Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Characterization, process analysis, and recycling of a benzoxazine-epoxy resin for structural composites
(USC Thesis Other)
Characterization, process analysis, and recycling of a benzoxazine-epoxy resin for structural composites
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Characterization, Process Analysis, and Recycling of a Benzoxazine-Epoxy
Resin for Structural Composites
by
Jonathan Lo
A Thesis Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(CHEMICAL ENGINEERING)
DECEMBER 2017
i
“Ce qui embellit le desert, c’est qu’il cache un puits quelque part.”
Antoine de Saint-Exupery
ii
Acknowledgements
Individual transformative achievements arise from the resolute, and oftentimes anonymous,
contributions of the myriad. I have been extremely lucky to belong to an amazing community of
friends and mentors which have helped me at every turn. My greatest thanks go to the members
of my thesis committee: Professor Steven Nutt, Professor Travis Williams, Professor Noah
Malmstadt, and Professor Andrea Armani. They have walked alongside me during the past six
years as a PhD student, placed opportunities in front of me that I could not imagine, dared me to
dream about achieving the impossible, and helped me open doors that I believed were sealed shut.
Without my committee, these scientific endeavors would never have occurred and for their
guidance, I am eternally grateful. It is truly upon their shoulders that I stand on today.
I am indebted to my family for their love and care all these years, as they have been supportive
in every conceivable manner. They have the patience and forbearance of saints over the numerous
years this research has taken. Of special importance is my fiancé as he has been boundlessly
benevolent, limitlessly loving, and for more times than I dare to admit, a psychological and
financial safety net. I am blessed that he decided to pick me out of the ethereal. For all that has
gone well in my life during my time as a graduate student, I dedicate to you. Indubitably, I take
full responsibility for all the things that fall somewhere in between.
I would like to extend a gracious thank you to all the present and past members of the Nutt Lab
for accompanying me during my time here. Your company has made early mornings and long
nights in the laboratory much more bearable. Of particular importance is Lessa Grünenfelder and
Timotei Centea for the multitude of ideas and productive conversations, inspiring me to dig deeper,
ask more questions, and expand my research. A huge thank you to Mark Anders, Lee Hamill,
Sarah Katz, Clifford Lester, Yijia Ma, Kristina Rojdev, Daniel Zebrine, Yixiang Zhang, and
iii
Yunpeng Zhang. All of you have truly made my PhD experience so much more enjoyable. From
sharing laughs with me in the office to exploring the vast concrete desert of Los Angeles, I truly
appreciate every minute I have spent in your company and I will treasure these memories wherever
my journey takes me. I am also indebted to Jennifer Gerson who was my guardian angel as the
Director of Doctoral Programs at USC. Without her behind-the-scenes help, I would never have
finished my degree. From the bottom of my heart – thank you.
A special appreciation to my undergraduate students whom I’ve had the honor and privilege
of mentoring: Matthew Thomas and Erynn Naccarelli. Your natural curiosity, enthusiasm for
learning, and insatiable appetite for knowledge will serve you far in life. In the time I have spent
with you, I am convinced that there is no research question too challenging for you to answer or
learning curve too steep for you to overcome. You two will go far in life – let nothing convince
you otherwise.
It has been said that friends are our chosen family, and I’m grateful for my family-away-from-
family for their support throughout these years. Thank you to Mary Boyd, Benjamin Decato,
Samantha Huyhh, James Joly, Michele Lee, John Mac, Sara Mc Carthy, René Zeto, and Lily Zhang
for the laughs, hugs, and sometimes metaphorical shoulder to cry on.
Last but not least, to complete a PhD, one must do something you truly love. Many years ago
I took AP Chemistry with Mrs. Tana Dearborn. My love of chemistry comes directly from her.
Her passion, energy, and enthusiasm put me on this very path today that led to my degree. Had I
not been in her class, I do not doubt that things would be very different today. Thank you for
starting me on this journey, for teaching me to be all that I could be, and for instilling in me the
belief that I should never, ever, ever, give up.
iv
Table of Contents
Acknowledgements ......................................................................................................................... ii
Table of Contents ........................................................................................................................... iv
List of Tables ................................................................................................................................. vi
List of Figures ............................................................................................................................... vii
Abstract ........................................................................................................................................... x
Chapter 1 : Introduction .................................................................................................................. 1
1.1 : Motivation ........................................................................................................................... 1
1.2 : References ........................................................................................................................... 7
Chapter 2 : The Effect of Processing Parameters on Volatile Release for a Benzoxazine/Epoxy
Resin ............................................................................................................................................... 9
2.1 : Abstract ............................................................................................................................... 9
2.2 : Introduction ......................................................................................................................... 9
2.3 : Materials and Methods ...................................................................................................... 14
2.4 : Results ............................................................................................................................... 19
2.5 : Conclusions ....................................................................................................................... 32
2.6 : References ......................................................................................................................... 34
Chapter 3 : A Method for In Situ Analysis of Volatiles Generated during Cure .......................... 36
3.1 : Abstract ............................................................................................................................. 36
3.2 : Introduction ....................................................................................................................... 36
3.3 : Materials and Methods ...................................................................................................... 39
3.4 : Results ............................................................................................................................... 42
3.5 : Conclusion ......................................................................................................................... 50
3.6 : References ......................................................................................................................... 52
Chapter 4 : Eliminating Porosity via Reformulation of a Benzoxazine-Epoxy RTM Resin ........ 54
4.1 : Abstract ............................................................................................................................. 54
4.2 : Introduction ....................................................................................................................... 54
4.3 : Materials and Methods ...................................................................................................... 56
4.4 : Results ............................................................................................................................... 62
4.5 : Conclusions ....................................................................................................................... 77
4.6 : References ......................................................................................................................... 79
Chapter 5 : Recycling Benzoxazine-Epoxy Composites via Catalytic Oxidation ........................ 81
5.1 : Abstract ............................................................................................................................. 81
5.2 : Introduction ....................................................................................................................... 81
v
5.3 : Materials and Methods ...................................................................................................... 84
5.4 : Results ............................................................................................................................... 88
5.5 : Conclusion ......................................................................................................................... 96
5.6 : References ......................................................................................................................... 97
Chapter 6 : Conclusions and Future Work .................................................................................. 100
6.1 : Modelling cure-induced volatile release ......................................................................... 100
6.2 : Catalytic Recycling of Composites via Electrolysis ....................................................... 102
6.3 : References ....................................................................................................................... 104
vi
List of Tables
Table 2-1: Summary of all volatile characterization experiments performed. “R” and “D”
designate ramps and dwells, respectively ..................................................................................... 15
Table 2-2: Summary of all “realistic cure cycle” volatile characterization experiments performed.
“R” and “D” designate ramps a dwells, respectively .................................................................... 15
Table 2-3: Total heats of reaction for non-isothermal scans (left) and specific exotherm energies
for isothermal runs (right) ............................................................................................................. 20
Table 2-4: Total weight loss versus ramp rate for degassed samples ........................................... 24
Table 2-5: Potential volatiles emitted during the realistic cure cycles and their methods of
elimination .................................................................................................................................... 26
Table 3-1: Resin systems considered using the RC/FTIR technique ............................................ 41
Table 4-1: Test matrix of samples considered to determine effect of formulation on final part
properties....................................................................................................................................... 60
Table 4-2: Weight percent of ethyl acetate remaining in neat resin samples ............................... 71
Table 4-3: Porosity and surface roughness measurements of carbon fiber panels manufactured
from Formulation A and B ............................................................................................................ 74
Table 4-4: Weight percent of ethyl acetate remaining versus their physical properties ............... 76
Table 5-1: Oxidant and catalyst screening for optimal homogenization conditions for
benzoxazine resin grounds ............................................................................................................ 89
vii
List of Figures
Figure 1-1: Representative structure of a benzoxazine monomer .................................................. 3
Figure 1-2: Schematic of the main steps of an RTM manufacturing process ................................. 4
Figure 1-3: Fiber impregnation mechanisms resulting in flow-induced porosity. Figure
reproduced from [18] ...................................................................................................................... 5
Figure 2-1: Mass loss over a cure cycle at atmospheric pressure for a high volatilizing resin .... 11
Figure 2-2: (A) Surface and (B) through-thickness porosity in an RTM sample ......................... 12
Figure 2-3: Cross-section view of reaction cell assembled (left) and exploded view of reaction
cell assembly (right) showing (A) N2 carrier gas inlet, (B) reaction cell fitting, (C) connection to
FTIR, (D) heating blocks with attached heating rods, and (E) sample holder with attached
thermocouple................................................................................................................................. 17
Figure 2-4: (A) Exotherms versus temperature from non-isothermal scans. (B) Exotherms versus
time from isothermal runs. ............................................................................................................ 19
Figure 2-5: Degree of cure evolution versus time for various cure dwell temperatures ............... 21
Figure 2-6: Weight loss data versus ramp rate, as seen in the TGA for (A) dynamic ramps and,
(B) isothermal dwells. ................................................................................................................... 22
Figure 2-7: Comparison of degassed versus non-degassed resin .................................................. 23
Figure 2-8: FTIR data for (A) 3°C/min ramp to 220°C and hold, (B) 2°C/min ramp to 220°C and
hold, and (C) 1°C/min ramp to 220°C and hold. The intensity of the color scale is in absorbance
units. .............................................................................................................................................. 25
Figure 2-9: FTIR data with DSC cure rate data superimposed for (A) 3°C/min ramp and (B)
1°C/min ramp to 220°C. The intensity of the color scale in in absorbance units. ........................ 27
Figure 2-10: FTIR data from the reaction cell showing volatile release as a function of cure
temperature for various cure cycles. Cure dwell at: (A) 175°C, (B) 185°C, (C) 195°C, (D)
Injection dwell for 3 hours at 130°C. The intensity of the color scale in in absorbance units. .... 29
Figure 2-11: Neat resin panels cured at: a) 185°C and 103 kPa of pressure, b) 185°C and
atmospheric pressure, c) 170°C and 69 kPa, d) 170°C and atmospheric pressure. ...................... 32
viii
Figure 3-1: (A, left) Cross section view of a reaction cell, showing (A) inlet port (B) sample
compartment (C) Heaters (D) Control thermocouple (E) Outlet port to FTIR. (B, right) Exploded
view of the reaction cell. ............................................................................................................... 40
Figure 3-2: Spectroscopic trace for an epoxy pre-preg sample .................................................... 42
Figure 3-3: Gram-Schmidt trace of the spectroscopic intensity for an epoxy pre-preg sample ... 43
Figure 3-4: (A) Spectroscopic trace for a benzoxazine/epoxy sample at ambient pressure. (B)
Spectroscopic trace for a benzoxazine/epoxy sample at 206 kPa (abs) (C) A corresponding IR
absorption spectrum for (A) at ~150 minutes. .............................................................................. 45
Figure 3-5: (A) Spectroscopic traces for the first 90 minutes of a polyimide pre-preg cured within
the reaction cell. (B) Spectroscopic traces for the last 180 minutes of a polyimide pre-preg cured
within the reaction cell. ................................................................................................................. 47
Figure 3-6: Mass loss during cure for a polyimide sample cured before and after lyophilization 49
Figure 3-7: (A) Spectroscopic traces for a 1-ply polyimide pre-preg (B) Spectroscopic traces for
a 3-ply polyimide pre-preg (C) Spectroscopic trace for the first 60 minutes of a 1-ply polyimide
pre-preg. (D) Spectroscopic trace for the first 60 minutes of a 3-ply polyimide pre-preg ........... 50
Figure 4-1: CAD model of the mini-RTM mold .......................................................................... 59
Figure 4-2: Comparison of cure-induced volatilization for Formulations A and B ..................... 63
Figure 4-3: A potential mechanism for the generation of aniline ................................................. 64
Figure 4-4: Solvent evaporation and degradation for formulations A and B ............................... 65
Figure 4-5: Proton NMR results for sample 1 .............................................................................. 67
Figure 4-6: Gradient-selected correlated spectroscopy (gCOSY) NMR results for sample 1 ...... 68
Figure 4-7: Quantification of ethyl acetate in neat resin samples of benzoxazine for a) sample 1,
b) sample 2, c) sample 3, and d) sample 4. The y-axis is in dimensionless units of intensity ...... 71
Figure 4-8: Photographs and binary maps of Formulation A and B showing surface porosity of
carbon fiber panels manufactured using Formulation A and B .................................................... 73
Figure 4-9: Surface roughness measurements of carbon fiber panels manufacturing using
Formulation A and B. Refer to Figure 8 for scale marker ........................................................... 74
ix
Figure 4-10: Storage modulus as a function of temperature for various benzoxazine neat resin
samples .......................................................................................................................................... 76
Figure 5-1: Chemical structure of the monomers and the crosslinked product ............................ 83
Figure 5-2: Exploded view of lab-scale RTM tool [34] ............................................................... 85
Figure 5-3: Composite panel pre- and post-depolymerization (left and right, respectively) ........ 90
Figure 5-4: Recycled carbon fibers under A) 1,000x magnification B) 10,000x magnification C)
A cross section of a carbon fiber part that has been partially digested ......................................... 91
Figure 5-5: Energy dispersive X-ray spectroscopy (EDS) on a partially digested fiber. Left:
Micrograph. Each hatch is 50 μm. Right: corresponding [Ce] as a function of depth. ................ 92
Figure 5-6: (left) Particle size after 24 hours of depolymerization. Note that the x-axis is in log-
scale due to the large order-of-magnitude difference in size distributions within the sample.
(Right). Particle sizes after 72 hours of depolymerization ........................................................... 93
Figure 5-7: A potential mechanism for the depolymerization of a benzoxazine/epoxy resin via
hydride abstraction and its predicted oxidative polymer degradation products ........................... 94
Figure 5-8: NMR data of the extract showing tetraester 4 and degradation products of epoxide 2.
Refer to Figure 5-7 for structures.................................................................................................. 95
x
Abstract
Carbon fiber reinforced polymers (CFRPs) are structural materials used in the aerospace,
automotive, and sporting goods industries due to their high specific properties, which typically
outperform traditional metallic counterparts. CFRP composites are a combination of high-strength
fibers bonded together by a polymeric matrix. While there are a wide variety of methods for
fabricating CFRP composites, such as resin transfer molding, pultrusion, or automated fiber
placement, controlling processing parameters is critical, regardless of the manufacturing method
employed. Poor control of the processing parameters for instance temperature, pressure, or
material handling (i.e. out-time, storage conditions, or resin pre-treatment procedures), can cause
defects, ultimately reducing the material properties of the final manufactured part. The focus of
this dissertation is to consider chemistry and engineering approaches for optimizing both the
manufacture and recycling of CFRP composites.
The first chapter of this thesis considers a method to characterize volatile release in a resin
transfer molded composite part. By understanding the dynamics and kinetics of volatile release
in situ, we can prescribe methods for controlling cure-induced byproducts and eliminate volatile-
induced porosity. In the second chapter, we expand this method and characterize a variety of
resins for use in composites manufacturing to further understand the correlation between volatile
release and porosity formation, as well as to understand the utility of a reaction cell/FTIR
technique. In the third chapter, we consider optimizing the chemistry of the resin, as opposed to
the cure cycle, and characterize the differences in final part properties that arise as a consequence
of each optimization system. In the last chapter, we propose a method for recycling
benzoxazine/epoxy parts via catalytic oxidation, which is ultimately important for opening up
emerging markets for composite applications in both Europe and Asia.
1
Chapter 1: Introduction
1.1: Motivation
Carbon-fiber reinforced polymer (CFRP) composites are replacing traditional materials used
in aerospace, automotive, defense, and marine industries due to high specific properties. For
example, plain-weave CFRP composites have stiffness to weight ratio 18% greater than
comparative aluminum parts, and 14% greater than steel. Additionally, CFRP composites have
excellent fatigue tolerance, toughness, and strength, typically better than their metallic
counterparts. When one takes into account the possibility of more complex fiber orientations, the
benefits of CFRP composites become apparent [1,2].
CFRP composites are produced by impregnating stands of pre-woven carbon fibers with a
polymeric liquid resin. The fibers are the primary load bearing members and they provide strength
and stiffness to the structure. The polymeric resin, sometimes referred to as the matrix, distributes
the load between the fibers, protects them from environmental damage, and keeps them oriented
in the proper direction. This latter point is important, as the properties of a CFRP composite are
anisotropic. For example, the tensile modulus and strength of a uni-directional CFRP composite
are at a maximum when they are measured in the direction of the fibers and at a minimum when
measured in the transverse direction. Contrast this with traditional isotropic metallic alloys, which
exhibit nearly equal properties independent of the direction of measurement [1,3].
When characterizing composites, the temperature is just as important as the direction of
measurement. The resins used in CFRP composites are thermosetting polymeric networks formed
by an interconnected and cross-linked web of covalently linked molecules. However, the
properties of polymeric networks depend strongly upon the temperature of measurement. Near
the glass transition temperature, or Tg, the Young’s modulus of the resin can drop by as much as
2
five orders of magnitude. Additionally, the resin changes from a glass-like material with a given
stiffness to one with rubbery, highly viscoelastic properties. In this state, whenever a load is
applied, there is an instantaneous elastic response, subsequently followed by a slower viscous
deformation. As a result, the Tg of the resin often determines the maximum allowable service
temperature of a CFRP composite part [1,4,5].
Consequently, a significant amount of research has been performed for the development of
high Tg resins, especially for use in the aerospace and defense industries. For example, the engine
cowlings, battery storage compartments, and thrust reversers on the Boeing 787 require service
temperatures in excess of 200°C. NASA is also currently developing a new resin, with a Tg greater
than 400°C, for use in manufacturing nozzle flaps, bushings, bearings and engine bypass ducts in
its next generation of launch propulsion systems [6]. Contrast this to a typical bisphenol-A
diglycidyl ether (BADGE or DGEBA) epoxy resin, which has a maximum service temperature of
125°C [1]. One class of resins in particular, known as benzoxazines, has shown great promise due
not only its high Tg, but also its high impact resistance. Additionally, benzoxazines (molecular
structure shown in Figure 1) often exhibit low cure shrinkage, do not release any toxic byproducts
during cure, and copolymerize very well with other species, making them very suitable for use as
a matrix for CFRP composite parts [7].
However, benzoxazine homopolymers can often be quite brittle due to the low molecular
weight of the final molecular structure which arises from the strong hydrogen bonding between
monomers. These hydrogen bonds restrict the mobility of the monomers during the free radical
polymerization reaction, which ultimately impedes network formation. As a result, benzoxazines
are often copolymerized with epoxies. This is because the ring-opening polymerization reaction
of benzoxazines yields phenolic groups; these phenolic groups in turn react with the epoxies,
3
producing additional crosslinking points. This extra crosslinking leads to significant increases in
the flexural stress and ultimate strain as compared to benzoxazine homopolymers [8–10].
Figure 1-1: Representative structure of a benzoxazine monomer
One drawback arising from high temperature resin systems is the complicated cure
mechanism. Additionally, some high temperature systems such as bismaleides, as well as the
benzoxazine/epoxy system that we consider here, emit volatiles during the polymerization
reaction. Understanding the specific origin of these volatiles is often problematic given their
complicated cure mechanism. However, allowing unchecked cure-induced volatilization,
especially in a closed system, can ultimately lead to porosity in the final manufactured
part. Porosity is the most common manufacturing defect in composites processing [11] and can
have detrimental effects on the shear strength [12–15], compressive strength [16], bending
strength, as well as fatigue life [17] of the ultimate part. Therefore, it is imperative that porosity
be eliminated at all parts of the composites manufacturing processes.
Historically, composites manufacturing was performed via a hand layup technique. While this
technique was reliable, it is by nature very labor intensive and slow, resulting in very expensive
parts. Recently, there has been interest in manufacturing methods that are more amenable to mass
production. These methods include filament winding, liquid injection molding, pultrusion, and
compression molding. Of particular importance is resin transfer molding (RTM), which is an
injection molding technique that allows for the production of composite parts with complex shapes
at low to moderate (5,000 to 50,000 parts per year) production rates [1]; consequently, this
4
manufacturing process has gained widespread attention in the aerospace, automotive, and defense
industries.
In an RTM manufacturing process cycle, several layers of a dried preform are loaded into a
two-part mold. Then, the mold is sealed and resin is injected. The resin then spreads throughout
the mold, soaking the dry fiber preform and displacing any air within the system. After the fibers
are completely saturated, the curing process begins. Depending on the cure temperature of the
resin and the resin system being used, curing can be done at room temperature, in situ if the mold
is equipped with heaters, or in an oven. After the resin has cured, the mold is opened, the part is
removed, and the cycle repeats. A schematic of this process is shown in Figure 1-2. In terms of
production costs, RTM is often a cost-saving alternative to other composite manufacturing
techniques such as vacuum bagging or standard compression molding due to the low labor and
capital costs involved.
