (b) Hill, M. D.;
Movassaghi, M. Chem.−Eur. J. 2008, 14, 6836−6844.
5. CONCLUSIONS
The real-time spectroscopic approaches herein described
allowed us to uncover a new and interesting sequence of
events in the mechanistic pathway of the alkylpyrimidine
formation from aliphatic ketones and alkyl nitriles promoted by
triflic anhydride. This was largely thanks to the advent of new
ultrafast 2D NMR techniques that, operating under low
gradients made possible by the use of selective band excitations,
succeeded in monitoring the organic reaction using unlabeled
compounds. Using this new selective and low-gradient-strength
based UF sequence, it was possible to confirm the presence of
intermediates reported in earlier studies such as (8) and (14) as
well as intermediates (11−13) that were here detected for the
first time. Estimations and calculations of NMR chemical shifts
are in very good agreement with the proposed structures. These
findings have permitted us to propose for the first time a new
́
(4) (a) Herrera, A.; Martínez-Alvarez, R.; Ramiro, P.; Molero, D.;
Almy, J. J. Org. Chem. 2006, 71, 3026−3032. (b) Herrera, A.;
́
Martínez-Alvarez, R.; Martín, N.; Chioua, M.; Chioua, R.; San
́
chez-
́
Vaz
́
quez, A.; Molero, D.; Almy, J. Tetrahedron 2009, 65, 1697−1703.
́
(c) Herrera, A.; Martínez-Alvarez, R.; Martín, N.; Chioua, M.; Chioua,
́
́ ́
chez-Vazquez, A.; Molero, D.; Almy, J. Tetrahedron 2009, 65,
R.; San
5817−5823. (d) Herrera, A.; Martínez-Alvarez, R.; Ramiro, P.;
San
́
́
chez, A.; Torres, R. J. Org. Chem. 2004, 69, 4545−4547.
́
(e) Martínez-Alvarez, R.; Herrera, A. Targets Heterocycl. Syst. 2008,
12, 59−85.
(5) (a) Movassaghi, M.; Hill, M. D. J. Am. Chem. Soc. 2006, 128,
14254−14255. (b) Movassaghi, M.; Hill, M. D. J. Am. Chem. Soc. 2006,
128, 4592−4593. (c) Hill, M. D.; Movassaghi, M. Synthesis 2007,
1115−1119. (d) Movassaghi, M.; Hill, M. D. Nat.Protoc. 2007, 2,
2018−2023. (e) Ahmad, O. K.; Hill, M. D.; Movassaghi, M. J. Org.
Chem. 2009, 74, 8460−8463. (f) Menor-Salvan, C.; Ruiz-Bermejo, M.;
2714
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715

Journal of the American Chemical Society
Article
́
Guzman, G. I.; Osuna-Esteban, S.; Veintemillas-Verdaguer, S. Chem.−
Eur. J. 2009, 15, 4411−4418. (g) Bannwarth, P.; Vailleix, A.; Gree, D.;
Gree, R. J. Org. Chem. 2009, 74, 4646−4649. (h) Yan, S.; Tang, Y.; Yu,
F.; Lin, J. Helv. Chim. Acta 2011, 94, 487−490. (i) Lin, M.; Chen, Q-z.;
Zhu, Y.; Chen, X-l.; Cai, J-j.; Pan, Y-m.; Zhan, Z-p. Synlett 2011,
1179−1183.
(6) Lai, P-S; Taylor, M. S. Synthesis 2010, 1449−1452.
́ ́
(7) (a) García Martínez, A.; Herrera Fernandez, A.; Moreno Jimenez,
F.; García Fraile, A.; Subramanian, L. R.; Hanack, M. J. Org. Chem.
́
1992, 57, 1627−1630. (b) Herrera, A.; Martínez-Alvarez, R.; Chioua,
́
́
M.; Chioua, R.; Sanchez, A. Tetrahedron 2002, 58, 10053−10058.
Chemical transformations induced by triflic anhydride were reviewed,
see: (c) Baraznenok, I. L.; Nenajdenko, V. G.; Balenkova, E. S.
Tetrahedron 2000, 56, 3077−3119.
(8) (a) Kupce, E.; Nishida, T.; Freeman, R. Prog. Nucl. Magn. Reson.
Spectrosc. 2003, 42, 95−122. (b) Kupce, E.; Freeman, R. J. Magn.
