Content uploaded by Qing Min Wang
Author content
All content in this area was uploaded by Qing Min Wang on Oct 27, 2017
Content may be subject to copyright.
Article
Monitor Polyimide Production from
Diamine and Dianhydride Reactions
Using a Combination of In Situ
Infrared and Raman Spectroscopy
Xiaoyun Chen
1
, Qing Min Wang
2
, and Luke Bu
2
Abstract
A two-step synthesis of polyimide reaction from dianhydride and diamine was followed by both infrared (IR) and Raman
spectroscopy in situ in this case study. In the first step, the two reactants form poly(amide acid), which resulted in
substantial spectral changes easily detected by both spectroscopic techniques. The reaction was found to occur rapidly
and reached completion within tens of minutes at room temperature. In the second step, trimethylamine (catalyst) and
acetic anhydride (dehydration agent) were added to form the five-member imide ring. Interestingly, different spectral
changes were observed using IR and Raman spectroscopic methods due to their different responses toward the various
functional groups involved. Conclusive evidence was also obtained based on the Raman results to demonstrate a long-lived
intermediate species, which was harder to observe in the IR results. This work provides a good case study of the combined
use of different vibrational spectroscopy techniques to extract maximum information from a reaction system.
Keywords
Raman, in situ, infrared, IR, polyimide
Date received: 4 January 2017; accepted: 22 February 2017
Introduction
Colorless polyimide materials and their thin films have been
widely considered as a top choice of flexible substrates for
a foldable or bendable display owing to their exceptionally
high thermal stability and high transparency.
1
A variety of
technologies are being investigated to develop such flexible
display devices among top display industry leaders. The over-
all requirements for this material are very strict and include
thermal, optical, mechanical, and chemical properties. Many
requirements are seemingly contradictory to each other in
terms of structure/property relationship. For example, the
low coefficient of thermal expansion behavior of the poly-
imide is achieved due to the charge transfer complex inter-
action between polymer chains.
1
However, the charge
transfer complex interaction results deep color, low trans-
parency, and high optical anisotropy, which are not desired.
Therefore, optimizing and balancing the properties by select-
ing the suitable chemistry and synthesizing the material
under a controlled manner is the key.
Polyimide synthesis requires two steps.
2
A dianhydride
reacts with a diamine in Step 1 (i.e., polycondensation or
step-growth polymerization) to form a poly(amide acid)
(PAA). Poly(amide acid) undergoes dehydration in Step 2
(i.e., imidization) to form polyimide. One key challenge
facing the synthesis of both steps is the lack of analytical
methods to quantify the reaction conversion. This is espe-
cially important for Step 2 because high conversion may
result in an insoluble product. Besides the monomer chem-
istry, a number of reaction conditions affect the properties
of final products. These conditions can be reaction tem-
perature, monomer addition sequence, dissolved oxygen
in the solvent, residual moisture in the solvent, etc. The
material’s properties are also highly dependent on the reac-
tion solvents. Therefore, understanding the evolution of the
structure change during the reaction is critical not only for
the robust material, but also for reducing the batch-to-
batch variation.
1
Analytical Sciences, The Dow Chemical Company, Midland, MI, USA
2
Electronic Materials, The Dow Chemical Company, Marlborough, MA,
USA
Corresponding author:
Xiaoyun Chen, Analytical Sciences, 1897 Building, The Dow Chemical
Company, Midland, MI 48667, USA.
Email: xchen4@dow.com
Applied Spectroscopy
2017, Vol. 71(9) 2128–2135
!The Author(s) 2017
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0003702817700427
journals.sagepub.com/home/asp
In this study, in situ infrared (IR) and Raman spectros-
copy were applied to monitor both steps. Semi-quantitative
methods were developed based on various peaks, such as
the anhydride C¼O stretch, PAA C¼O stretch, imide
C¼O stretch, and various backbone ring modes. Such
methods enabled real-time monitoring of the extent of
reaction. Recent years have witnessed an increasingly
wider range of in situ IR and Raman.
3–8
The monitoring
of organic reaction progress continues to be one of the
most frequently applications for both spectroscopy tech-
niques. Each technique has its own advantages and disad-
vantages. Near-infrared (NIR) is also a powerful technique,
but it offers less chemical specificity and is more often used
as a process analytical tool for known reactions and less
frequently as a tool for R&D reaction monitoring.