Figure 1-2: Schematic of the main steps of an RTM manufacturing process
Traditionally, porosity in RTM processes is caused by improper filling during the injection
process, which has been the conventional research focus [18–21]. This is frequently caused by the
resin flowing through the fiber preform at unequal rates due to competing viscous and capillary
effects, ultimately resulting in macro- or microvoids in the final part, as seen in Figure 1-3. Aside
from flow-induced porosity, volatile-induced porosity, caused by volatiles generated during the
resin polymerization reaction, can occur as well. The literature concerning volatile-induced
porosity is sparse, as typical RTM resins are low volatilizating (<1% volatilization during cure at
5
ambient pressure). However in some cases, such as in the benzoxazine/epoxy resin system we
consider, there can be up to 15% weight loss (under ambient pressure) during cure. This
manuscript will describe methods to quantify, reduce, or possibly even eliminate this cure-induced
volatilization.
Figure 1-3: Fiber impregnation mechanisms resulting in flow-induced porosity. Figure reproduced from [18]
Chapter 2 provides an overview of how volatilization during cure can be quantified using
thermal and chemical analysis techniques such as differential scanning calorimetry (DSC), thermal
gravimetric analysis (TGA), rheology, and Fourier-Transform Infrared (FTIR) spectroscopy, and
how these volatiles can ultimately result in both surface and through-thickness porosity. A novel
method to detect volatilization in situ using a reaction cell will be introduced here as well and
processing changes will be suggested in an effort to reduce ultimate part porosity.
Chapter 3 describes how the method of volatile characterization described in Chapter 2 can be
generalized to other resin systems such as polyimides, pre-pregs, and epoxies. By quantifying the
cure-induced volatile release in these systems, we can develop strategies for volatile mitigation,
prescribe techniques for material handling, as well as suggest additional processing steps to reduce
6
cure-induced volatilization. We also show that the method described in Chapter 2 is superior to
traditional methods for characterizing in situ volatile release.
Chapter 4 describes a chemistry-based method for optimizing final part properties. In contrast
to Chapters 2 and 3, wherein engineering-based approaches are used to maximize final part
properties, the chemistry of the resin is optimized. By using a hypothesis-driven approach, as
opposed to a phenomenological one, we develop a foundational understanding of defect formation
mechanisms and the sources for porosity. We also demonstrate that with proper formulation
changes, final part properties can be increased and porosity can be eliminated.
Lastly, Chapter 5 describes a catalytic approach for recycling carbon fiber composites.
Presently, there are no viable methods for recycling carbon fibers at the end of their lifecycle.
Current techniques include mechanical grinding or pyrolysis of composites, however, these are
not sustainable techniques. In addition to the high cost associated with pyrolysis and mechanical
grinding, the lack of a recycling option has prevented the use of composites within automotive
markets in Europe, thus preventing wider-scale adoption. In this chapter, we demonstrate a
potential pathway for the safe and environmentally sustainable approach for recycling CFRP
composites.
7
1.2: References
[1] Mallick PK. Fiber-Reinforced Composites: Materials, Manufacturing, and Design, Third
Edition. 3rd ed. Boca Raton: CRC Press; 2007.
[2] Ashby M, Shercliff H, Cebon D. Materials: Engineering, Science, Processing and Design.
2nd ed. Oxford: Elsevier; 2010.
[3] Hedley CW. Mold Filling Parameters in Resin Transfer Molding of Composites. Montana
State University, 1994.
[4] Seymour RB, Carraher Jr. CE. Mechanical Properties of Polymers. Struct. Relationships
Polym. SE - 5, Springer US; 1984.
[5] Kaufman HS, Falcetta JJ. Introduction to polymer science and technology: an SPE
textbook. Wiley; 1977.
[6] Xie W, Pan W-P, Chuang KC. Thermal characterization of PMR polyimides. Thermochim
Acta 2001;367-368:143–53.
[7] Ghosh NN, Kiskan B, Yagci Y. Polybenzoxazines—New high performance thermosetting
resins: Synthesis and properties. Prog Polym Sci 2007;32:1344–91.
[8] Rao BS, Rajavardhana Reddy K, Pathak SK, Pasala AR. Benzoxazine–epoxy copolymers:
effect of molecular weight and crosslinking on thermal and viscoelastic properties. Polym Int
2005;54:1371–6.
[9] Demir KD, Kiskan B, Aydogan B, Yagci Y. Thermally curable main-chain benzoxazine
prepolymers via polycondensation route. React Funct Polym 2013;73:346–59.
[10] Ishida H, Ning X. Phenolic Materials via Ring-Opening Polymerization: Synthesis and
Characterization. J Polym Sci Part A Polym Chem 2003;32:1121–9.
[11] Strong AB. Fundamentals of composites manufacturing: materials, methods and
applications. Society of Manufacturing engineers; 2008.
[12] Bowles KJ, Frimpong S. Void effects on the interlaminar shear strength of unidirectional
graphite-fiber-reinforced composites. J Compos Mater 1992;26:1487–509.
[13] Mueller de Almeida SF, CHAVES DE MAS SANTACREU A. Environmental effects in
composite laminates with voids. Polym Polym Compos 1995;3:193–204.
[14] Jeong H. Effects of voids on the mechanical strength and ultrasonic attenuation of
laminated composites. J Compos Mater 1997;31:276–92.
8
[15] Stone DEW, Clarke B. Ultrasonic attenuation as a measure of void content in carbon-fibre
reinforced plastics. Non-Destructive Test 1975;8:137–45.
[16] Suarez JC, Molleda F, Guemes A. Void content in carbon fibre/epoxy resin composites
and its effects on compressive properties. ICCM/9 Compos Prop Appl 1993;6:589–96.
[17] De Almeida SFM, Neto Z dos SN. Effect of void content on the strength of composite
laminates. Compos Struct 1994;28:139–48.
[18] Leclerc JS, Ruiz E. Porosity reduction using optimized flow velocity in Resin Transfer
Molding. Compos Part A Appl Sci Manuf 2008;39:1859–68.
[19] Rodriguez E, Giacomelli F, Vazquez A. Permeability-Porosity Relationship in RTM for
Different Fiberglass and Natural Reinforcements. J Compos Mater 2004;38:259–68.
[20] Rohatgi V, Patel N, Lee LJ. Experimental investigation of flow-induced microvoids during
impregnation of unidirectional stitched fiberglass mat. Polym Compos 1996;17:161–70.
[21] Sadiq TAK, Advani SG, Parnas RS. Experimental investigation of transverse flow through
aligned cylinders. Int J Multiph Flow 1995;21:755–74.
9
Chapter 2: The Effect of Processing Parameters on Volatile Release for a
Benzoxazine/Epoxy Resin
2.1: Abstract
Volatile release during cure is a potential cause of void formation during the resin transfer
molding of complex thermosetting resins. In this study, a blended benzoxazine-epoxy resin system
is analyzed to determine the rate at which volatiles are evolved, as well as the dependence of that
rate on process parameters. The evolution of thermophysical and thermochemical resin properties
is characterized using differential scanning calorimetry (DSC) and thermogravimetric analysis
(TGA). The identity and rate of evolution of the gaseous byproducts released during cure are
determined at ambient pressure using a Fourier transform infrared spectrometer (FTIR) linked to
a reaction cell. The results show that gas release during cure can be reduced but not eliminated by
degassing at elevated temperature. Furthermore, the results indicate that the nature and rate of
volatile release can be modified by judicious selection of cure cycle, as shown by a preliminary
analysis of manufactured neat resin panels.
2.2: Introduction
Resin transfer molding (RTM) is a closed-mold process traditionally used to manufacture
aerospace and automotive composite components with complex geometric features, good surface
finish, and low porosity. During RTM, microstructural defects can be caused by incomplete resin
flow or the entrapment of gaseous species. Flow-induced porosity arises when the infiltrating resin
gels before the fiber preform has been fully saturated. Incomplete infiltration may be caused by
low preform permeability, high resin viscosity, or an improper selection of gate or vent locations.
It may, however, be effectively mitigated by judicious material and process parameter selection
10
and mold design. Gas-induced porosity forms if gas bubbles remain entrapped within the part at
the time of resin gelation. Air-induced voids are the most common, and can be eliminated by
degassing the resin prior to injection to remove entrapped or dissolved air, drawing vacuum within
the mold cavity during injection, controlling the resin infiltration velocity such that, in conjunction
with the preform morphology and permeability, both intra-tow and inter-tow gas entrapment are
minimized, and applying sufficient hydrostatic pressure during cure.
Gas-induced voids may also be caused by gaseous by-products of the resin polymerization
reaction. However, most traditional RTM resins are formulated to release minimal volatiles during
cure. The most common thermoset resins used in RTM, epoxies, are easily processed and exhibit
relatively high mechanical properties and chemical/solvent resistance, although their
fire/smoke/toxicity (FST) behavior is poor [1]. Phenolic resins have also been used in RTM
applications because of excellent FST characteristics and high dimensional stability. However,
phenolics are comparatively brittle and have a limited shelf life. Additionally, the synthesis of
phenolics often requires the use of harsh acids or bases, which can ultimately increase
manufacturing and consumer costs [2,3]. Generally, both materials release few volatiles during
cure owing to the limited use of solvents during the fabrication of the resin and the reaction
chemistry itself [4].
Recently, interest in composite applications involving high temperature and extreme
environments has motivated the development of new polymer blends that seek to combine the
desirable mechanical and thermal properties of their constituents while mitigating their drawbacks.
One such combination consists of a benzoxazine and epoxy blend, which retain epoxy-like solvent
resistance and mechanical properties, while exhibiting phenolic-like FST behavior [5,6].
Furthermore, these blends have shown to be stable at room temperature, exhibit minimal cure
11
shrinkage, along with low moisture uptake under hot and humid conditions [7]. The potential
drawback of such systems is a complex chemical cure process, which can consist of both the
polymerization/cross-linking of the individual constituents as well as possible interactions. In
some cases, the progression of the cure reaction can lead to volatile release during cure, and cause
microstructural defects such as bulk and surface porosity.
The extent of this volatile release is exemplified in Figure 2-1, which shows the substantial
mass loss exhibited by the benzoxazine/epoxy resin considered in this study when cured in a
thermogravimetric analyzer (TGA, TA Instruments Q5000 IR) under a ramp to 185°C at 2°C per
minute. At ambient pressure, the mass loss is nearly 15%. Figure 2-2 shows examples of the
surface and through-thickness porosity that can occur within RTM samples produced with this
resin in certain process conditions, even with no entrapped air bubbles initially present.
Figure 2-1: Mass loss over a cure cycle at atmospheric pressure for a high volatilizing resin
Resin transfer molding has been the subject of research for more than two decades, and a
significant body of experimental and modeling literature has been developed [6–12]. However,
only limited studies have been carried out on voids induced by other volatiles, which may arise
12
well after injection, during hydrostatic cure. This relative scarcity is understandable, as
traditionally, little or no weight loss during cure has been one of the requirements for a high-
performance RTM resin. For example, resin systems such as epoxies, poly-ether amide resins
(PEAR), allyl phenol-formadehyde novolacs (AP), phenyl ethynylphenol-formaldehyde novolac
(PEPFN), and bisoxazoline-phenolics are commercial RTM resin types with low-volatilizing
behavior [8].
Figure 2-2: (A) Surface and (B) through-thickness porosity in an RTM sample
Within liquid molding literature, the issue of volatile release during resin cure has been
acknowledged to some extent via the development of specialized RTM tools. Some of these tools
allow the withdrawal of volatiles produced during cure [9] using modular inserts that absorb
offgasses which are subsequently removed [10], while others use a glass window in an effort to
observe and tailor the cure cycle to prevent it. In addition, Ghose, Watson et al. [11] concluded
that monomer degradation can have a significant effect on final part porosity in a vacuum-assisted
RTM, but that tailoring the cure cycle can prevent degradation and void formation. This study
demonstrates that for such materials, successful processing will rely on effective protocols for
reducing defect formation. Such protocols must be in turn derived from a fundamental
1 mm 5 mm
A. Surface porosity B. Bulk porosity
13
understanding of the relationships between material properties, process parameters and the rate
and identities of volatile released.
The issue of porosity induced by sources other than entrapped air has also been studied in the
context of prepreg processing. For example, Grünenfelder and Nutt [12] examined the effect of
dissolved moisture on final part porosity in vacuum bag-only processing, showing that the pressure
of vaporized water can create exponential increases in void content with cure temperature.
Naganuma and co-workers [13] examined the relationship between residual prepreg solvent and
voids, showing that residual solvent can drastically affect void topology, and Agius and co-workers
[14] showed that cure cycle modifications could decrease void content by up to a third without any
effect on the final degree of crosslinking. There has also been extensive modelling of void growth
in thermoplastic polymers, as seen in the work by Roychowdhury, Gillespie and Advani [15]. In
addition, Hou and Jensen [16] have examined the use of a novel double-vacuum bag method for
void reduction when volatilization during cure, whether because of water, solvents, or other
reaction byproducts, is a critical concern.
In general, the literature confirms that volatile release during cure complicates defect
suppression. During RTM, numerous process adjustments may be used to modify the resin
degassing and infiltration phases, but only the cure temperature and pressure remain available for
the cure phase, when the resin is stationary and under hydrostatic pressure. As a result, a detailed
understanding of the relationships between these process parameters, the polymerization reaction
of the resin and the nature and rate of volatile release is essential.
In this study, we analyze the factors governing volatile release for a benzoxazine-epoxy
blended resin. The cure kinetics of the resin are determined using standard thermochemical
methods, while volatile release is characterized using thermogravimetry and Fourier transform
14
infrared spectroscopy (FTIR). The combined data is used to identify predominant influences of
volatile release and indicates that volatile release can be mitigated or limited by specific changes
to the process conditions. The results provide viable guidelines for defect reduction and process
optimization.
2.3: Materials and Methods
The resin selected for this study consists of a benzoxazine-epoxy blend developed specifically
for RTM manufacturing. This particular system is a pre-commercial resin system, whose resin
chemistry is specifically formulated to emphasize defect formation phenomena such as volatile
release. Accordingly, issues such as cure-induced volatilization may be more problematic than in
commercially available resin systems. The system is a low-viscosity, one-part thermoset with a
recommended cure cycle consisting of injection at 110°C and three-hour cure at 185°C. The
minimum viscosity of the resin is approximately 0.1 Pa∙s.
The resin was analyzed in both idealized (non-isothermal or isothermal) and realistic cure
conditions using different experimental methods. The resin cure kinetics were measured using
differential scanning calorimetry (DSC) in order to track the polymerization/cross-linking process.
The resin viscosity was tracked throughout cure using parallel plate rheometric dynamic analysis
(RDA). The relationship between cure temperature and volatile release was investigated using a
gas-phase FTIR spectroscopy system coupled to a large-mass reaction cell. Finally, TGA
measurements were performed to determine the mass loss throughout cure. Two sets of conditions
were investigated. First, characterization data was obtained during an idealized non-isothermal
ramp, isothermal dwell or combined ramp and dwell conditions. Afterwards, realistic cure cycles
consisting of both injection and cure dwells were used to simulate realistic thermal conditions.
15
Table 2-1 and Table 2-2 summarize the characterization experiments, and each method is
described in detail as following.
Table 2-1: Summary of all volatile characterization experiments performed. “R” and “D” designate ramps and
dwells, respectively
Type Degassed Cure Type
Ramp Rate
(°C/min)
Nominal Dwell
Time (min) Cure Temp (°C)
TGA Yes R + D 1, 2, 3 120 175, 185, 195
TGA No R + D 2 60 185
TGA No R 1, 2, 3, 5, 10
TGA No D 60 165, 175, 185, 195, 205
DSC No R 1, 2, 3, 5, 10
DSC No D -- 360 165, 175, 185, 195, 205
Cell/FTIR No R + D 1, 3 60 220
Rheology No R + D 1, 3 120 220
Table 2-2: Summary of all “realistic cure cycle” volatile characterization experiments performed. “R” and “D”
designate ramps a dwells, respectively
Type Degassed Cure Type
Ramp Rate
(°C/min)
Dwell
Temp (°C)
Dwell Time
(h)
Cure Temp
(°C)
Cell/FTIR Yes R+D+R+D 2 110 1 175, 185, 195
Cell/FTIR Yes R+D+R+D 2 130 1 185
Rheology Yes R+D+R+D 2 110 1 175, 185, 195
Rheology Yes R+D+R+D 2 130 3 185
Resin de-gassing. When desired, 20 mL samples of the resin were degassed in a 50 mL beaker
within a vacuum oven (Yamato LDP21) at 110°C and approximately 10.1 kPa (0.1 atm). As
described in later sections, a small subset of FTIR characterization experiments was carried out on
non-degassed resin in order to maximize signal intensity. Subsequently, TGA, RDA, and FTIR
analyses were performed on degassed resin using realistic cure cycles to simulate industrial
16
processes. In addition, a separate TGA experiment was conducted to measure the difference in
weight loss between degassed and non-degassed resin.
Cure Kinetics - Differential Scanning Calorimetry. The heat of reaction at elevated temperatures
was measured in calorimetric DSC experiments (TA Instruments Q2000). Neat resin samples
weighing approximately 10 mg were placed in sealed hermetic aluminum pans. Non-isothermal
scans were performed at rates of 1, 2, 3, 5, and 10 ˚C/min from 35˚C to 315˚C, and isothermal runs
were performed at 10˚C increments between 165˚C and 205˚C. After each isothermal run, a non-
isothermal scan at 2˚C/min from 35˚C to 315˚C was performed to measure any residual heat
release. The dynamic scans were used to determine the maximum heat of reaction, while the
isothermal runs were used to estimate the thermodynamically allowable maximum degree of cure
possible at a given cure temperature.
Mass Loss - Thermogravimetric Analysis. Mass loss experiments were carried out by TGA (TA
Instruments Q5000 TGA). The temperature was ramped from 25°C to 175, 185, or 195°C at 1, 2,
3°C per minute and held isothermally for 90 min until there was no further change in mass loss.
These cure temperatures and ramp rates were chosen because they were small changes from the
manufacturer’s recommended cure cycle. In addition to these experiments, the manufacturer’s
recommended cure cycle consisting of a one hour injection dwell at 110°C followed by a 2°C/min
ramp to an isothermal 185°C hold for 3 hours was also performed for samples of degassed and
non-degassed resin. The TGA was purged with nitrogen at a rate of 50 mL/min throughout all
tests to prevent condensation of volatiles within the furnace.
17
Volatile Release - Fourier Transform Infrared Spectroscopy. FTIR analysis was performed on the
gas phase volatiles released during cure. Resin samples weighing approximately 10 g were cured
within a heated reaction cell (Advise Sensing) (Figure 2-3) connected to a gas-phase FTIR
(Thermo Electron Nicolet 4700). The amounts used in the reaction cell are orders of magnitude
greater than in typical TGA-FTIR samples, allowing for greater signal-to-noise ratios and
unambiguous identification of evolved species. The control software within the reaction cell
maintained the temperature to within 1°C of the set temperature, while nitrogen gas swept the
volatiles from the reaction cell headspace into the FTIR. A mass flow controller (Alicat Scientific)
maintained a constant carrier gas flow rate of 5 mL/min (± 0.02 mL/min), as measured at 25˚C.
Figure 2-3: Cross-section view of reaction cell assembled (left) and exploded view of reaction cell assembly (right)
showing (A) N 2 carrier gas inlet, (B) reaction cell fitting, (C) connection to FTIR, (D) heating blocks with attached
heating rods, and (E) sample holder with attached thermocouple.
B
D
D
E
18
The FTIR was equipped with a HgCdTe detector as well as a multi-pass spectroscopic
absorption gas cell (TGA/IR Interface, Thermo Electron) designed to detect and analyze low-
concentration vapor products. The spectroscopic absorption cell was maintained at 240°C to
prevent volatile condensation within the system. Note that in these experiments, the resin is
exposed to ambient pressure, rather than to the higher pressures used during actual molding. As
such, the volatile release behavior represents a worst-case scenario. However, in this manner, we
can detect all volatile species released, assess the relative amounts, and develop suppression
conditions.
While the DSC experiments clarify the cure reaction of the resin, the FTIR experiments were
used to identify the volatile species being released throughout cure, and to determine relative
intensity. Consequently, the effects of specific cure cycle characteristics were studied. Non-
isothermal ramps were performed at 1 - 3°C per minute, as well as realistic cure cycles of various
ramps and dwells. A summary of FTIR experiments performed are provided in Tables 1 and 2.
Rheology. Resin viscosities and gel times were determined by RDA using a rheometer (TA
Instruments ARES 2000EX) outfitted with 40 mm bottom plates and 25 mm top plates, separated
by a gap of 1 mm. Tests were carried out at 0.10% strain with a frequency of 1 Hz within the
linear viscoelastic regime. Resin gelation was defined as a viscosity of 1000 Pa∙s, which was close
to where G’ and G’’ intersected. The rheology experiments are summarized in Table 2-1 and
Table 2-2.
