The stability of N, N -carbonyldiimidazole toward atmospheric moisture

Article abstract of DOI:10.1021/op400281h

N,N-Carbonyldiimidazole (CDI) is known to be sensitive to degradation by atmospheric moisture. This work details some mechanistic aspects of CDI degradation by atmospheric moisture along with the major contributing factors to degradation rate. Also, several analytical techniques for the measurement of CDI purity that are less cumbersome than the traditional gas-capture assay are described.

Full text of DOI:10.1021/op400281h

Article
pubs.acs.org/OPRD
The Stability of N,N-Carbonyldiimidazole Toward Atmospheric
Moisture
Kenneth M. Engstrom,* Ahmad Sheikh, Raimundo Ho, and Robert W. Miller
GPRD Process Research and Development, AbbVie, Inc., 1 North Waukegan Road, North Chicago, Illinois 60064, United States
ABSTRACT: N,N-Carbonyldiimidazole (CDI) is known to be sensitive to degradation by atmospheric moisture. This work
details some mechanistic aspects of CDI degradation by atmospheric moisture along with the major contributing factors to
degradation rate. Also, several analytical techniques for the measurement of CDI purity that are less cumbersome than the
traditional gas-capture assay are described.
followed over time. The traditional method for assaying the
purity of CDI involves measuring the amount of carbon dioxide
evolved on hydrolysis.4 Development of other more convenient
methods was pursued and resulted in development of distinct
1H NMR- and HPLC-based assays. The HPLC-based assay
utilizes derivatization of CDI with benzyl alcohol as shown in
Scheme 1, followed by quantitation of dibenzyl carbonate in the
INTRODUCTION
■
N,N-Carbonyldiimidazole, commonly referred to as CDI, is
most often used in chemical synthesis to convert a carboxylic
acid functional group into a reactive acylating agent, particularly
for the formation of amide bonds.1 CDI is widely used in the
pharmaceutical industry for the preparation of both clinical
candidates and in the manufacture of several marketed drugs.2
It is an attractive reagent for use on large scale because it is
relatively inexpensive, safe, and commercially available. Also,
the carbon dioxide and imidazole reaction byproducts resulting
from its use are innocuous and readily purged from process
streams.
Scheme 1. Quantitation of CDI through derivatization with
benzyl alcohol
CDI can be prepared via reaction of 4 equiv imidazole with 1
equiv phosgene in THF.3 The hygroscopic imidazole hydro-
chloride reaction byproduct is removed via filtration. CDI is
then isolated from the filtrate via crystallization by evaporation
of the solvent and/or use of a nonpolar antisolvent. Several
notes in the synthetic procedure hint that CDI is sensitive to
atmospheric moisture, although no data quantifying this
sensitivity is provided.
Recently, we experienced a significant drop in efficiency of
acyl imidazolide formation in a CDI-mediated amide bond-
forming reaction at scale during transfer of the process from
our development facilities to the commercial manufacturing
site. Circumstantial evidence suggested degradation of CDI
during charging by atmospheric moisture to form imidazole and
carbon dioxide as one potential cause. This work details our
attempt to quantify the major contributing factors to the
degradation rate of CDI on exposure to atmospheric moisture.
This work, enabled by our development of several analytical
techniques for the measurement of CDI purity, led us to a
partial understanding of some of the mechanistic aspects of
CDI degradation, along with discovery of CDI attributes which
determine the relative degradation rates of individual lots of
CDI. This knowledge has practical implications regarding the
effective use of CDI in a manufacturing setting and may enable
more efficient processes by others using CDI at manufacturing
scale.
reaction mixture using a purchased dibenzyl carbonate
standard. In our case, direct quantitation through H NMR
was more expedient.
5
1
Degradation Rates and Mechanistic Considerations.
The purity, water content, and weight loss of samples of CDI
exposed to 35% relative humidity at 22 °C were followed over 8
1
h. The degradation rate as measured by H NMR appears to
follow zero-order kinetics, suggesting that CDI itself is not
involved in the rate-determining degradation step (Figure 1).