This case study compares the two techniques. There
have only been limited in situ IR and Raman studies to
monitor polyimide formation and usually with only one
technique.
9,10
Infrared and Raman spectroscopy have long
been known to provide complimentary information due to
their different selection rules. In the results section below,
we will show that while IR and Raman results offered very
similar type of information for Step 1, they tracked the
Step 2 conversion via very different functional groups.
More importantly, conclusive Raman spectroscopic evi-
dence for a long-lived intermediate species for Step 2 was
obtained (Fig. 1).
Experimental Section
Reagents and Materials
Due to the proprietary nature of the reaction and process
investigated here, only limited information can be provided.
The reaction was carried out in a 500 mL thee-neck flask
with mechanical stirring at room temperature. Raman and
IR probes were inserted into the reactor through two of
the three ports. All reagents were purchased from com-
mercial sources and used as received.
Raman Spectroscopy
A Kaiser RXN1 Raman system with 785 nm laser excitation
was used. Signals were collected using an immersion optic
in back-scattering geometry. Typical experimental param-
eters were: full laser power (400 mW at laser head and
300 mW at sample), 5cm
–1
resolution, 3 s exposure,
ten accumulations for averaging, which led to a total acqui-
sition time of 30 s per spectrum. Spectra were continuously
collected. Peak area analysis was carried out using iCIR
software from Mettler Toledo Auto Chem. Inc.
Chemometric analysis was carried out in Matlab.
Attenuated Total Reflection–Fourier
Transform Infrared
A Mettler Toledo ReactIR 15 system was used to collect
spectra from the reaction mixture. It was equipped with a
fiber optically coupled diamond attenuated total reflection
(ATR) crystal and a liquid-nitrogen cooled mercury–cad-
mium–telluride (MCT) detector. Peak area analysis was
carried out using iCIR software from Mettler Toledo
Auto Chem. Spectra were collected with 4 cm
–1
resolution
and 1 min acquisition time per spectrum. Spectral collection
interval was in the range of 1–10 min, with the longer inter-
val used for multi-day experiments.
Results and Discussion
Step 1: Amide Acid Formation
Infrared Results. In the first reaction, a 10 wt% dianhydride
solution in dimethylacetamide (DMA) was first added to a
reactor. An equal molar equivalent of diamine was then added
into the reactor in a single shot to start the Step 1 reaction.
Selected IR spectra from this reaction are shown in Fig. 2.
Spectrum (a) was collected prior to the diamine addition and
it shows many peaks from the dianhydride in comparison to
neat DMA, which is also shown in Fig. 2 as a reference. For
example, both the anti-symmetric C¼O stretch band at
Figure 1. The two-step reaction used in this study to form polyimide.
Chen et al. 2129
1787 cm
–1
and the symmetric C¼O stretch at 1859 cm
–1
could be clearly observed. Upon addition of diamine, the
reaction occurred rapidly as evidenced by both the exotherm
(see the right y-axis of Fig. 2 bottom panel), the rapid
decrease of the anhydride C¼O peaks, and the rise of the
product amide acid, which showed a characteristic acid C¼O
stretch peak at 1718 cm
–1
. Overall, the reaction appeared to
be complete within tens of minutes, in contrast to tens of
hours that were originally believed to be necessary. There are
many other peaks that could be used to track the reaction,
which all showed the same kinetics behavior and the results
are not shown here.
Raman Results
Representative Raman spectra from the Step 1 of another
reaction are shown in Fig. 3. This reaction was run at a
different condition to gain further insights into the reaction.
The solvent was changed to anhydrous n-butyl pyrrolidine
(NBP) and the addition sequence was reversed. Initially a
diamine solution was added to the reactor as shown in spec-
trum (a) in Fig. 3. Diamine had relatively weak Raman feature
in comparison to the NBP solvent and its most prominent
feature was a peak at 468 cm
–1
. Four shots of anhydride,
each containing approximately 0.25 molar equivalent, were
then sequentially added to the starting diamine solution with
a 20–70min reaction time following each shot. A fifth shot of
dianhydride was added which resulted in a slight molar
excess. The #1–5 in the Fig. 3 bottom panel marked the
time for these additions.
Similar to the observation made above in the IR experi-
ment, two peaks from the anhydride C¼O stretch modes
manifested themselves, though only transiently following
each shot as shown in spectrum (b) in Fig. 3. The higher
Figure 2. (Top) Representative IR spectra of Step 1. See the dashed vertical lines in the bottom panel for the collection time of spectra
(a)–(d). (Bottom) Peak area and temperature profiles. RTD, resistance temperature detectors.
