University of Southern California Dissertations and Theses

Developing efficient methods for the manufacture and analysis of composites structures

(USC Thesis Other)

Developing efficient methods for the manufacture and analysis of composites structures

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

DEVELOPING EFFICIENT METHODS FOR THE MANUFACTURE AND
ANALYSIS OF COMPOSITES STRUCTURES

by
Lee Hamill

A 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
(MATERIALS SCIENCE)

December 2017
i

DEDICATION

With all my heart, I dedicate this manuscript to my partner Steph Mirsky.

ii

ACKNOWLEDGEMENTS

There are many people who have impacted my work over the last five years, and it gives
me great pleasure to acknowledge them. Without the support and guidance of my family, friends,
and colleagues, this work would not stand where it is today. For that I am eternally grateful.
I would first like to acknowledge my advisor, Professor Steven Nutt. Five years ago, he
recognized an enthusiasm in me for an unattended project in his group and decided to take me on
as a graduate student. My projects have covered a wide variety of topics over the years, and each
one has been fascinating and peaked my interest in a different way. I am grateful to have had an
advisor who works so hard to connect students with opportunities that fall in line with their
interests. Throughout my time here, Professor Nutt always pushed me to develop a good story
about my work and to take steps to improve my public speaking skills. I leave here with improved
scientific reasoning and confidence in speaking, and this is entirely thanks to Professor Nutt.
There are several colleagues at USC that I would like to thank. Professor Timotei Centea
has provided considerable guidance that helped me lay the groundwork for several projects.
Yunpeng Zhang is capable of fixing anything and was a calm presence in many moments of panic
when equipment failed or experiments were not going as planned. I would like to acknowledge
Professor Jane Emerson, who collaborated with me on my last project and was an invaluable source
of physics knowledge. Finally, I would like to thank my undergraduate research assistants: Karissa
Hendrie, Ellen Emerson, and Christina Cabble. Without their help, I would still be running
experiments.
I had the great fortune of collaborating with many scientists outside of USC, each of whom
was incredibly kind and inspiring. I worked with Dr. Douglas Hofmann of the Jet Propulsion
iii

Laboratory on several projects, and he was a constant source of motivation and enthusiasm for the
work. I have enjoyed working with Doug Decker of Northrop Grumman Corporation, who came
to me with an industry challenge which inspired my first project. He also was a part of my last
project, in which we collaborated with Kevin McGushion of Exel Orbital Systems Inc. Finally, I
am grateful for the time I spent working with Stephanie O’Keefe of Liquidmetal Technologies.
None of this would have been possible without the funding and donations that supported
my work. Funding sources included the M. C. Gill Composites Center, the G8 Research Councils
Multilateral Initiative on Materials Efficiency and the “Sustainable Manufacturing Through Out-
of-Autoclave Prepregs project (NSF CMMI-1229011), the NASA SBIR Z2.01, and Northrop
Grumman Corporation. Materials were donated by Solvay Inc., Airtech International, and
Liquidmetal Technologies. Access to scientific equipment was provided by Zygo Corporation and
Ametek Inc.
Though they are affiliated with USC, there are two colleagues I would like to acknowledge
specifically for their friendship. Dr. Kamia Smith and I have gone through everything in the
program together, from spending countless hours studying for our screening exam to defending
within a month of each other. Thank you for always being willing to drop everything to be there
for me in times of need. I am eternally grateful to have just barely crossed paths with Professor
Lessa Grunenfelder at the end of her PhD and the beginning of mine. Lessa, thank you for being
the best friend I could ask for, for pushing me to follow my passion, and for always seeing the best
in me. I can’t thank you enough for your friendship.
Finally, I would like to extend my deepest gratitude to my family. I thank my parents, Pat
and Debbie Hamill, for always supporting me and for raising me in a house free from gender
stereotypes. I am the confident person I am today entirely because of you two. My brother, Grant
iv

Hamill, and my sisters, Lucy and Katie Hamill, have been constant sources of light in my life. I
am thankful to Grant for his unwavering, unconditional love, to Lucy for setting an incredible
example of personal strength and altruism, and to Katie for her warm hugs and kind spirit. I would
like to acknowledge my four grandparents: Bonnie and Denny Hamill, Mary and Tag Hobert.
Thank you to my grandfather, Denny, for paving the PhD pathway for me and for always being a
phone call away for scientific guidance. Thank you to my grandmother, Bonnie, for paving the
education pathway, a path on which I have just begun to embark. Thank you to my grandparents
Mary and Tag for the much needed relaxation each summer at your lake house. Last, but certainly
not least, I would like to thank my partner, Steph Mirsky. Thank you for always making me laugh
and for always being “on it!” when I was overwhelmed. You made this all possible, and I am
forever grateful for your love and support.

v

TABLE OF CONTENTS

Dedication ….………….……….…………………………………………………………... i
Acknowledgements ………………………………………………………….……………... ii
List of Tables ……………………….………………………………………………............ viii
List of Figures ……...………………………………………………………………………. ix
Abstract …………………….………………………………………………………………. xiii
Introduction ………………………………………………………….……………………... 1
CHAPTER 1 – Elimination of Surface Porosity in Out-of-Autoclave Composite
Manufacturing …………………………………………………………….………………... 6
1. Introduction ……………………………………………………………………... 7
1.1 Surface Porosity ……………………………….…………………………. 7
1.2 Objectives ……………………...…………………….…………………... 10
2. Materials and Methods …………………………………………………………. 10
2.1 Test Matrix ………………………………………….……………………. 10
2.2 Materials …………………………………………………………............. 12
2.3 Laminate Manufacturing …………………………………………............ 13
2.4 Surface Porosity Measurement ................................................................... 14
2.5 Visual Analysis …………………...……………………….……………... 15
3. Results and Discussion …………………………………………………………. 16
3.1 Source of Surface Porosity ……………….………………………………. 16
3.2 Material and Process Effects ……………….……………………………. 20
3.2.1 Room Temperature Vacuum Hold ……...………………………... 21
3.2.2 Out-Time …………………...……….……………………………. 21
3.2.3 Freezer Storage Time ….…………….……….….….……............. 24
3.2.4 Processing Effects …….……………….……….…….…………... 24
3.2.5 Statistical Significance ….……….….….……….………………... 25
vi

3.3 Mitigation Strategies ……………………………………………………... 26
4. Conclusions ……………………………………………………………………... 29
CHAPTER 2 – Bulk Metallic Glass and Composite Structures for Spacecraft Shielding … 31
1. Introduction ……………………………………………………………………... 32
2. Materials and Methods …………………………………………………………. 35
2.1 Materials ……….….………….……….…………………………………. 35
2.2 Hypervelocity Impact Tests ….….…...…….……….……………………. 35
3. Results and Discussion …………………………………………………………. 37
3.1 Penetration Depth ………………………………………………………... 37
3.2 Ballistic Limit Equation Development ………………....………............... 38
3.3 Detached Spall Behavior …...…….……….……………………………... 40
3.4 Shielding Geometries …………………………………….………............. 45
4. Conclusions ……………………………………………………………………... 46
CHAPTER 3 – Adhesion of Metallic Glass and Epoxy in Composite-Metal Bonding ….... 48
1. Introduction ….…………………………………………………………………... 49
2. Materials and Methods …………….……………………………………………. 50
2.1 Materials and Curing Methods …………………………………………... 50
2.2 Metal Surface Treatment …………………………………………............. 51
2.3 Characterization Methods ………………………………………………... 52
2.4 Lap Shear Preparation and Testing …….………………………………... 53
3. Results and Discussion ……………………….…………………………………. 55
3.1 Metal Surface Characterization ………….………………………………. 55
3.2 Lap Shear Results …….…………………………………………………... 58
3.3 Failure Classification ………………………………………….…………. 60
4. Conclusions ….………….………………………………….……………............. 62
CHAPTER 4 – Galvanic Corrosion Resistant Fiber Metal Laminates of Metallic Glass
and Carbon Composites ……………………………………...…………...………………... 65
1. Introduction ……………………………………………………………………... 66
vii

2. Materials and Methods …………………………………………………………. 68
2.1 Materials and FML Fabrication …………………………………............. 68
2.2 Corrosion Testing ………………………………………………………... 69
2.3 Mechanical Testing ………………………………………………............. 70
3. Results and Discussion …………………………………………………………. 71
3.1 Corrosion Behavior ………………………………………………............ 71
3.2 Mechanical Behavior ………….…………………………………………. 76
4. Conclusions ……………………………………………………………………... 81
CHAPTER 5 – Low Frequency Eddy Current Testing of Insulators and Composites ……. 83
1. Introduction ……………………….………………………….…………………. 84
1.1 Resonance-Tuned ECT …………………………………………………... 85
2. Materials and Methods …………………………………………………………... 86
2.1 Sensor Configuration and Testing Method ………………………………. 86
2.2 Materials …………………………………………………………………. 87
3. Results and Discussion …………………………………………………………... 88
3.1 Representative Scan ………………………………………………............ 88
3.2 Effect of Operating Parameters on Detectability ….……………………... 89
3.2.1 Transmit Coil Frequency …………………………………............ 89
3.2.2 Lift-Off …...………………………………………………............. 90
3.3 Effect of Permeability on Detectability …………………………………... 92
3.4 Applicability to CFRP ……………………………………………............. 92
4. Conclusions ……….……………………………………………………………... 94
Concluding Remarks ………………………………………………………………………. 96
Recommendations for Future Work ……………….………………………………………. 98
References ……………………………………………………….…………………………. 99

viii

LIST OF TABLES

Table 1-1 Parameters, ranges, prepreg choice and number of repeat tests included in
parametric study.

10
Table 1-2 Prepreg and tool plate material properties.

12
Table 2-1 Overview of the 9 hypervelocity tests. NP response stands for “no
penetration”, while DS denotes “detached spall”.

36
Table 3-1 Water contact angle results

57
Table 3-2 Surface roughness of BMG and Al samples after treatment

57
Table 3-3 Overview of failure modes observed on the failure surface of lap shear
samples. The letter “A” indicates that an adhesive film was used in the lap
shear sample.

61
Table 4-1 Equilibrium current densities for various metal/fiber pairs in a saltwater bath
(3.5% by weight NaCl).

72
Table 4-2 Steady-state electrode potentials of various materials in free-flowing
seawater. Potentials given in reference to a saturated calomel electrode
(SCE).

75
Table 4-3 Material properties for BMG-based and Al-based FML. 76

ix

LIST OF FIGURES

Figure 1-1 Vacuum bag assembly for laminate manufacture.

13
Figure 1-2 Sequence of actions used to quantify surface porosity. (a) Raw image taken
at 20  magnification is uploaded into image analysis software; (b) areas
of surface porosity are manually outlined; and (c) image is converted into
binary image where black pixels represent areas of surface porosity. The
black pixels in this image represent 1.43% of the total pixels in the image,
translating to 1.43% of this area of laminate surface being surface porosity.

14
Figure 1-3 Typical shape and distribution of surface porosity observed on laminates
made from (a) woven and (b) unidirectional prepreg. Note that no surface
porosity is present on laminates made from unidirectional prepreg,
whereas surface porosity is widely present on laminates made from woven
prepreg.

16
Figure 1-4 Surface topography of the (a) woven and (b) unidirectional prepreg used
in this study. Roughness lines given in (c) and (d) correspond to the white
lines in (a) and (b) respectively.

17
Figure 1-5 Time-lapse images of a glass plate-prepreg interface under vacuum at
room temperature. Images were taken (a) immediately after application of
vacuum as well as (b) 10 minutes, (c) 30 minutes, and (d) 1 hour in to the
vacuum hold. Select air bubbles are circled to highlight the progression of
shape, size and location of trapped air.

18
Figure 1-6 SEM micrographs of the woven prepreg surface. An open slit in the resin
is shown at (a) 22x and (b) 75x. (c) SEM micrograph taken at 25x of a
cross section of two stacked woven prepreg plies. Dry fiber bundles are
observed directly below the surface resin.

19
Figure 1-7 (a) Tool-side surface of a laminate made from woven prepreg plies
conditioned in a humidity chamber prior to layup and cure alongside a (b)
magnified view, which highlights the difference between porosity due to
trapped air and surface defects resulting from moisture in the prepreg.

20
Figure 1-8 The effect of (a) room temperature vacuum hold time, (b) out time at room
temperature, and (c) freezer time on surface porosity. (d) The effect of
room temperature aging method on surface porosity. “Control” laminates
22
x

were made from prepreg with no out time; “On Tool” laminates were made
from prepreg that was laid up on the tool at zero days of out time and
allowed to age for five days prior to cure; “Off Tool” laminates were made
from prepreg that was aged for five days prior to being laid up and were
cured immediately upon layup.

Figure 1-9 The effect of additional processing parameters on surface porosity was
considered. These included performing an intermittent debulk on the first
ply down, spiking the first ply down, and reducing the vacuum level.

25
Figure 2-1 Hypervelocity test set up including (a) the sample fixture, (b) the BMG
Vitreloy 1 plate with sectioning lines shown, (c) a representative long
exposure of a hypervelocity impact into a Vitreloy 1 plate, (d) the
configuration of the two high-speed cameras and the test chamber at the
NASA Ames Vertical Gun Range (AVGR), and (e) the long exposure
camera filming down on the test specimen from a port in the chamber.

37
Figure 2-2 Plot of penetration depth versus particle velocity for an aluminum
projectile impacting sheets of the BMG Vitreloy 1. The black data is from
Yang et al., and the grey data is from the current work. Dotted lines
represent the BLE of Ti 15-3-3 for each projectile diameter. Estimated
BLEs for Vitreloy 1 are represented by the solid lines.

38
Figure 2-3 Penetration depth versus particle velocity for a 3.17 mm Al projectile
impacting a single-wall BMG shield. Material properties of the metallic
glass were inserted into known equations for Ti, Al, and steel (shown as
dotted lines).

39
Figure 2-4 Hypervelocity impacts into single wall BMG sheets of Vitreloy 1 at (a)
0.82 km/s, (b) 1.25 km/s, (c) 1.56 km/s, (d) 2.34 km/s. The front and back
of the plates after impact are shown at the right after the Al projectiles
were removed.

41
Figure 2-5 Sheet thickness versus projectile velocity for the BMG Vitreloy 1 and the
BMG composite DH1. The closed data points represent no-penetration or
detached spall while the open data points represent penetration or detached
spall. The ballistic limit is the curve that divides the two regions. The
dotted curve represents the theoretical BLE for each material, and the solid
curve represents the actual BLE. The one square point is from an impact
into DH3.

42
Figure 2-6 Hypervelocity impacts into single-wall sheets of BMG composites. (a)
Long exposures of the impacts shown in (c–f), respectively. (b) Three
44
xi

plates of the BMG DH1 are shown before impact. (c) Impact into a DH1
plate at 0.91 km/s showing a catch. The front of the plate is shown at the
right after impact. (b) Impact into DH1 plate at 1.25 km/s showing
detached spall from the back of the sample. (c) Impact into a thin plate of
DH1 at 2.34 km/s, showing complete penetration. (f) Impact into a DH3
showing a catch.

Figure 2-7 (a) 2.79 km/s impact into a 4-wall Vitreloy 1 Whipple shield showing
penetration of only the first layer. (b) The sample holder after impact and
(c) a long exposure of the entire impact. (d) The hole generated in the back
of the first layer and (e) the debris embedded in the second layer.

46
Figure 3-1 Vacuum bag assembly for lap shear laminate manufacture. (a) The tool
plate includes a step to ensure that the composite lays flat as it cures. (b)
Complete vacuum bag assembly.

54
Figure 3-2 SEM micrographs of treated metal surfaces. BMG samples are shown on
the left with their Al counterparts on the right: (a) BMG control, (b) Al
control, (c) BMG abrasion, (d) Al abrasion, (e) BMG PAA, (f) Al PAA,
(g) BMG silane, and (h) Al silane.

56
Figure 3-3 Peak load values of lap shear samples tested in tension. Each column
height is the average of three identically produced samples with the
standard deviation shown through error bars. Results are grouped by
surface treatment method, with the letter “A” indicating that an adhesive
film was used. Within each group, metal type and processing method are
separated.

59
Figure 4-1 Galvanic corrosion current density exhibited by galvanic cells consisting
of (a) aluminum and carbon fiber and (b) BMG and carbon fiber in a
saltwater electrolyte solution.

72
Figure 4-2 Microscope images of (a) Al-based and (b) BMG-based FMLs prior to
submersion in saltwater bath. (c) Al-based and (d) BMG-based FMLs after
two weeks in saltwater bath.

74
Figure 4-3 Microscope images of BMG-based and Al-based FMLs after short beam
shear testing. Interlaminar shear of the composite layers (a) and (b), and
cracking of the metal shown in (c) and (d) for BMG-based and Al-based
FMLs, respectively.

76
Figure 4-4 Microscope images of FMLs after 3-point bend testing. Compression
failure shown in (a) and (b) for Al-based and BMG-based FMLs,
77
xii

respectively. (c) Multiple failure modes including compression and
interlaminar shear in the composite for a Al-based FML. (d) Failure by
tension in the outer metal layer for a BMG-based FML.

Figure 4-5 Load-displacement curves for BMG-based and Al-based FMLs during 3-
point bend tests.

78
Figure 4-6 Typical stress-strain curves for a BMG-based and Al-based FML that
exhibited metal-dominated failure.

80
Figure 5-1 Schematic of the sensor consisting of an outer (transmit) coil, radius 𝑟 1
,
and inner (receive) coil, radius 𝑟 2
. The coil axes are perpendicular to the
sample surface offset laterally by a distance 𝑑 . Lift-off, ℎ, is the distance
between the sensor face and the sample surface.

86
Figure 5-2 Receive coil output voltage as a function of traverse time. The dotted line
indicates sensor path with signal time points corresponding to traverse
distances to air/acetal and acetal/air interfaces.