Additionally, the water content of the material as measured by
Karl Fischer titration remained constant at 0.03 wt % (roughly
0.003 equiv) throughout the experiment, implying that any
water adsorbing onto the solids was reacting immediately.
Finally, the degradation rate as measured from sample weight
1
loss closely matches that determined by H NMR, suggesting
that release of carbon dioxide also follows zero-order kinetics
and is not involved in the rate-determining degradation step.
Interestingly, after 24 h the final CDI sample from the
1
experiment in Figure 1 contained no detectable CDI by H
NMR assay and 0.3% CDI according to weight loss. This result
demonstrates that the initial CDI content of a lot may be
measured simply from total sample weight loss after full
degradation. Although this weight loss assay does not provide a
1
purity determination as quickly as the H NMR- and HPLC-
RESULTS AND DISCUSSION
■
Analytical Methods. We first determined how the
degradation of CDI to imidazole and carbon dioxide could be
Received: October 9, 2013
Published: March 28, 2014
© 2014 American Chemical Society
488
dx.doi.org/10.1021/op400281h | Org. Process Res. Dev. 2014, 18, 488−494

Organic Process Research & Development
Article
Figure 1. Degradation of CDI with time at 35% RH and 22 °C.
Figure 2. Degradation rate of CDI with variable relative humidity.
based assays, its simplicity and accuracy should make it an
attractive method for measurement of CDI purity.
Figure 4, an increase in exposed surface increased degradation
rate. In all cases the rate of degradation increases when the
variables related to exposure of CDI to atmospheric moisture
are increased, suggesting water is involved in the rate-
determining step of CDI degradation.
In order to further understand the degradation reaction
pathway, an attempt was made to follow the transformation of
CDI to imidazole through single-crystal X-ray diffraction by
exposing a crystal of CDI to a controlled RH environment over
time. However, no intermediate structures were detected
through changes in cell parameters, and at the end of the
experiment parameters consistent with those of imidazole were
calculated. SEM images of the samples before and after
degradation shown in Figure 5 reveal substantial reduction in
particle size and deterioration in morphology. Crystal structure
Experiments were carried out on the same lot of CDI under
conditions in which the amount of atmospheric moisture (i.e.,
water concentration) supplied to react with the CDI samples
was varied. In the first experiment, measurement of the
degradation rate at different relative humidities was performed.
As shown in Figure 2, an increase in relative humidity increased
degradation rate. In a second experiment, measurement of the
degradation rate with variable air flow was performed. As
shown in Figure 3, an increase in air flow increased degradation
rate. In the final experiment, measurement of the degradation
rate for a constant mass of material with variable exposed
surface was performed by placing the same amount of CDI in
separate circular containers of variable diameter. As shown in
489
dx.doi.org/10.1021/op400281h | Org. Process Res. Dev. 2014, 18, 488−494

Organic Process Research & Development
Article
Figure 3. Degradation rate of CDI with variable air flow.
Figure 4. Degradation rate of CDI with variable exposed surface.
descriptors in Table 1 reveal that the true density value of
imidazole crystals is greater than that of CDI crystals, primarily
because of the H-bonding interactions that exist in the former
(Figure 6). The increase in true density helps explains breakage
and deterioration in morphology.
However, as shown in Table 2, no direct relationship between
specific surface area and degradation rate was found. Therefore,
counter to expectations, specific surface area is not the sole
variable which determines the relative degradation rates of
different lots of CDI.
Lot-to-Lot Degradation Rate Variability. The studies
detailed above were all performed using the same lot of CDI.
However, we observed that different lots of CDI had variable
degradation rates under the same conditions as shown in Table
2. All lots of CDI displayed similar melting points of 118−122
°C and the same pattern by PXRD analysis, indicating the
crystal form was not the source of degradation rate variability.
Specific surface area is also a variable likely to impact
degradation rate, as the amount of exposed surface is likely to
increase the kinetics of moisture uptake. As surface area is
inversely related to the particle size, the particle size
distribution (PSD) was measured, and the data were used to
estimate the specific surface area of the different CDI lots.
