Design, validation, and implementation of a rapid-injection NMR system

Article abstract of DOI:10.1021/jo100837a

A Rapid Injection NMR (RINMR) apparatus has been designed and constructed to allow the observation of fast chemical reactions in real time by NMR spectroscopy. The instrument was designed to allow the rapid (<2 s) injection and mixing of a metered volume

Full text of DOI:10.1021/jo100837a

pubs.acs.org/joc
Design, Validation, and Implementation of a Rapid-Injection
NMR System
Scott E. Denmark,* Bruce J. Williams,*,† Brian M. Eklov, Son M. Pham, and
Gregory L. Beutner
Roger Adams Laboratory, Department of Chemistry, University of Illinois, Urbana, Illinois 61801
sdenmark@illinois.edu
Received April 29, 2010
A Rapid Injection NMR (RINMR) apparatus has been designed and constructed to allow the obser-
vation of fast chemical reactions in real time by NMR spectroscopy. The instrument was designed to
allow the rapid (<2 s) injection and mixing of a metered volume of a reagent into a spinning NMR
tube followed by rapid acquisition of the data resulting from the evolution of the chemical process.
The various design criteria for this universal system included the ability to deliver any chemical
reagent at any temperature and allow for the observation of any nucleus. The various challenges
associated with the construction and implementation of this instrument are documented along with
the validation of the accuracy of the apparatus with respect to volume and temperature. Finally, the
ultimate validation and reproducibility of the technique is presented in the form of three case studies
that used the instrument to elucidate various aspects of organic reaction mechanisms. The authors
urge interested parties to not embark on the construction of their own instrument and invite those
whose research problems might be amenable to this kind of analysis to contact the corresponding
author for access to the apparatus described herein.
Introduction
methodology should be evaluated in the context of current
mechanistic paradigms. If the new process can be understood
and rendered predictable within the framework of modern
theory, then the theory has survived a critical test. If the new
process cannot be understood, then the origin of the deviation
should be identified and current theory modified to accom-
modate the new data. In either scenario, a thorough under-
standing of the factors that influence the new chemical trans-
formation (rate, stereoselectivity, substrate scope, byproducts)
is needed. Regrettably, this is rarely the case despite the
spectacular advances over the past decades in spectroscopic,
analytical, and computational methods.
The foundation of modern synthetic organic chemistry rests
firmly on the bedrock that is physical organic chemistry. Without
the unifying principles that correlate structure with reactivity
and selectivity, the universe of synthetic transformations would
seem a bewildering collection of empirical experience. It
is therefore axiomatic that each new advance in synthetic
†Correspondence author for all inquires relating to design and construc-
tion of the instrument. Email: williams@scs.illinois.edu.
5558 J. Org. Chem. 2010, 75, 5558–5572
Published on Web 07/30/2010
DOI: 10.1021/jo100837a
r
2010 American Chemical Society

Denmark et al.
JOCArticle
In the late 1990s a new family of catalytic, asymmetric
transformations was developed in these laboratories that
involved the use of chiral Lewis bases in conjunction with
Lewis acidic silicon compounds to effect carbonyl addition
(allylation, aldolization) and epoxide opening reactions.1
Because the basis of this new type of catalysis was poorly
understood, a comprehensive program designed to elucidate
the critical features of (1) reaction scope and stereoselectivity,2
(2) Lewis base-Lewis acid structure in the solid state and in
solution,3 and (3) structure of reactive intermediates and
reaction kinetics (turnover-limiting and stereochemistry-
determining steps) was launched.4 The first two objectives
could be met by the use of well-established protocols, but the
third represented greater challenges in view of the extraordi-
nary rapid rates of reaction. Of course, many experimental
techniques have been developed to determine the kinetic
behavior of reactions in solution and to identify reactive
intermediates including stop-flow UV-vis spectroscopy5
and in situ IR spectroscopy.6 Although each of these
techniques has their individual advantages and limitations,
they can provide a wealth of information when applied
appropriately. Indeed, early in these studies, the use of in
situ IR spectroscopy allowed real-time monitoring of slower
reactions (section 7.1). This instrument performs quite well
for substrates with strong infrared absorbances and where
reaction concentrations can be adjusted to obtain an appro-
priate signal-to-noise ratio without extending analysis times.
However, for faster reactions executed at high dilution,
in situ IR analysis was ineffective in obtaining meaningful
kinetic data. Moreover, this technique pales in com-
parison to high field NMR spectroscopy for the structural
determination of reactive intermediates and determination
of kinetic profiles in temperature ranges from -150 to þ
150 °C. In addition, the availability of highly sensitive probes
made NMR spectroscopy an ideal analytical tool for inter-
rogating chemical transformations at high dilution in real
time.
amount of a temperature-equilibrated reagent into a tem-
perature-equilibrated substrate while the sample is spinning
in the NMR spectrometer and ready for acquisition, all
while coordinating the injection event with data acquisi-
tion. In addition, care must then be taken to ensure that the
data is collected in a manner that produces integral values
that are not only internally consistent (resonance to reso-
nance within a single spectrum) but also consistent between
the plethora of spectra collected over the course of an
individual experiment. An instrument that addresses
these challenges was introduced already 30 years ago by
McGarrity,7 and since then the technique has been imple-
mented in a number of laboratories. Our objectives were to
take the basic concept and redesign the instrument for use
in high sensitivity, modern NMR spectrometers and to
address 21st century problems in organic reaction mecha-
nisms.
Background
Since the initial reports by McGarrity and co-workers,7
Rapid Injection NMR (RINMR) has become an invaluable
tool for the study of reaction mechanisms. The original
McGarrity apparatus consists of an injection assembly con-
taining a gastight syringe coupled to a long injection capillary
terminated with a perforated bulb. Once completely lowered
into the magnet, the bulb rests just below the reaction sample
in the NMR tube. The syringe is driven by a pneumatic piston
whose course is set by a micrometer screw. Movement of the
piston triggers the spectrometer pulse and acquisition seq-
uence. A major drawback of this design is that the syringe
and capillary remain stationary throughout the experiment
allowing diffusion of the injection liquid from the capillary
bulb into the reaction medium inside the spinning NMR
tube. Additionally, the injection assembly and capillary
causes a perturbation of the magnetic field homogeneity
resulting in peak broadening. Although this effect is minimal
under slow acquisition conditions, peak broadening is un-
avoidable for fast reactions. Even with these limitations,
various groups have successfully utilized the McGarrity
technique to observe reactive intermediates as well as to
determine reaction kinetics.8 Further demonstrating its uti-
lity, Klein and Gawley reported using the RINMR technique
originally developed for proton observation to study the
kinetics of tin-lithium transmetalation using 119Sn NMR
analysis.9
Yet despite the exquisite sensitivity, ability to operate
over a wide range of temperatures, and wealth of structural
information, NMR spectroscopy is experimentally a much
more complicated technique. The manifold difficulties are
immediately apparent and include rapidly delivering a known
(1) (a) Denmark, S. E.; Beutner, G. L. Angew. Chem., Int. Ed. 2008, 47, 1560–
1638. (b) Denmark, S. E.; Fujimori, S. In Modern Aldol Reactions; Mahrwald,
R., Ed.; Wiley-VCH: Weinheim, 2004, Vol. 2, Chapter 7. (c) Denmark, S. E.;
Stavenger, R. A. Acc. Chem. Res. 2000, 33, 432–440. (d) Denmark, S. E.;
Barsanti, P. A.; Beutner, G. L.; Wilson, T. W. Adv. Synth. Catal. 2007, 349,
567–582.
(2) (a) Denmark., S. E.; Chung, W.-j. J. Org. Chem. 2008, 73, 4582–4595.
(b) Denmark, S. E.; Ghosh, S. K. Tetrahedron 2007, 63, 8636–8644.
(c) Denmark, S. E.; Fan, Y.; Eastgate, M. D. J. Org. Chem. 2005, 70,
5235–5248. (d) Denmark, S. E.; Beutner, G. L.; Wynn, T.; Eastgate, M. D.
J. Am. Chem. Soc. 2005, 127, 3774–3789. (e) Denmark, S. E.; Fujimori, S.;
Pham, S. M. J. Org. Chem. 2005, 70, 10823–10840.
(3) (a) Denmark, S. E.; Eklov, B. M. Chem.;Eur. J. 2008, 14, 234–239.
(b) Denmark, S. E.; Fu, J. J. Am. Chem. Soc. 2003, 125, 2208–2216.
(c) Denmark, S. E.; Su, X. Tetrahedron 1999, 55, 8727–8738.
(4) (a) Denmark, S. E.; Eklov, B. M.; Yao, P. J.; Eastgate, M. D. J. Am.
Chem. Soc. 2009, 131, 11770–11787. (b)Denmark, S. E.;Pham, S. M;Stavenger,
R. A.; Su, X.; Wong, K.-T.; Nishigaichi, Y. J. Org. Chem. 2006, 71, 3904–3922.
(b) Denmark, S. E.; Bui, T. J. Org. Chem. 2005, 70, 10393–10399. (c) Denmark,
S. E.; Pham, S. M. Helv. Chim. Acta 2000, 83, 1846–1853.
(7) (a) McGarrity, J. F.; Prodolliet, J.; Smyth, T. Org. Magn. Reson. 1981,
17, 59–65. (b) McGarrity, J. F.; Ogle, C. A. J. Am. Chem. Soc. 1985, 107,
1805–1810. (c) McGarrity, J. F.; Prodolliet, J. J. Org. Chem. 1984, 49, 4465–
4470. (d) McGarrity, J. F.; Ogle, C. A.; Brich, Z.; Loosli, H. R. J. Am. Chem.
Soc. 1985, 107, 1810–1815.
(8) (a) Frye, S. V.; Eliel, E. L.; Cloux, R. J. Am. Chem. Soc. 1987, 109,
1862–1863. (b) Chen, X.; Hortelano, E. R.; Eliel, E. L.; Frye, S. V. J. Am.
Chem. Soc. 1992, 114, 1778–1784. (c) Palmer, C. A.; Ogle, C. A.; Arnett,
E. M. J. Am. Chem. Soc. 1992, 114, 5619–5625. (d) Reetz, M. T.; Raguse, B.;
€
Marth, C. F.; Hugel, H. M.; Bach, T.; Fox, D. N. A. Tetrahedron 1992, 48,
5731–5742. (e) Geletneky, C.; Forsterling, F.-H.; Bock, W.; Berger, S. Chem.
€
Ber. 1993, 126, 2397–2401. (f) Ogle, C. A.; Johnson, H. C., IV; Wang, X. L.;
Strickler, F. H.; Bucca, D.; Gordon, B., III. Macromolecules 1995, 28, 5184–
5191. (g) Bertz, S. H.; Carlin, C. M.; Deadwyler, D. A.; Murphy, M. D.; Ogle,
C. A.; Seagle, P. H. J. Am. Chem. Soc. 2002, 124, 13650–13651. (h) Mayr, H.;
Ofial, A. R.; Sauer, J.; Schmied, B. Eur. J. Org. Chem. 2000, 2013–2020.
(i) Bertz, S. H.; Cope, S.; Murphy, M.; Ogle, C. A.; Taylor, B. J. J. Am. Chem.
Soc. 2007, 129, 7208–7209.
(5) Wang, R.-Y. In Applications of Physical Methods to Inorganic and
Bioinorganic Chemistry; Scott, R. A., Lukehart, C. M., Eds.; John Wiley &
Sons: Chichester, 2007; pp 469-487 .
(6) Rein, A. J.; Donahue, S. M.; Pavlosky, M. A. Curr. Opin. Drug
Discovery Dev. 2000, 3, 734–742.
(9) Klein, R.; Gawley, R. E. J. Am. Chem. Soc. 2007, 129, 4126–4127.
J. Org. Chem. Vol. 75, No. 16, 2010 5559