88
Figure 5-3 (a) General relationship between output voltage of the receive coil and
frequency of the alternating current going through the transmit coil. (b)
Signal deflection amplitude as a function of transmit coil frequency. (c)
Amplitude of signal deflection as a function of frequency for an acetal/air
interface at a lift-off of 0.762 mm.

90
Figure 5-4 Signal deflection amplitude as a function of lift off for an air/acetal
interface.

91
Figure 5-5 Signal deflection amplitude as a function of magnetic permeability for a
ferrofluid/acetal interface.

92
Figure 5-6 Output voltage of the receive coil as a function of time, resulting from a
scan across a CFRP sample containing defects labeled 1 through 6. Scans
of the defect side and the control (no defects) side are shown for the (a)
small diameter and (b) large diameter sensor. A scaled outline of the part
and defects is placed above the scans in (b).

93

xiii

ABSTRACT

The increase in demand for commercial and space travel has motivated the development
of more efficient materials selection and manufacturing practices. More and more people are
engaging in air travel, and space exploration is constantly pushing the boundaries of current
capabilities. Both efforts require materials that maximize mechanical performance and minimize
weight. In addition, shorter processing times, cheaper manufacturing methods, and more versatile
evaluation techniques are desired. In the following chapters, five projects are described in which
such efforts are addressed.
In chapter 1, a particular kind of defect found in composites produced from out-of-
autoclave (OoA) prepregs, surface porosity, is investigated. Currently, the composites industry
employs several strategies to eliminate surface porosity including the use of a release film in
between the composite part and tool plate, using a resin-rich ply at the composite-tool interface,
or painting the cured composite part to fill in the voids. Each solution involves added time, cost,
and/or weight. With a goal of developing manufacturing methods that more efficiently eliminate
surface porosity, experimental efforts were made to identify the source of surface porosity and
possible mitigation methods. Results confirmed that surface porosity was primarily associated with
air that was trapped at the tool-prepreg interface during layup. The magnitude and distribution of
surface porosity was affected by multiple parameters, including vacuum hold time, freezer storage
and out-time, and material and process modifications that affect air evacuation. The results
indicated that prepreg out-time (and thus tack) and vacuum quality were the primary drivers of
surface porosity.
xiv

Chapters 2 – 4 explores new, high-performance materials for the application of
hypervelocity impact shielding for spacecraft. High-velocity impacts with debris are a major cause
of concern for spacecraft and satellites. Developing new materials that can protect against these
threats, while still providing low-density and sufficient toughness to survive launch loads, is
important for future spacecraft design. In chapter 2, hypervelocity impacts are used to estimate the
ballistic limit for bulk metallic glass (BMG) and their composites and to investigate spalling
behavior. The composites are shown to have excellent combinations of hardness and toughness for
use as shields. Chapters 3 and 4 expand on the potential for BMG to be used in high performance,
low density structures by pairing it with carbon fiber reinforced polymer (CFRP) composites in a
fiber metal laminate (FML) structure.
A first step in this effort is determining best practices to promote adhesion between BMG
and CFRP; investigation into this topic is discussed in chapter 3. Metal surface treatments for
aluminum which are known to promote adhesion were performed on both BMG and aluminum
samples, and surfaces were subsequently characterized via scanning electron microscopy (SEM)
and measurements of water contact angle (WCA) and roughness. Results of characterization
experiments indicated that the BMG alloy responded to surface treatments much like aluminum,
and that these treatment methods when applied to BMGs can promote adhesion. Lap shear tests
were also performed to evaluate the bond quality at composite-metal interfaces. The bond strength
at the BMG-composite interface was as strong or stronger than aluminum-composite interfaces,
with the exception of samples pre-treated by anodization.
The possibility of galvanic corrosion prohibits the pairing of carbon fiber and aluminum in
a fiber metal laminate (FML) structure. In chapter 4, a BMG-based FML is presented as a pathway
to using carbon fiber in FMLs without risk of galvanic corrosion. The galvanic coupling and
xv

mechanical behavior of an Al-based and a BMG-based CFRP FML were compared. Results
showed that when paired with CFRPs, BMG exhibits far less galvanic corrosion than aluminum
paired with CFRP. In fact, the corrosion between BMG and CFRP was similar in magnitude to the
corrosion between aluminum and glass fiber, the two constituent materials of GLARE, the most
widely used commercial FML. While interlaminar shear strength and flexural strength were similar
for both FML types, the bending stiffness, tensile strength and tensile modulus of BMG-based
FMLs were greater than those of Al-based FMLs.
Finally, in chapter 5, a new type of non-destructive testing (NDT) method is evaluated
which expands current capabilities. Eddy current testing (ECT), a NDT method widely used to
evaluate defects within conductive materials is explored in this study as it applies to insulators and
non-uniformly conductive materials. Previous work has shown that at high frequencies, differences
in electric permittivity can be detected with ECT. In this study, a new design of an ECT sensor
that employs two resonance-tuned coils is evaluated. Results show that material inconsistencies in
insulators are detectable due to spatial variations in permittivity and magnetic permeability, and
that detection is possible at lower frequencies than previously demonstrated. In addition to
determining signal dependence on individual electromagnetic parameters, sensitivity for defect
detection in a carbon fiber-reinforced polymer (CFRP) composite is qualitatively determined.
Although low signal-to-noise ratio is observed with a small-diameter coil, by increasing the coil
diameter, the signal-to-noise ratio is increased while preserving adequate spatial resolution to
detect defects in the sample. This study expands on previous studies of the application of ECT to
insulators, and demonstrates that defect detection is possible in CFRPs.

1

INTRODUCTION

In recent decades, the aerospace industry has seen a major shift in material selection for
aircraft structures. Traditional metals are being replaced by more advanced, high-performing
materials such as fiber-reinforced polymer (FRP) composites. This is directly observed in the
material make-up of the new Boeing 787 Dreamliner aircraft. By weight, 50% of the aircraft is
composed of composite materials, while aluminum only accounts for 20%. The Boeing 777, by
comparison, uses 12% composites and 50% aluminum. Boeing is not alone in this shift; the latest
Airbus aircraft, the A350-XWB, is composed 53% out of composite materials.
Motivating this shift is the improved mechanical properties (high strength, high stiffness,
fatigue resistance), corrosion resistance, and reduced weight that composite materials provide.
1

Weight is a critical factor for aircraft; the lighter an airplane is, the less fuel it requires to travel.
Reducing aircraft weight, and therefore required fuel, can lead to substantial cost savings. In the
aircraft design process, material selection is a critical step where engineers seek efficiency in
material selection by choosing low-density materials that meet the mechanical constraints of flying
conditions.
In addition to material selection, the composites industry is engaging in the development
of efficient processing methods. For example, in the last two decades the industry has seen a major
shift from autoclave processing toward out of autoclave practices.
2
Autoclaves are large,
pressurized ovens that provide the heat and pressure required to consistently produce defect-free
composite parts. The downsides of autoclave processing include large capital investments, high
operating costs, and long cycle times, which limit production rates and imposes a manufacturing
bottleneck. An alternative to autoclave processing is the vacuum bag only (VBO) method, in which
2

only an oven and a vacuum bag is required. VBO processing addresses all challenges presented by
autoclave processing. However, the absence of autoclave-level pressure on the parts during cure
increases the probability of defects, thereby decreasing process robustness.
2–5
Chapter 1 focuses
on one such defect: surface porosity. Surface porosity is a type of defect that arises on the tool-
side surface of a composite part. Through prepreg characterization and experimental analysis, the
primary source of surface porosity was determined to be air that is trapped at the tool-composite
interface during layup, and mitigation strategies were identified.
Chapters 2 – 4 describe research performed on bulk metallic glass (BMG) and the pairing
of BMG with carbon fiber reinforced polymer (CFRP) composites, motivated by challenges
experienced by the aerospace industry. Over the last 50 years, the number of space missions has
substantially increased, as more countries have become space capable and as smaller, cheaper
spacecraft have become more widely available. With this increase in space activity comes a
dramatic increase in space debris, which has been exponentially increasing for the last 15-20 years.
The debris takes the form of metal particles orbiting Earth, traveling at velocities ranging from 8
to 18 km/s.
6–9
This poses a serious and ever increasing threat to spacecraft, satellites, and manned
space stations, as impacts with debris can cause loss of functionality or loss of life.
9

Although debris of all sizes is a concern for spacecraft collision, it is small, undetectable
debris that poses the most serious threat to space missions. Micro-meteoroid and orbital debris
(MMOD) is generally classified as debris less than 100 mm in diameter, and is too small to be
detected by ground-based radar systems, which track larger objects.
7
Shielding is therefore an
integral component of spacecraft structural design to ensure the highest probability of survival
during an MMOD impact. Spacecraft shielding is applied in such a manner as to vaporize an
incoming projectile without penetration or detached spall occurring from the back wall of the
3

shield. Standards for testing and evaluating shield materials for the International Space Station
(ISS) define shield failure as a complete penetration, a through-hole, a through-crack, or detached
spall in the rear wall of the shield.
8
In chapter 2, BMGs are investigated as a possible candidate for
hypervelocity shielding. Ballistic limit equations (BLEs) are developed, which allow for
calculation of the critical shield thickness required to avoid penetration or detached spall.
Chapters 3 and 4 expand on chapter 2 by exploring a new high-strength, low-density
material for applications such as spacecraft shielding and aircraft: BMG-CFRP fiber metal
laminates (FMLs). The development of FMLs began at Delft University in the 1970s as part of an
effort to develop high strength, low density materials to replace the traditional choice of monolithic
aluminum for certain aircraft parts.
10
FMLs are laminate structures that consist of alternating layers
of metal and composite. The first rendition, ARALL, consisted of aluminum as the metal layer and
aramid fiber reinforced epoxy as the composite layer. ARALL was succeeded by GLARE, in
which the aramid fibers were replaced with glass fibers, a stronger, less costly material that was
more versatile.
11

In addition to high strength and low density, GLARE exhibited fatigue resistance superior
to monolithic aluminum; any cracks initiated in the aluminum layers are arrested upon reaching
the composite layer, greatly reducing the risk of fatigue failure. Fatigue resistance is the primary
design driver for airframes and skins, as the pressurization and depressurization cycles imposed
for each flight introduce a cyclic loading pattern, and stochastic cyclic loading arises from air
turbulence and vibrations. The well-known Aloha Air accident, in which the top of the fuselage
ruptured during flight, was attributed to high cycle fatigue, and led airplane companies to seek
materials with greater fatigue resistance.
10,12

4

Airbus incorporated GLARE into aircraft design – over half of the fuselage of the A380
consists of GLARE.
13
In addition to capturing desired properties such as high strength, high elastic
modulus, improved toughness, corrosion resistance, and fatigue resistance, the shift resulted in a
30% weight reduction.
10,12,14,15
Though GLARE is the only FML commercially available,
researchers have investigated the possibilities of using different composite and metal
combinations.
14,16–20
One particular combination that has proven unfeasible is carbon fiber
composite and aluminum because of the galvanic corrosion that results from contact between two
materials of different electric potential.
10,11,21–23
Research efforts have aimed to reduce the galvanic
corrosion of carbon fiber-aluminum FMLs by using barriers and coatings to isolate the two
materials.
11,21,23
However, this combination of materials in FMLs has yet to achieve commercial
viability.
BMG presents one alternative to aluminum that can be paired with CFRPs in a FML
structure and possibly avoid galvanic corrosion. Studies investigating this possibility are presented
in chapters 3 and 4. Chapter 3 focuses on metal surface treatment methods to promote adhesion.
Surface treatments commonly used on aluminum to promote adhesion were performed on BMG
samples, and the resulting surfaces were characterized and compared to treated aluminum surfaces.
Treated surfaces were characterized, and lap shear specimens were constructed and tested to
determine the resulting bond strength at the composite-metal interface. In a separate study
(presented in chapter 4), BMG-based and Al-based FMLs were fabricated, and the galvanic
corrosion and mechanical properties of each are compared.
The final section of this dissertation (chapter 5) responds to efficiency desires of the
composites industry regarding non-destructive testing (NDT), a set of analysis methods used to
evaluate the quality of structures without causing damage. It is a popular tool used in industry, as
5

it provides a way to ensure that manufactured parts do not have hidden defects. These methods
can be used for a wide range of applications from weld inspection to x-ray imaging of the inside
of a human body.
24,25
For FRP composites, ultrasonic testing (UT) is the most commonly used
method to detect defects within the part. However, UT has limitations which include (1) the
inability to detect certain types of defects within composites, and (2) the required use of a coupling
agent (typically water or gel). For cured composites, coupling agents are not necessarily an issue.
However, this type of testing would not be available for analysis of uncured prepreg layups. Errors
can occur during layup where foreign objects, such as release film or paper, accidentally get stuck
in between plies during layup. Currently, there is no way to inspect for the presence of these
objects. In chapter 5, a new type of NDT method is presented based on electromagnetic principles
and detectability of defects based on electric permittivity and magnetic permeability is
demonstrated. Results also show that detection of embedded defects in a CFRP panel is possible.

6

CHAPTER 1

Elimination of Surface Porosity in Out-of-Autoclave Composite
Manufacturing

7

1. Introduction
Traditionally, fiber reinforced polymer (FRP) composite materials for aerospace structures
have been fabricated using autoclave processing methods. Autoclave processing is robust, utilizing
above-ambient pressures during high temperature cure to suppress the formation of defects,
particularly voids, within the polymer resin. However, with the growing use of composite in
aircraft and the rise in demand for air travel, the high capital and operating costs, long cycle times,
and process flow restrictions associated with autoclave processing motivate the demand for more
cost-effective and flexible alternatives.
2

Vacuum bag only (VBO) prepreg processing is one such approach. VBO prepregs are
vacuum bag-cured in conventional ovens, and are therefore compacted only by an atmospheric
pressure differential of 101.3 kPa (1 atm, compared to ~ 5 atm typical in autoclaves). To achieve
autoclave quality levels under this reduced processing pressure, VBO prepregs feature a partially
impregnated microstructure (by design) that allows the evacuation of a majority of the air extant
between and within the laminate plies. This distinctive characteristic, along with several bagging
arrangements and cure cycle modifications, allows the manufacture of defect-free parts with high
microstructural quality. Nevertheless, the lack of positive pressure during processing renders
laminates cured through VBO methods more susceptible to certain defect-causing phenomena.
1.1 Surface Porosity
Surface porosity often arises on the tool-side surface of composite laminates made from
VBO prepregs. While it is generally not detrimental to mechanical properties, surface porosity
degrades the aesthetic quality of the part and must often be remedied, incurring additional time
and cost. Several solutions to eliminate it currently exist. For example, during layup, a resin-rich
surfacing ply can be added between the first prepreg ply and the tool plate to produce a smooth,
8

resin-rich surface. However, this adds parasitic weight to the final product and may not be viable
in weight-critical applications. Post-cure operations include gel coating and painting, which also
add weight to the final product as well as time and cost to the manufacturing process. The ability
to consistently produce void-free surfaces without unnecessary materials or process steps, while
much needed, is presently unavailable.
Surface porosity has been addressed in previous studies, but the primary causes are not
comprehensively understood, and few truly effective mitigation strategies have been proposed,
particularly for non-autoclave processes. Herring et al. reported that when using a non-autoclave
process based on bladder-induced consolidation (the Quickstep method), a decrease in compaction
pressure resulted in an increase in surface porosity, indicating that resin pressure is an influential
process parameter and pointing to why surface porosity arises more often in non-autoclave than in
autoclave processes.
26
Darrow et al. used a design of experiments (DoE) approach to identify
sources of surface porosity in autoclave-processed parts.
27
Several factors were considered
including prepreg supplier, moisture on the tool surface, and loss of vacuum during autoclave
staging. None of the parameters were identified as primary sources of surface porosity, but the
authors inadvertently discovered that use of a release film at the tool-part interface resulted in a
void-free surface. This solution works well for flat laminates and is commonly used for such
applications
28,29
, but is not viable for contoured parts due to the difficulty of draping a polymer
film over tooling with multiple curvatures.
Recent studies have suggested that surface porosity results from air that becomes trapped
between the prepreg and tool plate during layup.
30–32
In a previous study, we reported that
composite laminates made from woven prepreg exhibit more surface porosity than laminates made
from unidirectional (UD) prepreg, an observation attributed to the initial morphology of the
9

prepreg.
30
Air entrapment is less likely with UD prepregs because of the more uniform surface
topography. In the same study, we investigated the effect of tool surface roughness on surface
porosity to determine if a rougher tool plate created a more permeable tool-part interface, but
results showed no identifiable trend.
Grigoreiv et al. attempted to reduce surface porosity during VBO processing by
manipulating other tool plate properties. First, they studied the effect of tool surface energy on
surface porosity by treating tool plates with atmospheric pressure plasma.
31
The results did not
indicate a relationship between surface energy and surface porosity, but tool plates with higher
surface roughness resulted in laminates with reduced surface porosity. Then, the same group
investigated the effect of tool topography on surface porosity.
32
Different microstructural patterns
were created on the tool surface, confirming that surface topography, microstructural spacing, and
material cure cycle have an effect on the resulting surface porosity quantity. While our study
showed no relationship between tool roughness and surface porosity, this study suggested that
surface porosity can be reduced by optimizing the shape and size of microstructures present on the
tool surface.
A review of the literature indicates that surface porosity may be governed by both prepreg
material properties and tool properties, but does not clarify the relative importance of each factor.
The literature also contains numerous studies focused on the nucleation, growth and migration of
bulk porosity
33,34
, as well as on the material and processing factors governing internal voids
5
.
However, none of these works consider surface defects. In addition, few studies have directly
considered the role of other process parameters, including pressure and room temperature vacuum
hold time, on such surface defects. The mitigation strategies that have been proposed thus far are
either only viable in specific cases (e.g. the use of a release film) or remain relatively exotic (e.g.
10

tool surface patterning) and unlikely to be quickly implemented in an existing production
environment.
1.2 Objectives
In this work, we clarify the fundamental causes of surface porosity during VBO prepreg
processing and identify avenues for mitigating surface defects that avoid adding significant
manufacturing time or part weight. We describe a systematic experimental study consisting of
material characterization, manufacturing trials and surface porosity quantification. The resulting
data clarify the dominant material-process-property relationships, and identify science-based
approaches for effectively and consistently reducing surface defects.
2. Materials and Methods
2.1 Test Matrix
Table 1-1 outlines the factors considered, the ranges investigated, and the prepreg material
used. These factors include both material properties and manufacturing parameters.
Table 1-1. Parameters, ranges, prepreg choice and number of repeat tests included
in parametric study.