The process for CDI manufacture involves filtration to
remove imidazole hydrochloride. We suspected that variability
in removal of this byproduct and its incorporation into the CDI
batch could also contribute to the difference in degradation
rate, given that imidazole hydrochloride is hygroscopic and is
also known to increase the reactivity of acyl imidazolides.6 As
shown in Table 2, inclusion of the chloride content of each lot
as measured by ion chromatography allows for explanation of
the trends in relative degradation rates. For example, even
though CDI lot 1 has specific surface area similar to that of CDI
lot 2, its relatively low chloride content results in a slower
degradation rate. Also, even though the specific surface area of
CDI lot 4 is three times less than that of CDI lot 5, its
490
dx.doi.org/10.1021/op400281h | Org. Process Res. Dev. 2014, 18, 488−494

Organic Process Research & Development
Article
Figure 5. (a) SEM image of starting CDI particles; (b) SEM image of
imidazole particles obtained after complete CDI degradation.
Table 1. Crystal lattice parameters for CDI and imidazole
CDI
imidazole
space group
a, Å
P1
P21/n
7.604
5.382
9.056
90
̅
5.781
7.27
b, Å
c, Å
18.13
78.9
α, deg
β, deg
87.43
73.798
4
108.876
90
γ, deg
Z
4
volume Å3
density g/cm3
717.983
1.107
350.748
1.289
degradation rate is similar due to relatively high chloride
content.
To establish the relative importance of the CDI particle size
(or surface area) and chloride content to the degradation rate, a
chemometric model was developed on the basis of the partial
least-squares (PLS) method. PLS was employed here due to its
effectiveness in projecting highly correlated dimensional data
matrices into low dimensional subspace while providing good
estimates on the response. The input variables for PLS are the
cumulative PSD at 10%, 50%, and 90% volume basis (Dv10,
Dv50 and Dv90) and chloride content. A two-component PLS
model was found to give the minimum root mean predicted
residual sum of squares (root mean PRESS) whilst providing a
good R2 of 0.9711 (R2 is the fraction of the cumulative sum of
squares of response explained by the model fit) (Figure 7). As
shown from the model coefficients in Figure 8, chloride content
has a positive correlation with degradation rate and is the most
sensitive variable. All PSD numbers have a negative correlation
with degradation rate, confirming that the kinetics of moisture
uptake/exchange is increased by a higher specific surface area.
From the variable importance for projection (VIP) plot, all
variables have values close to or above 0.8, confirming that
Figure 6. (a) CDI unit cell showing no H-bonding network, (b)
imidazole structure with H-bonded chains.
these variables are important to the degradation rate of CDI.7
The capability of estimating the CDI degradation rate simply by
using PSD and chloride content provides practical utility in
estimating degradation reaction kinetics for any lot of CDI.
Practical Implications. On the basis of our findings, we
analyzed various lots of CDI purchased from several vendors.8
Alarmingly, multiple unopened drums of CDI whose
certificates of analysis claimed greater than 97% purity
contained material of only 50−80% purity, presumably due to
degradation. We did note that two of these drums with the
same vendor lot number had degraded to differing extents. This
suggests that the packaging of CDI in the individual drums by
vendors needs to be done appropriately to ensure the material
does not degrade during packaging, transport, or storage.
491
dx.doi.org/10.1021/op400281h | Org. Process Res. Dev. 2014, 18, 488−494

Organic Process Research & Development
Article
Table 2. Degradation rates, PSD, and chloride content of various lots of CDI
CDI
lot
degradation rate
(%/h)
Dv10
(μm)
Dv50
(μm)
Dv90
(μm)
PSD
span
specific surface area estimated from PSD
(m2/g)
chloride content (w/w%)
1
2
3
4
5
8.0
12.7
12.7
17.0
18.3
52
60
66
58
19
155
180
219
188
67
517
595
479
428
212
3.00
2.97
1.89
1.97
2.88
0.047
0.041
0.035
0.041
0.124
0.02
0.09
0.07
0.12
0.03
Additionally, several lots of CDI known to be of acceptable
quality at the time of their initial use in several of our processes
had also partially degraded on storage. This suggests that
resealing of the packaging can be critical if the same batch of
CDI is to be used multiple times.