Denmark et al.
JOCArticle
Within the past few years, more sophisticated RINMR
instruments have been developed to obviate the limitations
of the original design as well as to incorporate features
needed for a specific application. One instrument, developed
by Mok and co-workers, includes the option for the insertion
of an optical fiber for use in photo-CIDNP experiments.10
Aside from the optical fiber, this device consists of two major
parts, a coaxial glass insert fitted inside the NMR tube and a
pneumatic injector positioned outside of the magnet. Unique
from other devices, the insert is made from a piece of glass
capillary tubing fused on one end with a micropipet. The
micropipet is positioned just below the surface of the solution
in the NMR tube. During injection, the solution held within
the capillary is jetted through the micropipet to improve
mixing. The use of 5-mm Shigemi tubes further enhances
mixing efficiency and greatly reduces the required reaction
volume. Because the entire insert is made from glass, it is
reasonable to assume that its presence causes minimal dis-
turbance to the field homogeneity.
More recently, an advanced device was reported by Reich
and co-workers.11 In this design, the solvent delivery is
similar in practice to the McGarrity device, a syringe housed
above the magnet coupled to a long flexible Teflon needle
inserted into the NMR tube, but this instrument is designed
to house two syringes allowing for multiple injections during
a single experiment. Perhaps the most unique aspect of this
device is the presence of a mechanical stirrer. The motor is
mounted above the magnet, between the syringe housings,
and provides efficient mixing down to the solvent freezing
point. The raising and lowering of the stirrer paddle as well as
injections from either or both syringes can be controlled
manually or by the spectrometer pulse program. The authors
comment that the instrument was designed to operate with
10-mm NMR tubes to enhance sample sensitivity in multi-
nuclear experiments. While the increased sample size may be
advantageous in some situations, it is an unfortunate draw-
back when more precious reagents are required or when
minimal sample volumes are ideal.
FIGURE 1. Rapid Injection NMR apparatus. (A) Optical sensors
for injector travel stops. (B) Pneumatic valve system to actuate injec-
tor piston. (C) Ivek metering pump. (D) Reservoir flask and inert
atmosphere valves. (E) Electronics module. (F) Metering pump
controller. (G) Injector tip. (H) Injector rinse adapter.
2. Construction. The extreme demands on performance set
out above and the highly interdisciplinary nature of the project
would require the coordinated efforts of many individuals.
Fortunately, the design, fabrication, and spectroscopic faci-
lities at the University of Illinois are outstanding. A team was
assembled that included an electrical engineer (B.J.W.),
spectroscopist (Paul Molitor), machinist (Bill Knight), and
designers/users (S.E.D. and S.M.P.). The first step in the
process was to secure the RINMR apparatus developed by
McGarrity, which was kindly provided by Eliel, who has
employed it for many studies.8 Although inspirational, this
apparatus was deemed too crude for the objective outlined
above, and accordingly, a totally new design was implemen-
ted. The final, functional apparatus first came on line in 1999
and is depicted in Figure 1. What follows is a detailed des-
cription of the design, optimization, construction, validation,
and application of this apparatus.
Results
1. Design Criteria. Although the basic concept and design
of an RINMR apparatus has already been implemented, we
envisioned the development of a tool that would far surpass
the existing technology in a number of ways. First, to be a
truly universal tool, the apparatus must be able to deliver a
known amount of any reagent at any temperature under an
inert atmosphere and allow for the observation of any nucleus
within seconds of injection. Such demands required accurate
metering, rapid mixing, and high chemical resistance. Moreover,
to be compatible with high field instruments (500-600 MHz 1H)
and deliver the high resolution available from such machines,
an injection/mixing mechanism was needed that removed the
injector from the sample. At the outset of our studies (ca. 1997),
no such apparatus was commercially available or described
in the literature.
3. Components. 3.1. Design of Fluid Systems. 3.1.1. Reservoir.
The reagent storage system was designed around a standard,
100-mL pear-shaped flask (Figure 2). To ensure that a wide range
of materials including air- and moisture-sensitive reagents could
be safely stored, the reservoir was designed to include a gas inlet
that allowed the reservoir to be attached to a standard vacuum
(10) Mok, K. H.; Nagashima, T.; Day, I. J.; Jones, J. A.; Jones, C. J. V.;
Dobson, C. M.; Hore, P. J. J. Am. Chem. Soc. 2003, 125, 12484–12492.
(11) (a) Jones, A. C.; Sanders, A. W.; Bevan, M. J.; Reich, H. J. J. Am.
Chem. Soc. 2007, 129, 3492–3493. (b) Jones, A. C.; Sanders, A. W.; Sikorski,
W. H.; Jansen, K. L.; Reich, H. J. J. Am. Chem. Soc. 2008, 130, 6060–6061.
5560 J. Org. Chem. Vol. 75, No. 16, 2010