2130 Applied Spectroscopy 71(9)
wavenumber symmetric peak around 1850 cm
–1
(note that
the instrument artifact at this region was due to imperfect
splicing of the low and high wavenumber regions) was stron-
ger than the anti-symmetric peak around 1790 cm
–1
(note
that the sharp line at this region was due to room light) due
to the different selection rules of IR and Raman spectroscopy.
In addition to the spectral changes associated with the dia-
mine and dianhydride, the growth of several new peaks
attributed to the formation of the PAA product were
observed, such as the peaks at 1330 and 1616 cm
–1
. The
bottom panel of Fig. 3 shows that the consumption rate of
the diamine and anhydride matched the product formation
rate well. It is also interesting to note that almost all the
468 cm
–1
peak of diamine was consumed prior to the third
shot of dianhydride. This indicated that the 468 cm
–1
peak is
likely from the aromatic ring mode instead of from the NH
2
group because only half of the starting NH
2
groups should
have been consumed by this point. Complete consumption of
anhydride peak was observed at the end of the first four
anhydride addition. The firth shot of anhydride led to a
slight accumulation, indicating that all NH
2
groups had been
converted by this point.
Step 2: Imidization
In general, in situ IR and Raman results yielded very similar
information for Step 1. In contrast, the two spectroscopic
methods were found to monitor Step 2 based on functional
groups from different molecules.
Infrared Results
Figure 4 shows a series of representative in situ IR spectra
from Step 2. Initially a 12 wt% PAA in NBP solution was
prepared as described above and its spectrum was simply
a combination of NBP and PAA as shown in Fig. 4 spectrum
(a). At about 36 min a 2 mol equivalent of triethylamine
(TEA) and acetic anhydride were added to the solution to
Figure 3. (Top) Representative Raman spectra of Step 1. See the dashed vertical lines in the bottom panel for the collection time of
spectrum (a)–(e). (Bottom) Peak area and temperature profiles.
Chen et al. 2131
start the dehydration reaction to form imide. This led to
immediate appearance of characteristic acetic anhydride
bands in spectrum (b). Triethylamine had relatively weak
features that were dominated by the rest of the mixture.
All the peaks associated with acetic anhydride exhibited a
slow decrease as it was converted to acetic acid, which was
then immediately neutralized by TEA. Concurrently several
new peaks gain intensity. For example, a peak at 1720 cm
–1
gradually increased and its position is consistent with that of
imide C¼O stretch. Two other peaks at 1367 and 725 cm
–1
also increased, though their exact assignments were not
known and thus generally referred to as product. Overall,
the majority of the spectral changes were associated with
acetic anhydride due to its strong IR absorbance. It is inter-
esting to note that the growth of imide C¼O peak appeared
to lag behind that of the 1367 cm
–1
peak, which indicated
that an intermediate species might have been formed before
the imide formation. However, due to the stronger interfer-
ence of the solvent C¼O peak, this was not conclusive.
Raman Results
Figure 5 shows representative Raman spectra collected in
Step 2 of the reaction. During this reaction, the solution
clarity underwent substantial changes, which led to change
in the absolute Raman signal intensity. To facilitate compari-
son, all spectra in Fig. 5 are normalized by the solvent
NBP C¼O stretch peak around 1680 cm
–1
. Start spectrum
(a) in Fig. 5 contains features from both amide acid and
NBP. Four shots of TEA and acetic anhydride (each con-
taining approximately 0.5 molar equivalent) were then
sequentially added with roughly 15 min between each
shot. Spectrum (b) in was collected immediately after
these four shots. In contrast to the IR results where signifi-
cant peaks were observed from acetic anhydride, very weak
Raman signals from acetic anhydride and TEA were
observed due to their relatively small Raman cross-sec-
tions. Spectra (c), (d), and (e) were collected 0.5, 1.5, and
26 h after spectrum (b), respectively. Consistent with the
Figure 4. (Top) Representative IR spectra of Step 2. See the dashed vertical lines in the bottom panel for the collection time of
spectrum (a)–(e). (Bottom) Peak area profiles.