RTM Instrumentation. Neat resin panels were produced at appropriate temperatures and pressures,
as described in the volatile analysis section. These panels were produced in an instrumented, lab-
19
scale RTM mold that afforded accurate control of temperatures and pressures, unlike typical RTM
molds used for part production in the composites industry. The purpose of these experiments was
to confirm the conditions for volatile suppression developed in the previous section. In both
experiments, we used a ramp rate of 2°C/min to 185°C with an intermediate, one-hour injection
dwell at 110°C.
2.4: Results
Cure Kinetics. Figure 2-4A shows results from the non-isothermal DSC experiments. The
evolution of the heat of reaction exhibited a traditional single-peak specific heat flow curve for all
studied conditions. The total heat of reaction, HT, was obtained by integrating the area under the
exothermal peak and dividing by the sample mass. The final results of the calculations are shown
in Table 3. The average HT was calculated to be 448.9 J/g ± 7.2 J/g. The heats of reactions from
isothermal runs are provided in Table 2-3, and data plotted in Figure 2-4B. The total heats of
reaction indicate that with lower temperature dwells, the resin cured only partially. However, the
total heats of reaction obtained by summing the dwell and residual components are consistent with
results from non-isothermal scans.
Figure 2-4: (A) Exotherms versus temperature from non-isothermal scans. (B) Exotherms versus time from
isothermal runs.
A B
20
Table 2-3: Total heats of reaction for non-isothermal scans (left) and specific exotherm energies for isothermal runs
(right)
Non-Isothermal DSC Scans Isothermal DSC Runs
Ramp Rate H T
Temp. Heat of
Reaction
Residual
amount
Total Heat
1°C/min 451.6 J/g
165°C 363.6 J/g 83.28 J/g 446.88 J/g
2°C/min 459.6 J/g
175°C 400.5 J/g 43.02 J/g 443.52 J/g
3°C/min 447.0 J/g
185°C 438.6 J/g 15.15 J/g 453.75 J/g
5°C/min 440.2 J/g
195°C 448.4 J/g 10.14 J/g 458.54 J/g
10°C/min 445.9 J/g
205°C 449.8 J/g 6.39 J/g 456.19 J/g
The relatively consistent values obtained for the total heat of reaction indicates that despite the
presence of multiple thermosetting constituents within the blended resin, the nature of the
polymerization/cross-linking process is largely independent of cure cycle. Furthermore, the
presence of a single heat of reaction peak indicates that the primary constituents of the resin
combine in a single reaction, or react independently but simultaneously.
The degree of cure (α) can be defined from the heat of reaction using Eqs. (1) and (2), and is
shown in Figure 2-5. This figure suggests that the resin is more sensitive to dwell temperature
changes at lower temperatures rather than at higher temperatures.
21
Figure 2-5: Degree of cure evolution versus time for various cure dwell temperatures
𝑑𝛼 𝑑𝑡 =
1
𝐻 𝑇 𝑑𝐻 𝑑𝑡
1
𝛼 =
1
𝐻 𝑇 ∫ (
𝑑𝐻 𝑑𝑡 ) 𝑑𝑡 𝑡 0
2
Thermogravimetric Analysis. The total mass loss as a function of degree of cure for both
isothermal and non-isothermal experiments is shown in Figure 2-6. For non-isothermal
experiments, larger amounts of volatile release were observed for lower ramp rates. Conversely,
for isothermal experiments, the mass loss was more consistent, with 13 to 14% mass loss observed
in all experiments with no degradation observed up to 350°C. The decrease in total mass loss at
high ramp rates may be a consequence of reaction kinetics – at higher ramp rates, the rapidly
increasing degree of cure and viscosity “lock in” solvents and other potential volatiles, preventing
evolution, bubble nucleation and migration towards the free surface of the sample. Because
isothermal dwells at higher temperatures are associated with faster cure, this data suggests that
greater instantaneous rate of mass loss should be observed for higher cure temperatures.
22
Figure 2-6: Weight loss data versus ramp rate, as seen in the TGA for (A) dynamic ramps and, (B) isothermal
dwells.
For cure cycles consisting of a ramp and hold, the ramp rate and cure temperature are
summarized in
A B
23
Table 2-4. At ambient pressure without degassing, volatile release is consistent and
independent of cure cycle, with all samples exhibiting between 14% and 16% mass loss. This
constancy is due to the cure reaction being dominated by the isothermal cure hold, rather than the
non-isothermal ramp. When the resin is degassed (1 h at 110°C, 0.1 atm), volatile release is
reduced by approximately 2%, as shown in Figure 2-7. The consistent weight loss behavior
indicates that the measured mass loss is either concurrent with or caused by the resin cure process.
Figure 2-7: Comparison of degassed versus non-degassed resin
Volatile Analysis and Identification. Figure 2-8 shows FTIR spectral data from the reaction cell
experiments for a 1°C, 2°C and 3°C per minute ramp to a 220°C hold until cure, along with the
cell temperature. To maximize the FTIR signal, the resin in these experiments was not degassed.
The color intensity map within the contour plots indicates the FTIR signal intensity in unit-less
“absorbance units,” and is an indicator of instantaneous volatile concentration. The dashed line in
Figure 2-8 indicates gelation, as determined from the RDA experiments.
0 30 60 90 120 150
85
90
95
100
Mass (%)
Time (min)
Un-degassed
Degassed
0
50
100
150
200
Temperature (°C)
24
Table 2-4: Total weight loss versus ramp rate for degassed samples
Cure Temperature (°C) Ramp Rate (°C/min) Mass Loss (%)
175 1 16.0
175 2 14.5
175 3 14.2
185 1 14.5
185 2 14.0
185 3 15.4
195 1 14.3
195 2 13.6
195 3 13.4
The peak absorbance value in Figure 2-8A indicates that the maximum instantaneous volatile
intensity (which corresponds to the volume- of volatiles within the FTIR) during the 3°C/min ramp
was an order of magnitude greater than during the 1°C/min ramp. This difference is explained by
the DSC data in Figure 2-4A, which shows that the cure reaction occurs at a slower rate, and thus
over a longer time span, when a slower ramp rate is used. The FTIR analysis thus confirms that
volatile release is concurrent with resin cure.
The FTIR data also provides insight into the volatile species released during cure. In our
analysis, we define onset of volatile release as the point when the absorbance of a particular species
exceeds 0.025, which is the signal-to-noise threshold of our detector. In Figure 8A, the onset of
ethyl acetate volatilization occurs at 30 minutes (105°C) as indicated by the C-H alkyl stretch
(2870 wavenumbers), a C=O stretch (1750 wavenumbers), and C-O stretches (1250 and 1055
wavenumbers). Subsequently, carbon dioxide evolves at 210°C (58 minutes) as indicated by the
spike at 2350 wavenumbers, which indicates C=O asymmetric stretching. At the same time, water
appears. Finally, the evolution of aniline occurs during the latter half of the 220°C hold, indicated
25
by absorbencies at 1280, 3360 and 3442 wavenumbers, corresponding to the stretching of C-N and
N-H bonds. Because aniline evolves after gelation, it may involve thermochemical phenomena
unrelated to cure. Indeed, studies have shown that aniline is a degradation byproduct associated
with benzoxazines [17]. Table 2-5 summarizes the species identified by FTIR, along with the
onset times and temperatures during the 3°C/min ramp and hold.
Figure 2-8: FTIR data for (A) 3°C/min ramp to 220°C and hold, (B) 2°C/min ramp to 220°C and hold, and (C)
1°C/min ramp to 220°C and hold. The intensity of the color scale is in absorbance units.
When the resin cure cycle is modified, the volatilization behavior changes accordingly. During
the 1°C/min ramp and hold shown in Figure 2-8C, the evolution of ethyl acetate begins earlier (at
85°C rather than at 105°C). Additionally, the peak absorbance is much lower – 0.25 in the 1°C/min
ramp-and-hold - versus 2.0 in the 3°C/min ramp-and-hold. Finally, as shown in Figure 2-8B, the
results of the 2°C/min ramp is an intermediate between the 1°C/min and the 3°C/min ramp-and-
0 30 60 90 120
4000
3200
2400
1600
800
Wavenumbers (1/cm)
Time (min)
0.0
0.20
0.40
0.60
0.80
1.0
1.2
1.4
1.6
1.8
2.0
25
50
75
100
125
150
175
200
225
Temperature (°C)
A
0 60 120 180 240
4000
3200
2400
1600
800
Wavenumbers (1/cm)
Time (min)
0.0
0.025
0.050
0.075
0.10
0.13
0.15
0.18
0.20
0.23
0.25
25
50
75
100
125
150
175
200
225
Temperature (°C)
C
0 30 60 90 120 150
4000
3200
2400
1600
800
Wavenumbers (1/cm)
Time (min)
0.00
0.04
0.08
0.12
0.16
0.20
0.24
0.28
0.32
0.36
0.40
0
25
50
75
100
125
150
175
200
225
250
Temperature (°C)
B
26
hold. This is evidenced by the maximum absorbance and time-frame in which the volatiles are
emitted during cure. Superimposing the 3°C/min and 1°C/min FTIR data on the DSC data, as
shown in Figure 2-9, supports the assertion that volatile release is related to cure cycle, but also
shows that secondary effects may exist. For the 3°C/min ramp-and-hold, the dα/dt peak occurs
near the peak volatile intensity, whereas in the 1°C/min ramp-and-hold, the same peak appears
~15 min later. These observations indicate that transient volatile release behavior can be affected
by the cure cycle.
Table 2-5: Potential volatiles emitted during the realistic cure cycles and their methods of elimination
Species name
Range of
Evolution
Condition Observed Boiling Point P
vap
at Onset
Elimination
Method
Grease 60-110°C All 69°C 0.75 atm [19] Mold Pressure
Ethyl acetate 60-Gel All 77°C 0.55 atm [20] Mold Pressure
Phenol 170-185°C Realistic cure 182°C 0.83 atm [21] Mold Pressure
Water 190-200°C Characterization 100°C >> 10 atm [22] Degassing
Carbon dioxide 190-200°C Characterization -57°C >> 10 atm [23] Degassing
Aniline > 200°C Characterization 185°C 1.98 atm [24] Temp. Control
27
Figure 2-9: FTIR data with DSC cure rate data superimposed for (A) 3°C/min ramp and (B) 1°C/min ramp to
220°C. The intensity of the color scale in in absorbance units.
The hydrostatic resin pressures achieved in a closed RTM mold can potentially suppress certain
volatiles. However, among the species detected using FTIR, the water and carbon dioxide are
problematic, because their vapor pressures at evolution exceed 1 MPa (10 atm), and therefore
might require high-pressure RTM processing. As a result, volatile mitigation may require
additional strategies. In the next section, we investigate the influence of degassing and selection
of cure cycle on the identity, amount, and rate of released volatiles.
Realistic Cure Cycle Analysis. Experiments consisting of single ramps or isothermal holds
provide useful insights into basic material behavior. However, in practice, cure cycles generally
include both ramps and holds. Here, we analyze four cure cycles representative of typical
manufacturing practice. The first is based on a standard cure cycle (ramp from room temperature
at 2°C/min to 110°C, hold for one hour, ramp at 2°C/min to 185°C and hold until fully cured). The
second and third cycles differ in the cure temperature (dwells at 175°C and 195°C), while the
fourth consists of a modified injection dwell (injection dwell for 3 hours at 130°C rather than
110°C, and cure dwell at 185°C). The ramp rate of 2°C/min was selected due to the high sensitivity
and specificity of volatiles from our previous set of characterization experiments. Prior to each
0 60 120 180 240
4000
3200
2400
1600
800
Wavenumbers (1/cm)
Time (min)
0.00
0.03
0.05
0.07
0.10
0.13
0.15
0.18
0.20
0.23
0.25
0
1
2
3
4
5
6
7
d /dt (1/min)
0 30 60 90 120
4000
3200
2400
1600
800
Wavenumbers (1/cm)
Time (min)
0.0
0.20
0.40
0.60
0.80
1.0
1.2
1.4
1.6
1.8
2.0
0
4
8
12
16
20
d /dt (1/min)
A
B
28
cycle considered, the resin is degassed for 30 minutes at 110°C, at an absolute pressure of 10.1
kPa (0.1 atm).
Figure 2-10 shows that for all four cure cycles, the maximum absorbance values decreased
relative to the 1°C/min and 3°C/min ramp to 220°C, but volatiles were not completely eliminated.
The hypothesis that the rate of volatile generation depends directly on the rate of cure is supported
in this experiment, as the maximum FTIR signal intensity decreased with lower dwell temperature
(and thus lower cure rate). Additionally, the release of aniline is eliminated because the cure
temperature does not exceed 185°C, and with degassing, the presence of carbon dioxide and water
are completely eliminated. Indeed, the sudden spike at 2400 wavenumbers disappears, as do the
peaks at 1280, 3360 and 3442 wavenumbers. Because of the high vapor pressure of both water
and carbon dioxide, eliminating these species is potentially critical for suppressing volatile-
induced porosity. Additionally, removal of aniline to avoid degradation is equally important for
producing high quality parts [17].
29
Figure 2-10: FTIR data from the reaction cell showing volatile release as a function of cure temperature for various
cure cycles. Cure dwell at: (A) 175°C, (B) 185°C, (C) 195°C, (D) Injection dwell for 3 hours at 130°C. The intensity
of the color scale in in absorbance units.
When the cure temperature is reduced from 185°C to 175°C, two additional species are visible
in the FTIR trace – grease, which appears at the beginning of the cure cycle and corresponds to an
absorbance at 3000 wavenumbers (C-H bond), and phenol, which appears at 150 minutes,
corresponding to the three absorbance bands at 750, 1500, and 1600 wavenumbers. From
additional tests, grease appears to be a byproduct of the release agent (Frekote 770-NC), and
appears in FTIR spectra intermittently, only after the release agent is freshly applied in the reaction
cell. The appearance of phenol only in lower-temperature cure dwells, coupled with the shift in
peak volatilization from well into the dwell at 185°C to the end of the ramp at 175°C, indicates
that the release of volatiles can be path dependent with respect to the cure cycle. However, the
0 30 60 90 120 150 180
4000
3200
2400
1600
800
Wavenumber (1/cm)
Time (min)
0.00
0.03
0.05
0.08
0.10
0.13
0.15
0.18
0.20
0.23
0.25
25
50
75
100
125
150
175
200
225
Temperature (°C)
0 30 60 90 120 150 180
4000
3200
2400
1600
800
Wavenumbers (1/cm)
Time (min)
0.00
0.02
0.04
0.05
0.07
0.09
0.11
0.13
0.14
0.16
0.18
25
50
75
100
125
150
175
200
225
Temperature (°C)
0 60 120 180 240 300
4000
3200
2400
1600
800
Wavenumber (1/cm)
Time (min)
0.00
0.01
0.03
0.04
0.06
0.07
0.09
0.10
0.12
0.13
0.15
25
50
75
100
125
150
175
200
225
Temperature (°C)
0 30 60 90 120 150 180 210
4000
3200
2400
1600
800
Wavenumber (1/cm)
Time (min)
0.00
0.02
0.03
0.05
0.07
0.09
0.10
0.12
0.14
0.15
0.17
25
50
75
100
125
150
175
200
225
Temperature (°C)
A B
C D
30
nature and consequences of these volatile shifts on final part quality requires an additional study
in the future.
The FTIR data for the 3-hour dwell at 130°C indicate that volatiles evolve during two periods.
The first occurs at the start of the injection dwell as the temperature reaches 130°C. The second
arises during the cure dwell, and is less pronounced than in previous cycles. The presence of
phenol, previously observed only in the 175°C dwell, is also observed here, indicating that both
the identity and the release rate of the volatiles are affected by temperature and temperature history.
The temperatures of evolution and the boiling point and vapor pressures (P
vap
) of the various
species detected during cure are shown in Table 2-5. Amongst these, the carbon dioxide and water
species have the highest vapor pressures at the temperature of evolution, but can be eliminated
using degassing and appropriate cure cycle selection. The vapor pressures of the volatiles that
could not be eliminated via degassing or cure cycle changes, particularly ethyl acetate, phenol and
grease, can suggest the minimum pressure that must be applied to suppress volatile-induced
porosity. While the actual combination of species released during molding and the resulting void
nucleation and growth phenomena are not yet fully understood, process modifications that mitigate
the release of high pressure volatiles will decrease the total void-generating potential, as well as
reduce the resin pressure required to suppress volatiles.
RTM Instrumentation. To confirm these conclusions, neat resin panels measuring 7.62 cm by 12.7
cm by 0.32 cm (or 3 inches by 5 inches by 1/8
th
inch) were cured in a lab-scale RTM mold featuring
a transparent mold wall. The resin was degassed for 30 min at 110°C and injected at the same
temperature into a vacuum-evacuated mold cavity. The mold was flushed with resin until no
entrapped air bubbles remained within the cavity. Then, the resin outlet was sealed, and two
realistic cure cycles were imposed, consisting of a 2°C/min heat-up ramp to 110°C, a one-hour
31
hold at 110°C, and a 2°C/min ramp to 170°C or 185°C, respectively. Two cure pressures were
used for each dwell temperature: 0 kPa and 103 kPa (gauge) at 185°C, and 0 kPa and 69 kPa
(gauge) at 170°C.
Cure pressure selection was not arbitrary, but was guided by Antoine’s equation [18]. 103 kPa
and 69 kPa are the vapor pressures of the highest volatilizing species at each cure temperature,
185°C and 170°C respectively. Because the vapor pressure is defined as the pressure exerted by
the gaseous species on its surrounding medium, the hydrostatic pressure of the resin required to
collapse each bubble must be greater than or equal to the vapor pressure. Experimental
observations (including those in Figure 2-11) indicate that the critical pressure required to produce
bubble-free neat resin panels is approximately equal to the vapor pressure of the highest
volatilizing species. Note that this agreement does not necessarily imply that the vapor pressure
is a direct indicator of whether void formation can be suppressed, as void nucleation and growth
are complex phenomena that depend on multiple material properties and process factors. However,
because the vapor pressure of the volatile species is a significant driver of void nucleation, this
semi-quantitative analysis shows that it can be used as a first-order approximation.
Photographs of both panels are shown in Figure 2-11. The neat resin plates cured at
atmospheric pressure exhibited substantial porosity. While the morphologies of the bubbles
formed at 170°C differed from those formed at 185°C owing to the voids having time to coalesce,
in both cases substantial amounts of volatiles were generated during cure, indicating insufficient
hydrostatic pressure. Conversely, the panel cured at higher pressure exhibited no porosity,
indicating that void nucleation and growth were effectively suppressed. This experiment confirms
that cure-induced volatilization can be limited with proper process modifications, and validates the
utility of the characterization methodology used in this work.
32
Figure 2-11: Neat resin panels cured at: a) 185°C and 103 kPa of pressure, b) 185°C and atmospheric pressure, c)
170°C and 69 kPa, d) 170°C and atmospheric pressure.
2.5: Conclusions
We have investigated the relationships between the resin cure reaction and the release of
volatile gases for a blended benzoxazine-epoxy RTM resin. DSC experiments were performed to
determine the bulk degree-of-cure and rate of cure as functions of time and temperature. In
addition, TGA and an FTIR spectrometer coupled to a heated reaction cell provided insights into
the identity of volatiles and the volatile release behavior of the resin as functions of both cure cycle
and degree of cure. These methods provided complementary datasets and insights into the
polymerization reaction and the effect of process parameters on volatile release rates.
A
10 mm
B
C
D
33
The results indicate that for resin blends such as the one selected, the identity and rate of release
of volatiles can be tailored by judicious selection and control of cure cycle, and degassing prior to
cure. These process modifications can reduce the driving forces behind defect formation in the
manufacture of composite parts by decreasing the instantaneous rate of volatile nucleation and
growth, and process modifications based on them were shown to inhibit void formation during the
RTM processing of four neat resin plates.
All characterization experiments within this study were conducted at ambient pressures
(because of instrument capabilities) and consequently represent a worst-case scenario in which all
evolved gasses are present. In an RTM mold, high applied pressures can dissolve some evolved
gasses and suppress void growth, as well as potentially influence the cure reaction itself. A clearer
understanding of these phenomena is desirable. In this study, we also did not account for the role
of other defect-forming phenomena, such as entrapped air, inadequate fiber wetting, or cure
shrinkage. Such factors could potentially interact with volatile release, further complicating defect
suppression, and require further study. Nevertheless, an accurate understanding of basic resin
properties and behavior during cure is a necessary first step, and can guide process development
and contribute to manufacturing guidelines for successful and efficient manufacture of high-
quality composite parts.
34
2.6: References
[1] Auad ML, Zhao L, Shen H, Nutt SR, Sorathia U. Flammability properties and mechanical
performance of epoxy modified phenolic foams. J Appl Polym Sci 2007;104:1399 –407.
[2] Knop A, Scheib W. Chemistry and application of phenolic resins. Springer-Verlag Berlin;
1979.
[3] Lee Y-K, Kim D-J, Kim H-J, Hwang T-S, Rafailovich M, Sokolov J. Activation energy and
curing behavior of resol- and novolac-type phenolic resins by differential scanning
calorimetry and thermogravimetric analysis. J Appl Polym Sci 2003;89:2589 –96.