Denmark et al.
JOCArticle
FIGURE 2. Close-up of the injectant reservoir. (A) Three-way valves
controlling liquid and gas flow. Left line is for equilibration makeup,
and right goes to the metering pump. (B) Cover to flask access port.
(C) Gas inlet. Three-way valve allows gas to be sent into the reservoir
flask or either line (for cleaning). (D) Kovar-to-glass union. (E) Reser-
voir flask and dip tubes.
FIGURE 3. Close-up of the reservoir and Ivek metering piston
pump. (A) Three-way valve to allow pumping or siphoning from
the reservoir flask, E. (B) Metering pump. (C) Three-way valves that
control liquid flow from the reservoir. (D) Inert gas inlet. (E) Reser-
voir flask and dip tubes.
manifold, and the headspace could be purged of oxygen and
moisture using standard techniques. The reagent pickup lines,
which feed into the metering system, were manufactured from
stainless steel and extend into the bottom of the flask to ensure
that a maximum of the solution stored in the reservoir could be
used. Perhaps the most challenging part of designing the reser-
voir system was envisioning a safe and reliable method for
attaching the glass flask to the stainless steel cap, which contains
the gas inlet and solvent delivery lines. The final solution was to
fit the flask with a glass-to-Kovar to stainless steel union. The
necessary threads could then be carefully machined onto the
stainless steel extension providing an airtight seal (O-ring) bet-
ween the cap and flask. The entire reservoir system was mounted
onto a brushed aluminum plate. The plate was then easily atta-
ched or removed from the RINMR cabinet, allowing the reser-
voir to be mobile without having to move the entire apparatus.
This modular design allowed for easy cleaning and when
necessary, a rapid exchange of different reservoirs. The gas and
solvent lines connected to the reservoir were fitted with three-way
valves to provide an easy shutoff and to allow for the solution to
bypass the pump. The importance of this bypass will be discussed
in detail below.
3.1.2. Injection Pump. The reservoir feeds directly to the meter-
ing pump, the first stage of the delivery system (Figure 3).
Although the initial intent was to build the metering device, it
quickly become apparent that it would be extremely difficult to
manufacture a device that could accurately meter microliter
quantities of a wide range of reagents. Instead, a third-party
device was available that surpassed the design criteria.12 Although
the metering pump is available in various piston sizes, the largest
piston was chosen for ease of priming. This system is quite
flexible, allowing the selection of both the injection volume and
the rate of injection.
3.1.3. Hydraulics. The most mechanically complex aspect
of the RINMR apart from the metering pump is the reagent
delivery system. In the resting position, the injector tip was
located inside the NMR tube, ca. 20 mm above the solvent
level. Injection requires the downward movement of the injec-
tor tip to the bottom of the NMR tube with simultaneous
delivery of the reagent as the injector descends and as the
solution and tube spin around it. Thus, two pneumatic actua-
tors were installed to drive the injector tip into the solution
and then return it to its resting position above the sample
(Figure 4). The actuators were designed to operate separately
so that the rates of downward and upward movement could
be independently controlled.13 Additionally, three optical sen-
sors were installed: one at the upper end of travel (UEOT),
one at the bottom end of travel (BEOT), and one in the middle.
The middle optical sensor initiates the injection sequence
for the downstroke. The BEOT optical sensor reverses the
movement of the inject piston, while the UEOT optical
sensor ends the process and sends an acquisition pulse to
the spectrometer. The electronics control module allows for
the injection to take place on either the down or upstroke.
If upstroke is selected, the bottom optical sensor initiates
the upstroke injection sequence, and the middle sensor is
then rendered inactive. The whole of the operation is covered
in detail in Section 3.2 (Design of the Electronic Control
Module).
(12) Digispense 2000 fitted with a size A alumina ceramic piston and an
optional gland seal. The size A piston is rated from 1 to 2000 μL with 0.1 μL
resolution. Contact IVEK Corp. at (800) 356-4746 or visit www.ivek.com.
(13) The rates of travel are controllable, but increasing or decreasing them
too far can cause the pistons to not seat properly, leaving them in an
intermediary position and unable to move without resetting them manually.
J. Org. Chem. Vol. 75, No. 16, 2010 5561

Denmark et al.
JOCArticle
FIGURE 5. (A) Polycarbonate guide tube with spacer (black).
(B) Outer titanium guide tube. (C) Kel-F sleeve. (D) Ten-millimeter
injector tip, near full extension.
FIGURE 4. Pneumatic systems. (A) Electro-optical gates to con-
trol actuators. (B) Pneumatic piston location. (C) Pneumatic valve
system. (D) Independent pressure adjustment controls for both the
up and down strokes.
connected to the titanium tube. The housing for the linear
actuator is machined from aluminum and built to be wider
than the spin stack, preventing the guide tube from contacting
the sample spinner. Thus, the actuator housing rests at the very
top of the spin stack and transfers the entire weight of the
delivery apparatus to the magnet housing.
Although every care was taken to minimize the effects of a
strong magnetic field on the RINMR apparatus, it was soon
discovered that the NMR magnet could paralyze the actuators.
Through trial and error, an operating distance of roughly
5-6 ft was found to be necessary to obviate the effects of the
magnet.
The injector must travel precisely down the center of the
tube to ensure consistent rates of travel and tube spin.
Because of the very tight tolerances inherent in this system,
a physical alignment of the injector tube was needed to correct
for deviations from trueness. In practice, the alignment of
each new injector tube was checked before use and then
periodically thereafter, especially if the exposed portion of
the injector tube came into contact with anything during
handling. This process was accomplished by placing the injec-
tor tube into an upper stack of the magnet in a custom-built
stand discussed in section 4.3 and extending the injector tip
until it was flush with the bottom of the stack. The alignment
was then noted, and the injector tube was carefully bent by
hand to compensate. This process was repeated until the
injector was situated in the center of the opening at the
bottom of the stack. Then, the alignment could be verified by
placing an NMR tube in the stack. The tube was then spun at
a constant rate and the injector was actuated. No change in
the spin rate should be observed if the alignment is proper.
3.1.5. Evolution of the Injector Tip. Perhaps the most cru-
cial and highly developed portion of the RINMR apparatus
is the injector tip. During the initial period of development,
the importance of proper tip design for accurate reagent delivery
as well as for rapid mixing was not appreciated. Throughout
the development process, each tip was milled from titanium
3.1.4. Design of Injector. The injector tip and tube are manu-
factured from high-grade titanium,14 which was chosen for its
low magnetic susceptibility and chemical resilience.15 The
titanium injector tube is housed in a 1.04-m long high density
polycarbonate tube fitted with guides to prevent accidental
bending (Figure 5). The outer diameter of the guide tube is
milled slightly smaller than the inner diameter of the upper
stack and extends out above the magnet can. This tight fit and
the internal guides are crucial in maintaining the trueness of the
injector tube. Any curvature along the length of the titanium
tube results in an off-axis movement of the injector tip leading
to contact between the injector tip and the wall of the NMR
tube during an injection cycle. In the best case, this contact
slows the spin rate of the sample; in the worst case, this contact
can shatter the tube.16
The pneumatic valve that actuates the injector mechanism
is built directly onto the top of the guide tube and is physically
(14) Titanium tubing was made of A40 (<0.04% iron), 0.08300 o.d., 0.06300
i.d., 6000 long.
(15) Although more costly and more difficult to machine than stainless
steel or brass, initial testing with stainless steel tubing resulted in poor line
shape, even at room temperature.
(16) In practice, small changes to this lateral alignment can be made
manually as needed.
5562 J. Org. Chem. Vol. 75, No. 16, 2010