The fiber bed architecture we considered included typical woven fabric and unidirectional
product forms to determine the influence of prepreg surface topology on surface void formation
Parameter Range Prepreg Number of repeat tests
Material
Prepreg fiber bed architecture Unidirectional; woven UD; woven 3
Prepreg humidity conditioning 99%RH Woven 3
Prepreg out-time 0 – 49 days Woven 3
Out-time aging method On tool; off tool Woven 2
Prepreg freezer storage time 0 – 12 months Woven 3
Processing
Release method Liquid agent; film UD; woven 3
Room temperature vacuum hold 0 – 16 h Woven 3
Debulk 1st ply down 0 – 10 min Woven 3
Spike 1st ply down None; spiking Woven 3
Reduced vacuum level 80-99 kPa Woven 3
11

during VBO processing. Prepregs were conditioned in humid ambients to determine the effect of
absorbed moisture within the resin on surface porosity. Prepreg plies were conditioned for 30 hours
on a tray within a sealed bell jar partially filled with water below the tray. The water gave rise to
a constant relative humidity level of 99%, confirmed by a digital humidity sensor. A four-hour
vacuum hold at 20 °C was performed before cure to allow sufficient time for all air trapped during
layup to be evacuated, thus ensuring that any surface defects present were a result of moisture, not
trapped air. A wide range of out-times was selected to determine the effects of increased resin
viscosity and decreased prepreg tack on the formation and evolution of surface pores. Prepreg plies
were aged at ambient conditions while stored within sealed plastic bags, as well as after being
placed on the tool, to determine whether the state of the prepreg at the moment of layup was
significant. Finally, the influence of the freezer storage time was investigated by fabricating
laminates with material at the beginning and end of the stated life (12 months). In all cases, frozen
prepreg was stored in sealed bags provided by the material manufacturer at a temperature below -
23 °C (–10 °F).
Layup was performed using both liquid and film release methods. The room temperature
(RT) vacuum hold time was varied by subjecting parts to vacuum for zero to sixteen hours prior
to high temperature cure to evaluate the influence of air evacuation on surface pore formation. In
addition to RT vacuum holds, two additional methods of extracting air from the tool-part interface
were performed: a ten-minute intermediate debulk on the first ply and spiking of the first ply.
Debulking involved assembling a temporary vacuum bag around the first ply and applying vacuum
consolidation before laying up the remaining plies. Spiking entailed rolling a spiked roller tool to
create a regular pattern of transverse perforations within the prepreg. Finally, we investigated the
influence of vacuum quality by fabricating laminates using full vacuum (denoted as 99 kPa, and
12

corresponding to an absolute bag pressure of ~ 2.3 kPa) and 80% vacuum (denoted as 80 kPa, and
corresponding to a bag pressure of ~ 21.3 kPa).
Within a given set of results, we varied only one parameter. Any baseline case that is
presented was a baseline unique to the individual parameter being discussed. This ensured that any
parameter not considered in each individual case did not play a role in the results. For example,
when we investigated the effects of prepreg out-time, we used the same room temperature vacuum
hold time, freezer time, moisture content, fiber architecture, and release method for all laminates.
The only parameter we varied was out-time, and we fabricated the baseline laminates using unaged
prepreg (no out-time).
2.2 Materials
The experiments described below were performed on two representative carbon
fiber/epoxy VBO prepregs designed for primary structure applications. They consisted of a
toughened epoxy (Cytec Industries, Inc. 5320-1) and two fiber beds: an eight-harness satin woven
fabric and a unidirectional tape (UD). The fiber bed areal weight, resin content and cured ply
thickness of each material are provided in Table 1-2. The 5320-1 resin is designed for cure at 93
°C or 121 °C, for three or twelve hours, respectively. A RT vacuum hold ranging from four to
sixteen hours is also recommended, depending on the part size. The specified out life of the resin
is 30 days, and the freezer life is one year from date of manufacture.
Table 1-2. Prepreg and tool plate material properties.

Short
Name
Fiber
Architecture
Fiber Fiber Bed Areal
Weight (g/m
2
)
Resin Resin
Content
(wt%)
Cured Ply
Thickness (mm)
Woven 8 harness satin Carbon 370 Epoxy 36 0.39
UD Unidirectional Carbon 145 Epoxy 33 0.14
13

All laminates were fabricated on the same aluminum tool. Prior to use, the tool surface was
prepared through orbital sanding with 80-grit sandpaper. The tool surface roughness was 0.76 µm,
measured using a surface profilometer (Zygo NexView 8000).
2.3 Laminate Manufacturing
Laminates were cured by VBO processing in an air-circulating oven (Thermal Product
Solutions Blue M). The cure cycle was consistent with manufacturer recommendations, and
consisted of a RT vacuum hold of varying length, 1.5 °C/min ramp rates, and a three hour dwell
at 121 °C. The vacuum bag assembly, shown in Figure 1-1, consisted of a tool treated with a release
agent (liquid or film), the laminate, a bag-side non-perforated release film, edge breathing dams
made of sealant tape wrapped in fiberglass boat cloth, a layer of breather and the vacuum bag.

Figure 1-1. Vacuum bag assembly for laminate manufacture.
For liquid release, a commercial agent (Frekote 770-NC) was applied to the tool three times
in succession, allowing each coat to dry five minutes before applying the next layer. After the third
coat, the tool plate was dried for thirty minutes before laying up the laminate. After every third
laminate produced with the tool plate, the tool was wiped with acetone and release agent was
reapplied. For film release, a single ply of non-perforated fluorinated ethylene propylene (FEP)
was placed at the tool-part interface.
14

The size and stacking sequence of each laminate was 127 mm  127 mm and 4 plies stacked
[0/90]s respectively. In total, 106 laminates were produced for this study; the number of repeat
laminates produced for each testing condition is outlined in Table 1-1. The surface porosity values
reported below are averages.
2.4 Surface Porosity Measurement
Surface porosity was measured using a low-magnification, handheld microscope (Dino-
Lite Premier2 Digital Microscope). Sixteen images of each laminate surface were recorded at 20 
magnification and analyzed using software (ImageJ). Each image spanned an area of 180 mm
2
.
The procedure used to analyze the images is shown in Figure 1-2. Raw images were imported into
the program and all areas of surface porosity were manually selected. Using the software, the
micrographs were converted to binary images with areas of surface porosity black, and the
remaining area white. A percent value for surface porosity was determined by dividing the number

Figure 1-2. Sequence of actions used to quantify surface porosity. (a) Raw image
taken at 20  magnification is uploaded into image analysis software; (b) areas of
surface porosity are manually outlined; and (c) image is converted into binary
image where black pixels represent areas of surface porosity. The black pixels in
this image represent 1.43% of the total pixels in the image, translating to 1.43% of
this area of laminate surface being surface porosity.
15

of black pixels by the total number of pixels in the image. The sixteen values calculated for each
laminate were averaged to obtain an average surface porosity value for each part. Finally, the
average surface porosity values of a set of repeat laminates were averaged to determine the average
surface porosity for a given experimental condition.
2.5 Visual Analysis
The surface topography of uncured prepreg was measured using an optical surface
profilometer (Zygo NexView 8000). Three-dimensional topography maps were obtained, with x-
y dimensions of 9.1 mm  7.8 mm, and a surface roughness (Sa) value for each prepreg was
measured. Micrographs of prepreg surfaces and cross sections were acquired using a scanning
electron microscope (JEOL 6610) in backscatter mode, at an accelerating voltage of 20 kV.
To better understanding the mechanism of air evacuation and air entrapment, four 75 mm
 75 mm prepreg plies were laid up with a [0/90]s stacking sequence on a glass plate to simulate
a tool-prepreg interface that could be observed directly. This procedure has been used in previous
studies to evaluate air entrapment, distribution, and migration, both for internal porosity
33
and
surface porosity
35
. Time-lapse images were recorded with a portable microscope (Dino-Lite
Premier2 Digital Microscope, 20 ) while the assembly was held under vacuum. These images
clearly showed the location, size and evolution of gaps (or bubbles) between the resin-rich surface
of the tool-side prepreg ply and the surface of the glass tool under vacuum compaction. These
gaps, and their gradual collapse, are likely correlated to the presence and evacuation of air
entrapped at the tool-part interface.

16

3. Results and Discussion
3.1 Source of Surface Porosity
We investigated the effect of prepreg fiber architecture to explore the possibility that
surface porosity might stem from air trapped at the tool-prepreg interface during layup. Laminates
fabricated from woven prepreg exhibited extensive surface porosity, including both circular and
square-like pores ranging in size from 0.004 to 1.220 mm
2
(Figure 1-3a). In contrast, no surface
porosity was observed in any laminates produced with UD prepreg, as shown in Figure 1-3b. The
distinction is attributed to the different surface topography of the prepregs associated with the two
fiber architectures.

Figure 1-3. Typical shape and distribution of surface porosity observed on
laminates made from (a) woven and (b) unidirectional prepreg. Note that no
surface porosity is present on laminates made from unidirectional prepreg,
whereas surface porosity is widely present on laminates made from woven
prepreg.
Figure 1-4 shows the quantitative topography of the woven and UD prepregs prior to
laminating. The linear roughness profiles of each prepreg (Figures 1-4c and 1-4d) highlight the
differences between the two types of prepregs. Woven prepregs exhibit a periodic pattern of
hillocks and valleys corresponding to the overlaps and underlaps of tows in the weave, the height
17

differences ranging from 100-200 μm. When placed on the flat tool plate, the overlaps of the
prepreg will touch the tool, while the underlaps will remain slightly raised, creating regions in
which air can be trapped. Conversely, the UD prepreg presents a much smoother surface, with an
areal roughness of only 12.7 μm (compared to 58.6 μm for the woven prepreg), leading to less air
initially trapped at the tool-prepreg interface.

Figure 1-4. Surface topography of the (a) woven and (b) unidirectional prepreg
used in this study. Roughness lines given in (c) and (d) correspond to the white
lines in (a) and (b) respectively.
To directly observe air entrapment, the mechanism of air evacuation, and the conformation
of the prepreg to the tool surface, we laid up a small laminate on a glass plate, thereby simulating
a tool-prepreg interface while affording transparency. Time-lapse images of the system under
vacuum are shown in Figure 1-5. Figure 1-5a shows the interface immediately after applying
vacuum. The gaps (or bubbles) outlined in black are largely located within the gaps formed by
underlap regions of the woven prepreg. Under vacuum compaction, these gaps become air bubbles
18

trapped at the tool-prepreg interface. Figures 1-5b through 1-5d show that the sizes and positions
of the bubbles changed over time under vacuum. During the room temperature vacuum hold, the
bubbles decreased in size and remained relatively stationary. Occasionally, bubbles moved
slightly, always toward sites of tow underlap, indicating that the trapped air was evacuated out-of-
plane at tow underlaps. As the bubbles decreased in size, the resin-rich surface of the prepreg
gradually contacted the surface of the tool. Any bubbles remaining after resin gelation would have
formed surface porosity.

Figure 1-5. Time-lapse images of a glass plate-prepreg interface under vacuum at
room temperature. Images were taken (a) immediately after application of
vacuum as well as (b) 10 minutes, (c) 30 minutes, and (d) 1 hour in to the vacuum
hold. Select air bubbles are circled to highlight the progression of shape, size and
location of trapped air.
19

SEM micrographs reveal open slits in the woven prepreg that provide an evacuation route
for trapped air (Figures 1-6a and 1-6b). A cross section shows the dry fiber tows present in the
woven OoA prepregs (Figure 1-6c), known as engineered vacuum channels (EVaCs). In the
manufacture of OoA prepregs, resin films are applied to dry fabric to partially impregnate fibers
from top and bottom surfaces. Consequently, dry fiber tows (as shown in Figure 1-6c) reside
directly beneath surface layers of resin, and these dry fibers are specifically designed to provide
pathways for air evacuation from the laminate while under vacuum. Thus, the open slits shown in
Figures 1-6a and 1-6b provide external access to the dry fiber tow cores (EVaCs), facilitating the
evacuation of air trapped at the tool-prepreg interface. In particular, we observed bubbles moving
toward these pinhole openings during the vacuum hold, indicating that the vacuum within the dry
EVaCs can contribute to the evacuation of air from the tool-part interface. Note, however, that air
flow was not directly measured in this study.

Figure 1-6. SEM micrographs of the woven prepreg surface. An open slit in the
resin is shown at (a) 22x and (b) 75x. (c) SEM micrograph taken at 25x of a cross
section of two stacked woven prepreg plies. Dry fiber bundles are observed
directly below the surface resin.
We also investigated the effects of absorbed moisture on surface porosity, which reportedly
can lead to porosity in cured laminates produced from OoA prepregs.
3,36
Inspection of tool-side
surfaces of laminates conditioned at different humidity levels revealed defects distinctly different
from the shape and distribution of typical surface porosity (Figure 1-7). The surface defects took
the form of pits uniformly distributed across the entire laminate, unlike surface pores in “dry”
20

laminates, which were typically restricted to resin-rich areas between tows. Moisture in the prepreg
did not eliminate surface porosity due to trapped air, as this type of surface porosity was observed
on conditioned laminates (Figure 1-7). However, a magnified view highlights the difference
between the new, pit-like surface defects that were observed and the bubble-like surface porosity
known to result from trapped air (Figure 1-7b). A more comprehensive study of this parameter is
required to draw definitive conclusions about moisture-related surface porosity. However, these
results indicate that the most common type of surface porosity – large surface pores in resin-rich
tow overlap locations – arises from entrapped air.

Figure 1-7. (a) Tool-side surface of a laminate made from woven prepreg plies
conditioned in a humidity chamber prior to layup and cure alongside a (b)
magnified view, which highlights the difference between porosity due to trapped
air and surface defects resulting from moisture in the prepreg.
3.2 Material and Process Effects
The effects of select material and processing parameters were investigated to determine the
effect of each factor on air entrapment and/or removal. The parameters selected included RT
vacuum hold time, prepreg out-time and freezer storage time, prepreg permeability, and vacuum
level. These parameters are inevitable consequences of the need for freezer storage and common
layup processes used in laminate fabrication, all of which can increase the risk of porosity in
21

general. Mitigation strategies were identified through systematic investigation of the parametric
effects.
3.2.1 Room Temperature Vacuum Hold
We observed an inverse correlation between surface porosity and RT vacuum hold time.
In particular, as vacuum hold time increased, the amount of surface porosity exhibited by the cured
laminate decreased (Figure 1-8a). These results confirmed the original hypothesis that was based
on visual observation – that surface porosity is linked to entrapped air. Because air evacuation is a
time-dependent process, decreasing the room temperature vacuum hold decreased the opportunity
for trapped air to escape.
3.2.2 Out-Time
The effect of prepreg out-time on surface porosity was pronounced, as shown in Figure 1-
8b. The amount of surface porosity decreased by 83% after just four days of out-time and by 99%
after 14 days of out-time. The decrease in surface porosity with increased out-time was attributed
to the concurrent decrease in prepreg tack and ply compliance, and to the increase in ambient
temperature resin viscosity associated with RT aging of prepregs.
37
When prepreg tack decreases,
the adhesive bond between the prepreg and other objects, particularly the tool plate, also decreases.
This has been confirmed in previous studies through energy of separation tests, performed by
pressing together two plies of prepreg, and subsequently measuring the energy required to separate
them.
37,38
Prepreg tack affects adhesion strength
38
and varies with fluidity of the resin
39
. A tackier
resin forms a stronger adhesive bond to an opposing surface or substrate. The increase in resin
viscosity and decrease in ply compliance with out-time have also been previously documented,
and decrease the capacity (1) of the resin to flow, and (2) of the ply to deform under load.
39,40

These factors affect surface porosity as described below.
22

Figure 1-8. The effect of (a) room temperature vacuum hold time, (b) out-time at
room temperature, and (c) freezer time on surface porosity. (d) The effect of room
temperature aging method on surface porosity. “Control” laminates were made
from prepreg with no out-time; “On Tool” laminates were made from prepreg that
was laid up on the tool at zero days of out-time and allowed to age for five days
prior to cure; “Off Tool” laminates were made from prepreg that was aged for five
days prior to being laid up and were cured immediately upon layup.
When a prepreg with low out-time (high tack, low resin viscosity and high compliance) is
compacted under vacuum at ambient temperature, the resin-rich surface can adhere and conform
to the tool surface. In particular, the resin can flow locally within the gaps formed by the tow
overlaps, potentially closing off in-plane air evacuation pathways within the interface and out-of-
plane pathways between the interface and the EVaCs located within the dry fiber tows. As a result,
23

air bubbles can remain entrapped at the tool-part interface during cure, leading to surface porosity.
Conversely, for a prepreg with higher out-time (low tack, high resin viscosity and low compliance),
vacuum compaction does not lead to the same degree of conformation to the tool surface. The
lower flow capacity of the resin-rich regions, coupled with the stiffer saturated fiber bed regions,
allow the prepreg to retain an uneven surface topology. As a result, comparatively more air
evacuation pathways remain open, increasing the interface permeability relative to the low out-
time case
41
and enhancing the evacuation of entrapped air. During subsequent high temperature
ramp, the resin viscosity decreases by orders of magnitude, allowing the resin-rich ply surface to
conform to the tool and resulting in a uniform, void-free surface. Thus, increasing out-time leads
to a decrease in surface porosity. While contrary to intuition, these trends are nevertheless
consistent with previous reports that increased out-time led to a decrease in internal (macro) void
content in laminates, a phenomenon attributed to increased in-plane permeability.
37,41,42

To specifically separate the effects of reduced tack from the other consequences of out-
time (i.e. resin viscosity and ply compliance), four laminates were cured at the same time on the
same tool plate from prepreg with equal out-time. The only difference between the laminates was
the aging process prior to cure. Two of the laminates were fabricated from prepreg aged for five
days in a sealed bag prior to lay-up on the tool plate. The other two laminates were fabricated using
prepreg laid up on the tool plate at zero days out-time and allowed to age on the tool plate for five
days prior to oven cure. Figure 1-8d shows that the two laminates laid up on the tool plate at zero
days and allowed to age on the tool exhibited greater surface porosity than the laminates aged prior
to lay-up on the tool. The observations indicate that the observed decrease in surface porosity
resulted primarily from the reduced tack associated with out-time.