Reson. 2003, 162, 300−310.
(9) Hoch, J. C.; Stern, A. S. Method Enzymol. 2001, 338, 159−78.
(10) (a) Gal, M.; Frydman, L. In Multidimensional NMR methods for
the Solution State; Morris, G. A., Emsley, J. W., Eds.; Wiley&Sons:
Chichester, U.K., 2010; pp 43−60. (b) Mishkovsky, M.; Frydman, L.
Annu. Rev. Phys. Chem. 2009, 60, 429−448.
(11) Frydman, L.; Scherf, T.; Lupulescu.. Proc. Natl. Acad. Sci. U.S.A.
2002, 99, 15958−15862.
(12) (a) Gal, M.; Schanda, P.; Brutscher.; Frydman, L. J. Am. Chem.
Soc. 2007, 129, 1372−1377. (b) Gal, M.; Lee, M. K.; Varani, G.;
Frydman, L. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 9192−9197.
(13) (a) Gal, M.; Mishkovsky, M.; Frydman, L. J. Am. Chem. Soc.
2006, 128, 951−956. (b) Herrera, A.; Fernan
́
dez-Valle, E.; Martínez-
z, E.; Gal, M. Angew. Chem.,
́
Alvarez, R.; Molero, D.; Pardo, Z. D.; Sae
́
Int. Ed. 2009, 48, 6274−6277. (c) Herrera, A.; Fernan
́
dez-Valle, E.;
rez, E. M.; Martínez-Alvarez, R.; Molero, D.; Pardo, Z. D.; Saez,
E. Org. Lett. 2010, 12, 144−147.
́
Gutier
́
́
(14) (a) Pelupessy, P.; Duma, L.; Bodenhausen, G. J. Magn. Reson.
2008, 194, 169−174. (b) Shrot, Y.; Frydman, L. J. Chem. Phys. 2009,
131, 224516/1−224516/11. (c) Giradeau, P.; Shrot, Y.; Frydman, L. J.
Am. Chem. Soc. 2009, 131, 13902−13903. (d) Giradeau, P.; Akoka, S.
J. Magn. Reson. 2010, 205, 171−176.
(15) Brunner, P.; Ernst, R. R. J. Magn. Reson. 1979, 33, 83−106.
(16) Wolinski, K.; Hinton, J. F.; Pulay, P. J. Am. Chem. Soc. 1990,
112, 8251−8260.
(17) M06−2X method: Zhao, Y.; Truhlar, D. G. Acc. Chem. Res.
2008, 41, 157−167.
(18) PCM method: (a) Miertus, S.; Scrocco, E.; Tomasi, J. Chem.
́
Phys. 1981, 55, 117−129. (b) Pascual-Ahuir, J. L.; Silla, E.; Tunon, I. J.
̃
Comput. Chem. 1994, 15, 1127−1138. (c) Barone, V.; Cossi, M. J.
Phys. Chem. A 1998, 102, 1995−2001.
(19) Frisch, M. J. et al. Gaussian 09, revision B.01, Gaussian, Inc.:
Wallingford, CT, 2009.
(20) 13C NMR data for carbocations: (a) Breitmeier, E.; Voelter, W.
In Carbon-13 NMR Spectroscopy, 3rd ed.; VCH: New York, 1987.
(b) Stadler, D.; Goeppert, A.; Rasul, G.; Olah, G. A.; Prakash, S.; Bach,
T. J. Org. Chem. 2009, 74, 312−318.
(21) Kanemasa, S. In Science of Synthesis; Murahashi, S., Ed.; Thieme:
Stuttgart, 2001; Vol 19, pp 53−65.
(22) (a) Darbeau, R. W.; Gibble, R. E.; Pease, R. S.; Bridges, D. E.;
Siso, L. M.; Heurtin, D. J. J. Chem. Soc., Perkin Trans. 2 2001, 1084−
1090. (b) White, E. H.; DePinto, J. T.; Polito, A. J.; Bauer, I.; Roswell,
D. F. J. Am. Chem. Soc. 1988, 110, 3708−3709.
(23) Shrot, Y.; Shapira, B.; Frydman, L. J. Magn. Reson. 2004, 171,
163−170.
2715
dx.doi.org/10.1021/ja210154g | J. Am. Chem.Soc. 2012, 134, 2706−2715