2132 Applied Spectroscopy 71(9)
observations made based on IR results, the Step 2 reaction
proceeded much more slowly. A strong peak at 1784 cm
–1
continued to gain strength during this period and could be
attributed to the symmetric C¼O stretch of the imide
product. The spectral changes around 1600 cm
–1
showed
an interesting pattern. The amide acid 1614 cm
–1
peak
showed a monotonic decrease, while a 1602 cm
–1
peak
first grew and eventually disappeared. Spectrum (c) was col-
lected when the 1602 cm
–1
peak reached its maximum. At
this point, the amide acid 1614 cm
–1
peak had decreased to
roughly half its starting intensity, while the product imide
peak at 1784 cm
–1
just started to become noticeable (note
that a sharp room light feature was there at the beginning).
The 1784 cm
–1
peak growth was substantial as can be seen
in spectra (d) and (e) in Fig. 5, together with several other
peaks (e.g., peaks at 630 and 1390 cm
–1
, assignments
not known). The peak area profiles shown in the bottom
panel of Fig. 5 clearly demonstrated that an intermediate
species exist in the reaction from amide acid to imide. The
apparent decrease and subsequent increase of the amide
acid peak profile was due to the interference of the
1602 cm
–1
peak, which could influence the integrated area
of the baseline-corrected 1614 cm
–1
peak. Classical least
squares (CLS) has been successfully utilized to overcome
such spectral overlapping.
8,11–15
In order to overcome the
abnormal peak profiles observed in Fig. 5, a CLS model
was developed which included the following components:
NBP solvent, PAA which was obtained by subtracting the
solvent contribution form the starting solution, polyimide
product (end spectrum after solvent subtraction), and the
Figure 5. (Top) Representative Raman spectra of Step 2. See the dashed vertical lines in the bottom panel for the collection time of
spectrum (a)–(e). Inset shows an expanded view of the 1500–1650 cm
–1
region. (Bottom) Peak area profiles. The gray vertical lines
indicate the four shots of TEA and acetic anhydride.
Chen et al. 2133
intermediate (spectrum (c) in Fig. 5 after subtracting solv-
ent, PAA, and polyimide). The CLS responses, after being
normalized by the solvent CLS response, are shown in
Fig. 6. It is clear that the strange behavior observed for
the PAA area profile was successfully resolved by using
CLS. Reaction profiles based on the CLS analysis can then
be readily used to understand how various process condi-
tions affect the reaction kinetics, which is beyond the scope
of this work. The exact identify of the intermediate remains
to be further investigated (e.g., by NMR).
It is also worth pointing out that a similar approach was
attempted to analyze the IR data shown in Fig. 4 by CLS
model, but the results (not shown here) were much less
satisfactory. This is believed to be due to the following
reasons. The largest spectral changes observed in IR were
most associated with the consumption of acetic anhydride,
which together with TEA require the inclusion of two more
pure component spectra in the IR CLS model. This resulted
in a more complex CLS model for IR data than the three-
component model for Raman. Additionally, in the IR spec-
tra, the C¼O stretch region was highly crowded with
a maximum peak absorbance larger than two absorbance
units, which was outside the detector dynamic range.
Multivariate curve resolution (MCR) and principle com-
ponent analysis (PCA) were also tested. While both
methods yielded interesting trends that could potentially
be correlated to concentrations of various species, the
interpretation of such results is less straightforward than
the peak area or CLS analysis employed here.
Conclusion
Monitoring reactions in situ using IR and Raman spectros-
copy is becoming increasingly more prevalent as the instru-
mentations become more mature, robust and user friendly.
It is often not straightforward to judge which technique
will be more effective. In this case study, we monitored
an industrially important reaction to make polyimide from
diamine and dianhydride with both IR and Raman spectros-
copy in situ. The results demonstrate that the decision
on which technique is the preferable one should be made
on a case by case basis. While both techniques can track
Step 1 effectively, they are sensitive to different functional
groups in Step 2. For Step 2, Raman offers the advantage to
directly track the transient formation of an intermediate
species, which cannot be sensitively monitored using IR,
while IR was able to track non-aromatic species whose
signals were overwhelmed by the aromatic species in the
Raman spectra.
Acknowledgments
The authors thank Dr. Iou-Sheng Ke for help with Raman spectral
collection.
Conflict of Interest
The authors report there are no conflicts of interest.
Funding
This research received no specific grant from any funding agency in
the public, commercial, or not-for-profit sectors.