At lab scale, the existence of degradation of CDI by
atmospheric moisture is of relatively low importance to the
success or failure of a given reaction. Laboratory environments
are, more often than not, well controlled with respect to
humidity and air flow. In addition, there is usually not
significant surface exposure during material dispensing, and
reagent addition is generally carried out promptly. However,
these factors can all change significantly when moving from the
lab to a manufacturing environment. For example, the presence
of blow out walls as part of the structural design of many
processing suites containing manufacturing-scale reactors can
lead to wide fluctuation in humidity, even within the same
processing suite. Additionally, many processing suites operate
at a high rate of air flow designed to limit buildup of dust and
vapor which could result in an unsafe condition from both an
operator exposure or explosion hazard perspective. Finally,
many methods exist for charging a material from its container
to the reactor. While addition directly to the reactor in a timely
manner is certainly preferable as a charge method from the
perspective of limiting the exposure of the CDI to atmospheric
moisture, there are clearly situations where this is not viable.
For example, some facilities may require use of secondary
charge containers for additions of solids to reactors, which
could result in significant exposure of the CDI surface to
atmospheric moisture as the material is poured from its original
container into the secondary charge container. Additionally,
safety considerations may require charging of CDI using an
inerted solids addition funnel. Given the limited capacity of
most solids addition funnels, multiple charges of CDI to the
solids addition funnel may be required, necessitating some
forethought into how the contents of the CDI drum will be
protected from atmospheric moisture during the charge time.
All these factors indicate that degradation of CDI by
atmospheric moisture should not be ignored in a manufacturing
setting for processes in which a tight CDI operating range is
required.
Figure 7. Relationship between experimental degradation rate and
predicted degradation rate by PLS.
CONCLUSIONS
■
1
Measurement of CDI degradation rate by HPLC, H NMR,
and weight loss assays led to the determination that
atmospheric moisture is involved in the rate-determining step
of CDI degradation. As such, degradation rate increases with
increasing humidity, air flow, and exposed surface. Additionally,
particle size distribution and chloride content are the most
critical variables that impact the relative degradation rates of
different lots of CDI. These findings have practical implications
with respect to the assay, packaging, and storage of CDI. With
respect to charging of CDI in a manufacturing setting, these
Figure 8. Model coefficients of scaled and centered variables (Dv10,
Dv50, Dv90, and chloride content) for the response (degradation rate)
by PLS, and variable importance plot.
492
dx.doi.org/10.1021/op400281h | Org. Process Res. Dev. 2014, 18, 488−494

Organic Process Research & Development
Article
results indicate that precautions should be undertaken for
minimization of moisture contact with the reagent and suggest
that reaction completion assays may be critical to successful use
of CDI at scale.
with vacuum grease to slow mixing of dessicator atmosphere
with outside atmosphere. The dessicator contained a relative
humidity meter and a saturated aqueous K2SO4 solution. At the
time intervals shown in Figure 2, one dish was removed from
the dessicator, the material in the dish mixed with a spatula, and
1
EXPERIMENTAL SECTION
the CDI purity measured by H NMR assay. The relative
■
humidity was recorded every 5 min. Fluctuations in humidity in
Figure 2 correspond to the opening and closing of the
dessicator upon first adding and then sequentially removing
samples.
This experiment was repeated twice using the same CDI lot
and experimental design, with the exceptions of the use of
saturated aqueous NaCl and saturated aqueous LiCl solutions
in place of the saturated aqueous K2SO4 solution.
Assay of CDI by Derivatization to Dibenzyl Carbo-
nate. A tared flask is charged with benzyl alcohol
(approximately 10 equiv) and CDI (1 equiv) and heated to
80 °C for 30 min. The solution is cooled and weighed. CDI
purity is equivalent to the reaction yield. Reaction yield is
calculated using the w/w% dibenzyl carbonate in the reaction
mixture and the reaction mass.