Denmark et al.
JOCArticle
FIGURE 6. Several RINMR tip designs. (A) Intermediate design without mixing paddle. (B) Final design with mixing paddle. Note that the
beveled edge allows for reseating inside the Kel-F sleeve. (C) Cutaway view of the final tip design and actual picture.
and then gold brazed onto the end of the titanium injector
rod. In all, over 10 different tips were machined and tested.
The designs were evaluated by metered injection of ethanol
into a standardized solution of CHCl3 in CDCl3 at room
temperature and if successful, then again at -60 °C. The level
1
of mixing efficiency was determined by comparison of H
integrations after a single injection and then after shaking the
NMR tube. The initial design consisted of a simple flared cir-
cular tip with a single delivery port at the bottom (Figure 6).
Evaluation of this design revealed accurate delivery albeit
extremely poor mixing efficiency after a single injection. To
improve mixing, the single port was modified to three deli-
very ports each 0.025 mm (0.01000) in diameter spaced radi-
ally at 120° around the waist of the tip. A Teflon cap was also
added to cover the ports to minimize the effects of diffusion
during equilibration. It was believed that the cap would slip
down and away from the tip during fluid delivery. Although
the results improved somewhat, the mixing efficiency was by
no means ideal, and it was quickly discovered that the amount
of fluid delivered was not sufficient to create enough pressure
to remove the Teflon cap. In a subsequent design, a fourth
delivery port was added to the bottom of the tip to redistri-
bute the pressure on the cap. The results were a significant
improvement in mixing efficiency at room temperature.
However, in the final design, an S-shaped mixing paddle
was added to the bottom of the tip containing three radially
spaced delivery ports. The Teflon cap was replaced with a
Teflon sleeve attached to the guide tube that fitted around
the ports to provide a seal during setup and equilibration.
When delivering reagent, the tip slides down and away from
the stationary sleeve and then resets back inside the sleeve at
the end of the injection cycle. Evaluation of this design showed
excellent mixing efficiency at room temperature. Surpri-
singly, similar tests at -60 °C resulted in little or no delivery
of ethanol. Further testing confirmed that equilibration of
the injector assembly at subambient temperatures resulted in
the contraction of the solvent up the injector tip as it cooled.
To prevent this, a separate line was connected to the delivery
line, bypassing the metering pump.17 This “make-up” line
was attached to the delivery line via a 3-way valve and
allowed for fluid to be directly removed from the reservoir
during equilibration. Although Teflon was initially used for
the injector tip sleeve, its pliability forced the use of new sleeves
after only 4 or 5 injections. Kel-F was found to be a superior
FIGURE 7. Electronic control module (ECM).
replacement to Teflon, lasting nearly 100 cycles before requi-
ring replacement. With the tip, sleeve, and makeup line in
place, fluid could be routinely delivered with excellent mixing
over a broad temperature range.
Although the Kel-F sleeve effectively limited atmospheric
exposure, proper cleaning (rinsing with solvents used, followed
by abs EtOH, and rigorous air drying) is required. Over time,
the tip can become clogged, and sonication while pulling a
vacuum on the interior of the injector tube was effective in
clearing clogged ports.18 It should be noted that hundreds of
injections have been run with solutions containing SiCl4 (which
gives SiO2 upon hydrolysis) before this rigorous cleaning was
necessary when the injector was properly rinsed after use.
3.2. Design of the Electronic Control Module. The signals
necessary to operate the linear actuator and the Ivek fluid
pump are generated in the electronic control module (ECM,
Figure 7) whose design is illustrated as a block diagram in
Figure 8. The process begins with a trigger pulse generated by
the spectrometer (from a BNC connector labeled SP1 on the
back of the Inova equipment console). This trigger pulse (T)
is sent through a coaxial cable to an INPUT TRIGGER BUFFER
located in the ECM. The buffer generates a local trigger that
is generated either by the spectrometer pulse or by a button
on the ECM (labeled “calibrate”). This button allows for
local initiation of the inject procedure for purposes of cali-
brating inject volumes, linear actuator speeds, or cleaning
(17) Later designs also required a slight modification: A small amount of
the inside of the sleeve was removed to compensate for the braise point where
the tip was attached to the injector tube.
(18) The wires used for cleaning microliter syringes can also be useful in
clearing clogged ports.
J. Org. Chem. Vol. 75, No. 16, 2010 5563

Denmark et al.
JOCArticle
FIGURE 8. Wiring diagram (A) and waveforms (B) for the electronic control module.
the system tubing. The trigger pulse then goes to the DOWN
SOLENOID DRIVER (DSD), which starts the linear actuator
moving down at velocity vdown (waveform K). The speed of
this velocity is controlled by a valve located between the gas
source and the linear actuator, which has a thumb-screw
is called the upper end of travel (UEOT) optical sensor and
senses when the actuator is at the upper end of travel.
Initially, the UEOT phototransistor is turned off as the beam
interrupter is resting in the UEOT gap. As the linear actuator
begins its travel downward, the beam interrupter moves out
of the gap, which causes the phototransistor to turn on and
waveform A to go from high to low.
The beam interrupter then approaches the inject optical
sensor (IOS) causing waveform B to go from low to high and
back to low again as the beam interrupter passes through.
This pulse is sent to the DOWNSTROKE INJECT LOGIC’s “trigger”
input that generates waveform H. If the INJECT MODE switch
on the front panel of the ECM is set for downstroke inject,
waveform H is sent to the INJECT DRIVER’s “begin inject” input
adjustment (Figure 4) for effecting changes to vdown
.
Three optical sensors are located at the top of the sam-
ple inject tube (Figure 1A). These sensors consist of a light-
emitting diode (LED) shining on a phototransistor across a
gap that allows the beam to be interrupted, thus turning the
phototransistor on or off. These sensors are mounted on a
circuit board that remains fixed while the beam interrupter,
which is attached to the linear actuator, moves through the
gap between the LED and the phototransistor. The top sensor
5564 J. Org. Chem. Vol. 75, No. 16, 2010

Denmark et al.
JOCArticle
to begin the inject sequence (waveform Ndown). The volume
of the fluid injection and the rate at which it is delivered are
controlled by the Ivek pump controller, which is a separate
unit. There are UP/DOWN timers on the ECM that individually
time the up and down stroke. This helps facilitate the setting
for the fluid injection rate on the Ivek pump.
The beam interrupter then arrives at the bottom end of
travel (BEOT) optical sensor causing waveform C to go from
low to high. This simultaneously disables the DSD and en-
ables the UP SOLENOID DRIVER (USD, waveform K going from
high to low and waveform P going from low to high) causing
the linear actuator to instantly reverse direction and wave-
form C to go from high to low. At this point the injection is
finished. As the linear actuator begins its upward travel at
velocity vup, controlled by a valve similar to that for vdown
shown in Figure 4, mixing by the S-shaped paddle located on
the injector tip (Figure 6) begins in earnest. The pulse
generated at waveform C disables the DOWNSTROKE INJECT
LOGIC, so that it cannot send an inject pulse as the beam
interrupter passes through the IOS on the return trip to the
UEOT. (Waveform C is also sent to the INJECT MODE switch.
If the INJECT MODE switch on the front panel of the ECM is
set for upstroke inject, the DOWNSTROKE INJECT LOGIC is by-
passed, and waveform C is sent to the INJECT DRIVER to begin
the inject sequence, waveform Nup.)
The beam interrupter then arrives at the UEOT optical
sensor causing waveform A to go from low to high. This then
disables both the USD and the INJECT DRIVER. Waveform A is
also sent to the ACQUISITION OUTPUT’s “acquisition trigger”
input. This new trigger, generated by the ACQUISITION OUTPUT
LOGIC, is then sent to the “external clock” input on the DATA
ACQUISITION BOARD of the spectrometer via a coaxial cable
(waveform Q). At this point the rapid sample inject sequence
is complete. The spectrometer then resumes the experiment
and the acquisition process begins.
4. Validation and Calibration. 4.1. Volume. Initial injec-
tion volume calibrations were performed outside of the magnet
by injecting a solution of ethanol in CDCl3 into a solution of
dichloromethane in CDCl3 at room temperature. The sam-
ples were mixed, the integrated NMR spectra of these samples
were acquired, and the integral data was compared. Once
assured that the metering pump was accurately and reprodu-
cibly delivering the desired amount of solution under nominal
conditions, the volume calibrations were repeated at -60 °C
inside the magnet, and integrals were measured repeatedly
over ca. 1 min after injection. The samples were then removed
from the magnet and shaken, and the NMR data were reac-
quired. Unfortunately, this comparison showed that the
samples had to be shaken before the correct integral values
were obtained. This disparity indicated that the metering
pump was delivering the appropriate volumes but that in-
adequate mixing was occurring inside the magnet. This solu-
tion to this problem is described in Section 4.3 (Mixing).
After optimization of the injection sequence to provide a
homogeneous sample (see below), the volume calibration
was repeated in quadruplicate at -60 °C inside the magnet,
and the integration data were again compared (Figure 9).
The comparison of the measured to injected volumes was
linear as expected, and the data was stable over ca. 1 min.
Thus, the desired volumes of injectant were being accurately
and reproducibly delivered to the sample tube, and mixing
was complete and rapid.
FIGURE 9. Plot of the measured injection volume vs the volume
selected for injection.
4.2. Temperature. A variety of temperature parameters
were studied. First, the thermocouple in the probe was cali-
brated against a standard Varian methanol temperature cali-
bration standard to ensure the accuracy of the readings. Next,
the temperature gradient needed to be considered. While the
temperature at the thermocouple was controlled at -70 °C,
the temperature farther from the thermocouple was uncer-
tain. To understand the change in temperature gradient up
the stack, a thermocouple was inserted into the probe through
the guide tube. The temperature at various locations could
then be measured by varying the position of this thermo-
couple. Placing this thermocouple at the location of the
injector tip above a sample that was maintained at -70 °C
revealed a temperature ca. 10 °C warmer, even over this short
distance (ca. 50 mm). Cooling gas flows were increased to
20 L/min to minimize the gradient. It should be noted that
even with a 10 °C difference in temperature, a typical injec-
tion would increase the sample temperature by less than 1 °C.
4.3. Mixing. Whereas injecting a given volume rapidly and
reproducibly was straightforward to achieve, creating a homo-
geneous sample thereafter proved more difficult. Initial
calibration runs demonstrated that injection did not give a
homogeneous solution, and integral values tended to drift
with time after the injection as further (slow) mixing occurred.
To solve this problem required direct observation of the mix-
ing process. Thus, to facilitate this evaluation, a stand was
constructed for the upper stack of the instrument so that an
injection into a spinning sample outside the magnet could be
accomplished and the results observed visually (Figure 10).
Because injections of a colored dye solution occurred too
rapidly to evaluate with the unaided eye, high-resolution
video recording was required. Video data collected from
each run could then be inspected in slow-motion to observe
the course of the injection and the adequacy of sample
mixing. Some freeze-frame images acquired during this
process can be seen in Figure 11, and a sampling of the video
clips is included in the Supporting Information.
A problem that was discovered while analyzing the video
data is a delay in the onset of injection. The inject optical
sensor (IOS, Section 3.2) was positioned to actuate the Ivek
metering pump just as the injector tip reached the top of the
liquid in the NMR tube. However, for unknown reasons, the
delivery of reagent was delayed, resulting in a concentration
gradient at the bottom of the tube. The video data showed
J. Org. Chem. Vol. 75, No. 16, 2010 5565