[4] Yan Y, Shi X, Liu J, Zhao T, Yu Y. Thermosetting resin system based on novolak and
bismaleimide for resin-transfer molding. J Appl Polym Sci 2002;83:1651 –7.
[5] Tyberg CS, Bergeron K, Sankarapandian M, Shih P, Loos AC, Dillard DA, et al. Structure –
property relationships of void-free phenolic –epoxy matrix materials. Polymer (Guildf)
2000;41:5053 –62.
[6] Lee LJ, Young WB, Lin RJ. Mold filling and cure modeling of RTM and SRIM processes.
Compos Struct 1994;27:109 –20.
[7] Demir KD, Kiskan B, Aydogan B, Yagci Y. Thermally curable main-chain benzoxazine
prepolymers via polycondensation route. React Funct Polym 2013;73:346 –59.
[8] Nair C. Advances in addition-cure phenolic resins. Prog Polym Sci 2004;29:401 –98.
[9] Christensen IS, Walker MA. Resin transfer molding of composite materials that emit
volatiles during processing. 5686038, 1997.
[10] Pupin C, Loiselle-Shebib V, Ruiz E, Ross A, Dauchier M, Dambrine B, et al. Formation of
Porosities in RTM by VOCS and Water Vapour. Proc. 19th Int. Conf. Compos. Mater.,
2013, p. 8447–57.
[11] Ghose S, Watson KA., Cano RJ, Britton SM, Jensen BJ, Connell JW, et al. High
Temperature VARTM of Phenylethynyl Terminated Imides. High Perform Polym
2009;21:653 –72.
[12] Grünenfelder LK, Nutt SR. Void formation in composite prepregs – Effect of dissolved
moisture. Compos Sci Technol 2010;70:2304 –9.
[13] Naganuma T, Naito K, Kyono J, Kagawa Y. Influence of prepreg conditions on the void
occurrence and tensile properties of woven glass fiber-reinforced polyimide composites.
Compos Sci Technol 2009;69:2428 –33.
35
[14] Agius SL, Magniez KJC, Fox BL. Cure behaviour and void development within rapidly
cured out-of-autoclave composites. Compos Part B Eng 2013;47:230 –7.
[15] Roychowdhury S, Gillespie JW, Advani SG. Volatile-Induced Void Formation in
Amorphous Thermoplastic Polymeric Materials: I. Modeling and Parametric Studies. J
Compos Mater 2001;35 :340 –66.
[16] Hou T, Jensen B. Double vacuum-bag technology for volatile management in composite
fabrication. Polym Compos 2008;29:906 –14.
[17] Ishida H, Sanders DP. Improved thermal and mechanical properties of polybenzoxazines
based on alkyl-substituted aromatic amines. J Polym Sci Part B Polym Phys 2000;38:3289 –
301.
[18] Antoine, C. Tensions des vapeurs; nouvelle relation entre les tensions et les températures.
C. R. Acad. Sci. 1888; 107: 681-684
[19] Carruth GF, Kobayashi R. Vapor pressure of normal paraffins ethane through n-decane
from their triple points to about 10 mm mercury. J Chem Eng Data 1973;18:115 –26.
[20] Polák J, Mertl I. Saturated vapour pressure of methyl acetate, ethyl acetate, n-propyl acetate,
methyl propionate, and ethyl propionate. Collect Czechoslov Chem Commun
1965;30:3526 –8.
[21] Dreisbach RR, Shrader SA. Vapor pressure –temperature data on some organic compounds.
Ind Eng Chem 1949;41:2879 –80.
[22] Bridgeman OC, Aldrich EW. Vapor Pressure Tables for Water. J Heat Transfer
1964;86:279 –86.
[23] Giauque WF, Egan CJ. Carbon Dioxide. The Heat Capacity and Vapor Pressure of the Solid.
The Heat of Sublimation. Thermodynamic and Spectroscopic Values of the Entropy. J
Chem Phys 1937;5:45.
[24] Hatton WE, Hildenbrand DL, Sinke GC, Stull DR. Chemical Thermodynamic Properties of
Aniline. J Chem Eng Data 1962;7:229 –31.
36
Chapter 3: A Method for In Situ Analysis of Volatiles Generated during
Cure
3.1: Abstract
Conventional epoxies used in composites are formulated to release minimal amounts of
volatiles during cure, aside from moisture, which is a perpetual concern. Other matrix polymers,
particularly those designed for high temperature service, such as benzoxazines and polyimides,
generally feature more complex chemistries, and such formulations can release volatiles during
cure, ultimately causing unacceptable porosity. Here we demonstrate a general method for in situ
analysis of volatile release during cure. The method improves upon traditional characterization
techniques of TGA/MS and TGA/FTIR (thermogravimetric analysis, mass spectroscopy, and
Fourier transform infra-red spectroscopy). We characterize a commercial epoxy pre-preg typical
of those used in aerospace, a benzoxazine/epoxy blended liquid resin intended for high-
temperature service, and finally, a high temperature polyimide. By understanding the identity,
origin, and kinetics of volatile release in these three materials, we can prescribe additional pre-
treatments and define process windows to minimize and even eliminate cure-induced porosity.
We demonstrate the utility of a reaction cell/FTIR approach for characterizing cure-induced
volatiles and mitigating final part porosity.
3.2: Introduction
In composites manufacturing, volatile release during cure in any molding environment can be
problematic. While traditional epoxies such as bisphenol A (DGEBA) or tetraglycidal
diaminodiphenyl methane (TGDDM) are designed to release little or no volatiles during cure,
studies have shown that moisture release in epoxy pre-pregs are a source of porosity [1–3]. Other
37
polymers, particularly those formulated for high temperature service such as benzoxazine and
polyimides, have solvent additions and complex chemistries which has been shown to generate
volatiles during cure [4,5].
Regardless of the origin, unmitigated solvent release in both open- and closed-molding
environments has been shown to cause porosity in the final manufactured part, ultimately
decreasing performance and material properties. Additionally, understanding both the kinetics and
the identity of the volatiles released is essential to develop and implement mitigation strategies,
which will depend on when during cure the volatiles are generated (i.e., pre- or post-gelation). For
example, high vapor pressure volatiles that evolve pre-gelation can be removed via degassing,
while volatiles that are generated due to post-gelation can be controlled by judicious temperature
control. These possibilities dictate a need for an in situ method to analyze the volatiles generated
during cure [6].
Current strategies for characterizing cure-induced volatiles in situ rely on conventional
analytical tools. These tools typically include a thermogravimetric analyzer coupled with a mass
spectrometer (TGA/MS), or a thermogravimetric analyzer coupled with a Fourier transform
infrared spectrometer (TGA/FTIR). Yet, these techniques are both limited in scope and
applicability. The applicability of a mass spectrometer generally depends on the particular
instrument configuration, although typical mass spectrometers struggle with identifying mixtures
[7,8]. A mass spectrometer determines the mass-to-charge ratio (m/z) of a particular molecular
fragment, then works backwards to piece the fragments together. When mass spectrometry is
applied to mixtures, analysis is complicated by the fact that the resulting m/z is generally a
convolution of all components in the mixture. While mass spectrometry can provide valuable
insight on the molecular weight of a volatilized species, m/z ratios do not reveal direct information
38
about particular chemical bonds or structures, and thus can be less useful than FTIR spectra for
identification of compounds. While the inclusion of a gas chromatograph (GC) before a mass
spectrometer to aid in the separation of mixtures for easier identification can be useful, this
approach carries the potential for oversaturation of the GC column, and precludes in situ
identification of volatiles.
A TGA/FTIR setup is similar to the TGA/MS technique, although samples used in TGA/FTIR
are generally small (~50 milligrams), and the low signal-to-noise ratios in spectra often make
detection/identification of trace volatiles impossible [9]. Here, we describe how this limitation can
be overcome by simply replacing the TGA with a reaction cell (RC) that accommodates larger
samples (~10 gram samples versus ~50 milligram samples). Larger sample sizes generate much
greater signal-to-noise ratios in FTIR spectra and enable definitive identification of trace volatile
species. Furthermore, the RC can be operated under conditions that replicate actual composite
processing conditions, including high temperature and pressure. These conditions cannot easily
be achieved in a standard TGA without risking instrument damage. For these reasons, the method
we will describe has some distinct advantages over conventional techniques for in situ volatile
characterization.
In this paper, we consider four resin systems commonly encountered in composite processing:
a conventional epoxy pre-preg used for aerospace applications, a liquid benzoxazine/epoxy resin
designed for RTM processing, and two forms of a high-temperature polyimide. These systems
were selected to demonstrate the broad utility of the reaction cell/FTIR technique (RC/FTIR) for
the various material types used in composite manufacturing, along with the range of process
conditions, both temperatures and pressures, that the technique can replicate. We outline how the
39
technique can be used to develop strategies for volatile mitigation, techniques for material
handling, and processing steps to reduce cure-induced volatilization.
3.3: Materials and Methods
Four resin systems were selected for analysis of volatiles released during cure. The resins were
selected to represent the range of forms and types of polymers encountered in the production of
composites. The first resin system was a commercial epoxy pre-preg (Cycom 5320-1, Solvay)
designed for out-of-autoclave, vacuum-bag only cure. The recommended cure cycles are 177°C
for 2 hours or 93°C for 10 hours. Prior to analysis, the pre-preg was conditioned for 24 hr at 90%
relative humidity and 23°C (± 1°C).
The second resin was a benzoxazine-epoxy blended resin designed for liquid molding by RTM
(resin transfer molding). The thermoset resin features a low-viscosity, one-part liquid formulation
with a recommended process cycle of injection at 110°C and a three-hour cure at 185°C. The
minimum viscosity of the liquid resin is approximately 0.1 Pa∙s. Because this resin is intended for
use in molding processes where pressure is a key processing parameter, volatile analysis was
performed by curing at two applied pressures: 101 and 206 kPa (15 and 30 psi absolute).
The final resin system was a high-temperature polyimide resin (TriA-X, Kaneka) for
compression molding of prepreg. The cure cycle consists of a dwell at 250°C for four hours to
allow for imidization, removal of residual solvent, and crystallization, followed by an intermediate
dwell at 325°C for 30 minutes, and a cure dwell at 371°C for two hours. Volatile characterization
experiments were performed on the polyimide neat resin in addition to the polyimide pre-preg. In
both sets, the same cure cycle was used.
The experimental setup (Figure 3-1) for in situ analysis of gaseous volatiles released during
cure consisted of a heated reaction cell (Advise Sensing) connected to a gas-phase FTIR (Thermo
40
Electron Nicolet 4700). The reaction cell consisted of a sample cavity (5 cm in diameter by 1 cm
in height) that could be heated to 400°C and pressures of 450 kPa. A carrier gas (typically
nitrogen) carried evolved volatiles to an FTIR spectrometer via a heated gas line. Note that the
amounts used in the reaction cell (~1-10 g, depending on rate of volatile release) are orders of
magnitude greater than typical TGA-FTIR samples (~50 mg), yielding greater quantities of
volatiles that produced led to increased signal-to-noise ratios in FTIR spectra and unambiguous
identification of evolved species. The control software within the reaction cell maintained the
temperature to within 1°C of the set temperature, while nitrogen gas carried volatiles generated in
the reaction cell headspace into the FTIR. A mass flow controller (Alicat Scientific) maintained a
constant carrier gas flow rate of 5 mL/min (± 0.02 mL/min), as measured at 25˚C. For pressurized
experiments, the mass flow controller was connected downstream of the FTIR, allowing for equal
and uniform pressure to be applied across the entire system.
Figure 3-1: (A, left) Cross section view of a reaction cell, showing (A) inlet port (B) sample compartment (C) Heaters
(D) Control thermocouple (E) Outlet port to FTIR. (B, right) Exploded view of the reaction cell.
10 mm
A
C
C
E
B D
E
B
C
C
D
41
The FTIR was equipped with a HgCdTe detector and a multi-pass spectroscopic absorption
gas cell (TGA/IR Interface, Thermo Electron) designed to detect and analyze low-concentration
vapor species. In all experiments, the Gram-Schmidt intensity, which represents the total
spectroscopic absorbance within the cell, was collected in addition to the interferograms. The
spectroscopic absorption cell was maintained 20°C higher than the reaction cell to prevent volatile
condensation within the system. For analyses of the polyimide, a high-temperature gas cell (TGC-
M-LV from Harrick Scientific), equipped with silver O-rings (Parker Hannifin) and a temperature
controller (ATC-024-2, Harrick Scientific) was used to achieve the high temperatures required.
The resin systems, experimental conditions, and FTIR setup are summarized in Table 3-1.
When needed, mass loss experiments were also carried out by TGA (TA Instruments Q5000
TGA) to complement the qualitative spectroscopic data from FTIR with quantitative weight loss
data. Like the reaction cell experiments, the manufacturer’s recommended cure cycle was used to
define the temperature profile. In all experiments, the TGA was purged with nitrogen flow (50
mL/min) to prevent condensation of volatiles within the furnace.
Table 3-1: Resin systems considered using the RC/FTIR technique
Resin system Sample Size (g) Pressures (kPa, abs) FTIR Accessory
Benzoxazine resin ~10 101, 206 TGA/IR interface
Epoxy pre-preg ~10 101 TGA/IR interface
Polyimide resin ~0.75 101 High temperature cell
Polyimide pre-preg (1-ply) ~1.25 101 High temperature cell
Polyimide pre-preg (3-ply) ~1.25 101 High temperature cell
42
3.4: Results
Resin system 1: Epoxy pre-preg
Figure 3-2 shows time-resolved FTIR spectral data generated during cure conditions for an
epoxy pre-preg that had been conditioned with moisture as described earlier (90% RH). The
temperature profile imposed was identical to the recommended cure cycle. The color intensity map
indicates the instantaneous volatile concentration within the spectrometer, with high absorbances
depicted in red. Volatiles are detected starting at ~30 min and the concentration peaks at ~45 min.
The spectroscopic data indicate that the primary volatile species produced during cure was water,
as expected, and that water release occurred primarily from 30-60 minutes into the cycle [10].
Figure 3-2: Spectroscopic trace for an epoxy pre-preg sample
43
Figure 3-3 shows the Gram-Schmidt intensity corresponding to Figure 2A. The Gram-Schmidt
intensity provides the total absorbance within the FTIR gas cell which can be correlated to the
instantaneous concentration of volatiles within the cell. The correlation is achieved by considering
the total amount of light being absorbed by the sample, as opposed to light at an arbitrary
wavenumber. Therefore, the Gram-Schmidt intensity curve can be interpreted as a
“comprehensive summary” of the spectroscopic waveform data at any given point in time [11]. In
Figure 3-2B, the highest rate of volatile release occurs at ~40 minutes, coincident with the intensity
maxima. At this point, the cell reaches 100°C, and moisture is being released. Afterwards, there
is a slow decay in the intensity signal, although the non-zero intensity indicates that moisture is
still evolving from the pre-preg sample, even when the cell temperature reaches 177°C.
Figure 3-3: Gram-Schmidt trace of the spectroscopic intensity for an epoxy pre-preg sample
44
Note that in Figure 3-2 and Figure 3-3, the spectral intensity peak arises after the cell has
reached 100°C because of the attractive bond between water molecules and epoxy within the pre-
preg. Absent such an attractive force, water would no longer be present after the cell reached
100°C. Because water and epoxy molecules are held by hydrogen bonding, water is difficult to
remove, even at temperatures >100°C and under vacuum. These observations support the
widespread practice of limiting humidity exposure during handling of prepregs. Even small
quantities of moisture can generate unacceptable porosity in parts because of the large volume
expansion of the vapor phase.
Resin system 2: Benzoxazine/Epoxy blend.
Figure 3-4A shows spectroscopic traces generated from a benzoxazine/epoxy blend (liquid) at
ambient pressure. Following the recommended cure cycle, the evolution of three major species
was detected: (a) ethyl acetate (as indicated by the C-H alkyl stretch at 2870 wavenumbers, a C=O
stretch at 1750 wavenumbers, and C-O stretches at 1250 and 1055 wavenumbers), (b) hexanes (C-
H bond stretching at 3000 wavenumbers), and (c) phenol (para-substituted out-of-plane bending
at 850 wavenumbers, aromatic C-C stretch at 1500 wavenumbers, and O-H bond stretching at 3400
wavenumbers) [10]. Note that liquid molding of benzoxazine/epoxy blends requires addition of a
solvent, in this case, ethyl acetate, and hexane is a constituent of common mold release compounds.
By identifying the temperatures at these species begin to evolve, one can determine an approximate
pressure required to suppress volatilization using Antoine’s equation [5,12].
45
Figure 3-4: (A) Spectroscopic trace for a benzoxazine/epoxy sample at ambient pressure. (B) Spectroscopic trace for
a benzoxazine/epoxy sample at 206 kPa (abs) (C) A corresponding IR absorption spectrum for (A) at ~150 minutes.
The pressure exerted by species volatilizing from the polymer contributes to the overall vapor
pressure. Thus, applying gas pressure over the polymer can suppress the evolving species, and if
the applied pressure exceeds the vapor pressure, the species should simply re-solubilize into the
polymer. This effect is manifest in Figure 3-4A and Figure 3-4B. The volatile characterization
experiment in Figure 3-4A is repeated at 206 kPa (abs), and results are shown in Figure 3-4B. The
applied pressure almost completely suppresses the volatiles, and only trace amounts of volatiles
are detected in the FTIR spectrum.
In reality, void nucleation, bubble growth, and the solubility of gasses within a reacting
medium are complex and multi-faceted phenomena. Bubble growth can depend strongly on
material properties, processing conditions, and other factors. However, because vapor pressure is
a significant driver for void nucleation, one can use vapor pressure as a first-order indicator to
determine if bubble nucleation will occur at any time within a specific processing window [13,14].
46
Furthermore, if experiments using the RC/FTIR to simulate process conditions indicate that
volatilizing species cannot be suppressed by applying pressure alone, one can design/prescribe
pre-treatments such as degassing or temperature control, and/or modified cure cycles, to suppress
these species [15].
Resin system 3: Polyimide neat resin
Figure 3-5A and Figure 3-5B show FTIR traces generated from a polyimide powder sample
cured at ambient pressure. Because of the large disparity between the quantities of gasses evolved
during the first 90 minutes and the subsequent cure ramps, the data are displayed in two traces
with different color scales (Figure 3-5A and Figure 3-5B) for clarity. During the first 90 minutes
of cure, the primary volatiles generated are ethanol (O-H stretch at 3400 wavenumbers, C-H stretch
at 3000 wavenumbers, and C-O stretching at 1100 and 1050 wavenumbers) and acetaldehyde
(similar to the ethanol spectrum, with the inclusion of a small aldehyde peak at 3500 wavenumbers)
[10,16]. The significance of these results is threefold. First, we are able to replicate the extreme
temperature typically used for polyimide processing, and to quantify volatile release during cure
using the RC/FTIR system. Second, the volatilization of acetaldehyde is notable - it is a group 1
carcinogen and poses health and safety risks [17]. Industrial-scale processing of this particular
formulation of polyimide could require safety measures to mitigate these risks. Finally, the
evolution of sulfur dioxide (S=O bond doublet peak at 1380 wavenumbers) and carbon dioxide
(C=O doublet at 2200 wavenumbers) is detected during the two subsequent dwells at 350°C and
371°C [10]. The polymer formulation does not contain sulfur molecules by design, raising the
possibility of a residual contaminant acquired during synthesis of the polyimide monomers [18].
47
Figure 3-5: (A) Spectroscopic traces for the first 90 minutes of a polyimide pre-preg cured within the reaction cell.
(B) Spectroscopic traces for the last 180 minutes of a polyimide pre-preg cured within the reaction cell.
The second major component detected, ethanol, is most likely a solvent. The polyimide resin
is solid at room temperature, and thus degassing under vacuum cannot be used to remove solvent
remaining from resin synthesis. Nevertheless, large amounts of ethanol are generated during
imidization, and this requires mitigation. From TGA experiments, more than 20% by weight of
solvent evolves during the initial imidization ramp and hold. The large weight loss from solvent
volatilization can pose challenges, particularly when producing thicker, multi-ply polyimide
laminates, as volatiles can be trapped within the layers. This requires (a) longer processing times
to remove the volatiles, (b) porosity if the solvent is not sufficiently removed, or (c) blistering
when ramp rates and temperatures are excessive.
To illustrate how the analysis can be used, we describe one approach to reducing the amount
of ethanol released during cure of polyimide. In particular, by simply reducing the concentration
of ethanol present, we can relax the processing restrictions, ultimately allowing for higher ramp
rates, shorter dwells, and a faster overall cycle time. Because the monomer is solid at room
temperature and begins to crosslink upon heating, vacuum degassing is not an option. Instead, the
monomer can be lyophilized to remove much of the ethanol, simply by addition of water and
subsequent freeze-drying. Because water and ethanol are completely miscible in all amounts
48
(whereas the polyimide is insoluble in water), the ethanol can be removed with the water during
sublimation. This process also has the advantage of not affecting the out-life of the resin, as there
is no heat step.