24

3.2.3 Freezer Storage Time
The effect of freezer storage time on surface porosity was investigated by aging prepreg
sealed in bags in a freezer (less than –23 °C) for six and twelve months before being laid up on a
tool plate to cure. The results, shown in Figure 1-8c, demonstrated that surface porosity generally
decreased with freezer storage time – freezer storage for six months resulted in a 93% decrease in
surface porosity. The finding is consistent with the observed decrease in surface porosity resulting
from out-time, and is again attributed to reduced tack. Previous studies have shown that even at
freezer temperatures, epoxy resins undergo cross-linking and slowly cure, thus reducing tack.
39
At
12 months and zero vacuum hold time, the trend reverses, and surface porosity appears in laminates
produced at this testing condition. However, the amount of surface porosity is still less than
observed when using fresh prepreg, and 12 months marks the end of the specified shelf life of the
prepreg, the maximum allowable storage life at –18 °C or lower.
3.2.4 Processing Effects
The effects of several processing variations on surface porosity were also considered and
are summarized in Figure 1-9. The layup of large or complex parts often includes intermittent
debulking periods. These short, periodic applications of vacuum help individual plies conform to
the tool shape and can also evacuate entrapped air. In the present study, intermittent debulks of the
first (tool-side) ply produced no change in surface porosity. A second process variation involved
spiking the prepreg plies with a spiked roller tool before laying up the laminate to increase through-
thickness permeability. As a result, surface porosity was reduced by 40%, and the variance
decreased five-fold. Although a more comprehensive study of the effect of out-of-plane
permeability on surface porosity is required to draw definitive conclusions about the effects of this
25

parameter, the procedure of prepreg spiking does not appear to completely eliminate surface
porosity. Spiking can also lead to fiber breakage and reduce mechanical performance.
Finally, we investigated the effects of reduced vacuum on surface porosity. Vacuum level
is a key parameter in the manufacture of composite parts, and in production settings, bagged parts
commonly undergo changes in vacuum level resulting from bag size, bag leaks, and the fact that
different bags are often connected and disconnected from the same pump. Figure 1-9 compares the
results from three laminates cured at 80 kPa (reduced vacuum) to the results of control samples
cured at 99 kPa (full vacuum). A 20% decrease in vacuum level resulted in an almost 200%
increase in surface porosity. This result aligns with findings of previous studies and demonstrates
the importance of maintaining high vacuum when curing composite parts, particularly out of
autoclave.
4,43–46

Figure 1-9. The effect of additional processing parameters on surface porosity
was considered. These included performing an intermittent debulk on the first ply
down, spiking the first ply down, and reducing the vacuum level.
3.2.5 Statistical Significance
The t-test is a common method used to determine if two sets of data are significantly
different from each other. However, there is disagreement in the literature over whether or not t-
tests can be performed on sample population sizes less than five.
47
Due to time and material
restrictions, our sample sets were typically restricted to smaller population sizes (N = 3).
26

Nevertheless, we present the results of unpaired, two-sample t-tests between the control conditions
and each respective deviating condition. For example, a t-test was performed between the mean of
the control (zero out-time) and the mean of the laminates produced at six days of out-time to
determine the significance of the decrease in surface porosity observed with six days of out-time.
The difference in means of two sample sets is considered statistically significant if the p-
value determined from the t-test is less than the chosen threshold value (usually 0.05).
47
All out-
time levels were deemed statistically significant (p < 0.05) except for the four-day level, where the
p-value was 0.053. The results presented for aging method (allowing prepreg to age off the tool vs.
on the tool) were less significant, with a p-value of 0.076. However, these particular data sets had
an extremely small sample size of N = 2. The increase in surface porosity with decreased vacuum
level in the bag was statistically significant with a p-value of 0.018, indicating that vacuum quality
was a critical factor for defect suppression. Similarly, the length of RT vacuum hold had a
statistically significant effect on surface porosity, as a t-test between laminates fabricated from
fresh prepreg with no RT vacuum hold and those that underwent a 16-hour hold resulted in a p-
value of 0.042. Finally, a p-value of 0.038 was obtained when comparing laminates fabricated
from fresh prepreg to laminates made from prepreg with six months of freezer time, marking the
result as statistically significant. The remaining results associated with freezer time were
statistically less significant due to the coupled effect of the room temperature vacuum hold, which
prevented, in some cases, the formation of surface porosity in both control and aged laminates.
3.3 Mitigation Strategies
In this study, we investigated surface porosity as a function of material and process factors.
For the chosen materials and part size, surface porosity was observed only for woven fabric prepreg
laminates manufactured on tools coated with liquid release agents. These results, in conjunction
27

with visual observations of laminates compacted on a glass tool plate, indicated that surface
porosity was typically caused by air entrapped at the tool-part interface, within gaps formed by the
uneven surface of woven fabric prepregs.
Prepreg out-time was identified as a major factor affecting surface porosity due to its
influence on air evacuation. High out-time leads to reduced tack and ply compliance and increased
resin viscosity. The consequent decrease in conformation between the prepreg and tool surface
increases the ability to evacuate air entrapped at the interface relative to the baseline case with no
out-time. For this study, surface porosity was largely eliminated for out-times greater than five
days. The effects of freezer time were shown to be statistically less significant.
The quality of the vacuum drawn in the bag was also identified as critical: higher bag
pressures reduce the driving force for air evacuation while directly causing more air to be present
in the bag, leading to higher surface porosity. The results also confirmed that, as intuited, longer
room temperature vacuum hold times reduced surface porosity levels by extending the time
available to evacuate air prior to cure.
Two modifications designed to improve air evacuation were investigated: debulking and
spiking the first ply. Both protocols improved surface porosity levels but had low statistical
significance. Finally, the effect of humidity conditioning was studied for a single sample, which
exhibited surface porosity with different characteristics (pore size, shape and distribution) than
that observed in all other samples, further supporting the conclusion that typical surface porosity
is mostly associated with air entrapment at the tool-part interface.
Of the various measures we investigated to mitigate the effect of entrapped air on surface
porosity, two were shown to be effective: (1) maintaining a high vacuum level in the bag, and (2)
using prepreg with reduced tack for the first (tool-side) ply. Both measures reduce air entrapment
28

and/or foster air evacuation. The first method increases the driving force for air evacuation, while
the second reduces the barrier to evacuation. Using prepreg with reduced tack raises two concerns.
First, reduced tack will increase the risk of plies slipping out of position as they are laid up. Despite
this concern, using prepregs with four days of out-time significantly reduced surface porosity, and
this out-time falls well within both the thirty-day out life and the twenty-day tack life of the
material. Second, the use of reduced tack prepreg increases the risk that unacceptable amounts of
internal porosity will result in the laminate. As prepreg out-time and/or freezer time increases,
resin viscosity also increases
48
, impeding tow impregnation and potentially leading to micro-voids
in cured laminates. However, previous research has shown that the void content does not begin to
noticeably increase until 14 days of out-time, whereas surface porosity decreases to nearly zero by
the 14-day mark.
39

In the commercial production of composite parts, surface porosity is often encountered
intermittently – some laminates exhibit extensive surface porosity, while others fabricated by a
similar process exhibit negligible surface porosity. The results described here provide insight into
this phenomenon. Nearly all of the data exhibit large coefficients of variation (roughly 20-100%),
which reflects the variability from panel to panel and the inconsistent and localized nature of air
entrapment. In addition, a multitude of other factors can contribute to intermittent surface defects
in a production environment, most notably, variability in vacuum quality due to multiple parts
connected to the same vacuum pump, and minor but not insignificant differences in prepreg out-
times, particularly between zero and five days. Consistent control of surface porosity in laminates
fabricated with OoA prepregs requires both an understanding of the fundamental causes and a
practice of tracking requisite processing and material parameters.

29

4. Conclusions
In this study, we demonstrated the roles of several material and processing parameters in
surface porosity formation and mitigation. Air trapped during the layup process at the tool-prepreg
interface was identified as the source of large surface pores located in resin-rich regions of the
cured composite tool-side surface. Moisture present in the prepreg prior to cure also played a role
in surface defects, but these defects were distinct from the surface porosity typically caused by
entrapped air. Moisture effects were not the primary focus of this particular study, and further
investigation is required to draw conclusions about the formation and mitigation of surface defects
that arise from absorbed moisture in the prepreg. For surface porosity that arises from entrapped
air, results confirm that vacuum level in the bag and prepreg freezer and out-time are the primary
drivers of effective mitigation.
As composite applications expand into new sectors of industry, surface appearance will
continue to play a major role in consumer acceptance (e.g automotive, recreational equipment,
wind energy, gas storage, etc.). Non-autoclave cure will only be successful if the defects normally
suppressed by autoclave pressure continue to be absent, and both performance and appearance are
comparable to parts processed in the autoclave. Because of the absence of high pressure to suppress
the multiple sources of defects, imperfections, and variability, OoA prepregs are particularly
susceptible to air entrapment and gas-induced voids. However, all sources of porosity can be
controlled/eliminated by developing prepregs with increased air evacuation (effective
permeability). A key finding of this study is that prepreg out-time increases in-plane permeability
of the tool-prepreg interface, and that using prepreg with just a few days of out-time, at least for
the tool-side surface ply, can effectively eliminate surface porosity. These defect-reduction
30

strategies can improve manufacturing efficiency and sustainability by reducing the number of steps
required to produce a part, a widespread objective within the composites industry.

*This study was published in Composites: Part A in August 2015
49

31

CHAPTER 2

Bulk Metallic Glass and Composite Structures for Spacecraft Shielding

32

1. Introduction
Assessing the performance of spacecraft shields requires ground-based hypervelocity
impact testing, where various geometries of shields are subjected to impacts of particles that range
in size and velocity. Due to limitations of ground-based hypervelocity guns, firing velocities are
restricted to 0.5 – 10 km/s. These velocities are typically below that which would be seen during
an actual MMOD impact, so ground-based testing must be used as a predictor for real impacts. To
predict the hypervelocity impact response of materials at higher velocities (8-18 km/s), ballistic
limit equations (BLEs) are developed using the slower, ground-based tests. These equations allow
for calculation of the critical shield thickness required to avoid penetration or detached spall, and
they are a function of physical properties of the shield material (i.e. density, hardness, etc.) and the
projectile (i.e. density, diameter, velocity, etc.) BLEs are developed through a combination of
empirical hypervelocity impact test results and analytical models and simulations.
9

Ballistic impact responses of common shielding materials such as aluminum, titanium, and
stainless steel have been widely studied through extensive hypervelocity testing and computational
simulation. The developed equations allow for the calculation of penetration depth as a function
of projectile velocity, assuming a semi-infinite target. Critical shield thickness is then calculated
based on the desired failure mode. Equations 2-1 – 2-3 give the well-established BLEs of
aluminum, titanium and stainless steel.
50

𝑃 Al
≥ 5.24(𝑑 )
19 18 ⁄
(BHN)
−0.25
(
𝜌 𝑝 𝜌 𝑡 )
0.5
(
𝑉 cos𝜃 𝐶 𝑡 )
2 3 ⁄
(2-1)
𝑃 Ti
≥ 5.24(𝑑 )(BHN)
−0.25
(
𝜌 𝑝 𝜌 𝑡 )
0.5
(
𝑉 cos𝜃 𝐶 𝑡 )
2 3 ⁄
(2-2)
𝑃 Steel
≥ 0.434(𝑑 )
19 18 ⁄
(
𝜌 𝑝 𝜌 𝑡 )
0.5
(𝑉 cos 𝜃 )
2 3 ⁄
(2-3)
33

In these equations, 𝑃 is the penetration depth (cm), 𝑑 the projectile diameter (cm), BHN
the Brinell hardness of the target (in this case the shielding material), 𝜌 the density of the projectile
(𝜌 𝑝 ) and target (𝜌 𝑡 ) (g/cm
3
), 𝐶 𝑡 the speed of sound in the target (km/s), 𝑉 the projectile velocity
(km/s), and 𝜃 the impact angle from the target normal. Critical thickness, in Eqn. 2-4, is determined
based on the desired failure mode, where 𝐾 = 1.8, 2.4, or 3.0 to prevent perforation, detached
spallation or incipient spallation, respectively.
50
Here, 𝑡 𝑠 refers to the critical thickness of the
shield, and 𝑃 ∞
refers to the ballistic limit of the shielding material assuming semi-infinite
thickness.
𝑡 𝑠 ≤ 𝐾 ∗
𝑃 ∞
(2-4)
Similarities between Eqns. 2-1 – 2-3 suggest that an ideal shield would consist of a material
with high hardness and high density, even though practicality indicates that a low areal density is
used to reduce the mass of spacecraft. As such, an optimal shielding material is one that combines
high hardness with the lowest density possible. A ceramic would appear to be an optimal choice,
but one material property that is not explicitly shown in these equations is that of low melting
temperature. Using a material with a low melting temperature results in vaporization of the shield
upon hypervelocity impact, decreasing the chance of solid particles making it through to the
spacecraft wall.
51
Ceramic shields may not melt during impact and could threaten the spacecraft
wall with debris. Moreover, the shielding material needs to possess a sufficient toughness to
survive launch loads. These observations demonstrate why high-performance crystalline metals
are typically used as shields.
Considering these requirements for effective shielding, amorphous metal (AM), also
known as bulk metallic glass (BMG) when made thicker than 1 mm, appears to be a strong
candidate for hypervelocity impact shielding. AMs are multicomponent alloys that are designed
34

around deep eutectics such that they can be quenched from the liquid state into a noncrystalline,
glassy state.
52–55
The amorphous microstructures created in these alloys allow for unique
mechanical properties compared to crystalline metals. BMGs can exhibit carbide-like hardness
(650 BHN, 60 RC), near theoretical yield strengths (>2 GPa), and high elastic limits (2%). They
also exhibit polymer-like processability, allowing for complex shapes to be constructed
56
. BMGs
can therefore have densities similar to other crystalline alloys (e.g., titanium) but with hardness
typically found in ceramics. Moreover, BMGs can have much higher toughness than ceramics,
which allows them to be used in structural applications. Monolithic BMGs are extremely hard, but
brittle due to their inability to arrest a growing crack. Recent work has therefore focused on a new
class of metal matrix composites. BMG matrix composites (BMGMCs) incorporate ductile,
crystalline dendrites into an amorphous matrix which prevent shear bands that form in the glass
matrix from evolving into opening cracks.
56–58
The result is high fracture toughness (>200MPa
m
1/2
), increased fatigue endurance limit (30% of yield strength), and improved ductility (>10%).
BMGMCs typically have a density comparable to that of titanium (5.1–5.8 g/cm
3
vs. 4.5–5.0
g/cm
3
, respectively), while exhibiting more than twice the hardness (600 BHN vs. 250 BHN,
respectively), leading to interest as a spacecraft shield.
59–63

The objective of this study is to evaluate the performance of BMG materials as
hypervelocity shields and compare the results to those of materials currently used as shields for
the ISS. Approximate, theoretical BLEs are developed and presented here. In addition, analysis
and discussion of practical aspects of shielding including detached spall and the effectiveness of
complex shielding geometries is also included. Prior to my joining the group at USC, hypervelocity
impact tests were conducted on two BMG materials at the NASA Ames Vertical Gun Range
(AVGR) in November, 2011 and consisted of nine total impact shots (outlined in Table 3). In the
35

fall of 2012, I joined the efforts to analyze the data and develop BLEs for the BMG materials. This
project was a collaboration between myself and Marc Davidson, a fellow PhD student at the USC
Composites Center, and Dr. Douglas C. Hofmann, a scientist at the Jet Propulsion Laboratory
(JPL).
2. Materials and Methods
2.1 Materials
The materials in this study included the BMG Vitreloy 1 (Zr41.2Ti13.8Cu12.5Ni10Be22.5) and
the BMGMCs DH1 (Zr36.6Ti31.4Nb7Cu5.9Be19.1) and DH3 (Zr39.6Ti33.9Nb7.6Cu6.4Be12.5). Both
BMGMCs contain ductile crystalline dendrites for increased toughness with volume fraction of 40
and 66% in DH1 and DH3, respectively. Single sheets of Vit. 1 were fabricated by die-casting in
a commercial process by Howmet. These sheets were donated for the study by Liquidmetal
Technologies of Rancho Santa Margarita, Ca. The BMGMCs were fabricated using a semi-solid
forging process developed jointly between Caltech and UCSD.
56
In this method, ingots of the
various alloys are held isothermally between the solidus and liquidus temperatures to allow
coarsening of the dendrites. The samples were then quenched by forging, allowing the remaining
eutectic liquid to form a glassy matrix around the dendrites. Two water-cooled copper plates were
used to produce thin sheets.
2.2 Hypervelocity Impact Tests
Table 2-1 gives a description of each of the nine samples used in the study along with the
corresponding ballistic response and penetration depth (to be discussed in subsequent sections).
Five tests were performed on monolithic BMG plates with the same thickness and four tests were
performed on BMGMCs ranging in thickness and composition. All tests were conducted with a
3.17 mm diameter aluminum projectile at an impact angle of 0 degrees. The single-sheet Vit 1
36

samples were impacted at four different velocities (0.82, 1.25, 1.56, and 2.34 km/s). Three single-
sheet DH1 samples were impacted at three different velocities (0.91, 1.25, and 2.34 km/s). Finally,
one single-sheet DH3 sample was impacted at a velocity of 0.8 km/s. One test consisted of four
stacked layers of Vit 1 sheets in a Whipple shield configuration and was impacted at a velocity of
2.79 km/s to simulate the impact of a multi-walled shield.
Table 2-1. Overview of the 9 hypervelocity tests. NP response stands for “no
penetration”, while DS denotes “detached spall”.