Denmark et al.
JOCArticle
FIGURE 11. Examples of poor (left) and good (right) mixing of an
injected dye solution into 10-mm (bottom) NMR tubes. The center
image shows the injector tip near full extension. Numbers below
each frame illustrate the injection program used: total injection volume
(μL) at injection rate (μL/s)-no. of seconds to travel down-no.
of seconds to travel up. Center and right-hand samples were spun at
20 Hz. Left-hand sample was not spun to illustrate the added mixing
that occurs when the spinning sample encounters the s-shaped paddle
on the bottom of the injector.
A paramagnetic relaxation agent (such as Cr(dpm)3) can be
added to decrease the T1 relaxation times. In general, acquir-
ing data in a quantitative manner slows data acquisition and
only needs to be done when quantification is required. Quali-
tative data can be acquired much more rapidly.
Data sets were acquired as arrayed experiments. This
protocol facilitated data storage and manipulation by creat-
ing a single file that could be processed as a group. That is,
one spectrum could be phased, line broadened, and inte-
grated, and then these properties could be propagated to all
of the spectra from the experiment. Once the data was collec-
ted, a spectrum would be phased, the appropriate peaks
would be integrated to at least (2.5 times the line width to
ensure the integral region contained all of the signal intensity
and, in some cases, baseline-corrected. In some cases, each
spectrum would be phased independently. Changes in phase
with time was not uncommon and could be corrected for. In
practice, a number of different workup scripts were written
so that they could be called in rapid succession to determine
which parameters needed to be adjusted to control phase and
baseline drift and give accurate integral values. These scripts
also automated retrieval and tabulation of integral data into
a format that could be easily imported into a spreadsheet
program for plotting and further analysis.
FIGURE 10. Stand manufactured to hold the upper stack and allow
visual inspection of the injection process. The RINMR instrument is in
storage configuration in the background. The injector is stored upright
to minimize warping along the length of the titanium injector tube.
that the mixing paddle effectively distributed the gradient,
but to ensure homogeneity, the timing of the injection was
advanced such that the pump was actuated slightly before the
tip entered the solution in the sample tube.
The RINMR system has some limitations with respect to
mixing speed, some of which are related to the technique, and
some of which are specific to the instrument. For example,
the time required to create a homogeneous sample in a 5-mm
NMR tube is longer than that in a 10-mm tube, which became
apparent when attempting to sample at ca. 3 s intervals using
a 5-mm tube. Additionally, the injection cycle for 10-mm
tubes was optimized at a total time of 4 s. This parameter
limits how rapidly the first data point can be collected to
about 3 s after the beginning of the injection.
6. Reliability. It should be noted that over the course of this
instrument’s lifetime (now ca. 12 years) no probe has been
damaged, no instrument time has been lost, and not even a
single sample tube has been broken in the magnet. This record
is a testament to the tight design tolerances and the reliability
in use this instrument demonstrates.
7. Implementation. 7.1. Trichlorosilyl Enol Ether Kinetics.
Previous publications from these laboratories illustrated the
utility of chiral Lewis base catalysis in asymmetric aldol addi-
tions of trichlorosilyl enol ethers.19 For example, the combi-
nation of cyclohexanone trichlorosilyl enol ether (1) with
benzaldehyde (2) in the presence of chiral phosphoramides
(Scheme 1) affords aldol products with high diastereoselec-
tivities and moderate to high enantioselectivities. During the
5. Data Acquisition and Manipulation. A standard Varian
pulse sequence (s2pul) was modified to include the produc-
tion of a trigger pulse sent to the ECM and to wait for a
trigger pulse from the ECM before beginning the acquisition
(see Section 3.2 above). A conditional statement was inclu-
ded so that this exchange of triggers occurred only prior to
acquisition of the first spectrum of the arrayed set. Acquisi-
tion parameters were modified to obtain quantitative inte-
gral data. Namely, 90° pulse widths and T1 times were
determined. The pulse width was set to 22.5°, 45°, or 90°,
and the recycle delay time was set to 3-5 times the longest
T1 time (3ꢀ for 22.5°, 4ꢀ for 45°, and 5ꢀ for a 90° pulse).
(19) (a)Denmark, S. E.;Fujimori, S. In ModernAldol Reactions; Mahrwald,
R., Ed.; Wiley-VCH: Weinheim, 2004; Vol. 2. Chapter 7. (b) Denmark, S. E.;
Stavenger, R. A. Acc. Chem. Res. 2000, 33, 432–440. (c) Denmark, S. E.;
Fujimori, S.; Pham, S. M. J. Org. Chem. 2005, 70, 10823–10840.
5566 J. Org. Chem. Vol. 75, No. 16, 2010