Figure 3-6 shows that after only one lyophilization cycle, the amount of weight lost during
cure drops by almost a factor of two, from 21.5% to 11.9%. The plot shows weight loss versus
time, and the cure cycle is plotted on the same graph. The data indicates that freeze-drying can be
an effective approach for removing residual solvent and demonstrates how volatile analysis can be
employed to address processing problems. Admittedly, the chemistry may be complex, and there
may be side reactions that generate ethanol during cure. Nevertheless, lyophilization can be an
effective measure to mitigate volatile release. Note that without the data from the spectroscopic
traces in Figure 3-4, the reduction in cure-induced volatiles shown in Figure 3-5 would not have
been possible. Unambiguous identification of volatiles, coupled with understanding of the kinetics
of volatile release, were required to diagnose and remove the residual solvent remaining from
monomer synthesis.
Resin system 4: Polyimide pre-preg
Figure 3-7A below shows the FTIR spectra when one ply of polyimide pre-preg is heated
within the RC. Not surprisingly, the main volatile components released during the imidization
dwell are ethanol and acetaldehyde, although concentrations were not as high as in the neat resin
experiments. Carbon dioxide was also detected during the high temperature cure, although sulfur
dioxide was not, supporting the hypothesis that it was a residual contaminant originating from the
synthesis of the imide monomer.
49
Figure 3-6: Mass loss during cure for a polyimide sample cured before and after lyophilization
Figure 3-7B shows spectroscopic traces generated from three plies of polyimide pre-preg cured
within the RC. To facilitate comparisons, the first 60 min of each plot are enlarged and presented
in Figure 3-7C and Figure 3-7D. Comparing Figure 3-7C and Figure 3-7D, it takes ~25% longer
for the ethanol and acetaldehyde to clear from the cell during the imidization dwell, even when the
sample sizes are similar. A vertical dotted line has been placed on Figure 3-7C and Figure 3-7D
showing approximate times when the volatiles have cleared the cell. The data indicate that the
volatiles are slow to egress from the pre-preg stack due to diffusion limitations, and that these
limitations are less severe when analyzing only a single ply of pre-preg. In other words, even
though the imidization reaction occurs quickly, as observed by the sudden spike in spectroscopic
intensity during the initial ramp, it takes much longer for volatiles to evacuate from the stack, thus
requiring a long imidization dwell. However, as discussed previously, proper lyophilization of the
polyimide resin before the resin is combined with the fibers in the pre-preg could reduce the dwell
time required by reducing the amount of volatile to evacuate.
50
Figure 3-7: (A) Spectroscopic traces for a 1-ply polyimide pre-preg (B) Spectroscopic traces for a 3-ply polyimide
pre-preg (C) Spectroscopic trace for the first 60 minutes of a 1-ply polyimide pre-preg. (D) Spectroscopic trace for
the first 60 minutes of a 3-ply polyimide pre-preg
3.5: Conclusion
We have demonstrated the utility of a temperature/pressure reaction cell coupled with an FTIR
spectroscometer for in situ quantification of cure-induced volatile release. Example studies were
described to illustrate the ability to quantify the kinetics of water release during cure in epoxy pre-
pregs, the effect of pressure on a benzoxazine/epoxy resin blend, and to identify volatiles released
from a polyimide resin powder and pre-preg. In all cases, we able to recommend processing
parameters or pre-treatments to either broaden the process window, decrease cycle time, or
mitigate/minimize defects in the composite product.
51
The RC/FTIR method described here overcomes some of the major limitations of conventional
characterization techniques, particularly TGA/FTIR or TGA/MS, when used for analysis of
volatiles generated during composite processing. Using only these conventional methods, we
would not have been able to quantify cure-induced volatile release and recommend process
improvements, as the volatiles emitted from TGA samples were well below the detection limits of
the FTIR, even with the most sensitive detectors. Only via the RC/FTIR approach were we able
to generate a sufficient amount of volatiles for quantification. Also, with traditional
characterization techniques, we could not simulate effects of applied pressure during processing,
which ultimately enabled us to identify a simple remedy to suppress volatile release during RTM
of benzoxazine/epoxy composites.
While the work described here describes a general method for in situ quantification of volatiles
and demonstrates the utility of the method, challenges remain. At present, the RC/FTIR setup
cannot be used under vacuum, a condition frequently encountered in composite processing.
Additionally, it is difficult to quantify the amounts of volatile being generated at a particular time
(i.e., in grams), especially when multiple species are present. Despite these limitations, we have
built a system capable of simulating open- and closed-mold environments, capable of reproducing
pressures up to 450 kPa and temperatures up to 400°C, while simultaneous quantifying any cure-
induced volatiles with high sensitivity.
52
3.6: References
[1] Grünenfelder LK, Nutt SR. Void formation in composite prepregs – Effect of dissolved
moisture. Compos Sci Technol 2010;70:2304–9.
[2] Koushyar H, Alavi-Soltani S, Minaie B, Violette M. Effects of variation in autoclave
pressure, temperature, and vacuum-application time on porosity and mechanical properties
of a carbon fiber/epoxy composite. J Compos Mater 2012;46:1985–2004.
[3] Anderson JP, Altan MC. Formation of voids in composite laminates: Coupled effect of
moisture content and processing pressure. Polym Compos 2015;36:376–84.
[4] Hou TH, Jensen BJ, Hergenrother PM. Processing and properties of IM7/PETI composites.
J Compos Mater 1996;30:109–22.
[5] Lo J, Anders M, Centea T, Nutt SR. The effect of process parameters on volatile release
for a benzoxazine–epoxy RTM resin. Compos Part A Appl Sci Manuf 2016;84:326–35.
[6] Lo J, Anders M, Centea T, Nutt SR. Characterizing Volatile Release in an RTM Resin via
Dielectric Cure Monitoring, FTIR, TGA, and DSC. Proc. 59th SAMPE Int. Tech. Conf.,
2014.
[7] Fenn JB, Mann M, Meng CK, Wong SF, Whitehouse CM. Electrospray ionization-
principles and practice. Mass Spectrom Rev 1990;9:37–70.
[8] Groenewoud WM, de Jong W. The thermogravimetric analyser - coupled - Fourier
transform infrared/mass spectrometry technique. Thermochim Acta 1996;286:341–54.
[9] Sibilia JP. A guide to materials characterization and chemical analysis. VCH; 1996.
[10] Chu PM, Guenther FR, Rhoderick GC, Lafferty WJ. The NIST quantitative infrared
database. J Res Natl Inst Stand Technol 1999;104:59.
[11] Sparks DT, Lam RB, Isenhour TL. Quantitative gas chromatography/fourier transform
infrared spectrometry with integrated gram-schmidt reconstruction intensities. Anal Chem
1982;54:1922–6.
[12] Antoine C. Tensions des vapeurs; nouvelle relation entre les tensions et les températures.
Comptes Rendus Des Séances l’Académie Des Sci 1888;107:681–4.
[13] Kardos JL, Duduković MP, Dave R. Void growth and resin transport during processing of
thermosetting—matrix composites. Epoxy resins Compos. IV, Springer; 1986, p. 101–23.
[14] Kardos JL, Duduković MP, McKague EL, Lehman MW. Void formation and transport
during composite laminate processing: an initial model Framework. Compos. Mater. Qual.
Assur. Process., ASTM International; 1983.
53
[15] Anders M, Lo J, Centea T, Nutt SR. Eliminating volatile-induced surface porosity during
resin transfer molding of a benzoxazine/epoxy blend. Compos Part A Appl Sci Manuf
2016;84:442–54.
[16] Evans JC, Bernstein H. The Vibrational Spectra of Acetaldehyde and Acetaldehyde-D1.
Can J Chem 1956;34:1083–92.
[17] National Toxicology Program. NTP 11th Report on Carcinogens. Rep Carcinog 2004;11:1-
A32.
[18] Ishida Y, Miyauchi M, Ogasawara T, Yokota R. Development of “TriA-X”
Polyimide/Carbon Fiber Composites Prepared by Imide Solution Prepregs. Proc. 18th Int.
Conf. Compos. Mater. ICCM-18 (ed Conf. Organ. Jeju, Korea, 21À26 August, 2011.
54
Chapter 4: Eliminating Porosity via Reformulation of a Benzoxazine-
Epoxy RTM Resin
4.1: Abstract
Use of benzoxazine resins in composites is limited by volatile-induced porosity, which
degrades the thermomechanical properties of the product. In the present study, we demonstrate
how to eliminate cure-induced volatilization and volatile-induced defects in benzoxazine
composite laminates, using a chemistry-based approach. Like most resins formulated for high-
temperature service, benzoxazine and benzoxazine/epoxy blends generally include solvent
additives. Consequently, composite parts produced with such resins exhibit higher levels of cure-
induced volatile release, often leading to porosity in the final manufactured part. Here, we develop
a method to eliminate porosity by analyzing volatile release and the effects of residual solvent in
a pre-commercial benzoxazine/epoxy system designed for liquid molding by RTM. Utilizing
thermogravimetric analysis (TGA), nuclear magnetic resonance (NMR) spectroscopy, and
dynamic mechanical analysis (DMA), we correlate the concentration of residual solvent remaining
within the final manufactured part with the Tg, degradation temperature, and dynamic modulus.
Lastly, a resin synthesis method is demonstrated that eliminates residual solvent in order to
produce composite parts with optimal surface finish and thermomechanical properties. The report
outlines a methodology for optimizing blended resin chemistry for production of high-quality
composite parts.
4.2: Introduction
Benzoxazines are an emerging class of phenol-like thermosetting resins which provide superior
fire, smoke, and toxicity properties, as well as excellent thermal and mechanical performance [1],
55
particularly in humid environments [2]. Additionally, they reportedly exhibit near-zero cure
shrinkage [3] and low heat of reaction, allowing the manufacture of large parts without the risk of
an uncontrolled exotherm [3,4]. These properties render benzoxazine-based resins desirable for
applications in electronics, aerospace, and automotive industries.
However, benzoxazine resins are not without drawbacks. They are brittle and can be difficult
to process, factors that present challenges for chemical and process optimization. For example,
some precursors used to synthesize benzoxazines, such as phenol, dioxane, or formaldehyde, often
carry significant health and environmental risks, thus increasing production costs [5]. Recently,
methods for synthesizing benzoxazines have been reported using ethyl acetate as a solvent,
eliminating the use of more hazardous solvents [6]. However, the ethyl acetate-based synthetic
method reportedly can lead to volatilization during RTM cure, which ultimately may result in
porosity in composite parts [7].
Another drawback of benzoxazine resins manifests during liquid molding of composite parts:
with most polymer resins, porosity in RTM-based composites arises from incomplete fiber
wetting, entrapment of air, or flow-induced porosity. However, benzoxazine resins generally
require a solvent such as ethyl acetate to reduce viscosity, and thus are susceptible to cure-induced
volatilization, which can cause porosity even when the injection is perfectly executed (without
entrapped air). Indeed, the tendency for benzoxazine resins to exhibit porosity caused by cure-
induced volatilization has prevented widespread use for industrial applications, despite the
inherent advantages. Anders et al. showed how cure-induced volatilization in benzoxazine-epoxy
blends could be eliminated via cure cycle optimization [8]. In the present study, we show that
even when porosity is suppressed by this approach, the thermochemical properties of the final part
decline because of ethyl acetate entrapped in the final part, which acts as an effective plasticizer.
56
We correlate the residual amount of solvent with thermomechanical properties of the cured
benzoxazine/epoxy part. Next, we consider alternative methods of benzoxazine/epoxy synthesis
that do not require ethyl acetate and compare the difference in thermochemical properties from the
two production methods. Finally, we demonstrate how optimizing resin chemistry enables
production of the highest quality parts.
We consider two formulations of a blended benzoxazine/epoxy resin and characterize the
relationship between resin synthesis and cure conditions, volatile release, and solvent entrapment,
in addition to final part properties, particularly dynamic modulus and glass transition temperature
(Tg).
4.3: Materials and Methods
The weight loss at ambient pressure of a characteristic resin cure cycle was determined by
TGA using the manufacturer-recommended cure cycle. Neat resin samples were cured in a
miniature RTM cell at select cure conditions, then ground and analyzed by NMR spectroscopy to
determine the effects of degassing, cure cycle, and cure pressure on the amount of residual solvent
in the cured resin. The glass transition temperature and stiffness of the cured neat resin panels
were determined by DMA. Next, the resin was reformulated without solvent, and the same
analysis repeated. Lastly, a composite panel was produced with the reformulated resin to
demonstrate ease of processing and the broader impacts of this work on composites manufacturing.
Altogether, the methodology presented here demonstrates (1) how solvents used in the
production of the benzoxazine/epoxy resin can adversely affect the properties of the final
manufactured part, and (2) how the same methodology can be used to guide modifications to the
resin formulation. Ultimately, the phenomena associated with volatile release can indicate how
modifications to the resin formulation can be used to improve part quality.
57
Materials
Two sources of benzoxazine-based resins were used in this study. The first was a commercially
available, pre-blended 3:1 benzoxazine and epoxy system that was used as-is without any
additional purification or modification. The resin was a low viscosity (100 mPa∙s at 110°C), pre-
catalyzed system intended for use in RTM manufacturing processes. The manufacturer’s
recommended cure cycle was a one-hour injection dwell at 110°C and a cure for 3 hours at 185°C
with 2°C/min ramps. Henceforth, this resin system will be referred to as Formulation A.
The second benzoxazine-based resin source was synthesized using a commercial benzoxazine
resin (Araldite CY-179, Huntsman Corporation), a cycloaliphatic epoxy resin (3,4-
epoxycyclohexylmethyl 3,4-epoxycyclohexanecarboxylate from Beantown Chemicals), and an
acid catalyst (A233 from King Industries). These components are the same constituents used for
the first pre-blended resin system, and the details of this synthesis are given below. Henceforth,
this resin system will be referred to as Formulation B.
Resin Synthesis
Formulation B was synthesized by combining benzoxazine and epoxy in a 3:1 ratio. The
benzoxazine and acid catalyst were dried under vacuum for 4 days. The epoxy was lyophilized
with 50 wt% benzene and was allowed to stand under vacuum at 60°C until no benzene was present
(verified by NMR). Afterwards, 750 g of benzoxazine, 250 g of epoxy, and 100 mg of acid catalyst
were combined in an N2 atmosphere (< 0.5 ppm H2O and O2). These reagents were allowed to
mix for 24 hours on a stir plate set to 120 RPM at 80°C under vacuum to ensure full
homogenization. Subsequent NMR analysis on the product confirmed the absence of residual
solvent.
58
Thermogravimetric Analysis (TGA)
Thermal stability studies were carried out using TGA (TA Instruments Q5000). The
temperature was increased from 25°C to 185°C at 2°C/min and held at 185°C for 90 minutes. In
addition, a separate experiment was performed in which the temperature was increased from 25°C
to 350°C at 5°C/min. The TGA was purged with nitrogen at a rate of 5.0 mL/min throughout all
experiments. The TGA data was used to evaluate the thermal stability of the resin [9], as well as
to understand the mass of cure-induced volatiles that were generated over various cure cycles.
Mini-RTM System
Figure 4-1 shows the miniature RTM tool used in this study. The main body of the min-RTM
cell was made of anodized aluminum with inlet and outlet ports, in addition to grooves used to
promote a one-dimensional flow front. A frame featured a rectangular mold cavity of 3.2 × 76 ×
127 mm, defining the dimensions of the molded part. A pressure transducer (GP:50 NY Ltd.
Model 131) along with a K-type thermocouple (Omega Engineering) were mounted in the middle
of the tool cavity. A data acquisition system (cRIO-9076, National Instruments) was used to
acquire temperature and pressure data. Two 300-watt heaters embedded through the length of the
main tool body supplied heat to the mold and were controlled by a PID controller (Watlow EZ-
Zone model PM6R1CA).
59
Figure 4-1: CAD model of the mini-RTM mold
Four neat-resin samples were produced using the mini-RTM cell, and the manufacturing
parameters are given in Table 4-1. Unless specified, the nominal manufacturer-recommended cure
cycle was used. Because Sample #3 was cured with one side of the mini-RTM mold exposed to
the ambient environment, this sample was later post-cured in an oven for 12 hours at 185°C to
ensure full degree of cure, despite radiative heat losses from the exposed side.
Two composite panels were also produced using the min-RTM system. The fiber bed used
was a five-harness satin carbon fiber fabric (Sigmatex Ltd.) with a 3000 fiber per tow count and
an areal weight of 364 g/m
2
. In total, 8 layers of carbon fabric were set in a quasi-isotropic layup.
Injections were performed using a pneumatic injection system (Radius 2100 cc). Vacuum was
applied to the mold cavity both before and during injection via the outlet port until no bubbles
were observed in either the mold or the outlet port. Post-injection, a constant pressure was applied
hydrostatically to the system by closing the outlet port, while the inlet port was left open, and
Outlet valve
Inlet valve
Heating elements
Spacer plate
Digital microscope
Molded sample
Transparent tool plate
60
applying the appropriate pressure using the injector. For consistency, both neat resin and
composite parts were produced using the same manufacturer’s recommended cure cycle,
irrespective of formulation.
Table 4-1: Test matrix of samples considered to determine effect of formulation on final part properties
Sample
Number
Sample
Type
Cure Cycle Degassed
Applied
Pressure
Mold Environment Formulation
1 Neat resin Nominal No 450 kPa Closed A
2 Neat resin Nominal Yes 450 kPa Closed A
3 Neat resin
½°C/min to
185°C
Yes 101 kPa Open to atmosphere A
4 Neat resin Nominal No 450 kPa Closed B
5 Composite Nominal Yes 450 kPa Closed A
6 Composite Nominal Yes 450 kPa Closed B
Nuclear Magnetic Resonance (NMR) Spectroscopy
After the neat resin panels were produced, a ~10 mm × 10 mm section was removed, weighed,
and ground to a fine powder. The grounds were weighed and soaked in ~2 mL of deuterated
chloroform (99.8%, Cambridge Isotope Laboratories) for 7 days while being stirred at room
temperature with a magnetic stir bar. The slurry was then filtered in a packed column with
diatomaceous earth (Celite) and cotton, then collected in an NMR tube for
1
H and gradient
correlation spectroscopy (gCOSY) analysis.
All spectra were obtained on an NMR spectrometer (Varian 500 MHz) with chemical shifts
reported in units of ppm. The acquisition involved a 2.56 minute acquisition sequence in which
8,192 complex points were recorded, followed by a 2 second relaxation delay. A total of 32 scans
61
were completed followed immediately by the start of another acquisition. Chemical shifts were
referenced to the residual
1
H solvent (relative to TMS) with spectra processed with software
(MesRe Nova v. 9.0.0-12821). In experiments where the total amount of ethyl acetate was being
quantified, 1 μL of 99.8% mesitylene (VWR) was also added to the NMR tube as an internal
standard in addition to the filtrate.
Dynamic Mechanical Analysis (DMA)
Dynamic mechanical properties of the cured samples were determined from 150°C to 315°C
via DMA (TA Instruments Q800) to measure Tg and stiffness. These samples were cut from the
mini-RTM neat resin panels, nominally 65 × 10 × 3.2 mm. DMA measurements were carried out
under dual cantilever mode at 0.2% strain, and the frequency of the dynamic force was maintained
at 1 Hz. Each sample was repeated for reproducibility and the reported values represent these
averaged values.
Surface Roughness Measurements
Surface roughness measurements were made with a 3D non-contact optical profilometer (Zygo
NexView) using coherence scanning interferometry. Individual measurements using a 2.75
objective lens with 0.5X zoom measuring 1.75 mm by 1.75 mm were stitched together until the
entire composite panel was imaged. The average surface roughness (Sa in equation 3-1, root mean
square surface roughness (Sq in equation 3-2), and maximum absolute deviation (Sz) were
subsequently calculated with software (Mx v. 6.3.0.31). In equations 3-1 and 3-2 below, Z is the
measured height, a is the sample area, and x and y are the horizontal and vertical dimensions of
the part.
62
𝑆 𝑎 = ∬ |𝑍 (𝑥 , 𝑦 )| 𝑑𝑥 𝑑𝑦 𝑎
3-1
𝑆 𝑞 =
√
∬ (𝑍 (𝑥 , 𝑦 ))
2
𝑑𝑥 𝑑𝑦 𝑎
3-2
4.4: Results
Thermogravimetric Analysis
Figure 4-2 shows results from the TGA studies on the neat resin for benzoxazine Formulation
A. The results indicate that when the resin is cured at ambient pressure with a 2°C/min ramp to
185°C, the resin weight loss is up to 12% due to cure-induced volatilization. In a separate study,
the volatiles were analyzed using in situ FTIR [6], and the primary volatile species was identified
as ethyl acetate, a byproduct of the benzoxazine monomer synthesis. However, when the resin is
cured under positive pressures typical of RTM molding scenarios, only a trace amount of weight
loss (less than 0.1% by weight) is observed. These data indicate that solvent evolution can be
suppressed by application of pressure, and in such cases, solvent will be trapped within the fully
cured polymer matrix, as it is not consumed during the polymerization reaction [7].