Figure 2-1 shows the AVGR facility and experimental setup. Using a 0.3 cal light-gas gun,
impact velocities of 0.8 – 2.79 km/s were achieved. Although the AVGR facility is capable of
firing 3.17 mm aluminum projectiles at 5.5 km/s, slower velocities were chosen, because the
determination of BLEs require that some of the plates are not penetrated during impact. Figure 2-
1a shows the test frame used to hold the samples used for hypervelocity testing. The cutout in the
top plate was added after a shadow was observed to have obscured an impact (as discussed below).
Figure 2-1b shows a large plate of the BMG Vitreloy 1 (Vit 1) that was cut into several 100 mm 
100 mm test pieces, to ensure a constant test sample thickness of 3.34 mm. Figure 2-1c shows a
long exposure image of the impact of a projectile into the center of the pate, with the clamps used
to hold the sample to the fixture visible. Figures 2-1d and 2-1e show the arrangement of the three
Alloy Sample
Description
Mass (g) Thickness
(mm)
Impact Velocity
(km/s)
Ballistic
Response
Penetration
Depth (mm)
Vit 1 Single Sheet 174.085 3.37 2.34 DS 1.21
Vit 1 Single Sheet 176.275 3.38 0.82 NP 0.39
Vit 1 Single Sheet 183.549 3.38 1.56 DS 0.79
Vit 1 Single Sheet 188.949 3.41 1.25 DS 0.74
Vit 1 4 Layer Sheet 151.822 2.82 2.79 DS 1.07
DH1 Single Sheet 7.869 0.76 2.34 DS 0.76
DH1 Single Sheet 41.916 2.13 0.91 NP —
DH1 Single Sheet 41.190 2.15 1.25 DS 1.3
DH3 Single Sheet 45.512 2.34 0.80 NP —
37

cameras used to capture the experimental data. Two are high-speed and one is setup for long
exposures of the impact.

Figure 2-1. Hypervelocity test set up including (a) the sample fixture, (b) the
BMG Vitreloy 1 plate with sectioning lines shown, (c) a representative long
exposure of a hypervelocity impact into a Vitreloy 1 plate, (d) the configuration of
the two high-speed cameras and the test chamber at the NASA Ames Vertical
Gun Range (AVGR), and (e) the long exposure camera filming down on the test
specimen from a port in the chamber.
3. Results and Discussion
3.1 Penetration Depth
The first experiments performed on the AMs were designed to determine the penetration
depth of the Al projectiles into the sheets of the BMGs at various velocities. In the current study,
sheets of BMGs were impacted whereas the only penetration experiments from literature were into
the ends of thick rods of Vit 1.
64
Figure 2-2 shows penetration depth data of the Al projectiles on
Vit 1 from the current study (grey circles) and from Yang et al. (black triangles) and compares it
to the titanium (Ti-15-3-3-3) BLE (Eqn. 2-2), shown as a dotted line for each projectile diameter.
65

To determine the penetration depth as a function of projectile velocity, the depth of the crater
created by the Al projectile was measured (Table 2-1). Solid, best fit curves through the data were
38

estimated by modifying the Ti BLE with a fitting parameter to match the data. In both the current
study and the study from Yang et al., Vit 1 exhibits an approximately 50% decrease in penetration
depth compared to Ti-15-3-3. This result is expected, since Vit 1 has both a higher hardness (653
BNH compared to 257 BNH) and density than Ti-15-3-3 (6.00 g/cm
3
compared to 4.73 g/cm
3
).

Figure 2-2. Plot of penetration depth versus particle velocity for an aluminum
projectile impacting sheets of the BMG Vitreloy 1. The black data is from Yang
et al., and the grey data is from the current work. Dotted lines represent the BLE
of Ti 15-3-3 for each projectile diameter. Estimated BLEs for Vitreloy 1 are
represented by the solid lines.
3.2 Ballistic Limit Equation Development
BLEs for commonly used shield materials, such as aluminum, steel, and titanium (Eqns. 2-
1 through 2-3) have been empirically developed through statistics resulting from large numbers of
impact experiments. In the current preliminary study on BMGs, the number of impacts was limited
due to cost and availability of materials so some assumptions were used to compare the
performance of the BMGs with other, well established materials. As a preliminary assessment, the
respective material properties of the BMG Vit 1 were purely plugged into the BLEs for Ti, Al, and
steel and plotted for a 3.17 mm Al projectile impacting between 0 and 4 km/s, shown in Figure 2-
3. What was immediately evident is that the three BLEs, which estimate the performance of Vit 1
39

Figure 2-3. Penetration depth versus particle velocity for a 3.17 mm Al projectile
impacting a single-wall BMG shield. Material properties of the metallic glass
were inserted into known equations for Ti, Al, and steel (shown as dotted lines).
using well established equations for other materials, were very similar with less than 10%
difference in penetration depth. To compare the three equations versus experimental data, four
impacts into sheets of Vit 1 were also plotted in Figure 2-3, at 0.82, 1.25, 1.56, and 2.34 km/s.
Surprisingly, the experimental data shows a penetration depth lower than all three estimates using
established BLEs, indicating that the hardness of the BMG is very effective at stopping
penetration. The four shots also verify that the shape of the BLEs for Vit 1 are similar to those
predicted from the BLEs for Al, Ti, and steel. Since Vitreloy 1 has a yield strength and a density
similar to titanium alloys, the form of the BLE developed for Ti was used as an approximation for
a BLE for Vit 1. This was done by adjusting the titanium BLE with a fitting parameter, 𝑁 , which
adjusted the shape of the Ti BLE to match the Vit 1 data. Using this method, the BLE determined
for Vit 1 is given in Eqn. 2-5 with 𝑁 = 4.3 or 3.8 (for a 3.17 or 5 mm Al projectile, respectively).
This is compared to 𝑁 = 5.4 in Ti.
𝑃 Vit.1
≥ 𝑁 ∗
(𝑑 )(BHN)
−0.25
(
𝜌 𝑝 𝜌 𝑡 )
0.5
(
𝑉 cos𝜃 𝐶 𝑡 )
2 3 ⁄
(2-5)
40

3.3 Detached Spall Behavior
The BLEs for penetration depth assume a semi-infinite sheet of material upon impact and
do not account for back surface affects.
9
The BLE for Vit 1 (Eqn. 2-5) was developed purely on
penetration depth and did not account for the behavior of BMGs subjected to impact-induced shock
waves. Spall is a type of fracture that occurs simultaneously over an area of material. In the case
of brittle materials, it is the result of nucleation and growth of many microcracks simultaneously.
In hypervelocity impacts, this occurs due to the travel of shock waves through the material. Several
groups have studied the spall strength, or resistance to spallation, in BMGs.
66,67
Yuan et al. found
that spall strength in the BMG Zr41.25Ti13.75Ni10Cu12.5Be22.5 decreases as impact force increases.
68

The difference in spall behavior between the BMG Vit 1 and the BMGMC 𝛽 -Vit
(Zr56.3Ti13.8Cu6.9Ni5.6Nb5.0Be12.5) was studied by Zhuang et al.
69
They determined that, due to the
brittleness of BMGs, spalling in Vit 1 was a product of the nucleation, growth and coalescence of
microvoids in the shear bands caused by the shock wave. The composite 𝛽 -Vit, on the other hand,
experienced ductile spalling, during which voids grew along the boundary between the ductile bcc
𝛽 phase and the amorphous matrix. Therefore, the more brittle a BMG is, the more likely it is to
experience severe spalling. Figure 2-2 suggests that Vit 1 out-performs Ti-15-3-3-3 in a ballistic
impact because of its increased hardness. However, this data represents only penetration depth and
does not encompass all factors important in MMOD shielding and ballistic impact resistance; it
does not account for detached spall.
Figure 2-4 shows a series of still images from videos of each Vit 1 hypervelocity impact in
order of increasing impact velocity. The samples are viewed from the side and are backlit to allow
for contrast (it should be noted that Figure 2-4d was the first sample impacted and contained a
shadow from the holding fixture. This shadow was removed during subsequent tests so that they
41

Figure 2-4. Hypervelocity impacts into single wall BMG sheets of Vitreloy 1 at
(a) 0.82 km/s, (b) 1.25 km/s, (c) 1.56 km/s, (d) 2.34 km/s. The front and back of
the plates after impact are shown at the right after the Al projectiles were
removed.
looked like Figure 2-4a through 2-4c). Figure 2-4a shows that at the lowest velocity of 0.82 km/s,
the Vit 1 sheet succeeds in catching the 3.17 mm Al projectile without resulting in detached spall
from the back side of the panel, so-called a “nopenetration” (NP). As the velocity is increased
(Figure 2-4b through 2-4d), spalling occurs from the back of the unsupported plate at velocities
much lower than are predicted from the BLE in Eqn. 2-4. Although the impacts in Figure 2-4b
through 2-4d appear as though the Al projectile has fully penetrated the plate, there is actually a
slight delay in the formation of the spall from the high-speed video, indicating that a propagating
wave is responsible for the spall, not the projectile penetrating the plate. In fact, the Al projectile
was stuck into the surface of all the Vit 1 plates in Figure 2-4, indicating that there was not full
penetration. These impacts are termed “detached spall” (DS), which is separate from “penetration”
(P). The still images from Figure 2-4 also show a debris cloud flying downrange and ejecta
splashing backwards off the face of the plate. At these relatively low velocities, the debris cloud
42

and ejecta are comprised of solid spalled material and no melting was observed (which was
confirmed by analyzing the debris).

Figure 2-5. Sheet thickness versus projectile velocity for the BMG Vitreloy 1 and
the BMG composite DH1. The closed data points represent no-penetration or
detached spall while the open data points represent penetration or detached spall.
The ballistic limit is the curve that divides the two regions. The dotted curve
represents the theoretical BLE for each material, and the solid curve represents
the actual BLE. The one square point is from an impact into DH3.
To adjust the estimated BLE for Vit 1 based on the penetration data, sheet thickness versus
projectile velocity was plotted; closed circles were used to represent samples that experienced no
penetration or detached spall and open circles to represent samples that experienced some
penetration or detached spall (Figure 2-5). The value of 𝑁 in Eqn. 2-4 was adjusted so that the
critical thickness line for no-detached spall (Eqn. 2-7) bisected the closed circle and the open circle
(since the true ballistic limit must be between those values). The value of 𝑁 was determined to be
9.2 instead of 4.3 as was previously determined from penetration measurements that were
independent of spall. In Figure 2-5, the new ballistic limit with 𝑁 = 9.2 is shown as a solid blue
line, and it is compared with the previously determined (theoretical) limit with 𝑁 = 4.3. The final
43

BLE for Vit 1 is given in Eqn. 2-6 (where it is noted that the equation has not been verified for
changes in impact angle or for velocities above 3 km/s).
𝑃 Vit.1
≥ 9.2(𝑑 )(BHN)
−0.25
(
𝜌 𝑝 𝜌 𝑡 )
0.5
(
𝑉 cos𝜃 𝐶 𝑡 )
2 3 ⁄
(2-6)
𝑡 Vit.1
≥ 2.4𝑃 Vit.1
(2-7)

The dotted black line in Figure 2-5 shows the BLE for Vit 1 that was developed using the
penetration experiments alone and the dotted grey line is the BLE for Ti-15-3-3-3 from the
literature. Using the data on penetration alone, it appears that Vit 1 outperforms Ti in terms of
penetration. However, once spalling was considered, the actual data shows that the performance
of Vit 1 is slightly worse than Ti (the solid black line compared to the dotted grey line). As expected
from the literature, Vit 1 is subject to poor shock spalling behavior, even resulting from impacts
with small Al projectiles. To address this problem, BMGMCs have been developed by reinforcing
the BMG with soft, crystalline dendrites for added toughness and ductility (the mechanical
performance of these alloys is well-established in the literature by the current authors). Three
single-wall plates of the BMGMC DH1 (a Zr-Ti BMGMC with >5% ductility in tension and a
fracture toughness greater than 100MPa m
1/2
) were tested (Figure 2-6).
The three single-sheet samples of the BMGMC DH1 were impacted at three different
velocities (0.91, 1.25, and 2.34 km/s), and the results from these tests are also plotted in Figure 2-
5 alongside the results of the Vit 1 impacts. Figure 2-6c shows the impact of the 3.17 mm Al
projectile at 0.91 km/s exactly at the ballistic limit for DH1. Although it appears as though no
spalling has occurred, a small piece of debris was shown to be traveling down range. As such, the
parameter 𝑁 in the BLE developed for DH1 was adjusted so that it passed just to the right of the
data point (the impact is shown as a closed grey triangle in Figure 2-5 and the BLE is shown as a
44

Figure 2-6. Hypervelocity impacts into single-wall sheets of BMG composites.
(a) Long exposures of the impacts shown in (c–f), respectively. (b) Three plates of
the BMG DH1 are shown before impact. (c) Impact into a DH1 plate at 0.91 km/s
showing a catch. The front of the plate is shown at the right after impact. (b)
Impact into DH1 plate at 1.25 km/s showing detached spall from the back of the
sample. (c) Impact into a thin plate of DH1 at 2.34 km/s, showing complete
penetration. (f) Impact into a DH3 showing a catch.
solid grey line). To verify this result, the velocity was increased to 1.25 km/s in a DH1 plate of the
same thickness (shown in Figure 2-6d) and detached spall was, in fact, observed. The BLE of the
BMGMC DH demonstrates that DH1 outperforms both Ti and Vit 1 substantially, indicating that
the much higher toughness of the BMGMC results in a higher resistance to spalling. Moreover,
the density of the composites can be tuned to be as low as 5.0 g/cm
3
by increasing the amount of
Ti in these alloys and their hardness is still significantly higher than Ti,, indicating that they should
be excellent alternatives to crystalline Ti as shields. A third impact was performed on a thin (0.76
mm) single wall sheet of DH1 at a much higher velocity (2.34 km/s) to verify that the material
does become penetrated. In this impact, shown in Figure 2-6e, the sheet was penetrated by the
45

projectile, leaving a circular hole. For the impacts in DH1, it was found that the value of 𝑁 equal
to 6.3 best described the data, and the final BLE for DH1 and the corresponding expression for
critical shield thickness to prevent detached spall are given by Eqns. 2-8 and 2-9, respectively.
𝑃 DH1
≥ 6.3(𝑑 )(BHN)
−0.25
(
𝜌 𝑝 𝜌 𝑡 )
0.5
(
𝑉 cos𝜃 𝐶 𝑡 )
2 3 ⁄
(2-8)
𝑡 DH1
≥ 2.4𝑃 DH1
(2-9)
One impact was also performed on the BMGMC DH3, which is comprised of the same
composition BMG matrix as DH1 but with a much higher volume fraction of crystalline phase
(66% vs. 40%). DH3 is notable for its ultra-high toughness (~200MPa m
1/2
). One impact was
performed at 0.8 km/s and no penetration or detached spall was observed [Figures 2-6 and 2-7f].
This verifies that increasing the volume fraction of soft crystals did not adversely affect the ballistic
limit. However, with only one impact, and no penetration, a comparison between DH3 and DH1
was not possible. Figure 2-6a shows long exposure images from the four BMGMC impact tests,
Figure 2-6b shows the samples prior to impact testing, and the column on the right of Figures 2-
7c through 2-7f shows images of the front surface of the plates after impacting.
3.4 Shielding Geometries
The range of velocities in the current study (0.8–2.8 km/s) was much lower than the
velocities of actual MMOD impacts (8–18 km/s). Using the final estimated BLEs for Vit 1 and
DH1 (Eqns. 2-6 and 2-8), it can be seen that the critical thickness to prevent detached spall upon
an impact of a 3.17 mm Al projectile traveling at a velocity of 15 km/s is equal to 19.4 and 13.1
mm for Vit 1 and DH1, respectively. As velocity increases, so does the critical thickness required
to prevent detached spall, and thus, areal density is added to the shielding. Therefore, to effectively
shield a spacecraft wall from perforation, multi-layer shield configurations (called Whipple
shields) are used to progressively diffuse and catch projectiles and debris. Figure 2-7a shows the
46

Figure 2-7. (a) 2.79 km/s impact into a 4-wall Vitreloy 1 Whipple shield showing
penetration of only the first layer. (b) The sample holder after impact and (c) a
long exposure of the entire impact. (d) The hole generated in the back of the first
layer and (e) the debris embedded in the second layer.
performance of a 4-wall Whipple shield constructed of four stacked sheets of Vit 1 with a gap
(called standoff) in between each layer (the standoff was approximately twice the thickness of the
sheets in the current test, 6 mm). The thickness of each sheet was 2.82 mm, and the impact velocity
was 2.79 km/s, the maximum velocity accessible using the current gun configuration. Equation 2-
6 and Figure 2-5 show that a single sheet of thickness 2.82 mm shot at a velocity of 2.79 km/s
would result in penetration or detached spall; however, the four-wall shield is able to completely
arrest the impact using only one additional layer and standoff. The debris cloud is sufficiently
diffuse after the impact with the bumper shield (or outer shield) that the second sheet does not
experience a localized impact. In the design of actual spacecraft shields, wall thickness, standoff
distance and material selection are all used to provide the maximum amount of shielding capability
at the lowest possible areal density. Figures 2-8b through 2-8e show the post-impact images of the
Whipple shield impact.
4. Conclusions
In this study, we determined approximate BLEs for the BMG Vit 1 and the BMGMC DH1.
We observed the hypervelocity impact resistance of Vit 1, DH1, and DH3 single-sheet samples.
47

In addition, we saw that multi-wall shields are effective in diffusing impact energy. Because there
was a limited number of hypervelocity runs, the BLEs derived in this study are not meant to
represent standard equations for the materials. Many more shots on each material are needed along
with extensive computational simulation before a comprehensive set of equations can be
developed. This work does, however, motivate the use of amorphous materials in spacecraft
shields. Ongoing work in this area is focusing on developing new laminate composites, low-
density panels using Ti-based BMGs, and higher velocity impacts.