Denmark et al.
JOCArticle
TABLE 1. Arrhenius Activation Energies for Aldol Additions of Trichlorosilyl Enol Ethers Determined by ReactIR and RINMR Analysisa
conditions
method
Ea (kcal mol-1
)
A (M-1 s-1
)
ΔHq (kcal mol-1
)
ΔSq (cal mol-1 K-1
)
ΔGq (kcal mol-1
)
1 þ 2, 5 mol % 5
1 þ 2, 5 mol % 5
1 þ 2, 10 mol % 4
ReactIR
RINMR
RINMR
0.9 ( 0.1
1.8
2.1 ( 0.5
5.1 ( 1.7
0.4
224 ( 100
0.3 ( 0.1
1.2
1.5 ( 0.5
-67.1 ( 0.7
-63.3
-51.9 ( 1.0
20.6 ( 0.2
20.4
17.3 ( 1.0
aActivation parameters were calculated for T = 303 K.
course of optimization it was discovered that the diastereo-
selectivity of the aldol addition is highly sensitive to the
bulkiness and loading of the catalyst. To explain this behavior,
a mechanistic rationale was formulated in which the reaction
proceeds through independent pathways to the two diastereo-
mers, such that the operative pathway is dependent upon
catalyst size and concentration. The central tenet of this hypo-
thesis is that the two pathways differed in the number of cata-
lyst molecules in the rate- and stereochemistry-determining
transition structures. Thus, pathways involving either one or
two phosphoramides are available, the former being opera-
tive with bulky catalysts or low catalyst concentration,
to afford syn-adducts from 1, whereas the latter is operative
with less sterically encumbered catalysts and higher catalyst
concentrations to give anti-adducts. To differentiate these
two pathways and validate the hypothesis, a series of kinetic
studies were undertaken.20
following injection of the desired amount of 2. Experiments
were generally conducted in triplicate to ensure reproduci-
bility. Using this protocol, Arrhenius parameters were de-
termined over a temperature range of -11 to -65 °C.
RINMR analysis provided good agreement with the pre-
vious IR studies, Table 1.21
Thus secured, the more challenging task of studying reac-
tions promoted using the more reactive phosphoramide 4
was undertaken. To determine the order in catalyst, the rate
of the aldol addition between 1 and 2 was measured as a
function of the concentration of 4. Injections were performed
at -80 °C using a 10-fold range in catalyst concentration.
Data points were collected at 1-s intervals for no less than
five half-lives. Plotting the data as log(kobs) vs log([catalyst])
revealed a second-order dependence (m = 2.113) on phos-
phoramide 4, Figure 12. This provided the direct evidence for
a mechanistic dichotomy between catalyst 4 and 5, namely,
that reactions using 4 proceed through a two phosphoramide
pathway whereas reactions using 5 proceed through a one
phosphoramide pathway. On the basis of these results, two
different rate equations could be formulated:
SCHEME 1
dð3Þ
2
¼ k½enol etherꢁ½aldehydeꢁ½4ꢁ
dt
and
dð3Þ
¼ k½enol etherꢁ½aldehydeꢁ½5ꢁ
dt
Arrhenius parameters were also determined using 10 mol % 4
over a 20 °C temperature range, between -70 and -90 °C,
Table 1. Attempts to extend this temperature range while
maintaining CD2Cl2 as the bulk solvent proved unsuccessful
because of extremely fast reaction rates for temperatures
above -65 °C. Conversely, decreasing the temperature below
-90 °C resulted in highly viscous solutions and presented the
possible risk of sample freezing. Comparing the Arrhenius
activation parameters for the two catalyzed reactions shows
a remarkable dependence of ΔGq on ΔSq. Reactions with
high entropic costs are often rationalized by virtue of a highly
Initial efforts using simple GC analysis and ReactIR
techniques were successful in studying the addition of 1 to
2 catalyzed by phosphoramide 5 and clearly established first-
order dependence on [1], [2], and [5]. Unfortunately, similar
studies using phosphoramide 4 were inconclusive because of
limited instrument sensitivity and the enhanced rates afforded
by this catalyst. To overcome these obstacles, the RINMR
instrument offered the opportunity to secure more accurate
kinetic measurements. To validate the utility of the newly
constructed RINMR, the addition of 1 to 2 catalyzed by 5
(previously studied by in situ IR methods) was reexamined.
These experiments were performed in a 5-mm NMR tube
containing enol ether 1 and the phosphoramide in CD2Cl2.
Benzaldehyde (2) was added via the injector as a solution in
CD2Cl2 solution. In general, a 1H NMR spectrum was acqui-
red immediately prior to injection to ensure that no leakage
had occurred. Data points were then rapidly recorded
FIGURE 12. Plot of log(kobs) vs log([catalyst]) for the addition of 1
to benzaldehyde catalyzed by 4 at T = -80 °C. The graph depicts
the linear fit to f(x) = mx þ b (m = 2.113, R2 = 0.992).
(20) (a) Denmark, S. E.; Pham, S. M. Helv. Chim. Acta 2000, 83, 1846–
1853. (b) Denmark, S. E.; Pham, S. M.; Stavenger, R. A.; Su, X.; Wong,
K. T.; Nishigaichi, Y. J. Org. Chem. 2006, 71, 3904–3922.
J. Org. Chem. Vol. 75, No. 16, 2010 5567

Denmark et al.
JOCArticle
ordered transition structure or a highly unfavorable pre-
equilibrium. In the case involving these aldol additions, one
or both of these may be operative.
exquisite ability to collect meaningful kinetic data at cryo-
genic temperatures with air-sensitive reagents.
The basics of this reaction are quite simple. The reaction
displays partial first-order kinetics in both aldehyde and silyl
ketene acetal, as expected for a simple aldol addition reac-
tion. Control experiments showed no rate dependence upon
the concentrations of the added acid scavenger (DIPEA) or
upon the added paramagnetic relaxation agent [Cr(dpm)3].
All of the systems studied also showed zero-order kinetics
with respect to SiCl4 concentration. While at first glance this
is counterintuitive, it simply illustrates that the actual cata-
lytic species is in fact the Lewis acid-Lewis base complex.
Because the Lewis base is catalytic and the Lewis acid is used
in stoichiometric quantities, the quantity of the complex in
solution is defined only by the concentration of the Lewis base.
Thus, RINMR kinetics unambiguously determined a second-
order dependence for aldol reactions with phosphoramide 4
and a remarkable entropic dependence on ΔGq for these reac-
tions. It is also interesting to note that the preparative results
indicate that the reactions involving catalyst 4 are more enantio-
selective as compared to phosphoramide 5, suggesting that
the 2:1 (phosphoramide/silicon) pathway is more discrimina-
ting. To observe the enhanced stereoselectivity of the active
2:1 complex, one must also conclude that the 2:1 complex is
more kinetically viable than the 1:1 complex. From a mecha-
nistic standpoint, reaction through the 1:1 pathway should
only be available when the 2:1 pathway is inaccessible. If so,
increasing the local concentration of the second Lewis basic
unit should greatly enhance the 2:1 pathway over the less
desirable 1:1 pathway. These conclusions led to the formula-
tion of highly active and selective bisphosphoramides as
catalysts for the aldol additions of silyl enol ethers derived
from aldehydes, ketones, esters, nitriles, and conjugated
ketones, esters, and amides to aldehydes.
7.2. Propionate Silyl Ketene Acetal Aldol Kinetics. A long-
standing question in Lewis base catalyzed aldol additions is
the mechanism of Lewis base activation of the weak Lewis
acid, SiCl4. Empirically, the combination of a Lewis base
with a weak Lewis acid led to rapid and highly selective
addition of silyl ketene acetals to aldehydes (e.g., Scheme 2).
A typical preparative reaction is complete within minutes at
-78 °C, necessitating the utmost rapidity in data collection
to ensure accurate kinetic data. As an example, Figure 13
shows a typical data set from a kinetic analysis of the reaction
shown in Scheme 2. Inspection of the data in Figure 13
demonstrates the versatility of the technique: both the dis-
appearance of 7 and the appearance of 8 can be tracked. No
intermediates are formed over the course of the reaction,
indicated by the lack of any transient signals in the NMR
spectrum over time.
FIGURE 13. Aliphatic region of the reaction of 6 and 7 in the
presence of SiCl4 and catalyst 9c. Spectra were recorded every 7 s for
ca. 30 min. In this example, the reaction t1/2 is ca. 3 min.
Of greater interest were the striking differences in the cata-
lytic Lewis base partial reaction orders. The hypothetical cata-
lytic cycle, drawn by analogy to the Lewis base catalyzed allyla-
tion reaction, proposed a 2:1 Lewis base to Lewis acid complex
as the active catalyst system (Scheme 3).22 According to this
hypothesis, monodentate phosphoramides (cf., 10, 11) would
display second-order partial rates, while the bidentate phos-
phoramides would display first-order kinetics. To our surprise,
the actual partial rate data was much more complex, point-
ing to the more complex nature of the catalyst resting state.
The monomeric catalyst 11 displayed a first-order partial
rate dependence, whereas HMPA (10) displayed a complex
2/3 partial order, each different from the expected second-
order behavior. It is a testament to the reproducibility and
sensitivity of the instrument that first and 2/3 partial reaction
orders could be accurately differentiated.
SCHEME 2
The dimeric catalysts 9a and 9b each display partial first-
order kinetic behavior, as expected. However, dimeric cata-
lysts 9c and 9d displayed 1/2 order kinetic behavior, when
(21) The RINMR numbers for 5 are a single measurement and therefore
do not include quantified errors. Nevertheless, all of the energy parameters
(Ea, ΔHq, and ΔGq) differ from the accepted ReactIR numbers by less than
1 kcal/mol.
(22) For a summary see: Denmark, S. E.; Eklov, B. M.; Yao, P. J.;
Eastgate, M. D. J. Am. Chem. Soc. 2009, 131, 11770–11787. and references
therein.
A full kinetic analysis of the Lewis base catalyzed aldol
addition was recently published.4a The study of this system
proved to be an extraordinarily practical showcase for this
instrument; a full kinetic analysis of this system with a number
of different catalyst systems demonstrated the instrument’s
5568 J. Org. Chem. Vol. 75, No. 16, 2010