63
Figure 4-2: Comparison of cure-induced volatilization for Formulations A and B
Even after the resin is treated with a degas step, volatiles evolve during cure. The degassed
resin shown in Figure 2A was pre-treated (per manufacturer’s recommendation) at 110°C for 60
minutes at 7 kPa (abs), yet still showed 11 percent weight loss over the same cure cycle. The
substantial weight loss occurs because the ethyl acetate is bound to the benzoxazine monomers by
intermolecular forces, making it difficult to remove by conventional degassing. Because ethyl
acetate is both polar and polarizable, it participates in strong van der Waals and dipolar
interactions, thereby leading to its strong intermolecular attractive forces. As we will see, these
forces ultimately cause solvent molecules to remain with the benzoxazine molecules past the point
of gelation and vitrification, even though the cured product is completely insoluble in ethyl acetate.
Indeed, benzoxazines are often selected for applications expressly because of the high wet-Tg
values and the resistance to solvents [3].
The strong intermolecular forces described above make it impossible to completely remove
ethyl acetate, even under more aggressive pre-treatment conditions. Because of this, we attempted
to reformulate the resin without solvent (ethyl acetate). Figure 2 also shows that when the resin is
reformulated without ethyl acetate, the cure-induced weight loss is reduced four-fold, from ~12%
50 100 150 200 250 300 350
80
85
90
95
100
Mass (%)
Temperature (°C)
Form. A
Form. B
0 30 60 90 120 150
85
90
95
100
Mass (%)
Time (min)
Form. A
Form. A (degassed)
Form. B (with water)
Form. B
25
50
75
100
125
150
175
200
Temperature (°C)
64
to ~3% (shown as Form. B (with water)). This 3% weight loss occurs from the evolution of aniline,
which is released as a consequence of the polymerization reaction. Aniline generation typically
occurs at temperatures greater than Tg because of degradation of the benzoxazine molecules [10],
although it can also occur at lower temperatures by the ring opening polymerization reaction.
In Figure 4-3, we propose a mechanism for the formation of aniline as a byproduct.
Polymerization of the resin initiates with thermal heterolysis of the benxoxazine C-O bond of (1)
to give zwitterion (2). Step 2 can proceed alternatively to productive crosslinking with an epoxide
co-monomer to produce resin (5)[11], or can decompose with elimination of N-phenylformimine
(4) to give (3) [12]. The fragmentation of (1) to (3) and (4) can be a stepwise reaction (as drawn)
or a concerted hetero-Diels Alder [4+2] cycloreversion reaction, although in the absence of an
advantageous reagent to decompose (4), this reaction can revert to return benzoxazine (1). In the
presence of water, (4) can hydrolyze to give aniline and a formaldehyde equivalent, both of which
we have observed by
1
H NMR when extracting ground powders of cured resin with chloroform-d.
Figure 4-3: A potential mechanism for the generation of aniline
65
Because water is a necessary precursor for the generation of aniline, synthesis of the resin in a
moisture-free and solvent-free environment should completely suppress the generation of all cure-
induced volatiles. Suppression of aniline is demonstrated in Figure 2. When the resin is processed
in a nitrogen atmosphere, the sample weight loss is reduced to approximately 0.5% over the
standard cure cycle, an amount comparable to traditional epoxy-based resins [13].
Figure 4-4 shows the mass loss behavior when the temperature is ramped past cure to 350°C,
at which point degradation begins to occur. In Figure 4-4, the cessation of cure-induced volatile
emission has been denoted with a solid arrow, and the onset of degradation has been denoted by
the dashed arrows. Note that the onset of degradation, (as determined by the inflection point),
occurs at different temperatures for the two formulations. Degradation begins at 299°C (±2.0°C)
for Formulation A, at 320°C (±1.4°C) for Formulation B (with water), and at 326°C (±1.8°C) for
Formulation B.
Figure 4-4: Solvent evaporation and degradation for formulations A and B
The differences between the degradation temperatures between Formulation A and
Formulation B (with water) indicates that the ethyl acetate remains trapped within the polymer
matrix past cure, even when cured at ambient pressure. Indeed, if all the residual solvent
25 50 75 100 125 150 175 200 225 250 275 300 325 350
80
85
90
95
100
Mass (%)
Temperature (°C)
Form. A
Form. B (with water)
Form. B
66
evaporated during cure, there would be no discernable difference between the two formulations
after cure, and their decomposition temperatures would be similar. In addition, the differences
between the hydrous and anhydrous degradation temperatures of Formulation B highlight the
necessity of a water-free synthesis method as well as support for the mechanism demonstrated in
Figure 4-3.
Although the TGA data indicates that ethyl acetate is present in the final composite part in
Formulation A, the amount of ethyl acetate cannot be determined by TGA alone. In particular,
within a TGA, one cannot exactly replicate the high-pressure molding conditions in an RTM. For
this reason, we chose to use the mini-RTM system to produce neat resin panels and then employ
NMR spectroscopy to detect and measure the presence of ethyl acetate, and to determine the
residual amount in the final product as a function of cure condition.
RTM and NMR Spectroscopy
Figure 4-5 shows 1-D
1
H NMR spectra for Sample 1 (cured within the Mini-RTM system)
with the corresponding peak assignments and integrations. The structure of ethyl acetate is also
given on the top right with labeled carbons. The locations of the peaks are shown at the top of the
plot. These peak assignments and splitting patterns all correspond with literature values for ethyl
acetate in deuterated chloroform [14]. The extraneous peak at δ = 1.56 arises from water.
67
Figure 4-5: Proton NMR results for sample 1
The peak integrations are given below the peaks in purple, with corresponding frequencies at
the top. In NMR spectra, peak intensities are proportional to the number of protons at the indicated
shift. This holds true for the protons bonded at carbons #1 and #4. However, the peak shift at δ =
1.26 ppm at position #5 is problematic, as it integrates to approximately 13 protons rather than the
3 expected from a terminal –CH3 group, because this ethyl acetate overlaps an additional material
with a coincidental chemical shifts at 1.26 ppm.
To distinguish between these possibilities, a gradient-selected correlation spectroscopy
(gCOSY) NMR experiment was performed. This experiment can determine sets of protons that
are bonded to adjacent carbons via the nuclear magnetization transfer phenomena, and are
correlated by scalar coupling. If ethyl acetate were to exist in the filtrate, and correspondingly in
68
the benzoxazine neat resin panel, the peak shift at δ = 1.26 ppm must correlate with the peak shift
at δ = 4.12 ppm.
Figure 4-6 shows the data from the gCOSY for Sample 1. The results from the 1D proton
experiments are shown on top and to the left as a reference. Any data point along the y = x line
implies merely self-correlation, or that a proton correlating with itself. However, a signal is evident
at (1.26, 4.12) ppm and the corresponding inverse (4.12, 1.26) ppm. This outcome indicates
correlation between the two sets of protons in the ethyl group (CH3CH2—) of ethyl acetate, or at
bond positions 4 and 5. Note that the protons at carbon 1 have no directly adjacent carbon-neighbor
protons to couple with, and thus the only signal we observe from these sets of protons is the self-
correlation signal.
Figure 4-6: Gradient-selected correlated spectroscopy (gCOSY) NMR results for sample 1
69
These results indicate that ethyl acetate is present in the final benzoxazine part, along with an
additional convoluting compound. From prior FTIR experiments [6], this convoluting compound
is identified as hexane, likely originating from the mold release agent. Hexanes have peak shifts
at δ = 1.26 and 0.88 ppm, therefore accounting for the “extra” signal. Indeed, a weak peak at δ =
0.88 ppm can be seen in Figure 5 [14].
With ethyl acetate positively identified within the final benzoxazine part, we now try to
quantify the amount, by weight, of ethyl acetate remaining as a function of the processing
parameters outlined in Table 4-1. This is accomplished by introducing a known amount of
mesitylene as an internal standard in the NMR sample and comparing the relative intensities of the
aryl mesitylene peak shifts against the ethyl acetate peak shifts. The mass of the ethyl acetate
trapped within the benzoxazine neat resin panel can then be deduced using Equation 3 below,
where 𝑆 𝐸𝑡𝑂𝐴𝑐 and 𝑆 𝑠𝑡𝑑 are the integrated areas under the peak of ethyl acetate and mesitylene
standard, 𝑁 𝑠𝑡𝑑 and 𝑁 𝐸𝑡𝑂𝐴𝑐 are the numbers of protons in the standard and ethyl acetate represented
by the particular peak shift, 𝑀 𝑊 𝐸𝑡𝑂𝐴𝑐 and 𝑀 𝑊 𝑠𝑡𝑑 are the molecular weights of the ethyl acetate
and standard, and 𝑚 𝑠𝑡𝑑 and 𝑚 𝐸𝑡𝑂𝐴𝑐 are the masses of the standard and ethyl acetate.
𝑆 𝐸𝑡𝑂𝐴𝑐 𝑆 𝑠𝑡𝑑 ∗
𝑁 𝑠𝑡𝑑 𝑁 𝐸𝑡𝑂𝐴𝑐 ∗
𝑀 𝑊 𝐸𝑡𝑂𝐴𝑐 𝑀 𝑊 𝑠𝑡𝑑 ∗ 𝑚 𝑠𝑡𝑑 = 𝑚 𝐸𝑡𝑂𝐴𝑐 3-3
Figure 4-7 shows spectra for the proton NMR experiments with added mesitylene. Because
ethyl acetate had been positively identified within Sample 1, 2D correlation experiments were not
performed on Samples 2 or 3 (the same benzoxazine resin was used for all three samples). In
contrast to Figure 4-5, the integrations shown in Figure 4-7 are relative integrated areas compared
to the mesitylene peak at δ = 6.80 ppm, rather than the number of protons.
70
Sample 1, which was cured at 450 kPa (65 psi) and received no pre-treatment prior to cure,
showed the most ethyl acetate (~0.067 weight percent) trapped within the final benzoxazine part,
as shown in Figure 4-7A. Note that this figure represents a lower bound on the ethyl acetate
remaining in the matrix, as solvent may still be present within the cured sample that could not be
extracted, or solvent that evaporated during the extraction process. From the TGA experiments,
one would expect that the actual amount of solvent trapped within the resin would be closer to
12%.
When the sample was degassed for 60 minutes at 110°C at 7 kPa, the amount of ethyl acetate
recovered decreased by approximately half (to 0.025%), as shown in Figure 4-7B. Lastly, when
the sample was degassed and cured at ambient pressure, with one side allowed to effervesce and
off-gas freely, the amount of ethyl acetate recovered decreased to 0.005% (Figure 4-7C). A
summary of the samples and their corresponding solvent content is given in Table 4-2. It is
remarkable that we were able to recover ethyl acetate in quantifiable amounts. First, ethyl acetate
has a boiling point of 77.1°C, and therefore, degassing the resin at high temperature should remove
it. Qualitatively, an open vial of ethyl acetate exposed to atmosphere will evaporate completely.
Additionally, the combination of high temperature and decreasing solubility in the liquid phase
during resin polymerization should also eliminate all dissolved solvent in principle. However,
NMR measurements reveal a small amount of residual solvent in Sample 3. These results indicate
that intermolecular interactions between the benzoxazine and ethyl acetate bind them together
through degassing, gelation, and ultimately vitrification.
71
Figure 4-7: Quantification of ethyl acetate in neat resin samples of benzoxazine for a) sample 1, b) sample 2, c) sample 3, and d)
sample 4. The y-axis is in dimensionless units of intensity
Table 4-2: Weight percent of ethyl acetate remaining in neat resin samples
Sample Number Ethyl acetate peak area Weight % Ethyl acetate
1 0.58 0.067
2 0.18 0.025
3 0.03 0.005
4 None present 0.000
As described in the following section, intermolecular interactions between benzoxazine and
ethyl acetate affect final part properties, as the ethyl acetate acts like a plasticizer, occupying
interstitial spaces between the polymer bonds. Indeed, the weight percentage of ethyl acetate in
the cured benzoxazine resin, as indicated by the TGA experiments, is within the range of
A B
C
D
72
plasticizers frequently used in plastics and composites manufacturing [15,16]. The ethyl acetate
molecules cause a decrease in stiffness and Tg, occupying spaces where a benzoxazine-
benzoxazine or benzoxazine-epoxy bond should be. Because the solvent molecule cannot
covalently bond to either of these two species, the properties of the final part are diminished
[17,18]. Because the solvent molecules are attracted to the polymer molecules through
vitrification, the only reliable way of removing the ethyl acetate is to reformulate the resin without
solvent.
Figure 4-7D shows the NMR results of Formulation B, which contains no ethyl acetate. Note
the absence of signal at δ = 4.12 and 2.04, which is a necessary indicator of the presence of ethyl
acetate. The other peaks correspond to water (δ = 1.54 ppm), chloroform (δ = 7.26 ppm), hexane
(δ = 1.26 and 0.88), and acetone (δ = 2.16) [14]. Note that the presence of water, chloroform,
hexane, and acetone are results of the sample preparation process rather than the polymerization
reaction. Consequently, the absence of ethyl acetate and other solvents demonstrate that the
composite part is both solvent-free and water-free.
Mini-RTM and Composite Parts
A comparison of composite parts manufactured using Formulation A and Formulation B is
shown in Figure 4-8, along with the corresponding surface porosity values. While both samples
showed surface porosity, Formulation A showed 8.6%, potentially due to the solvents that
volatilized and nucleated bubbles within the resin. Thus, we expect that elimination of volatilizing
species can reduce or eliminate surface porosity in benzoxazine-based composites.
73
Figure 4-8: Photographs and binary maps of Formulation A and B showing surface porosity of carbon fiber panels
manufactured using Formulation A and B
Because Formulation B contained no solvents or volatiles, nucleation of volatilizing species
was eliminated as a possible defect formation mechanism. Instead, the surface porosity that
ultimately forms is due to cohesive failure of the resin from the tensile stress caused by cure
shrinkage [19,20]. In other words, the stresses caused by cure shrinkage within the mold cause
the resin to pull apart, ultimately inducing surface porosity.
Due to the distinctive nature of these defect formation mechanisms, the quality of surface
porosity observed is different. In the composite sample produced with Formulation A, the porosity
manifests primarily as surface roughness. However, in Formulation B, the surface is nearly
smooth and porosity occurs near fiber tows. Only where the pores break through the surface is
roughness present, as shown in Figure 4-9. Note that the composite part produced with
Formulation B exhibited sub-micron surface roughness and less absolute height variation than
Formulation A, as shown in Table 4-3.
Formulation
A
Photograph ImageJ
A
B
10 mm
74
Figure 4-9: Surface roughness measurements of carbon fiber panels manufacturing using Formulation A and B.
Refer to Figure 8 for scale marker
Note that these samples represent a worst-case scenario in terms of surface porosity, with no
attempts to optimize cure cycle to reduce defects. Anders et al [8] showed that holds at various
temperatures can be used to eliminate surface porosity for Formulation A, and a similar approach
can be used to optimize the cure cycle for Formulation B. Furthermore, because Formulation B
shows markedly less surface porosity than Formulation A for a given cure cycle, Formulation B
should be more amenable to optimization. The samples produced with the mini-RTM cell
demonstrated reduction in surface defects due to the formulation change, and subsequently, the
thermomechanical properties of the composite samples were characterized using DMA to show
the effects of solvent on the final part.
Table 4-3: Porosity and surface roughness measurements of carbon fiber panels manufactured from Formulation A
and B
Formulation Surface Porosity (%) Sa (µm) Sq (µm) Sz (µm)
A 8.6 14.528 18.156 481.193
B 5.8 0.985 4.445 330.976
Formulation A Formulation B
75
Dynamic Mechanical Analysis (DMA)
Figure 4-10 shows the storage modulus for the four neat resin benzoxazine/epoxy samples, as
measured by DMA. The storage modulus decreases as the amount of residual ethyl acetate within
the neat resin increases. Sample 1 had no pre-treatment to remove ethyl acetate and thus exhibited
the lowest storage modulus, whereas Sample 3 was pre-treated and cured in a manner that
maximized solvent removal, and thus showed the highest storage modulus from Formulation A.
More importantly, Sample 4, which was taken from a neat resin panel reformulated with no
residual solvents, showed the highest storage modulus out of all samples tested. Like the NMR
experiments, the results, summarized in Table 4-3, show direct consequences of material and
process parameters on the storage modulus of the three samples.
The measured values of the glass transition temperature (Tg) show dependence similar to the
storage modulus, because solvent molecules interfere with the degree of crosslinking. In principle,
low molecular motion at high temperatures ultimately leads to a high Tg [21]. However, when the
degree of crosslinking and the length of polymer chains in a network is reduced, the Tg will
correspondingly decrease. By examining the onset of the E' drop-off, Sample 4 showed the highest
value of Tg (206°C), Sample 1 showed the lowest Tg (168°C), and Samples 2 and 3 showed
intermediate values. This ordering is attributed to solvent molecules occupying free volume that
would otherwise be occupied by a polymer bond. This data, along with the storage modulus data,
is summarized in Table 4-4.
76
Figure 4-10: Storage modulus as a function of temperature for various benzoxazine neat resin samples
Within the context of composites manufacturing, the glass transition temperature functions as
an upper bound on service temperature, and reducing the Tg limits the service temperature (and
applications) of the resin. Because benzoxazines are frequently considered for high-temperature
applications, reducing the maximum Tg achievable correspondingly limits utility. Only by
reformulating the benzoxazine/epoxy blend were we able to maximize the material properties of
the cured resin.
Table 4-4: Weight percent of ethyl acetate remaining versus their physical properties
Sample Number Weight % Ethyl acetate E' at 100°C Tg
1 0.067 3075 MPa 168°C
2 0.025 3080 MPa 174°C
3 0.005 3152 MPa 192°C
4 0.000 3341 MPa 206°C
100 125 150 175 200 225 250 275
0
250
500
750
1000
1250
1500
1750
2000
2250
2500
2750
3000
3250
3500
Sample 1
Sample 2
Sample 3
Sample 4
Stiffness (MPa)
Temperature (°C)
77
4.5: Conclusions
We have shown that ethyl acetate, a residual byproduct of the benzoxazine synthesis,
ultimately reduced the Tg and stiffness of the final cured part. While we were able to partially
improve the final part properties via degassing, it was only through reformulation that we were
ultimately able to maximize the Tg and stiffness.
The experimental techniques used in this study are complementary. In particular, (a) the mini-
RTM cell in conjunction with (b) NMR provide insights into how various processing parameters
directly affect the performance and usability of a final manufactured part. Combining these
methods constitutes a powerful method for optimizing cure cycles of high-volatilizing and high-
performance polymers such as benzoxazines to ensure maximum part quality. Indeed, the
combination of these two techniques could ultimately allow for process optimization via a
systematic scientific approach, as opposed to a phenomenological approach.
These results demonstrate the importance of optimizing the process chemistry of blended
resins used in liquid molding, as even trace amounts of ethyl acetate can adversely affect property
levels and ultimately degrade part quality. However, ethyl acetate can be difficult to remove
through traditional methods of degassing, indicating that a different pathway may be needed to
synthesize the benzoxazine, particularly to achieve a formulation free of solvents. In addition,
water was also identified as a contributor to cure-induced volatilization. While water was not a
solvent released during cure, it enabled an unwanted side-reaction that generated aniline. To
polymerize the benzoxazine/epoxy blend without the generation of volatiles, the reaction must
occur without water or other residual solvents.
In a broader context, findings presented here have implications for composite processing. In
particular, we have demonstrated a methodology useful for reducing porosity and eliminating
78
surface roughness in a blended benzoxazine resin for liquid molding. We have not optimized the
formulation or process, although our methodology reveals a viable pathway for optimization and
furnishes a foundational understanding of sources of porosity and other defect formation
mechanisms. The methods demonstrated in the present study illustrate the utility of analytical
tools for identifying sources of porosity in blended high-temperature resins, a necessary step before
optimizing formulations, process parameters, and protocols. In addition, our approach can also be
applied to any resin system, opening the door to formulations for composite resins that can
populate regions of the performance/cost space which are presently vacant.
In performing these experiments, we have shown that reducing and eliminating porosity in
benzoxazine-based composites is possible with proper formulation changes and manufacturing
protocols. At present, this comes at a cost of extra processing steps, which will increase both time
and cost. While these processing steps may not be feasible in a manufacturing environment, they
pinpoint locations where chemical optimization strategies can be applied. Future efforts will be
devoted to further reducing/eliminating surface defects and reducing processing time by
optimizing the cure cycle. Additionally, because water was a requirement for the formation of
aniline, opportunities to optimize performance may arise from using a catalyst (e.g., acid
anhydrides or chlorides) to consume water during initiation of the polymerization reaction, as
production and cure of the benzoxazine/epoxy resin under strictly water-free conditions will most
likely be cost-prohibitive in industrial settings.