*This study was published in Advanced Engineering Materials in January 2014.
70
A follow up
study was published in the same journal in February 2015.
71

48

CHAPTER 3

Adhesion of Metallic Glass and Epoxy in Composite-Metal Bonding

49

1. Introduction
The aerospace industry constantly seeks materials that provide some benefit in
performance and cost while minimizing or reducing weight. Aircraft companies in particular
increasingly use composite materials in aircraft components. This shift has ameliorated concerns
about fatigue issues normally associated with metallic aircraft while helping reduce aircraft weight
and thus fuel costs. One composite-based structure deployed in recent commercial aircraft design
is the fiber metal laminate (FML). FMLs consist of alternating layers of fiber-reinforced polymer
(FRP) and metal. The most common commercial FML product, GLARE, consists of alternating
layers of glass fiber-reinforced epoxy and aluminum. However, researchers have investigated the
possibilities of using different composite and metal combinations.
14,16–20
One particular
combination that has proven unfeasible is carbon fiber composite and aluminum because of the
galvanic corrosion that results from contact between two materials of different electric
potential.
10,11,21–23
Research efforts have aimed to reduce the galvanic corrosion of carbon fiber-
aluminum FMLs by using barriers and coatings to isolate the two materials.
11,21,23
However, this
combination of materials in FMLs has yet to achieve commercial viability.
Bulk metallic glass (BMG) presents one alternative to aluminum that can be paired with
carbon fiber FRPs (CFRPs) in a FML structure and possibly avoid galvanic corrosion; BMGs resist
corrosion because of the absence of grain boundaries in the microstructure.
72
BMGs exhibit other
appealing attributes, including high hardness, strength, elastic limits, and impact resistance,
making them candidates for certain high performance applications.
52,59,70,73
For example, recent
reports have shown that BMGs are viable candidates for hypervelocity impact shielding for
spacecraft and satellites.
70,74
Weight is an especially critical consideration for space applications,
as space structures must be as light as possible to minimize mission cost. FMLs offer an
50

opportunity to leverage the intrinsic low density of FRPCs combined with the high hardness and
impact resistance of BMGs.
The objective of this study was to evaluate the effects of surface treatments on adhesion
between BMG and FRPs in an FML structure. Surface treatments commonly used on aluminum
to promote adhesion were performed on BMG samples, and the resulting surfaces were
characterized and compared to treated aluminum surfaces. The treatments selected mechanically
and/or chemically altered the surface to promote adhesion. The metal surfaces in this study were
evaluated by scanning electron microscopy (SEM), and surface roughness values and water contact
angles were measured. For mechanical characterization, lap shear specimens were constructed and
tested to determine the resulting bond strength at the composite-metal interface. In the presentation
of results, we show that some of the surface treatments commonly used on aluminum also increase
adhesive bonding for BMGs and CFRPs, demonstrating the potential to introduce FML designs
that pair carbon fiber composites with corrosion resistant BMG alloys.
2. Materials and Methods
2.1 Materials and Curing Methods
Two alloys were selected for this study – a common precipitation-hardened aluminum
alloy (Al 6061-T6) and a BMG alloy (Zr44Ti11Ni10Cu10Be25). A carbon fiber reinforced epoxy
prepreg, consisting of a toughened epoxy (Solvay Industries, Inc. 5320-1) and a unidirectional
(UD) carbon fiber tape, was chosen as the composite layer.
Adhesion samples were produced by co-curing, a method commonly used to produce FML
components as well as in investigations of composite-metal adhesion.
75–78
During co-curing, the
composite cures as it simultaneously bonds to the metal substrate, and excess resin from the
composite prepreg serves as the adhesive between the two components.
75
Some samples included
51

an added adhesive layer (3M Scotch-Weld Structural Adhesive Film AF 163-2) at the composite-
metal interface to compare the adhesion that results from a bond with excess resin to a bond with
added adhesive.
2.2 Metal Surface Treatment
Researchers have evaluated the effects of various metal surface treatments on
adhesion.
72,75,76,79–86
The basic objective when preparing a surface for adhesion is to maximize
surface energy, and this can be achieved through several avenues.
86
For example, increasing
surface roughness increases the surface energy of a substrate and provides mechanical interlocking
between the substrate and polymer or adhesive, thus increasing bond strength.
79,86,87
A second way
to increase surface energy and bond strength is to modify surface chemistry.
79

Four surface conditions were investigated, including a control and three surface treatment
methods. Prior to surface treatment, all metal surfaces were cleaned with a common degreaser,
methyl ethyl ketone (MEK).
79,86
As a control, metal surfaces were not further treated after the
degreasing stage. The first surface treatment method was simple mechanical abrasion, a means of
increasing surface roughness.
75,79,80
Metal samples were abraded with 80-grit abrasive using a
hand-held orbital sander, followed by another round of degreasing with MEK prior to bonding.
Phosphoric acid anodizing (PAA) was used as the second surface treatment method.
Samples were initially degreased with MEK, then subjected to a sequence of electrochemical
treatments. First, samples were immersed for 10 minutes in an FPL deoxidizing solution held at
65-71°C (ratio of 14 g sodium dichromate, 300 ml DI water, and 94 ml sulfuric acid). Following
the FPL etch, samples were rinsed with DI water, then transferred to the anodizing solution (ratio
of 400 ml DI water to 45 ml phosphoric acid). The bottom portion of each sample was submerged
in the PAA solution, while the top portion of the sample was connected to the positive pole of an
52

external power source. The negative pole was connected to a copper wire, which was immersed in
the PAA solution. During the first three minutes of anodizing, the voltage of the power source was
gradually increased to 15 V, and held constant for the remainder of the test. Each sample was
anodized for a total of 20 minutes. After anodizing, the samples were rinsed with DI water and air
dried. Anodizing was performed immediately prior to bonding.
In a third surface treatment method, a silane film was applied to sample surfaces, a method
previously shown to promote polymer-metal adhesion,
76,83,84,88,89
including some BMG alloys.
85

Samples were prepared following the procedure outlined by Wang and Gupta.
89
The silane solution
consisted of 3 wt% silane (3-(Trimethoxysilyl)propyl methacrylate, Fischer Scientific) and 97%
methanol. Initially, each sample was ground with 80 and 120 grit abrasive and subsequently
degreased with MEK. The silane solution was then applied to each metal sample by pipette. After
one minute, excess solution was shed, and the samples were cured in an oven for one hour at
110°C.
2.3 Characterization Methods
Select methods were used to characterize treated metal surfaces to compare the effects of
each treatment on the two alloys. Characterization included measuring water contact angle and
surface roughness and imaging by scanning electron microscope (SEM).
Water contact angle is a common measure used to analyze the potential for a surface to
adhere to other substances, and has been used on both BMG and aluminum surfaces in the past.
90,91

The higher the energy of a surface, the more likely it is to form bonds with other substances to
lower its energy. Therefore, a low contact angle (more spreading of the water) indicates a high
surface energy, and a high contact angle (less spreading of the water) indicates a low surface
53

energy. A measurement system (Rame Hart 290F1) was used to measure the contact angle for a
water droplet on each treated surface. Tests were performed at 20 °C using 4 μL drops.
Profilometery was performed to measure the surface roughness of the samples using an
optical surface profilometer (Zygo NewView 8000), and SEM was used to image sample surfaces
before and after surface treatment. Electron micrographs were recorded with a low vacuum SEM
(JEOL JSM-6610LV) in scanning electron imaging (SEI) mode at 20kV. Evaluation of surface
chemistry was performed using x-ray photoelectron spectroscopy (XPS). XPS spectra were
acquired using a photoelectron spectrometer (Kratos Ultra X-ray Photoelectron Spectrometer) with
the analyzer lens in hybrid mode. A monochromatic aluminum anode was used to perform high
resolution scans with a 5mA operating current, 10 kV voltage, 0.1 eV step size, and 20 eV pass
energy. Pressure was maintained in the range 1-3  10
-8
torr. Binding energies of the spectra were
referenced to the C1s core level at 284.6 eV.
2.4 Lap Shear Preparation and Testing
Lap shear tests, commonly used to measure interface bond strength, were used in this study
to evaluate the effect of each surface treatment on adhesion.
75–78,80,86,90
Sample preparation and
testing were performed in accordance with ASTM D5868. Lap shear samples were produced by
co-curing metal and composite pieces. Each metal piece was 25.4  101.6  1.5 mm, while
composite pieces were produced by stacking four plies, each 25.4  101.6 mm. An individual lap
shear sample was fabricated by combining one metal and one composite piece with a 25.4 mm
overlap. Among the set of samples, surface treatment, processing method, and use of an adhesive
film between the composite and metal were varied. For each surface treatment, 12 samples were
produced. Of the 12 samples, six were cured in an autoclave, while the other six were cured by a
vacuum bag only (VBO) method in an oven. Within each set of six, three samples included a single
54

25.4  25.4 mm film of adhesive (3M Scotch-Weld Structural Adhesive Film AF 163-2) between
the metal and composite, while the other three did not.
Samples were cured on an aluminum tool plate and sealed with a vacuum bag, to which
vacuum was applied during the cure cycle. The tool plate included a step of equal thickness to the
metal (1.5 mm) to ensure that the composite lays flat as it cures, and edge dams were placed around
each composite edge to maintain the shape of the composite and provide a pathway for the trapped
air in the composite to evacuate (Figure 3-1a). Figure 3-1b shows the complete vacuum bag
assembly. A vacuum bag sealed the system, and air was evacuated through a valve. Underneath
the vacuum bag was a layer of porous breather, which allowed for even air evacuation. Finally,
release film was included in between the composite prepreg and breather to prevent adhesion of
the breather to the part. The manufacturer’s recommended cure cycle was used: 3 hour dwell at
121 °C, 1.5 °C/minute ramp rate. Samples cured in the autoclave were cured under a pressure of
80 psi, while oven-cured samples were subject to atmospheric pressure.

Figure 3-1. Vacuum bag assembly for lap shear laminate manufacture. (a) The
tool plate includes a step to ensure that the composite lays flat as it cures. (b)
Complete vacuum bag assembly.
55

Lap-shear samples were tested in tension mode (INSTRON 5585H). Each end of the
sample was gripped in fixtures with an initial grip separation of 75 mm, and load was applied with
an extension rate of 13 mm/min. Upon failure, each surface was examined and classified as one or
more of the failure modes outlined in ASTM D5573.
3. Results and Discussion
3.1 Metal Surface Characterization
SEM micrographs (Figure 3-2) revealed the effects of surface treatments on the surface
topography of BMG and aluminum alloys. In the control, the BMG sample was smooth, while the
Al sample exhibited greater initial roughness (Fig 3-2a and 3-2b). Figure 3-2d shows that after
abrasive grinding, the pattern of grooves on the Al surface remained, and the grooves appeared to
be deeper. Abrasive grinding also produced grooves on the BMG surface (Figure 3-2c), although
the grooves were shallower than the grooves on the Al surface for the same treatment, a
consequence of the greater hardness of BMG alloys. This distinction indicates that abrasion time
must be adapted to the specific BMG alloy.
PAA produced generally similar results for BMG and Al samples (Figures 3-2e and 3-2f),
with a few key differences. Both samples exhibited a surface film, an oxide layer that typically
forms when PAA is performed, indicating that PAA formed an oxide layer on BMG surfaces,
much like it does on Al. XPS results confirmed the presence of oxide compounds on the surface
of both BMG and Al samples after anodization. However, although a film is observed on both
samples, the appearance is distinct for the BMG and Al samples. The Al surface exhibits a uniform
surface with pits, while the BMG surface exhibits a pattern of dried cracks and no observable
depressions.
56

Figure 3-2. SEM micrographs of treated metal surfaces. BMG samples are shown
on the left with their Al counterparts on the right: (a) BMG control, (b) Al control,
(c) BMG abrasion, (d) Al abrasion, (e) BMG PAA, (f) Al PAA, (g) BMG silane,
and (h) Al silane.
57

The surfaces of silane treated samples (Figures 3-2g and 3-2h) resembled those of the
control samples. However, the grooves and surface defects on the control surfaces were not as
apparent on silane treated samples, indicating that a film was formed on the surface. The presence
of silane was revealed by XPS for both samples, confirming the formation of a silane film. Aside
from these minor distinctions, there are no signficant differences between the control samples and
the silane treated samples.
Water contact angle (Table 3-1) and surface roughness (Table 3-2) measurements provide
quantitative support to the SEM observations. The BMG and Al control samples differed - as-
received BMG samples were smooth (surface roughness of 0.146 µm), while the as-received Al
samples were rougher (surface roughness of 0.462 µm). The difference was attributed to the
processing method used to fabricate the samples. BMG samples were produced by casting, which
produced a surface that replicated the mold surface. Casting molds typically have low surface
roughness, so it is common for cast BMG alloys to have extremely smooth surfaces. Industrial Al
sheet metal, on the other hand, is typically produced by high volume forming/grinding processes
with higher tolerances on surface quality.
Table 3-1. Water contact angle results.
Water Contact Angle (°)
Control Abrasion PAA Silane
BMG (Zr44Ti11Ni10Cu10Be25) 67.26 ± 3.63 32.68 ± 4.04 0.00 ± 0.00 61.13± 3.00
Al 6061 – T6 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 72.87 ± 7.10

Table 3-2. Surface roughness of BMG and Al samples after treatment.
Surface Roughness ( μm)
Control Abrasion PAA Silane
BMG (Zr44Ti11Ni10Cu10Be25) 0.146 0.786 0.438 0.539
Al 6061 – T6 0.462 1.683 0.448 0.717

58

Abrasive grinding enhanced wetting in BMG samples. The WCA for BMG decreased from
67.26° in the control state to 32.68° after abrasive grinding. The Al samples, on the other hand,
yielded a WCA of 0° for both the control and abrasive ground states. Comparing WCA results to
roughness data, additional factors contributed to the different wetting behavior of aluminum and
BMG. Typically, the greater the surface roughness, the lower the WCA. However, when
comparing the BMG abrasion sample to the Al control sample, the BMG abrasion sample exhibited
a greater WCA, despite being rougher, indicating that factors other than roughness contribute to
the WCA values. In general, the BMG alloy exhibited lower wettability (greater WCA) than Al
samples.
The PAA treatment yielded WCA values of 0° for both BMG and Al. Roughness values
for the PAA treated BMG and Al samples were also similar - 0.438 µm and 0.448 µm, respectively.
The purpose of PAA treatment was to achieve a roughened as well as a chemically altered surface.
Chemical modification to the surface is the primary factor responsible for the increased wettability
- the WCAs for both BMG and Al were reduced to 0° after PAA treatment, while the surface was
less rough than it was after abrasive grinding in each case.
Finally, silane treatment did not substantially increase wettability or roughness of either
the BMG or Al surface. WCA values for silane-treated BMG and Al samples were 61.13° and
72.87°, respectively. The WCA for Al increased compared to all other states, indicating that the
silane film was more hydrophobic than the Al surface. Roughness values for silane-treated BMG
and Al surfaces were 0.539 µm and 0.717 µm, respectively.
3.2 Lap Shear Results
Figure 3-3 shows the results of lap shear tests, organized in groups corresponding to the
surface treatment used on the metal and whether or not an adhesive film was included (indicated
59

by an “A”). Within each group, the average peak load is shown for each metal type and processing
method. In all cases except for PAA, BMG samples yielded peak load values that matched or
exceeded those of Al samples for each set of conditions. The results indicate that the BMG can be
a viable substitute for aluminum in future FMLs and in applications involving metal-to-composite
co-curing. In addition, samples that included an adhesive film yielded greater values of peak load
than samples without an adhesive film. Samples with adhesive achieved the same approximate
peak load for both alloys, regardless of the surface treatment, indicating that in this case, at least,
the intrinsic performance of the adhesive outweighs the benefits obtained from a particular surface
treatment.