Denmark et al.
JOCArticle
SCHEME 3
SCHEME 4
these were also expected to display first-order rate dependencies.
These four catalysts differ only in the length of the alkyl linker
separating the two phosphoramide units. The fact that varying
the length of the tether was changing the partial reaction order
of the catalyst was very surprising to say the least.
7.3. meso Epoxide Opening Kinetics. The advantages of
rapid injection NMR as a noninvasive technique for the
study of reaction kinetics are nicely illustrated in the case of
the Lewis base catalyzed opening of meso epoxides with
silicon tetrachloride (SiCl4). An earlier publication from
these laboratories had shown that the chiral phosphoramide
3 gives good yields and enantioselectivities in the desymme-
trization process with a range of meso epoxides (Scheme 5).24
Since then, a variety of diverse catalyst structures including
n-oxides25 and phosphine oxides26 have been disclosed that
promote this transformation, but few have achieved signifi-
cant improvements in terms of scope or selectivity. To further
understand the catalytic process and find a logical path forward
for optimizing selectivity with these phosphoramide-containing
catalysts, kinetic studies were needed to elucidate the mecha-
nism of the reaction and understand the role of the catalyst.
Unfortunately, the rapid rate and moisture sensitivity of this
reaction make it an extremely difficult case for traditional
analytical techniques. Sampling by syringe for high perfor-
mance liquid chromatography (HPLC) or gas chromatogra-
phy (GC) would allow the aliquot to rapidly warm to room
temperature, leading to results that are not representative of
the true state of the reaction at -78 °C. Furthermore, remov-
ing an aliquot by syringe could lead to hydrolysis of SiCl4,
opening a competitive and unselective Brønsted acid cata-
lyzed pathway and further complicating analysis.
The exquisite granularity of this data allowed the deriva-
tion of a unified reaction mechanism that is consistent with
all of these different catalyst behaviors (Scheme 4). The data
is consistent with the earlier proposed 2:1 Lewis base to Lewis
acid complex as the active catalyst in all of these systems. It is
the nature of the catalyst resting states that required modi-
fication. For the simplest catalyst, 11, this resting state is a
2:1 complex, as a mixture of neutral and cationic species.
Because the 2:1 complex is already formed in the resting
state, no additional 11 is required to reach the transition
state, and a first-order rate is observed. The same 2:1
complex is required in the transition state with HMPA, but
in this case the resting state is a 3:1 complex.3a The observed
2/3 order rate dependence demonstrates that one of the three
HMPA must dissociate from the complex before reaching
the transition state.23 The dimeric catalysts 9a and 9b are
readily explained by our original hypothesis: these phos-
phoramides are bidentate in both the ground and transition
states and hence display partial first-order rate dependencies.
Dimeric catalysts 9c and 9d both displayed 1/2 order partial
rates. By postulating that the same intermediate is involved
for all four of the dimeric phosphoramides 9a-d, the 1/2
order rate demands a dimeric resting state. The longer tethers
in 9c and 9d allow these phosphoramides to bridge between
two silicon atoms and form large macrocyclic resting states,
which then break down to form two active catalysts. The
proposed resting states discussed above have all been sup-
ported by spectroscopic observations (29Si and 31P NMR),
and the large macrocyclic dimers are precedented.4a
Initial attempts to monitor the reaction using in situ infra-
red spectroscopy as a noninvasive technique did address the
precision and reproducibility issues associated with HPLC
(24) Denmark, S. E.; Barsanti, P. A.; Wong, K.-T.; Stavenger, R. A.
J. Org. Chem. 1998, 63, 2428–2429.
(25) (a) Tao, B.; Lo, M.-C.; Fu, G. C. J. Am. Chem. Soc. 2001, 123, 353–
354. (b) Nakajima, M.; Saito, M.; Uemura, M.; Hashimoto, S. Tetrahedron
ꢀ
Lett. 2002, 43, 8827–8829. (c) Malkov, A. V.; Gordon, M. R.; Stoncius, S.;
ꢀ
Mussain, J.; Kocovsky, P. Org. Lett. 2009, 11, 5390–5393.
(26) (a) Tokuoka, E.; Kotani, S.; Matsunaga, H.; Ishizuka, T.; Hashimoto,
S.; Nakajima, M. Tetrahedron: Asymmetry 2005, 16, 2391–2392. (b) Pu, X.; Qi,
X.; Ready, J. M. J. Am. Chem. Soc. 2009, 131, 10364–10365.
(23) In other words, the order in HMPA is the ratio of the molecularity in
the transition state divided by the molecularity in the ground state.
J. Org. Chem. Vol. 75, No. 16, 2010 5569

Denmark et al.
JOCArticle
SCHEME 5
SCHEME 6
to be zero-order. Upon going from 1 to 10 equiv of SiCl4,
only small changes in initial rate were observed, and overall
first-order behavior was maintained throughout. In order to
determine the order in HMPA, changes in the first-order rate
constant (kobs) were examined within the range of 10-20 mol %
catalyst (Figure 14D). This narrow range of catalyst con-
centrations was employed because of the rate of these reac-
tions. Despite these limitations, this concentration range still
contains the synthetically relevant catalyst loadings.
Plotting ln[HMPA] against ln(kobs) revealed that the order
in HMPA was 2.2 (R2 = 0.9769). On the basis of all these
results, the following rate equation for the epoxide opening
was determined: d(2)/dt = kobs[SiCl4]0[12b]1 (where kobs
k[HMPA]2).
=
and GC monitoring. However, because no strong IR absor-
bance was available in the molecule of interest, only the weak
absorbance of the C-O bond in the epoxide could be used. In
the end, this technique proved unreliable because of signal
overlap, and these studies were abandoned.
By comparison, RINMR was a more promising technique
for this study because it represents a noninvasive technique
that can detect more diagnostic signals for these structures.
Initial experiments confirmed that the high levels of signal
dispersion accessible on a 500 MHz (1H) NMR instrument
would allow for accurate monitoring of reaction conversion
without the kind of problems associated with signal overlap
encountered with ReactIR.
Having confirmed the viability of this method, the kinetic
experiments were performed in a 10-mm NMR tube with a
solution of 1.1 equiv of SiCl4 and HMPA as catalyst in
CD2Cl2 at -75 °C (Scheme 6).27 The epoxide 12b was then
added from the injector as a solution in CDCl3, generating a
solution with an overall concentration of 0.2 M. First, the
order of each reagent was determined using the integral
method with initial rate analysis.28 Each data point was
reproduced in triplicate. Minor variations between runs were
observed, likely due to a competitive Brønsted acid catalyzed
pathway, but the data was sufficiently reliable to draw some
important conclusions regarding the reaction mechanism.
Employing the integral method to analyze the raw data
revealed that the reaction exhibited good, overall first-order
behavior (Figure 14A and B). Overall second-order behavior
was ruled out by comparison of graphs of ln(12bt) and
1/(12bt) versus time out to 70% conversion (R2(ln(12bt)) =
0.9977 vs R2(1/(12bt) = 0.9421). The linearity of the ln(12b)
versus time plot was clearly higher than that of the 1/(12b)
versus time plot (Figure 14C). A similar conclusion could be
drawn from plots of ln(1 - 13bt) and 1/(1 - 13bt) out to 70%
conversion. Employing the method of initial rates with data
collected to 15% conversion, the order in SiCl4 was revealed
The results of the RINMR kinetic study of the epoxide
opening suggest a mechanism that is consistent with the
larger picture of Lewis base catalyzed-Lewis acid activated
reactions such as the silyl ketene acetal addition described in
Section 7.2. Furthermore, these observations are consistent
with studies by Fu on the N-oxide catalyzed opening of meso
epoxides with SiCl4.25a Foregoing studies have shown the
existence of a 3:1 HMPA/SiCl4 resting state (coordinatively
saturated complex i, Scheme 7).3a Dissociation of one mole-
cule of HMPA from i would then generate the coordinatively
unsaturated, cationic silicon complex ii. Because the rate
dependence on [HMPA] observed for the epoxide opening of
12b is second-order, the equilibrium between i and ii must be
extremely rapid compared to subsequent steps in the cataly-
tic cycle (k2 and k-2 . k3). This conclusion is supported by
the fact that the addition of silyl ketene acetal 7 is at least
20-fold faster than the epoxide opening.29 Therefore the dis-
sociation of HMPA, which is kinetically significant in the
case of the silyl ketene acetal addition (giving rise to a 2/3
order in HMPA), is simply a pre-equilibrium step and does
not factor into the rate equation in this case, leading to the
simple second-order dependence on HMPA. In other words,
the concentration of the coordinatively unsaturated complex
ii is at steady state. Attack of the chloride ion of the activated
complex iii would then occur to afford the product complex
iv, which would release the catalyst and the silylated product
14, allowing for catalyst turnover.
The knowledge that two phosphoramide molecules are
present in the active chiral Lewis acid complex suggested that
a dimeric phosphoramide, such as (R,R)-9ac, might provide
higher selectivity. However, when this hypothesis was tested,
it was found that the dimeric phosphoramide (R,R)-9c actu-
ally gave lower selectivity in the reaction with 12b, suggest-
ing that mechanistic differences between these two catalyst
structures may exist (Scheme 8). So even though these studies
did not lead to an improved phosphoramide catalyst for
(27) Denmark, S. E.; Barsanti, P. A.; Beutner, G. L.; Wilson, T. W. Adv.
Synth. Catal. 2007, 349, 567–582.
(28) Schmid, R.; Sapunov, V. N. Non-Formal Kinetics; Verlag Chemie:
Weinheim, 1982.
(29) Approximate half-lives for the two reactions under similar conditions
support this claim: aldolization of 7, t1/2=ca. 80 s (0.020 M in 7,-60 °C); epoxide
opening of 12b, ca. 1800 s (0.025 M in 12b, -75 °C).
5570 J. Org. Chem. Vol. 75, No. 16, 2010