79
4.6: References
[1] Kim HJ, Brunovska Z, Ishida H. Synthesis and thermal characterization of
polybenzoxazines based on acetylene-functional monomers. Polymer (Guildf)
1999;40:6565–73.
[2] Ishida H, Allen DJ. Mechanical characterization of copolymers based on benzoxazine and
epoxy. Polymer (Guildf) 1996;37:4487–95.
[3] Ishida H, Allen DJ. Physical and mechanical characterization of near-zero shrinkage
polybenzoxazines. J Polym Sci Part B Polym Phys 1996;34:1019–30.
[4] Wang Y-X, Ishida H. Cationic ring-opening polymerization of benzoxazines. Polymer
(Guildf) 1999;40:4563–70.
[5] Ning X, Ishida H. Phenolic materials via ring-opening polymerization of benzoxazines:
Effect of molecular structure on mechanical and dynamic mechanical properties. J Polym
Sci Part B Polym Phys 1994;32:921–7.
[6] Li WH, Jiang W. Methods for preparing benzoxazines using aqueous solvent, 2012.
[7] Lo J, Anders M, Centea T, Nutt SR. The effect of process parameters on volatile release
for a benzoxazine–epoxy RTM resin. Compos Part A Appl Sci Manuf 2016;84:326–35.
[8] Anders M, Lo J, Centea T, Nutt SR. Eliminating volatile-induced surface porosity during
resin transfer molding of a benzoxazine/epoxy blend. Compos Part A Appl Sci Manuf
2016;84:442–54.
[9] Astm. Standard Test Method for Thermal Stability by Thermogravimetry 1 2014;E37-1:1–
5.
[10] Ishida H, Sanders DP. Improved thermal and mechanical properties of polybenzoxazines
based on alkyl-substituted aromatic amines. J Polym Sci Part B Polym Phys 2000;38:3289–
301.
[11] Oral E, Godleski Beckos C, Malhi AS, Muratoglu OK. The effects of high dose irradiation
on the cross-linking of vitamin E-blended ultrahigh molecular weight polyethylene.
Biomaterials 2008;29:3557–60.
[12] Jubsilp C, Punson K, Takeichi T, Rimdusit S. Curing kinetics of Benzoxazine–epoxy
copolymer investigated by non-isothermal differential scanning calorimetry. Polym Degrad
Stab 2010;95:918–24.
[13] Shokralla SA, Al-Muaikel NS. Thermal properties of epoxy (DGEBA)/phenolic resin
(NOVOLAC) blends. Arab J Sci Eng 2010;35:7–14.
80
[14] Fulmer GR, Miller AJM, Sherden NH, Gottlieb HE, Nudelman A, Stoltz BM, et al. NMR
chemical shifts of trace impurities: Common laboratory solvents, organics, and gases in
deuterated solvents relevant to the organometallic chemist. Organometallics 2010;29:2176–
9.
[15] Hunrath C, Tran K. Flexible benzoxazine resin. US 2013/0317155 A1, 2013.
[16] Kumar SR, Dhanasekaran J, Mohan SK. Epoxy benzoxazine based ternary systems of
improved thermo-mechanical behavior for structural composite applications. RSC Adv
2014;5:3709–19.
[17] Levine H, Slade L. Water as a plasticizer: physico-chemical aspects of low-moisture
polymeric systems. In: Franks F, editor. Water Sci. Rev. 3, 1988, p. 79–185.
[18] Sears JK, Darby JR. The technology of plasticizers 1982.
[19] Eom Y, Boogh L, Michaud V, Sunderland P, Månson J. Stress-initiated void formation
during cure of a three-dimensionally constrained thermoset resin. Polym Eng Sci
2001;41:492–503.
[20] Merzlyakov M, McKenna GB, Simon SL. Cure-induced and thermal stresses in a
constrained epoxy resin. Compos Part A Appl Sci Manuf 2006;37:585–91.
[21] Fox TG, Flory PJ. Second-Order Transition Temperatures and Related Properties of
Polystyrene. I. Influence of Molecular Weight. J Appl Phys 1950;21:581.
81
Chapter 5: Recycling Benzoxazine-Epoxy Composites via Catalytic Oxidation
5.1: Abstract
Carbon fiber reinforced polymers (CFRPs) are structural composites used in the aerospace and
sporting goods industries. Their chief value lies in their high specific properties, which generally
outperform metallic counterparts. There is a contemporary need for viable methods for recycling
CRFPs at the end of their lifecycles and for utilizing the considerable production waste (c.a. 30%)
of CFRP part manufacturing [1,2]. The cost associated with these waste streams is a principal
economic driver inhibiting the penetration of CRFPs into larger-scale manufacturing, particularly
in the automotive industry [3]. Reported techniques for CRFP degradation involve pyrolysis or
mechanical grinding of the CFRP, which are respectively outlawed in some jurisdictions and can
reduce the thermomechanical properties of the recycled products [4,5]. In this study, we report a
conceptually different approach to degrading a commercial blended benzoxazine/epoxy resin
under mild, oxidative conditions. The thermosetting resin is polymerized, characterized, and then
catalytically de-polymerized via hydride abstraction with a ruthenium catalyst. These results
demonstrate a concept for sustainable, low-cost recycling of CFRP composites.
5.2: Introduction
CFRP composites are used increasingly in industry due to the characteristic high strength-to-
weight ratio [6,7]. Plain-weave CFRP composites can exhibit specific stiffness and strength in
excess of 300% greater than aluminum or steel. They have fatigue resistance, toughness, and ther-
mal expansion properties typically superior to metallic alloys [8]. Due to these properties,
composites are widely used in aerospace, recreational equipment, and wind energy industries, and
are emerging in automotive manufacturing. Despite this widespread use, strategies for recycling
82
of composites at the end of their life remain inefficient [9, 10]. This creates a serious sustainability
problem that impedes wider-scale adoption and must be addressed.
Current CRFP recycling strategies involve thermal or mechanical means to remove the polymer
matrix and recover the fibers [11,12]. While these strategies work to some extent, they damage the
fibers and can present pollution emission problems. Solvolytic approaches involving chemical
dissolution of the matrix have been reported, but these have low efficiency and often require
extreme operating conditions, up to 400 °C and 270 bar [13–16]. Recent effort has produced resin
systems that will depolymerize on demand, although these systems typically exhibit inferior
thermomechanical properties [17]. Additionally, a “degrade on demand” strategy does not address
the mass of CRFPs already in use [18–20]. We report here a new recycling strategy that is
applicable to a certain class of optimized CFRP composites already in use. The method
demonstrates two unexpected observations - that (a) it is possible to overcome the inertness of
benzoxazine resins, which are among the most thermally stable and chemically resistant CRFP
matrices known, and (b) it is possible for a homogeneous catalyst to effect bond cleavage within
the CRFP super-structure without a pre-treatment or forcing condition.
Chemical structures for the commercial benzoxazine resin system used in this study are shown
in Figure 5-1. This system is used for aerospace and automotive purposes [21–23]. The
polymerization sequence (the resin curing process) proceeds via intermolecular Friedel-Craft
alkylation by a thermally-generated iminium intermediate [23,24]. The role of the epoxy is to
increase the crosslink density and the thermomechanical properties of the cured polymer, as
benzoxazine/epoxy copolymers have greater Tg, strength, and ductility than the benzoxazine
homopolymer [25]. In our tests, CRFP composites of this system exhibit excellent
thermomechanical properties.
83
Figure 5-1: Chemical structure of the monomers and the crosslinked product
The di-n-benzylaniline linkage of the polymerized matrix maps well on to n-
phenyltetrahydroiosquiniline system popularized about 20 years ago as substrates for catalytic C—
H oxidation via mild, oxidative iminium formation, thus making it an attractive target for cleavage
under analogous conditions. [26–30] To test this strategy, we fashioned a screen of oxidants and
catalysts (Table 5-1) in which we measured the mass of polymer dissolved from a solid (fiber-free)
benzoxazine polymer grounds. Among oxidants, hydrogen peroxide was selected, because it is
found in some known CFRP digestion schemes; oxone was selected as a peroxide source. Ceric
ammonium nitrate (CAN) and periodate were chosen because they are known to be effective
surrogates for electrolysis in ruthenium-catalyzed water oxidation. Somewhat surprisingly, the
benzoxazine matrix is stable to most of these conditions, which attests to the chemical inertness of
benzoxazine composites, but CAN was able to effect dissolution of the polymer matrix in the
presence of RuCl3.
84
5.3: Materials and Methods
General Procedures. Air and water sensitive procedures were carried out either in a Vacuum
Atmosphere glove box under nitrogen (2-10 ppm O2 for all manipulations) or using standard
Schlenk techniques under nitrogen. Methanol-d4, D2O, chloroform-d, and dimethylsulfoxide-d6
were purchased from Cambridge Isotopes Laboratories and used as received. Dichloromethane,
diethyl ether, THF, and hexanes are purchased from VWR and dried in a J. C. Meyer solvent
purification system with alumina/copper(II) oxide columns.
SEM photographs were recorded on a Low-Vacuum electron microscope (JEOL JSM-6610)
with a tungsten filament. The working parameters of the recorded images, such as working
distance and accelerating voltage, are recorded in the images in the main text. All EDX spectra
were recorded in situ within the electron microscope during imaging. The spectrometer detector
used was a Sapphire Si (Li) unit and the spectra was recorded with Genesis software.
NMR spectra were recorded on a Varian VNMRS 500 or VNMRS 600 spectrometer, processed
using MestreNova. All chemical shifts are reported in units of ppm and referenced to the residual
1
H or
13
C solvent peak and line-listed according to (s) singlet, (d) doublet, (t) triplet, (dd) double
doublet, etc.
13
C spectra are delimited by carbon peaks, not carbon count.
Distilled water (Arrowhead); acetic acid and trifluoroacetic acid (Fisher); RuCl3 and CuOAc
(Strem); KRuO4, CAN, and Oxone (Aldrich); NaIO4 and CuI (Alfa Aesar; and hydrogen peroxide
(30% aq, Macron) were purchased from their respective suppliers and used as received.
Benzoxazine/epoxy resin (BZ 9102) was a gift from the commercial supplier, Henkel. This was
used as received.
1
H NMR analysis of this commercial resin is consistent with the structural
assignments illustrated in Scheme 1 of the main text.
85
Composite panels were made with the resin transfer mold (RTM) shown below in Figure 5-2.
More information concerning the dimensions and construction of the mold used can be found in
reference 34.
Figure 5-2: Exploded view of lab-scale RTM tool [34]
Protocol of Catalyst/Oxidant Screen. Distilled water (5 mL) was added to acetic acid (5 mL) in a
20 mL test tube with a Teflon stir bar (VWR), along with neat, cured, fiber-free benzoxazine matrix
grounds (ca. 50 mg) that had been pulverized in a desktop coffee grinder to give a particulate size
on the order of ca. 1 mm. Oxidant (0.5 molar equivalents to 1 molar equivalent of benzoxazine)
and catalyst (0.5 equiv), were added to each test tube. The test tubes were then capped with a
rubber septum, heated to 80°C and stirred at 300 RPM. After 24 hours of reaction time, the resin
86
grounds were dried in a vacuum oven for 2 hours to remove any excess solvents, then weighed.
The percent decomposition of resin was determined by Equation 5-1:
% 𝐷𝑒𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛 =
𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 𝑀𝑎𝑠𝑠 − 𝐷𝑖𝑠𝑠𝑜𝑙𝑣𝑒𝑑 𝑀𝑎𝑠𝑠 𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 𝑀𝑎𝑠𝑠 5-1
Manufacture of Neat Resin and CFRP Composite Panels. Molded neat-resin samples were
fabricated with the BZ 9102 formulation of Henkel resin. The tool surfaces of the resin transfer
mold (RTM) were coated with a liquid mold release agent (Frekote 770-NC, from Henkel) prior
to sealing the mold. Before injection, the resin was vacuum-degassed for 45 minutes at 110°C at
a reduced pressure of 6 kPa in order to remove residual solvents and entrapped bubbles within the
resin. During injection, two layers of fiberglass were wrapped around the RTM in order to reduce
thermal gradients.
Injections were performed at 300 kPa and 110°C, both per manufacturer’s recommendations,
with a pneumatic injector (Radius 2100 cc). Vacuum was applied to the cavity both before and
during injection via an external vacuum pump. Afterwards, excess resin was flushed through the
outlet port until no bubbles remained visible in the outlet tubing or within the cavity. Post-
injection, a constant hydrostatic pressure was applied to the system by closing the outlet valve,
leaving the inlet valve open, and applying the desired pressure via the injector. All samples were
cured at 450 kPa (absolute) pressure. The cure temperature cycles consisted of a 2°C/min ramp to
185°C, along with a 2 hour hold at 185°C. After 2 hours, the pressure was reduced, the mold
opened, and the sample removed.
Composite panels were manufactured using 8 layers of carbon fabric set in a quasi-isotropic
layup. Injection pressures and temperatures were identical to those used in the neat resin samples,
and the post-fill packing pressure was also set to 450 kPa (absolute). The fiber bed used to mold
87
composites samples consisted of a five-harness satin carbon fiber fabric from Sigmatex Ltd with
an areal weight of 364 g/m
2
and a 3000 fiber/tow count. Both neat resin and composite samples
measured nominally 76 mm by 127 mm (5 inches by 7 inches).
Protocol of Resin Depolymerization. A small sample, measuring approximately 30 mm by 15 mm,
was excised from the larger composite panel by a tile saw and placed in a 500 mL round bottom
flask along with of distilled water (100 mL), triflouroacetic acid (100 mL), RuCl3 (11.93 mg), and
(NH4)2Ce(NO3)6 (70.2 g). The flask was capped with a glass stopper, and the reaction was allowed
to proceed for 24 hours at 60°C while being stirred at 360 RPM. After 24 hours, the individual
fibers were removed via tweezers and washed in a 50% aqueous methanol solution, then patted
dry with a paper towel. After 3 washings, the fibers were immersed in a desktop ultrasonic
cleaning bath (VWR) in a 50% aqueous methanol solution water for 15 minutes, then allowed to
dry at 100°C overnight in an oven.
Three separate degradation procedures for materials isolation following CRFP panel
degradation were used. In the first degradation experiment, after the fibers were recovered, the
volatiles were distilled away from the crude digest by rotary evaporation, then, of methanol (250
mL) and with sulfuric acid (200 μL, concentrated, from Macron Chemicals) were added to give a
pH of ca. 1 as tested by pH strips. The resulting homogeneous solution was stirred at 85°C under
reflux for 24 hours. Afterwards, the solution cooled to room temperature, poured over
concentrated sodium bicarbonate at ethyl acetate (250 mL each) and extracted. The resulting
aqueous phase was extracted twice more with ethyl acetate (250 mL each). The combined organic
fractions dried with MgSO4 and concentrated by rotary evaporation, and resuspended in deuterated
methanol. A compliment of NMR spectra were recorded.
88
In the second degradation experiment, after the fibers were recovered as described above, then
the crude digest solution was allowed to react at 60°C for an additional 72 hours before the
methanolysis was performed. The subsequent methanolysis and extraction procedure was
identical.
In the third degradation experiment, after the fibers were recovered, the original solution was
allowed to react at 60°C for an additional 168 hours (7 days) before the methanolysis was
performed. The subsequent methanolysis and extraction procedure was identical.
Protocol of Dynamic Light Scattering. Dynamic light scattering (DLS) measurements were
performed at 25°C on a BI-200SM (Brookhaven Instrument Corp) with a 25 mW HeNe laser
(λ=632.8 nm, Melles Griot, model number 05-LHP-928) and a BI-9000AT digital correlator.
Diffusion coefficients (D) were determined from the autocorrelators measured at a scattering angle
of 90° using a non-negative least squares (NNLS) regression method. Hydrodynamic radii were
subsequently calculated according to the Stokes-Einstein equation.
DLS samples were prepared by aliquotting a 2 mL sample from the digestion solution, allowing
it to cool to room temperature, and then filtering it through filter paper (Whatman, #1) to remove
stray carbon fibers or crystals of oxidant. DLS measurements were taken for 5 minutes by which
nominally 3 – 4 million data points were observed.
5.4: Results
Table 5-1 below show the results of the catalyst and oxidant screen. Two key points are
important to note here. First, the chemical inertness of the benzoxazine is apparent as it was
unaffected by the majority of homogenization conditions. This is unsurprising as benzoxazines
are used commercially due to their high solvent and moisture resistance properties. The second is
89
that while CAN was strong enough to effect some dissolution of the polymer, due to its high
oxidative ability, it was only through the combination of ruthenium and CAN that homogenization
was achieved.
Table 5-1: Oxidant and catalyst screening for optimal homogenization conditions for benzoxazine resin grounds
Oxidan t
H2O2 Oxone CAN NaIO4
C atal y st
RuCl3 0% 0% 98% 0%
CuOAc 0% 0% ~5% 0%
CuI 0% 0% ~5% 0%
KRuO4 0% 0% ~5% 0%
With effective conditions for homogenization of neat polymer grounds, our next step was to
degrade resin cleanly away from carbon fiber in CFRP parts. This was demonstrated by treating a
sample measuring ca. 30 × 15 mm excised from a larger composite part (figure 2, left) with our
conditions from Table 4-1, entry 3. We found that degradation of solvent (acetic acid) by CAN
limited the efficiency of oxidation; switching to triflouroacetic acid removed this limitation and
enabled complete homogenization of the polymer matrix without damage to the reinforcing fibers.
The recovered fibers retain long-range order. Note that while the fibers have been freed from
the polymer matrix and are now unsupported, they remain entangled together. This indicates that
although the fibers are re-leased from their support, long-range order of the individual bundles, or
fiber tows, is preserved. The organization of the recovered fibers will ultimately facilitate their
reuse, as it obviates any subsequent post-processing to reorganize them [9].
90
Figure 5-3: Composite panel pre- and post-depolymerization (left and right, respectively)
The fibers are not damaged by the catalytic process. Figure 5-3 shows images of recycled carbon
fibers. The long-range order is evident in Figure 5-3. The images in Figure 5-3 show that the
recycling process causes no apparent damage to the recovered carbon fibers, despite the high
oxidative potential of CAN (+1.6 V v. NHE) and the acidity of the solvent (pKa = 0). Preserving
fibers undamaged will result in superior thermomechanical properties of recycled materials
relative to those made of fibers recovered via pyrolysis or high-pressure/temperature solvolysis:
fibers treated with these processes are often damaged during recovery, causing surface pits that
weaken the resulting composite products [13,31].
10 mm
91
Figure 5-4: Recycled carbon fibers under A) 1,000x magnification B) 10,000x magnification C) A cross section of a
carbon fiber part that has been partially digested
The reactive materials move sequentially through the multi-layer composite panel one ply at a
time. We view our reactive conditions as a homogeneous solution of acid (TFA), oxidant (Ce
IV
),
and catalyst (Ru), that degrades a solid, laminar material. Among these, the cerium is easily
detected by energy-dispersive X-ray spectroscopy (EDX). Thus, we designed an experiment in
which we degraded a panel to partial thickness, then cut a slice through it and examined the cross
section by electron microscopy and EDX. Figure 5-4 (left) shows the cross section, with the
partially degraded layer on the bottom of the image. The same figure (right) shows [Ce] by EDX
A
B
C
92
recorded along a red trace in the image. One sees lowest [Ce] in the interior ply of the laminate
and high [Ce] at the boundary of the interior ply and the degrading ply, with the greatest [Ce] at
the interface between the intact composite and the region where the matrix has been digested. At
this interface, individual fibers have started pulling away from the bulk as the matrix is being
digested. The cerium concentration decreases away from the interface where there is less matrix
to digest.
Figure 5-4 also shows the selectivity for benzoxazine/epoxy matrix degradation over fiber
damage. This cross section of a partially digested composite panel shows in the bottom half
individual carbon fibers that separated from the surrounding matrix as it depolymerized and
dissolved. The top half of the image shows a partial ply of carbon fibers that remains bound in the
matrix. Because the matrix is more reactive than the carbon fibers, it reacts preferentially. In
contrast, recycling by pyrolysis attacks both the matrix and the fibers.
Figure 5-5: Energy dispersive X-ray spectroscopy (EDS) on a partially digested fiber. Left: Micrograph. Each hatch
is 50 μm. Right: corresponding [Ce] as a function of depth.
350
300
250
200
150
100
50
0
0 20 40 60 80 100
[Ce] (K-ratio)
Distance (µm)
93
Figure 5-5 shows DLS data for the particle size after 24 and 72 hours of degradation. This data
can be used to estimate how far the polymer matrix has been depolymerized under the reaction
conditions. As seen in Figure 5A, there is a large order-of-magnitude difference within the sample,
ranging from 10’s nanometers to microns. This data indicates that while the matrix has been
homogenized within the solution, there are still oligomers remaining and depolymization is
incomplete. However, after 3 days of digestion, the particle size is much more homogenous, with
all the particle sizes measured at 10.9 nm ± 0.1 nm. This yields quantitative proof of the matrix
depolymerization. After more than three days of digestion, the particle sizes fall below the
detection capabilities of DLS.