Figure 3-3. Peak load values of lap shear samples tested in tension. Each column
height is the average of three identically produced samples with the standard
deviation shown through error bars. Results are grouped by surface treatment
method, with the letter “A” indicating that an adhesive film was used. Within
each group, metal type and processing method are separated.
60

There was no consistent trend in peak load when comparing the results of samples
processed by VBO and autoclave methods. In some cases, VBO samples resulted in greater peak
loads than autoclave samples, while in others the reverse was true. In most cases, differences in
adhesive strength fell within the error for each result, and thus the two processing methods yielded
no significant differences.
3.3 Failure Classification
Sample failure modes were classified according to ASTM 5573 and included adhesive, thin
layer cohesive, light fiber tear, fiber tear, and stock break. Adhesive failure refers to a clean failure
at the metal-polymer interface with no polymer residue at the metal surface, and indicates a weak
interface bond that typically corresponds to lower peak load values. Thin layer cohesive failure
indicates a slightly stronger bond than adhesive failure, and is identified by a thin layer of polymer
or adhesive remaining on the metal surface upon failure. Sample surfaces were classified as light
fiber tear failure mode if a thin layer of polymer was transferred to the metal surface along with a
small amount of fibers. If failure occurred exclusively below the composite surface, and fibers
were observed across the entire metal side of the failure surface, the failure mode was designated
as fiber tear. Finally, stock break failure mode was assigned to samples in which failure did not
occur along the bond line, but instead resulted in failure of the metal. A summary of the failure
classification results is shown in Table 3-3.
All samples that did not include a layer of adhesive film at the composite-metal interface
resulted in a combination of adhesive, thin layer cohesive, and light fiber tear failure modes. These
failure modes, which indicate weak bonds, were consistent with the low peak load values exhibited
by samples with no adhesive layer. On average, in BMG samples without adhesive, 56% of the
failure surface was classified as adhesive, while in aluminum samples without adhesive, 10% was
61

classified as such. This finding indicates that the epoxy in the prepreg bonded more readily to
aluminum than to BMG.
Table 3-3. Overview of failure modes observed on the failure surface of lap shear
samples. The letter “A” indicates that an adhesive film was used in the lap shear
sample.

Adhesive
Thin Layer
Cohesive /
Light Fiber
Tear
Fiber Tear /
Light Fiber
Tear
Stock Break
Control
Al 24% 76% -- --
BMG 72% 28% -- --
Control / A
Al 50% -- 50% --
BMG -- -- 100% --
Abrasion
Al -- 100% -- --
BMG 45% 55% -- --
Abrasion / A
Al 22% -- 78% --
BMG 1% -- 99% --
PAA
Al 2% 98% -- --
BMG 59% 41% -- --
PAA / A
Al -- -- 50% 50%
BMG 100% -- -- --
Silane
Al 15% 85% -- --
BMG 48% 52% -- --
Silane / A
Al 30% -- 70% --
BMG -- -- 100% --

Surface treatment of the BMG samples resulted in failure modes indicative of stronger
bonds. The BMG control samples resulted in 72% adhesive failure, while those treated by abrasion,
PAA, and silane exhibited ~50% adhesive failure. Aluminum samples exhibited a similar decrease
in adhesive failure after surface treatment, but a lower percent area covered by adhesive failure
overall relative to BMG.
The failure modes observed in samples that included an adhesive film indicated stronger
bonding, in accord with lap shear test results. Overall, in BMG samples with adhesive, 75% of the
failure area was characterized by light fiber tear and fiber tear, while only 25% area exhibited
adhesive failure. The Al samples with adhesive film yielded failure modes similar to those without
62

adhesives. While the percent area covered by adhesive failure increased slightly to 25%, the
remaining 75% exhibited light fiber tear and fiber tear, as well as some stock-break failure,
indicating that bonding was overall stronger, and in some cases, stronger than the aluminum itself.
Closer examination of the results of the individual surface treatment methods provides
support to the conclusion from the lap shear results that the BMG samples with the added adhesive
layer exhibited stronger bonding than aluminum counterparts. The BMG control samples in this
category yielded 0% adhesive failure, while the aluminum samples resulted in 50% adhesive
failure. Similarly, silane treated BMG samples produced 0% adhesive failure, while the silane
treated aluminum samples resulted in 30% adhesive failure. An exception to the conclusion that
BMG samples produced stronger bonding than aluminum is evident in the PAA surface treatment.
BMG samples with this surface treatment resulted in 100% adhesive failure, while aluminum PAA
samples exhibited 0% adhesive failure. This distinction is reflected in the lower bond strength
measured in lap shear tests and indicates that the PAA treatment method must be altered either
chemically or in application when performed on BMG alloys.
4. Conclusions
Using surface treatments that are well-established for aluminum alloys, we have
determined the effectiveness of these treatments on the adhesive bond strength between BMG
alloys and polymer matrices in composites. While the methods chosen for surface treatment
produced BMG surfaces similar to treated aluminum surfaces, intrinsic differences in the surface
chemistry of the two alloys contributed to differences in adhesion behavior. Lap shear results along
with analysis of the failure surfaces confirmed that under shear loading conditions, BMG samples
exhibited peak loads that matched or exceeded those of aluminum samples in all cases except for
63

those that were anodized. The latter distinction indicates that a different anodization recipe will be
required for BMG alloys.
Commercial FMLs are currently restricted to aluminum/glass-fiber-epoxy composite
combinations, primarily for two reasons: (1) to avoid galvanic corrosion between the metal and
composite layer, and (2) aluminum is readily available in sheet metal form. BMGs are highly
resistant to corrosion because of their non-crystalline microstructure and lack of grain boundaries.
The use of BMGs in FMLs could potentially eliminate the issue of galvanic corrosion and open
the design space to CFRPs, thereby leveraging both the superior performance of CFRPs (relative
to GFRPs), and of BMGs (relative to Al alloys). In addition, BMG production has long been
restricted to melt-spun ribbons less than 0.5 mm thick, or small castings of BMG, yielding
thickness greater than 1 mm. These processes were not suitable for production of large sheets of
BMG required for FMLs. However, recent efforts to produce BMG sheets by twin-roll casting
have demonstrated production of sheet stock with thicknesses of 0.1 – 1 mm
92
, potentially enabling
larger scale production of BMG-based FMLs.
This study is the first to evaluate the adhesion at BMG-composite interfaces and compare
it to the interface currently present in commercial FMLs. The results indicate that BMG is a viable
option for structures that involve composite to metal co-curing and bonding. While the high cost
of BMG may preclude use in commercial aircraft, the results support the use of BMG-based FMLs
for structural applications in which the superior mechanical properties of BMG yield distinct
benefits. One application in particular for which BMG-FMLs are being considered is spacecraft
shielding. BMGs have been shown to provide effective protection against hypervelocity impact
from micro-meteorites and debris, both of which constitute serious threats to spacecraft. FMLs
present an opportunity to leverage the high strength and hardness properties of BMGs while
64

minimizing weight through the incorporation of composite material in a laminate structure. Results
of this study demonstrate the possibility for BMGs to be bonded to composite materials in this
way.

*This study is under review with the journal Composites: Part B.

65

CHAPTER 4

Galvanic Corrosion Resistant Fiber Metal Laminates of Metallic Glass and
Carbon Composites
66

1. Introduction
Currently, the design of Al-based FMLs is restricted to the use of fiberglass, as the risk of
galvanic corrosion prohibits the use of carbon fiber reinforced polymer (CFRP) composites. Thus,
designers are unable to leverage the superior specific modulus and strength of CFRPs, limiting the
property space available to designers.
10–12,15,93–97
Galvanic corrosion occurs when two dissimilar
conductive materials are exposed and electrically connected by an electrolyte. In FMLs, the
polymer matrix of the CFRP in principle electrically isolates the two conductive materials;
however, the two components are exposed at free edges of the laminate and can be electrically
connected in the presence of an electrolyte.
Previous research has explored galvanic corrosion between fiber composites and various
metals. Tavakkolizadeh et. al reported that galvanic corrosion occurred when carbon fibers were
in electrical contact with steel in an electrolyte, and they showed that by coating the carbon fibers
with epoxy, the rate of galvanic corrosion was reduced, although not eliminated.
98
Ireland et. al
considered the effect of including carbon nanotubes in glass fiber (GF) reinforced epoxy
composites on galvanic corrosion with aluminum. They reported that galvanic corrosion occurred
for the composites modified with carbon nanotubes, despite the polymer barrier between the two
conductive materials.
99
Other studies demonstrated that certain metals (titanium, anodized
aluminum, and certain stainless steel alloys) resulted in various degrees of galvanic corrosion when
paired with CFRPs.
100,101

In this study, the pairing of bulk metallic glass (BMG) with carbon fiber reinforced polymer
(CFRP) composite in a FML structure was explored. BMGs are distinguished by a non-crystalline
microstructure and high strength and elastic modulus.
102,103
Although metallic glasses are well-
known commercial products, they have normally been used in thin magnetic ribbons for use in the
67

transformer industry or as large castings for electronic devices and golf clubs (thus called bulk
metallic glass). Until recently, tough and corrosion-resistant BMGs have never been manufactured
in a form factor suitable for the use in laminates, which would normally be in the range of 50-200
μm thick sheetmetal. In the current study, a tough Zr-based BMG was manufactured into a
continuous roll of sheetmetal 150 μm thick, allowing for the first experimentation on a BMG
material that has the potential to be scaled up into mass production.
BMGs also exhibit high corrosion resistance, due to the formation of passive oxide layers
and the absence of grain boundaries, and thus are suitable candidates for CFRP-based FMLs.
Previous studies identified factors that contribute to the corrosion resistance of BMGs. Peter et al.
investigated the corrosion behavior of Zr-based BMGs relative to crystalline counterparts and
discovered that while there was no statistically significant difference between corrosion potentials
for each, the difference in pitting potential and protection potential indicated that crystalline metal
is less immune to pitting corrosion than non-crystalline metal.
104
Pitting corrosion occurs
preferentially at grain boundaries, so the absence of grain boundaries increases a metal’s resistance
to corrosion. Other studies have cited the strong passive oxide layer that forms on BMGs as a
factor in their corrosion resistance.
102,105,106
Scully et al. specifically cites Zr-based BMGs as
having the best anodic passivation ability.
106

To date, only one study has investigated BMG-CFRP FMLs. Sun et al. evaluated the
mechanical performance of FMLs consisting of 30 μm thick Al-Ni-La metallic glass ribbons and
unidirectional CFRP.
107
The present study expands on Sun’s work by comparing BMG-based and
Al-based FMLs with a metal ply thickness that more closely matches those of commercial GLARE
(0.2 – 0.5 mm).
108
Analysis includes comparison of galvanic corrosion behavior and determination
of mechanical properties that have been investigated in previous studies of FMLs: interlaminar
68

adhesion
97,109,110
, bending
111,112
, and tensile
109,111–116
properties. The results presented in the
following sections show negligible galvanic corrosion between BMG and carbon fiber and
increased tensile strength and modulus for BMG-based FMLs relative to Al-based FMLs,
demonstrating a pathway to higher performance FMLs that incorporate CFRPs.
2. Materials and Methods
2.1 Materials and FML Fabrication
FMLs were fabricated using two metallic alloys: 2024-F aluminum and a bulk metallic
glass (Zr65Cu17.5Ni10Al7.5), each with a thickness of 150 μm. CFRP was used for the composite
layer, consisting of a toughened epoxy (5320-1, Solvay, Inc.) and unidirectional (HexTow IM7,
Hexcel Corporation) carbon fiber tape. Each FML employed the same general stacking sequence;
a four-ply, [0/90]s composite layer was placed between each pair of metal layers. Previous research
by the authors investigated adhesion methods for BMG-based and Al-based FMLs and determined
that the strongest bond was achieved when the metal was initially degreased and an adhesive layer
(3M Scotch-Weld Structural Adhesive Film AF 163-2) was included at each composite-metal
interface.
117
In this study, metal strips were cleaned with acetone prior to stacking, and the same
adhesive layer was included. The number of metal layers, and thus the overall stacking sequence
and thickness, varied for each test and is outlined in subsequent sections. Individual samples were
cut to size with a diamond blade and precision cutter (Buehler Isomet 4000).
Samples were cured in a vacuum bag assembly in an autoclave, and full vacuum (~95 kPa)
was maintained during cure. The cure cycle included a three-hour dwell at 121 °C with a 1.5
°C/min ramp rate (both up and down). Pressure was applied when the vessel reached 37.8 °C at a
rate of 68.9 kPa/min to a final pressure of 0.689 MPa which was held for the duration of the
temperature dwell.
69

Sheets of unidirectional dry fibers were used in the galvanic corrosion experiments: carbon
fiber (305 g/mm
2
, Fiber Glast Developments Corp.) and glass fiber (Saertex 955 g/m
2
, Fiber Glast
Developments Corp.). For one set of corrosion tests, the BMG was devitrified by heating under
vacuum to 400 °C (above the glass transition temperature of 375 °C) and holding for 30 minutes.
An x-ray diffraction (XRD) pattern was acquired from the devitrified samples to confirm
crystallinity (Bruker, D8 Advance).
2.2 Corrosion Testing
Galvanic corrosion between each metal alloy and fiber pair was evaluated by constructing
a galvanic cell and monitoring the resulting current. Each galvanic cell consisted of a 2.5 cm
2
area
of the metal and fibers submerged in an electrolyte (3.5% by weight NaCl in DI water), separated
by a distance of 2 cm. In this experimental setup, the metal acted as the working electrode, the
fibers were connected to ground, and a saturated calomel electrode (SCE) was used as the reference
electrode. A galvanostat (VersaStat 3 Potentiostat Galvanostat, Princeton Applied Research)
monitored the free-flowing current as a function of time for each galvanic cell.
Long term corrosion was qualitatively monitored. One Al-based and one BMG-based FML
were submerged in a salt water bath (3.5% by weight NaCl in DI water) for two weeks. Each FML
followed the stacking sequence outlined in the previous section, contained seven metal layers, and
were 16.3  14.2  7 mm in size. A polished cross-section of each FML was qualitatively evaluated
using optical microscopy (Keyence VHX-5000), and an image of each was recorded at 150 
magnification prior to submersion. The condition of each sample was re-evaluated after the two-
week period.

70

2.3 Mechanical Testing
Interlaminar, bending, and tensile properties were evaluated using the short beam shear
(ASTM D2344), 3-point bend (ASTM D7264), and tensile (ASTM D3039) tests, respectively. All
tests were performed on the same mechanical testing device (INSTRON 5567). Approximate
dimensions for each type of specimen are given below, with length and width varying by ± 0.5
mm and thickness varying by ± 10%.
Short beam shear samples contained seven metal layers and were approximately 39.3 
13.1  7 mm. The support span was adjusted for each sample, so that the support span-to-measured
thickness ratio was 4.0. Five samples were tested for each FML type, and tests were performed in
compression mode at rate of crosshead movement of 1.0 mm/min. According to the ASTM
standard, interlaminar shear strength was calculated using Eqn. 4-1, where ILSS is the interlaminar
shear strength [MPa], 𝑃 𝑚 is the maximum load [N], 𝑏 is the measured specimen width [mm], and
ℎ is the measured specimen thickness [mm].
ILSS =
3𝑃 𝑚 4𝑏 ℎ
(4-1)
Three-point bend tests were also performed in compression mode at a speed of 1.0
mm/min. Five samples for each FML type were tested, and each sample had four metal layers and
was approximately 78.8  13  3 mm. The support span-to-thickness ratio was chosen to be 19:1
to ensure that the length of each sample was at least 20% greater than the support span. Flexural
strength was calculated using Eqn. 4-2, where 𝜎 𝑚 is the maximum stress at the outer surface [MPa],
𝑃 𝑚 is the peak applied force [N], 𝐿 is the support span [mm], 𝑏 is the measured width of the beam
[mm], and ℎ is the measured thickness of the beam [mm].
𝜎 𝑚 =
3𝑃 𝑚 𝐿 2𝑏 ℎ
2
(4-2)
71

Tensile tests were performed on eight samples for each FML type at a displacement rate of
2.0 mm/min, and a clip-on extensometer was used to record strain information (Instron, 2630-100
series). Sample dimensions were approximately 95  9.5  1.3 mm, each sample containing two
metal layers. To reduce the stress concentration at the grips, end tabs were constructed by adhering
two layers of the aluminum ribbon to the top and bottom of each end of the sample with the same
film adhesive used within the laminate. Grip length was adjusted for each sample to maintain a
gage length of 50 mm. Tensile strength was calculated using Eqn. 4-3, where 𝜎 T
is the tensile
strength [MPa], 𝑃 𝑚 is the peak applied force [N], and 𝐴 is the average cross-sectional area of the
sample [mm
2
]. For each sample, the cross-sectional area was calculated at three locations within
the gage section by multiplying the measured width and thickness at that point; these three values
were averaged to obtain an average cross-sectional area for the sample. Tensile modulus of
elasticity (E) was calculated using Eqn. 4-4, where 𝜎 𝑖 is the stress at the 𝑖 𝑡 ℎ
point (GPa), and 𝜀 𝑖 is
the strain at the 𝑖 𝑡 ℎ
point. Points 1 and 2 were chosen by selecting strain values closest to 0.001
and 0.003.
𝜎 T
=
𝑃 𝑚 𝐴 (4-3)
𝐸 =
𝜎 2
−𝜎 1
𝜀 2
−𝜀 1
(4-4)
3. Results and Discussion
3.1 Corrosion Behavior
Results of the galvanic corrosion tests are displayed in Figure 4-1 and Table 4-1. In this
test, a greater current density indicates a greater rate of corrosion. The galvanic cell of aluminum
and carbon fiber (Al/CF) yielded a current density of 1792.6 A/m
2
, while the BMG/CF cell yielded
a current density was 96.7 mA/m
2
, (~ 19× less). The current density of the BMG/CF cell is similar
72

Figure 4-1. Galvanic corrosion current density exhibited by galvanic cells
consisting of (a) aluminum and carbon fiber and (b) BMG and carbon fiber in a
saltwater electrolyte solution.
Table 4-1. Equilibrium current densities for various metal/fiber pairs in a
saltwater bath (3.5% by weight NaCl).