Denmark et al.
JOCArticle
FIGURE 14. Kinetic plots for opening of 12b: (A) overall kinetic profile, (B) first-order analysis, (C) second-order analysis, (D) order in HMPA.
SCHEME 7
SCHEME 8
how mechanistic insights, even those that may seem beyond the
scope of traditional methods, can be vital to the design of new
and valuable synthetic methods.
desymmetrization, these results did lead to the understand-
ing that the complex i was a powerful in situ generated chiral
Lewis acid catalyst that might have broader applications to
other electrophiles. In subsequent years, these kinetic studies
served as the starting point for the development of a wide
array of highly selective, asymmetric Lewis base catalyzed-
Lewis acid promoted processes.30 This outcome nicely illustrates
Discussion
1. Realization of Design Goals. At the outset of this enter-
prise, we set ourselves ambitious goals for the capabilities
of this instrument, and these goals were all realized. The uni-
versality of the injection system allows for the use of any rea-
gent that is compatible with the glass reservoir, the ceramic
piston of the metering pump, the Teflon delivery lines, and
the titanium injection tube. Indeed, in addition to the rea-
gents described in the case studies above, we have also injec-
ted highly reactive compounds such as organomagnesium
and organolithium reagents as well as strong Lewis acids
(30) (a) Denmark, S. E.; Beutner, G. L.; Wynn, T.; Eastgate, M. D. J. Am.
Chem. Soc. 2005, 127, 3774–3789. (b) Denmark, S. E.; Fan, Y. J. Org. Chem.
2005, 70, 9667–9676. (c) Denmark, S. E.; Heemstra, J. R., Jr. J. Org. Chem.
2007, 72, 5668–5688. (d) Denmark, S. E.; Wilson, T. W.; Burk, M. T.;
Heemstra, J. R., Jr. J. Am. Chem. Soc. 2007, 129, 14864–14865.
J. Org. Chem. Vol. 75, No. 16, 2010 5571

Denmark et al.
JOCArticle
such as tin tetrachloride. The modular construction of the
reservoir allows for these reagents to be transferred in the
laboratory under an inert atmosphere and then transported
to the NMR laboratory and connected to the apparatus
without compromising the integrity of the reagent by expo-
sure to the atmosphere. Accurate calibration allowed access
to different reaction temperatures that are limited only by the
variable temperature controller in the NMR instrument.
The greatest challenge in the design was the ability to accu-
rately and reproducibly deliver a known amount of a reagent
for kinetic measurements. Although many iterations of the
injection tip were required, in the end, the collection of repro-
ducible kinetic data in a number of case studies provided
confidence that even that goal was achieved.
2. Limitations. A number of limitations are readily appa-
rent. First, the time delay after injection and before data
accumulation represents a built in limitation for the analysis
of extremely fast reactions. Currently, efficient injection and
mixing can be accomplished at a minimum of 1 s. Adding a
second of delay before acquisition means that a 2-s blockout
window is imposed. In addition, the pulse frequency is limited
by the relaxation times (T1) of the relevant nuclei. This
limitation is general for all NMR techniques and will be of
importance only when quantitative measurements are desired.
Finally, whereas multiple injections of a given reagent can
be made to observe, inter alia, catalytic turnover, reaction
robustness, and influence of stoichiometry, the current design
allows only two components to be mixed.
3. Design Modifications. Both up and down linear actuator
speeds are set with manually adjusted valves. Arriving at the
desired values can take some time as there is no way to know,
without trial and error, what the speeds are. If not correct on the
first try, a manual adjustment is made to the thumbscrew on the
valve. Several tries may be necessary for each direction. Also,
the actuators are gas-operated valves, and since gas is very com-
pressible, linear operational speed is more a goal than a reality.
The responsiveness of the actuators could be improved
with digitally controlled, hydraulically operated valves. These
valves would be accompanied by the necessary electronic
control to set the speeds to the desired value within a certain
resolution (0.1 s). This reproducibility could be achieved
with a hydraulically operated valve because fluid is not gene-
rally compressible.
A design element that is unique to this system, but not a
limitation, is the significant dead volume required to deliver
the reagent from the reservoir to the injector. Because the
reagent is delivered at this distance via a metering pump and
the injector has to be lifted vertically to insert into the mag-
net, the Teflon line connecting the pump to the titanium tube
is rather long (5-6 ft.) and the overall dead volume of the
system is ca. 6 mL. Thus it is prudent, if possible, to inject the
less precious component of a reaction since only several
microliters are actually used. A possible solution would be
to have a volume of the expensive reagent located on the
inject housing. The Ivek fluid pump could use an inexpensive
fluid to drive a mechanical cylinder located in the tubing that
would then drive the more expensive reagent. The two fluids
could be isolated in this way. Leakage and accuracy pro-
blems would have to be overcome but are not insurmoun-
table. This problem is also avoided with shielded magnets.
Conclusions
The design, construction, validation, and implementation
of a Rapid Injection NMR apparatus have been described.
The apparatus has proven reliable and robust in a number of
mechanistic studies wherein currently available methods for
the acquisition of accurate kinetic data were not suitable.
The instrument has been designed to provide universal appli-
cation for the real time analysis of fast chemical reactions.
The authors urge interested parties to not embark on the
construction of their own instrument and invite those whose
research problems might be amenable to this kind of analysis
to contact the corresponding author for access to the appa-
ratus described herein.
Acknowledgment. The authors are grateful to the National
Science Foundation for generous financial support (NSF
CHE-9803124, 0105205, 0414440 and CHE 0717989).
Supporting Information Available: Detailed schematics
and plans for construction of the instrument, volume and
temperature calibration data and spectra, and a selection of
movies (in QuickTime format) that illustrate the injection
and mixing process. This material is available free of charge
via the Internet at http://pubs.acs.org.
5572 J. Org. Chem. Vol. 75, No. 16, 2010