The w/w% dibenzyl carbonate in the reaction mixture is
determined by analysis of a sample diluted with MeCN using
the following HPLC method: Supelco Acentis Express C18,
100 × 3.0 mm, 2.7 μm particle size, 0.8 mL/min, 45 °C, 5 μL
injection volume, detection at 205 nm, 70:30 0.1 v/v% aqueous
H3PO4:MeCN to 5:95 0.1 v/v% aqueous H3PO4:MeCN over
15 min. Quantitation was accomplished using a standard
solution prepared by dissolving 13.9 mg purchased dibenzyl
carbonate in 100 mL MeCN.
Degradation Rate of CDI with Variable Air Flow
(Figure 3). To three different tared 45 mm diameter
crystallizing dishes were added 4.00 g portions of CDI of
1
95.8% purity as measured by H NMR assay. The dishes were
placed on the opening of a chemical fume hood with the
emergency air flow activated. Measurement of the air flow by an
air flow meter over several minutes showed an average air flow
of 270 ft/min. At the time intervals shown in Figure 3, the
material in one dish was mixed with a spatula and the CDI
Assay of CDI by 1H NMR Analysis. A tared vial is charged
with approximately 0.7 mL of CDCl3 or d6-DMSO of known
water content and then weighed. A known amount of CDI is
1
purity measured by H NMR assay.
This experiment was repeated once using the same CDI lot
and experimental design, with the exception of placement of
the CDI dishes on the benchtop. Measurement of the air flow
by an air flow meter over several minutes showed an average air
flow of 0−5 ft/min.
1
then added to the vial and the H NMR spectrum acquired.
CDI purity is calculated using the areas of the CDI (8.2, 7.5,
and 7.2 ppm in CDCl3) and imidazole (7.7 and 7.1 ppm in
CDCl3 peaks) with correction for the water content of the
solvent.
Degradation Rate of CDI with Variable Exposed
Surface (Figure 4). To three different tared 45 mm diameter
crystallizing dishes were added 8.00-g portions of CDI of 95.8%
Decomposition of CDI with Time (Figure 1) and Assay
of CDI by Weight Loss. To seven different tared 45 mm
diameter crystallizing dishes were added 4.00 g portions of CDI
1
purity as measured by H NMR assay. The dishes were placed
1
on the benchtop. At the time intervals shown in Figure 4, the
material in one dish was mixed with a spatula and the CDI
of 93.9% purity as measured by H NMR assay. At the time
intervals shown in Figure 1, one dish was weighed, the material
in the dish mixed with a spatula, and the CDI purity measured
by 1H NMR assay. For the final sample after 24 h, the dish was
weighed, the material in the dish mixed with a spatula, and the
1
purity measured by H NMR assay.
This experiment was repeated twice using the same CDI lot
and experimental design, with the exceptions of use of 25 mm
diameter vials and 75 mm diameter crystallizing dishes in place
of the 45 mm diameter crystallizing dishes.
1
CDI purity measured by H NMR assay was 0%. The initial
purity of the CDI lot could be calculated from the total weight
loss of this sample as shown below.
Measurement of PSD of CDI. The PSDs of different CDI
lots were measured by a laser diffraction technique using the
Malvern Mastersizer 2000 coupled with the Hydro 2000S wet
dispersion accessory (Malvern Instruments, Malvern, U.K.).
Measurement conditions including the choice of dispersant/
surfactant combination, sample mass, stirrer/pump speed,
obscuration concentration and measurement time were
determined by conducting method development studies prior
to actual data collection. Specific surface area values were
estimated from the PSD with a CDI true density of 1.39 g/mL
and the assumption of spherical particles.
4.00 g initial sample − 3.40 g final sample
= 0.60 g weight loss
0.60 g weight loss
(28.01 g/mol for CO atoms lost − 2.02 g/mol for 2H atoms gained)
= 0.023 mol gas lost
0.023 mol gas lost = 0.023 mol CDI in initial sample
Measurement of Chloride in CDI. The w/w% chloride in
CDI was determined by analysis of a sample diluted with 1:1
THF/water using the following ion chromatography method:
Metrosep A Supp 7, 250 mm × 4.0 mm, 5 μm particle size with
a Metrosep RP, 50 mm × 4.0 mm guard column, 0.7 mL/min,
23 °C, 20 μL injection volume, detection by conductivity with
current of 50 mA and range of 50 μS/cm, isocratic aqueous 3.5
mM Na2CO3 and 1.0 mM NaHCO3 mobile phase for 15 min.