Figure 5-6: (left) Particle size after 24 hours of depolymerization. Note that the x-axis is in log-scale due to the
large order-of-magnitude difference in size distributions within the sample. (Right). Particle sizes after 72 hours of
depolymerization
Figure 5-6 shows a proposed mechanism of polymer cleavage that is based on literature [26-30].
We posit that the depolymerization occurs via hydride abstraction from captodative methylene
groups in the polymer’s di-N-benzylaniline linkages, whereby the cerium oxidizes the complex to
a ruthenium oxo species represented diagrammatically as a Ru=O structure. This species pulls a
10 100 1000 10000
0
10
20
30
40
50
60
70
80
90
100
Intensity (--)
Size (nm)
10.5 10.6 10.7 10.8 10.9 11.0 11.1 11.2 11.3 11.4 11.5
0
10
20
30
40
50
60
70
80
90
100
Intensity (--)
Size (nm)
94
hydride from the polymer linkage to give a reduced metal species and an iminium cation. The
ruthenium is reoxidized by cerium. Polymer cleavage occurs by hydrolysis of the intermediate
iminium ion.
Figure 5-7: A potential mechanism for the depolymerization of a benzoxazine/epoxy resin via hydride abstraction
and its predicted oxidative polymer degradation products
If the mechanism proposed in Figure 5-7 is correct, then a product of the depolymerization
reaction should be a bisphenol-F tri- or tetra-carboxylate, with each of the carboxylate groups
deriving from a methylene group that previously served as a polymeric linkage; this is sketched as
3. The degradation should also release aniline (or nitrobenzene) from the cleavage of the linking
aniline groups.
We observe aniline, nitrobenzene, and tetracarboxylate 3 at early stages in our depolymerization
process. We ran our degradation to the point of homogenization of the polymer, then stopped it
and decanted off the resulting solution. The solvents were distilled from this solution for potential
re-use, and aniline and nitrobenzene were found in this volatile fraction. The remaining non-
volatile syrup was stirred in acidic methanol, and organics were extracted.
1
H NMR data on the
95
recovered organics are shown in Figure 4-8 We find 4, the tetramethyl ester of 3 and a collection
of peaks appropriate for degradation products of epoxide 2, each marked according to their
assignments. 3 is not stable to our degradation conditions and cannot be isolated when the solutions
exposed to the conditions for a prolonged period. This is expected, due to the high oxidation
potential of cerium(IV). Because conditions are known for decarboxylation of salicylates such as
3 and the conversion of the decarboxylated bis(phenol) back to 1, this discovery shows that our
technology will be a viable approach to recovery of both the reinforcing fibers and the matrix itself.
[32,33]
Figure 5-8: NMR data of the extract showing tetraester 4 and degradation products of epoxide 2. Refer to Figure 5-7
for structures
96
5.5: Conclusion
In sum, this work reports three key findings. Composite material recycling can be achieved under
conditions of homogeneous redox catalysis. This result alone is significant, as it demonstrates (1)
the intercalation of the homogeneous reaction milieu into the composite material construct in a
layered fashion that results in complete homogenization of the polymer matrix in a short time, and
(2) the conditions required of the catalysis are strikingly mild, compared to those of current
techniques for recycling composites. Lastly, (3) despite the incorporation of a high-potential
oxidant in this first proof-of concept demonstration, the recovered carbon fibers emerge
undamaged and ready for reuse. We further show that even the solvents and resin monomers can
be recovered in principle, and ultimately these can be reused, which is a unique feature that is
brought to the field by the use of catalysis under mild conditions.
These results demonstrate that oxidative recycling of benzoxazine-based composite materials is
potentially a viable strategy, although the chemistry of benzoxazine/epoxy blends (such as the one
studied here) is different from that of other composites, such as amine-linked epoxies, which
require different catalytic conditions for efficient degradation in our hands. A key drawback of
this demonstration technology is the large quantity of cerium required to drive this reaction to
completion. We chose cerium as a demonstration reagent, because it is a surrogate for electrolysis.
Once made electrolytic, this technology defines the basis for a process potentially suitable for
recycling industrial-scale quantities. In total, the results provide a proof of concept, demonstrating
that the recycling of some epoxy formulations is possible under mild conditions through a
homogeneous catalysis approach.
97
5.6: References
[1] Gardiner G. Recycled carbon fiber update: Closing the CFRP lifecycle loop. Compos World
2014.
[2] Barnes F. Recycled carbon fiber: Its time has come. Compos World 2016.
[3] Sims G, Bishop G. UK Polymer composites sector: Foresight study and competitive
analysis. NPL NetComposites 2001.
[4] Directive EU. Directive 2000/53/EC of the European Parliament and of the Council of Sep.
18, 2000 on End-of Life Vehicles. Off J Eur Communities, Artic 2000;7.
[5] Council EU. Council Decision 2003/33/EC of 19 December 2002 establishing criteria and
procedures for the acceptance of waste at landfills persuant to Article 16 of and Annex II to
Directive 1999/31/EC. Off J Eur Communities 2003;16:L11.
[6] Campbell Jr FC. Manufacturing technology for aerospace structural materials. Elsevier;
2011.
[7] Mallick PK. Fiber-Reinforced Composites: Materials, Manufacturing, and Design, Third
Edition. 3rd ed. Boca Raton: CRC Press; 2007.
[8] Ashby M, Shercliff H, Cebon D. Materials: Engineering, Science, Processing and Design.
2nd ed. Oxford: Elsevier; 2010.
[9] Pimenta S, Pinho ST. Recycling carbon fibre reinforced polymers for structural
applications: Technology review and market outlook. Waste Manag 2011;31:378–92.
[10] Asmatulu E, Twomey J, Overcash M. Recycling of fiber-reinforced composites and direct
structural composite recycling concept. J Compos Mater 2014;48:593–608.
[11] Oliveux G, Dandy LO, Leeke GA. Current status of recycling of fibre reinforced polymers:
Review of technologies, reuse and resulting properties. Prog Mater Sci 2015;72:61–99.
[12] Nahil MA, Williams PT. Recycling of carbon fibre reinforced polymeric waste for the
production of activated carbon fibres. J Anal Appl Pyrolysis 2011;91:67–75.
[13] Jiang G, Pickering SJ, Lester EH, Turner TA, Wong KH, Warrior NA. Characterisation of
carbon fibres recycled from carbon fibre/epoxy resin composites using supercritical n-
propanol. Compos Sci Technol 2009;69:192–8.
[14] Morin C, Loppinet-Serani A, Cansell F, Aymonier C. Near- and supercritical solvolysis of
carbon fibre reinforced polymers (CFRPs) for recycling carbon fibers as a valuable
resource: State of the art. J Supercrit Fluids 2012;66:232–40.
98
[15] Prinçaud M, Aymonier C, Loppinet-Serani A, Perry N, Sonnemann G. Environmental
Feasibility of the Recycling of Carbon Fibers from CFRPs by Solvolysis Using Supercritical
Water. ACS Sustain Chem Eng 2014;2:1498–502.
[16] Shibata K. FRP recycling technology by dissolving resins under ordinary pressure. JEC
Compos Mag 2011;48:50–2.
[17] La Rosa AD, Banatao DR, Pastine SJ, Latteri A, Cicala G. Recycling treatment of carbon
fibre/epoxy composites: Materials recovery and characterization and environmental impacts
through life cycle assessment. Compos Part B Eng 2016;104:17–25.
[18] Dang W, Kubouchi M, Yamamoto S, Sembokuya H, Tsuda K. An approach to chemical
recycling of epoxy resin cured with amine using nitric acid. Polymer (Guildf)
2002;43:2953–8.
[19] Dang W, Kubouchi M, Sembokuya H, Tsuda K. Chemical recycling of glass fiber
reinforced epoxy resin cured with amine using nitric acid. Polymer (Guildf) 2005;46:1905–
12.
[20] Pastine S. Can epoxy composites be made 100% recyclable? Reinf Plast 2012;56:26–8.
[21] Dogan Demir K, Kiskan B, Yagci Y . Thermally Curable Acetylene-Containing Main-Chain
Benzoxazine Polymers via Sonogashira Coupling Reaction. Macromolecules
2011;44:1801–7.
[22] Kim H-D, Ishida H. Study on the chemical stability of benzoxazine-based phenolic resins
in carboxylic acids. J Appl Polym Sci 2001;79:1207–19.
[23] Ghosh NN, Kiskan B, Yagci Y . Polybenzoxazines—New high performance thermosetting
resins: Synthesis and properties. Prog Polym Sci 2007;32:1344–91.
[24] Ning X, Ishida H. Phenolic materials via ring-opening polymerization of benzoxazines:
Effect of molecular structure on mechanical and dynamic mechanical properties. J Polym
Sci Part B Polym Phys 1994;32:921–7.
[25] Ishida H, Allen DJ. Mechanical characterization of copolymers based on benzoxazine and
epoxy. Polymer (Guildf) 1996;37:4487–95.
[26] Baslé O, Li C-J. Copper-Catalyzed Oxidative sp3 C−H Bond Arylation with Aryl Boronic
Acids. Org Lett 2008;10:3661–3.
[27] Guo S, Qian B, Xie Y, Xia C, Huang H. Copper-Catalyzed Oxidative Amination of
Benzoxazoles via C− H and C− N Bond Activation: A New Strategy for Using Tertiary
Amines as Nitrogen Group Sources. Org Lett 2010;13:522–5.
[28] Murahashi S, Komiya N, Terai H, Nakae T. Aerobic Ruthenium-Catalyzed Oxidative
99
Cyanation of Tertiary Amines with Sodium Cyanide Aerobic Ruthenium-Catalyzed
Oxidative Cyanation of Tertiary Amines with. Biochemistry 2003:19–21.
[29] Murahashi S-I, Zhang D. Ruthenium catalyzed biomimetic oxidation in organic synthesis
inspired by cytochrome P-450. Chem Soc Rev 2008;37:1490–501.
[30] Ghosh SC, Ngiam JSY , Seayad AM, Tuan DT, Chai CLL, Chen A. Copper-catalyzed
oxidative amidation of aldehydes with amine salts: synthesis of primary, secondary, and
tertiary amides. J Org Chem 2012;77:8007–15.
[31] Pickering SJ. Recycling technologies for thermoset composite materials—current status.
Compos Part A Appl Sci Manuf 2006;37:1206–15.
[32] Kaeding WW. Oxidation of Aromatic Acids. IV . Decarboxylation of Salicylic Acids. J Org
Chem 1964;29:2556–9.
[33] Fu Z, Liu H, Cai H, Liu X, Ying G, Xu K, et al. Synthesis, thermal polymerization, and
properties of benzoxazine resins containing fluorenyl moiety. Polym Eng Sci
2012;52:2473–81.
100
Chapter 6: Conclusions and Future Work
The work presented here in this thesis lays the foundation for future work in volatile
characterization in composite resins, as well as for recycling CFRP parts. The previous chapters
detail methods for characterizing cure-induced volatile release, optimizing the chemistry of the
resin, and recycling the final manufactured product. In this chapter, we propose potential future
projects for modelling cure-induced volatile release using thermodynamic principles, as well as
recycling carbon fiber composites via electrolysis.
6.1: Modelling cure-induced volatile release
Chapters 2 and 3 describe a quantitative approach for characterizing volatile release as a
function of cure parameters. However, pinpointing the mechanics of volatile release and
modifying the cure cycle in order to eliminate porosity can be challenging, especially considering
the large number of permutations possible (e.g. ramp rates, dwell times, cure pressures, etc.).
Therefore, it would be advantageous to model cure behavior and void nucleation beforehand to
determine whether a particular cure cycle for a given formulation is potentially susceptible to
generating porosity in the final part.
In theory, this model would predict final part porosity by being given the constitutive volatile
components released upon cure along with a proposed cure cycle. It would do so by first
computing the solubility of the gasses within the resin and determining whether or not they will
nucleate at any point during the cure cycle. The cure model will then compute the gelation time
from the cure cycle. By comparing the data from the bubble nucleation model with the cure model,
we could potentially determine if bubble nucleation was still ongoing during gelation, which would
ultimately lead to porosity.
101
One such model that has potential is a model proposed by Blander and Katz [1] that describes
bubble nucleation and growth in liquids. While more complicated bubble nucleation theories exist
that take into account the non-idealities which arise in dealing with a polymer solution [2], the
model parameters involved are often difficult to quantify experimentally. Additionally, specific
bubble nucleation models exist which deal with void nucleation in the context of composites
processing, but the key volatile that is modeled is water vapor, rather than solvents and reaction
byproducts [3]. Therefore, Blander and Katz’s model, to a first approximation, appears to be the
most reasonable for our application. While a complete overview of their work is outside the scope
of this proposal, the critical equations and parameters necessary for our applications of the model,
along with the proposed methods in which these parameters will be obtained, are described below.
First, the key results of Blander and Katz’s nucleation model are given in Equations 6-1
through 6-3.
𝐽 ≈ 3.73 ∗ 10
35
(
𝑑 2
𝜎 (𝑇 )
𝐵 𝑀 3
)
1
2
𝑒𝑥𝑝 (
−1.182 ∗ 10
5
𝜎 3
𝑇 [𝑃 𝐸 (𝑇 ) − 𝑃 𝐿 ]
2
𝛿 2
) 6-1
𝛿 = 1 − (
𝑑 𝑔 𝑑 ) +
1
2
(
𝑑 𝑔 𝑑 )
1
2
6-2
𝐵 = 1 −
1
3
(1 −
𝑃 𝐿 𝑃 𝐸 (𝑇 )
) 6-3
In these equations, J is the nucleation rate of bubbles per unit time, per unit volume, d is the
density of the liquid, σ(T) is the surface tension of the liquid as a function of temperature, PE(T) is
the equilibrium vapor pressure of the volatile as a function of temperature, PL is the hydrostatic
102
pressure of the liquid surrounding the bubble, T is the temperature of the system, dg is the density
of the gas, and M is the molar mass of the dissolved gas.
Our FTIR characterization experiments have already determined the identities of the dissolved
gasses being produced, therefore determining M, PE(T) and dg are trivial. T and d are also already
known variables by nature of performing these experiments – therefore, the only dataset that is
required for the implementation of this model is σ(T). In literature, this relationship of surface
tension as a function of temperature is usually measured by using a dynamic contact angle analyzer
setup within an oven, and temperature is varied over the range of interest. Additionally, this
procedure has been done before in an RTM processing context, where the effect of surface
modifications on resin impregnation and wetting was studied [4]. While our application will be
different, the technique will still be applicable. In conclusion, the development of a bubble
nucleation model for our resin could be both easily obtainable while still being very helpful.
While this model may need to be refined in the future to account for parameters that are
currently missing, such as the dependence of solvent solubility within a reacting thermoset, to a
first approximation it will help us understand the physiochemical processes in this complex resin;
ultimately, this model, combined with a cure model, could help contribute to the development of
manufacturing guidelines necessary for the manufacture of porosity-free composite parts.
6.2: Catalytic Recycling of Composites via Electrolysis
In Chapter 5, we demonstrated that it was possible to degrade a composite part via catalytic
oxidation, using cerium as our oxidant. However, the amount of cerium required to finish this
reaction is unsustainable. Nevertheless, the use of cerium indicates that the depolymerization
reaction is possible using electrolysis. Once the reaction is possible via an electrolytic pathway,
catalytic oxidation will be a cheap and sustainable pathway to recycle composite parts. In a
103
production environment, the carbon fiber part would be broken in half and inserted into a reaction
vessel containing the requisite catalyst. Once electrodes are clipped to both sides of the composite
part, the depolymerization reaction can begin. Because carbon by itself is an excellent cathode,
the reaction can proceed without any additional pre-processing steps.
104
6.3: References
[1] Blander M, Katz JL. Bubble nucleation in liquids. AIChE J 1975;21:833–48.
[2] Kim KY, Kang SL, Kwak H-Y. Bubble nucleation and growth in polymer solutions.
Polym Eng Sci 2004;44:1890–9.
[3] Kardos JL, Duduković MP, Dave R. Void growth and resin transport during processing of
thermosetting—matrix composites. Epoxy resins Compos. IV, Springer; 1986, p. 101–23.
[4] Lee G-W, Lee N-J, Jang J, Lee K-J, Nam J-D. Effects of surface modification on the
resin-transfer moulding (RTM) of glass-fibre/unsaturated-polyester composites. Compos
Sci Technol 2002;62:9–16.
[5] Leedy, D. W.; Adams, R. N. J. Am. Chem. Soc. 1970, 92 (6), 1646–1650.
Abstract (if available)
Abstract
Carbon fiber reinforced polymers (CFRPs) are structural materials used in the aerospace, automotive, and sporting goods industries due to their high specific properties, which typically outperform traditional metallic counterparts. CFRP composites are a combination of high-strength fibers bonded together by a polymeric matrix. While there are a wide variety of methods for fabricating CFRP composites, such as resin transfer molding, pultrusion, or automated fiber placement, controlling processing parameters is critical, regardless of the manufacturing method employed. Poor control of the processing parameters for instance temperature, pressure, or material handling (i.e. out-time, storage conditions, or resin pre-treatment procedures), can cause defects, ultimately reducing the material properties of the final manufactured part. The focus of this dissertation is to consider chemistry and engineering approaches for optimizing both the manufacture and recycling of CFRP composites. ❧ The first chapter of this thesis considers a method to characterize volatile release in a resin transfer molded composite part. By understanding the dynamics and kinetics of volatile release in situ, we can prescribe methods for controlling cure-induced byproducts and eliminate volatile-induced porosity. In the second chapter, we expand this method and characterize a variety of resins for use in composites manufacturing to further understand the correlation between volatile release and porosity formation, as well as to understand the utility of a reaction cell/FTIR technique. In the third chapter, we consider optimizing the chemistry of the resin, as opposed to the cure cycle, and characterize the differences in final part properties that arise as a consequence of each optimization system. In the last chapter, we propose a method for recycling benzoxazine/epoxy parts via catalytic oxidation, which is ultimately important for opening up emerging markets for composite applications in both Europe and Asia.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Chemical recycling of amine/epoxy composites at atmospheric pressure
PDF
Fabrication and analysis of prepregs with discontinuous resin patterning for robust out-of-autoclave manufacturing
PDF
Sustainable manufacturing of out-of-autoclave (OoA) prepreg: processing challenges
PDF
In situ process monitoring for modeling and defect mitigation in composites processing
PDF
Developing efficient methods for the manufacture and analysis of composites structures
PDF
In situ process analysis for defect control during composites manufacturing
PDF
Polymer flow for manufacturing fiber reinforced polymer composites
PDF
Material and process development and optimization for efficient manufacturing of polymer composites
PDF
Defect control in vacuum bag only processing of composite prepregs
PDF
Development of composite oriented strand board and structures
PDF
Processing and properties of phenylethynyl-terminated PMDA-type asymmetric polyimide and composites
PDF
Chemical depolymerization of amine-epoxy cured carbon fiber reinforced polymer composites and their re-use
PDF
Comprehensive uncertainty quantification in composites manufacturing processes
PDF
Ageing and mechanical failure of fiber reinforced polymers
PDF
Efficient manufacturing and repair of out-of-autoclave prepreg composites
PDF
Fabrication, deposition, and characterization of size-selected metal nanoclusters with a magnetron sputtering gas aggregation source
PDF
Void evolution in vacuum bag-only prepregs
PDF
Slurry based stereolithography: a solid freeform fabrication method of ceramics and composites
PDF
Impact of exposure to brine/CO₂ on the mechanical and transport properties of the Mt. Simon sandstone
PDF
Synthesis of high-quality nanoparticles using microfluidic platforms
Asset Metadata
Creator
Lo, Jonathan Nienpei
(author)
Core Title
Characterization, process analysis, and recycling of a benzoxazine-epoxy resin for structural composites
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Chemical Engineering
Publication Date
03/23/2018
Defense Date
09/07/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
benzoxazines,carbon fiber composites,evolved gas analysis,mechanical properties,OAI-PMH Harvest,Recycling,resin transfer molding,thermoset resins
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Nutt, Steven Ray (
committee chair
), Malmstadt, Noah (
committee member
), Williams, Travis Jesse (
committee member
)
Creator Email
jonathan_lo@huntsman.com,jonathnl@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-433093
Unique identifier
UC11264512
Identifier
etd-LoJonathan-5760.pdf (filename),usctheses-c40-433093 (legacy record id)
Legacy Identifier
etd-LoJonathan-5760.pdf
Dmrecord
433093
Document Type
Dissertation
Rights
Lo, Jonathan Nienpei
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
benzoxazines
carbon fiber composites
evolved gas analysis
mechanical properties
resin transfer molding
thermoset resins