References
1. J. Kang. ‘‘Flexible Display Technology and Market: the Opportunities
that Flexible OLED Brings’’. IHS Technology, March 2016.
2. K.L. Mittal. Polyimides: Synthesis, Characterization, and Applications.
New York: Springer Science and Business Media, 2013.
3. K. Reinhold-Lo
´pez, A. Braeuer, B. Romann, N. Popovska-Leipertz,
A. Leipertz. ‘‘Simultaneous in Situ Raman Monitoring of the Solid and
Gas Phases During the Formation and Growth of Carbon
Nanostructures Inside a Cold Wall CCVD Reactor’’. Carbon. 2014.
78: 164–180.
Figure 6. Classical least squares analysis results based on Raman spectra of Step 2.
2134 Applied Spectroscopy 71(9)
4. Y. Tong, P. Zhang, L. Dang, H. Wei. ‘‘Monitoring of Cocrystallization of
Ethenzamide–Saccharin: Insight into Kinetic Process by In Situ Raman
Spectroscopy’’. Chem. Eng. Res. Des. 2016. 109: 249–257.
5. K. Uz
ˇarevic
´, I. Halasz, T. Fris
ˇc
ˇic
´. ‘‘Real-Time and in Situ Monitoring of
Mechanochemical Reactions: a New Playground for All Chemists’’.
J. Phys. Chem. Lett. 2015. 6(20): 4129–4140.
6. C. Wang, T.J. Vickers, J.B. Schlenoff, C.K. Mann. ‘‘In Situ Monitoring of
Emulsion Polymerization Using Fiber-Optic Raman Spectroscopy’’.
Appl. Spectrosc. 1992. 46(11): 1729–1731.
7. S.E. Barnes, Z.T. Cygan, J.K. Yates, K.L. Beers, E.J. Amis.
‘‘Raman Spectroscopic Monitoring of Droplet Polymerization in a
Microfluidic Device’’. Analyst. 2006. 131(9): 1027–1033.
8. X. Chen, K. Laughlin, J.R. Sparks, L. Linder, V. Farozic, et al. ‘‘In Situ
Monitoring of Emulsion Polymerization by Raman Spectroscopy: A
Robust and Versatile Chemometric Analysis Method’’. Org. Process
Res. Dev. 2015. 19(8): 995–1003.
9. Y. Seo, S.M. Lee, D.Y. Kim, K.U. Kim. ‘‘Kinetic Study of the Imidization
of a Poly (ester amic acid) by FT-Raman Spectroscopy’’.
Macromolecules. 1997. 30(13): 3747–3753.
10. Y. Han, Y.-H. Yoo, J.-M. Rhee, M.-H. Lee, S.-C. Yu. ‘‘Two-Dimensional
Raman Correlation Spectroscopy Study of the Pathway for the
Thermal Imidization of Poly (Amic Acid)’’. Bull. Korean Chem. Soc.
2003. 24(3): 357–362.
11. D.M. Haaland, R.G. Easterling. ‘‘Application of New Least-Squares
Methods for the Quantitative Infrared Analysis of Multicomponent
Samples’’. Appl. Spectrosc. 1982. 36(6): 665–673.
12. D.M. Haaland, R.G. Easterling, D.A. Vopicka. ‘‘Multivariate Least-
Squares Methods Applied to the Quantitative Spectral Analysis of
Multicomponent Samples’’. Appl. Spectrosc. 1985. 39(1): 73–84.
13. D.M. Haaland, R.G. Easterling. ‘‘Improved Sensitivity of Infrared
Spectroscopy by the Application of Least Squares Methods’’. Appl.
Spectrosc. 1980. 34(5): 539–548.
14. X. Chen, C. Meece, M. Gonzalez, H. Chiang, X. Qiu, et al. ‘‘Isoclast
Active Manufacture: In Situ Spectroscopic Investigation of the
Unstable Products of Cyanamide and Sodium Hypochlorite
Reactions’’. Org. Process Res. Dev. 2015. 19(1): 139–144.
15. X. Chen, R. Pell, S. Sarsani, B. Cramm, C. Villa, et al. ‘‘In Situ
Attenuated Total Reflectance Fourier Transform Infrared (ATR FT-
IR) Spectroscopy Monitoring of 1,2-Butylene Oxide Polymerization
Reaction by Using Iterative Concentration-Guided Classical Least
Squares’’. Appl. Spectrosc. 2013. 67(8): 940–948.
Chen et al. 2135