Quantitation was accomplished by appropriate dilution of a
purchased 1000 ppm chloride standard with 1:1 THF/water.
Partial Least Squares Regression. PLS analysis was
carried out in JMP (version 10, SAS, NC). The optimum
0.023 mol CDI in initial sample × 162.15g/mol
= 3.73 g CDI in initial sample
⎛
⎜
⎝
⎞
⎟
⎠
3.73 g CDI in initial sample
4.00 g initial sample
× 100 = 93.2%purity
Degradation Rate of CDI with Variable Relative
Humidity (Figure 2). To three different tared 45 mm
diameter crystallizing dishes were added 4.00-g portions of
1
CDI of 95.8% purity as measured by H NMR assay. The
dishes were placed in a glass dessicator with the seal coated
493
dx.doi.org/10.1021/op400281h | Org. Process Res. Dev. 2014, 18, 488−494

Organic Process Research & Development
Article
number of PLS components was selected by minimizing the
root mean PRESS. All data from the input variables were mean
centered and scaled to unit variance to eliminate the influence
of any given variable with high variance on the model
parameters.
AUTHOR INFORMATION
Corresponding Author
*E-mail: kenneth.engstrom@abbvie.com
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The work presented in the article was sponsored by AbbVie.
AbbVie contributed to the design, research, and interpretation
of data, writing, reviewing, and approving the publication.
Kenneth Engstrom, Ahmad Sheikh, Raimundo Ho, and Robert
Miller are employees of AbbVie. The authors thank Paul
Brackemeyer for collecting single-crystal X-ray data along with
Michael Theine for the dispensing of numerous CDI samples
from the chemical warehouse. The authors also thank
colleagues involved in the transfer of processes to our
commercial manufacturing facility, particularly Samrat Mukher-
jee and Rory Delaney, for valuable discussions surrounding
CDI charging practices.
REFERENCES
■
(1) Montalbetti, C. A. G. N.; Falque, V. Tetrahedron 2005, 61, 10827.
(2) (a) Carey, J. S.; Laffan, D.; Thomson, C.; Williams, T. W. Org.
Biomol. Chem. 2006, 4, 2337. (b) Dale, D. J.; Dunn, P. J.; Golightly, C.;
Hughes, M. L.; Levett, P. C.; Pearce, A. K.; Searle, P. M.; Ward, G.;
Wood, A. S. Org. Process Res. Dev. 2000, 4, 17.
(3) (a) Staab, H. A. Leibigs Ann. Chem. 1957, 609, 75. (b) Staab, H.
A.; Wendel, K. Org. Syntheses 1968, 48, 44.
(4) Armstrong, A.; Li, W. N,N-Carbonyldiimidazole. In e-EROS:
Encyclopedia of Reagents for Organic Synthesis; John Wiley & Sons, Ltd:
New York, 2007; 10.1002/9780470842898.rc024.pub2.
(5) We anticipate quantification of the derivatization-based assay by
GC would be successful, although we have not evaluated it specifically.
(6) (a) Woodman, E. K.; Chaffey, J. G. K.; Hopes, P. A.; Hose, D. R.
J.; Gilday, J. P. Org. Process Res. Dev. 2009, 13, 106. (b) We have
observed that imidazolium triflate can accelerate the rate of reaction
between an aliphatic carboxylic acid and CDI. It is possible that
imidazolium hydrochloride similarly accelerates reaction between CDI
and moisture.
(7) Wold, S. PLS for Multivariate Linear Modeling 1994. In QSAR:
Chemometric Methods in Molecular Design: Methods and Principles in
Medicinal Chemistry; Wiley-VCH: New York, 1995.
(8) At the time of this analysis, all vendors from which we purchased
CDI were not the original manufacturers of the material.
494
dx.doi.org/10.1021/op400281h | Org. Process Res. Dev. 2014, 18, 488−494