to the current density between aluminum and glass fiber (GF) (16.8 mA/m
2
, ~ a factor of 5 less).
Thus, when Zr-based BMG is used with CFRP to produce FMLs, galvanic corrosion potential is
expected to be roughly commensurate with GLARE FMLs currently in use.
The long-term corrosion of BMG-based and Al-based FMLs was evaluated by examination
of polished sections (Figure 4-2). Micrographs in Figures 4-2a and 4-2b show each FML prior to
immersion in the saltwater bath, confirming identical starting conditions. Voids were present
Metal Fiber Current Density
(mA/m
2
)
Aluminum Carbon 1792.6
BMG (amorphous) Carbon 96.7
BMG (crystalline) Carbon 131.3
Aluminum Glass 16.8
BMG (amorphous) Glass 0.2
73

primarily within the adhesive layer, and there were few voids within the composite layers or
between layers. Fourteen days in the saltwater bath produced vastly different outcomes for
each FML (see Figure 4-2c). As expected from the results of the galvanic corrosion tests, the Al-
based FML showed more effects of corrosion than the BMG-based FML. Oxidation of the metal,
manifest by a white powdery substance, was clear on all exposed aluminum surfaces, while few
blemishes were observed on the BMG-based FML. In fact, the appearance of the BMG-based FML
is almost identical to the starting condition, indicating negligible corrosion rate (as expected from
Figure 4-1).
Galvanic corrosion is influenced by multiple factors, including electrode potential,
electrolyte chemistry, alloying, heat treatment, surface conditions, and geometry of the materials.
Electrode potential of the conductive materials is generally a good predictor for galvanic behavior;
faster rates of galvanic corrosion are expected for larger differences in electrode potential of the
paired materials. Electrode potentials can vary drastically with electrolyte, so one must also
consider the specific material-electrolyte pair of interest.
118
Table 4-2 gives the electrode potential
of relevant elements and alloys in free-flowing saltwater.
119
The aluminum alloy reported here is
different from the alloy used in this study, and as mentioned previously, alloying can affect
galvanic corrosion behavior. However, there is a large potential difference between the alloy listed
here and graphite (1.04 V), and the electrode potential values for aluminum alloys are known to
be much different from those of carbon and graphite in the galvanic series. The major constituent
of the BMG alloy here was zirconium, with smaller amounts of copper, nickel, and aluminum. The
electrode potential of zirconium is similar to that of graphite (potential difference of 0.29 V).
Nickel and copper have slightly larger potential differences from graphite (0.45 V and 0.61 V,
respectively), although those differences are less than that of aluminum. Because Zr, Cu, and Ni
74

Figure 4-2. Microscope images of (a) Al-based and (b) BMG-based FMLs prior
to submersion in saltwater bath. (c) Al-based and (d) BMG-based FMLs after two
weeks in saltwater bath.
75

are all closer to graphite in the galvanic series than aluminum is, alloy composition undoubtedly
contributes to the reduced galvanic corrosion exhibited by the BMG/CF pair.
Table 4-2. Steady-state electrode potentials of various materials in free-flowing
seawater. Potentials given in reference to a saturated calomel electrode (SCE).
119

Microstructure can strongly influence corrosion behavior, since corrosion occurs
preferentially at grain boundaries.
104
The effect of microstructure was investigated by comparing
the corrosion behavior of the devitrified (crystalline) BMG to that of the non-crystalline BMG.
Results were similar to those reported in a previous study by Peter et. al; the corrosion current
density from crystalline BMG was slightly greater than the non-crystalline BMG (Table 4-1).
However, the current density resulting from the crystalline BMG/CF pair was approximately 14
times less than that of the aluminum/CF pair, indicating that microstructure was a minor factor
affecting the corrosion resistance of this BMG.
Surface oxide layers can exhibit different potentials than the base metal, and thus affect
corrosion behavior. For example, the standard reversible electrode potential for titanium is
negative, but in practice, the electrode potential of titanium in the galvanic series is positive
because of the passive oxide layer on the surface.
118
BMGs generally have strong passive oxide
layers, and Zr-based BMGs in particular exhibit the greatest passivation capacity.
106
Thus, the
passive surface of the BMG undoubtedly contributed to the corrosion resistance observed here.

Material
Graphite …………………………………………………………………………………….. +0.25
Zirconium …………………………………………………………………………………… -0.04
Nickel ………………………………………………………………………………….. -0.20
Copper ………………………………………………………………………………… -0.36
Aluminum alloy ………………………………….…………………………………… -0.79
Steady-state
electrode potential, V
versus SCE
76

3.2 Mechanical Behavior
Interlaminar shear strength, flexural strength, and tensile strength of the FMLs are outlined
in Table 4-3. BMG-based FMLs exhibited a slightly greater interlaminar shear strength compared
to Al-based FMLs. The failure modes were similar for both FML types - primarily interlaminar
shear within the composite layers, as shown in Figures 4-3a and 4-3b. Cracks also propagated
through the metal layers on occasion (Figures 4-3c and 4-3d). Because failure occurred between
Table 4-3. Material properties for BMG-based and Al-based FML.

Figure 4-3. Microscope images of BMG-based and Al-based FMLs after short
beam shear testing. Interlaminar shear of the composite layers (a) and (b), and
cracking of the metal shown in (c) and (d) for BMG-based and Al-based FMLs,
respectively.
Interlaminar Shear Strength (MPa) 50.7 ± 1.3 2.6% 48.3 ± 0.7 1.4%
Flexural Strength (MPa) 808.3 ± 98.3 12.2% 827.2 ± 26.4 3.2%
Tensile Strength (MPa) 892.9 ± 83.3 9.3% 675.6 ± 52.2 7.7%
Tensile Modulus of Elasticity (GPa) 50.9 ± 2.6 5.1% 45.1 ± 2.3 5.1%
Al BMG
77

composite plies (which were identical for both FML types) the values measured for interlaminar
shear strength were essentially the same.
Flexural strength values for BMG-based and Al-based FMLs were statistically equivalent,
and failure occurred primarily within the composite plies. Typical failure modes included
compression of the composite (Figures 4-4a, 4-4b, and 4-4c) and interlaminar shear (Figure 4-4c).
Two of the BMG-based FML samples failed in tension of the outer metal layer (Figure 4-4d); this
failure mode was not witnessed in Al-based FML samples. Although flexural strength was similar
for both FML types, the two FMLs differed in bending stiffness, as manifest in the load-deflection
curves (Figure 4-5). The slopes of the BMG-based FML curves were greater than those of the Al-
based FML curves, indicating greater bending stiffness.

Figure 4-4. Microscope images of FMLs after 3-point bend testing. Compression
failure shown in (a) and (b) for Al-based and BMG-based FMLs, respectively. (c)
Multiple failure modes including compression and interlaminar shear in the
composite for a Al-based FML. (d) Failure by tension in the outer metal layer for
a BMG-based FML.
78

Figure 4-5. Load-displacement curves for BMG-based and Al-based FMLs
during 3-point bend tests.
Average values of tensile strength and modulus of elasticity are reported in Table 4-3. All eight
BMG-based FML samples failed in the gage section, and tensile strength values ranged from 799
– 1072 MPa. Two Al-based FML samples failed in the gage section, while the remaining six failed
at the grips. The two samples that failed in the gage section represented the high and low end of
the tensile strength range of approximately 614 – 751 MPa. BMG-based FMLs exhibited a greater
elastic modulus than aluminum counterparts (50.9 GPa vs. 45.1 GPa for BMG and aluminum,
respectively).
As reported by the prepreg manufacturer, the tensile strength and modulus of a [0/90] s
laminate are 1330 MPa and 83.3 GPa, respectively.
120
The reported tensile strength and modulus
of Zr-Cu-Ni-Al BMGs range from 1300 – 2000 MPa and 88 – 111 GPa, respectively.
121
The tensile
strength of Al 2024-F is roughly an order of magnitude less, ranging from 140 – 210 MPa, and the
modulus is reported as 72.4 GPa.
122
Therefore, the BMG-based FMLs are expected to be stronger
79

with greater modulus of elasticity than the Al-based FMLs, and this was confirmed by the results.
Contrary to expectation, the BMG-based FML was only about 23% stronger in tension, despite
BMG being 500% – 1300% stronger than aluminum. Inspection of the failed samples revealed
both CFRP-dominated and metal-dominated failure. CFRP-dominated failure manifested in
samples that separated into two pieces, indicating that the CFRP itself failed. Metal-dominated
failure was identified in samples that remained in once piece but failed in the metal. Five of the
eight BMG-based FMLs exhibited CFRP-dominated failure, while that was the case for only two
of the eight Al-based FMLs (of which one failed in the gage section and one failed at the grips).
In CFRP-dominated failure, fibers failed as load increased until ultimately just the metal
remained. Therefore, the tensile strength values for samples with CFRP-dominated failure are
slightly misleading – the metal was the remaining intact component just before failure, which had
a smaller cross-sectional area than the FML as a whole. The results indicate that strength is
primarily dominated by the metal in Al-based FMLs and by the composite in BMG-based FMLs.
In general, BMG-based FMLs were stronger than Al-based FMLs but did not reach the high
strength exhibited by BMG.
Figure 4-6 shows stress-strain curves for a BMG-based and Al-based FML that exhibited
metal-dominated failure. A key difference observed in the curves was that the BMG-based FML
exhibited purely elastic deformation, whereas the Al-based FML exhibited elastic-plastic
deformation (as indicated by the change in slope of the stress-strain curve). Yield strengths for Zr-
Cu-Ni-Al BMGs are not reported, because they have been shown to undergo only elastic
deformation before fracture at a strain of ~1.5%.
123
Likewise, yield strength for the [0/90]s CFRP
used in the FMLs is not reported by the manufacturer. Elongation at failure of the CFRP is also
not reported; however, the individual carbon fibers are reported to fail at an elongation of 1.9%.
124

80

Al 2024-F, on the other hand, has a reported yield strength of 75 MPa, and a strain to failure of
20%.
122

Figure 4-6. Typical stress-strain curves for a BMG-based and Al-based FML that
exhibited metal-dominated failure.
The BMG-based FML failed at a strain of ~1.5% and exhibited metal-dominated failure,
indicating that the FML failed because the BMG reached its maximum elongation. The Al-based
FMLs began plastically deforming at a stress of approximately 320 MPa and reached an elongation
of about 1.7%, demonstrating similar ductility to that of the BMG-based FML. Although Al 2024-
F on its own exhibits a strain to failure of 20%,
122
the Al-based FML exhibited reduced ductility
because of the stress concentration that formed in the aluminum sheets. The CFRP allowed the
FML to carry a larger stress at a lower strain than would aluminum on its own. As the FML
strained, more and more of the fibers within the CFRP failed, resulting in the aluminum sheets
carrying an increasing fraction of the load. The stress concentration caused the aluminum to reach
81

its tensile strength (140 – 210 MPa), and thus fail, at a much lower failure strain than exhibited by
monolithic aluminum.
4. Conclusions
In this study, we explored BMG as a candidate substitute for aluminum in FML structures.
Mechanical test results revealed that BMG-based FMLs exhibited equal if not superior mechanical
properties compared with Al-based FMLs. Interlaminar and bending properties were dominated
by the composite layers, but tensile strength and modulus were increased with the incorporation
of BMG. Galvanic corrosion results indicated that BMGs were markedly superior to aluminum in
terms of corrosion resistance, exhibiting a current density almost 20 times less than that exhibited
by aluminum when paired with carbon fiber. Furthermore, the current density of the BMG/CF pair
was only six times greater than that of the Al/GF pair, while the current density of the Al/CF pair
was over 100 times greater than that of the Al/GF pair. Thus, the galvanic corrosion exhibited by
BMG-based FMLs is expected to be on the order of that exhibited by GLARE, making it a
promising candidate for CFRP-based FMLs.
This study demonstrated that BMG-CFRP FMLs can be constructed with negligible
galvanic corrosion between constituents, and furthermore, the thickness of the Zr-based BMG
sheet is similar to the Al sheet metal used to produce GLARE. The cost of BMGs in general will
most likely preclude its use in commercial applications for which lower-strength materials are
sufficient. However, BMG alloys exhibit some unique mechanical properties, and as such they can
meet the needs of highly specialized applications. For example, previous research by the authors
in particular has identified BMG as a viable option for hypervelocity shielding for spacecraft
because of its strength and hardness.
70
Spacecraft shielding is a weight-critical application,
motivating the use of composite structures. By pairing BMG with CFRP in a FML structure, the
82

low density of the CFRP and the high strength and hardness of the BMG can be simultaneously
leveraged. Thus, BMG-based FMLs can potentially expand the material options that exist for
FMLs and provide a pathway to incorporating carbon fiber without risk of corrosion.

*This study is under review with the journal Advanced Engineering Materials.

83

CHAPTER 5

Low Frequency Eddy Current Testing of Insulators and Composites

84

1. Introduction
Electromagnetic non-destructive testing (NDT) is widely used for the evaluation of
conductive materials.
125–127
One common approach, eddy current testing (ECT), employs a system
of coils that induces eddy currents in a conductive sample and subsequently monitors changes in
the secondary magnetic field generated by the eddy currents.
126–128
Defects in the sample distort
the eddy currents, resulting in a change in the magnetic field detected by the coils. ECT has
limitations, including shallow penetration depth, an inability to detect defects with interfaces
parallel to the surface, and the requirement that the probe maintain minimal lift-off from the test
material. Despite these limitations, ECT is widely used for NDT and inspection because of the
advantages offered in terms of cost, portability and ease of interpretation of results. Currently, the
practical utility of ECT is restricted to conductive materials. However, with the growing use of
non-conductive materials in industrial applications that require non-destructive evaluation, there
is a need for an approach for these materials with comparable advantages to ECT.
Recent work has explored the use of ECT on non-conductive materials. High frequency
eddy current (HFEC) devices in the range of 1 to 500 MHz are capable of permittivity
characterization of insulators.
129,130
Gäbler et. al reported the theoretical influence of sample
permittivity on the induced magnetic field and demonstrated experimentally that differences in
sample permittivity are detectable by a HFEC device.
129
In that study, exposed holes in a
polymethyl methacrylate (PMMA) block are shown to be detectable when the coils are operated
between 1.75 and 3.5 MHz, and sub-surface holes at varying depths are detectable when operated
at 6 MHz. Mizukami et. al provided a theoretical derivation of permittivity effects and
experimental results in which holes in a glass fiber reinforced plastic (GFRP) were detectable by
an eddy current device operated at 10 MHz.
130

85

In addition to insulators, there is a specific need for expanded NDT options for carbon fiber
reinforced polymer (CFRP) composites, a material in which the conductive carbon fibers are
embedded in a non-conducting matrix. CFRPs are increasingly used in applications requiring
materials with high strength-to-weight ratios. As the use of CFRPs continues to expand, so will
the demand for low cost NDT techniques that are portable and capable of accessing structurally
complex regions without the use of a coupling agent. ECT satisfies these requirements and has
been shown to be successful in exploiting the conductivity of the carbon fibers to detect various
CFRP features. Researchers have demonstrated that the following features of CFRPs are detectable
by ECT operated at 1-500 MHz: fiber orientation, fiber fracture, fiber volume fraction, and
artificial defects up to 8mm deep.
131–136

1.1 Resonance-Tuned ECT
In the present work, we report an approach to NDT testing of insulators using an eddy
current type device with transmit and receive coils that are tuned such that their respective resonant
frequencies are close in value and the quantity 𝑄 𝑉 𝑚 is maximized, where 𝑄 (quality factor) is the
ratio of resonant frequency to bandwidth, and 𝑉 𝑚 is the maximum output voltage. With this system,
flaw detection sensitivity is optimized.
137

The sensor is capable of detecting defects based on permittivity, similar to those studied
by Gäbler and Mizukami, but at lower frequencies (600-900 kHz).
129,130
In addition to
demonstrating detection based on permittivity, we demonstrate the feasibility of using magnetic
permeability as an additional parameter providing image contrast.

86

2. Materials and Methods
2.1 Sensor Configuration and Testing Method
Two sensors were used in this study, each consisting of a transmit coil (TC) and a receive
coil (RC). In both sensors, the RC was a 1 mm diameter 1000 turn coil. For one sensor, the TC
was 12.7 mm diameter, 100 turns, and for the second sensor, the TC was 24.5 mm diameter, 85
turns. The RC was placed inside the TC, such that their axes were parallel but offset by 4.3 mm
and 11 mm, for the small diameter and large diameter sensors, respectively (Figure 5-1). The TC
was driven with a function generator (Hewlett Packard 33120A) operating from 940 to 980 kHz.
Voltage obtained from the RC circuit was displayed on an oscilloscope (Tektronix TDS2014B),
and data sets were processed (LabView Signal Express, National Instruments Tektronix Edition).
For one-dimensional scans, the coil assembly was mechanically rastered across samples at
variable lift-off (ℎ = 0 − 4 mm) with a lab-developed pulley-based system. Scanning velocity was
constant with the exception of start and stop points, which were distant from defect regions of
interest (Figure 5-1).

Figure 5-1. Schematic of the sensor consisting of an outer (transmit) coil, radius
𝑟 1
, and inner (receive) coil, radius 𝑟 2
. The coil axes are perpendicular to the
sample surface offset laterally by a distance 𝑑 . Lift-off, ℎ, is the distance between
the sensor face and the sample surface.

87

2.2 Materials
Defects were simulated in acetal resin (Delrin) blocks by drilling 12.7 mm diameter holes
of varying depths. Scans were performed on either air-filled or fluid-filled gaps to examine the
effects of electromagnetic properties in the defect. Tests involving air-filled gaps were performed
on an acetal block with two holes, 24.75 mm and 25.5 mm deep, for the first and second hole
respectively. For magnetic permeability, a second acetal block with 19.5 mm deep holes was filled
with four different ferrofluids with manufacturer-stated magnetic permeability values of 1.88,
4.57, 5.68, and 24.63 (×10
-6
H/m
2
) (Ferrotec EMG 911, EFH 1, EFH 3, and EMG 900
respectively).
Carbon fiber epoxy prepreg was used to construct a composite defect sample. The prepreg
material consisted of a toughened epoxy (Solvay, Inc. CYCOM 5320-1) and woven fabric of
carbon fibers (Cytec Industries, Inc. T650-35 3K 8HS Fabric). The 305 mm × 232 mm × 2.75 mm
laminate consisted of eight prepreg plies in a [0/90]2s stacking sequence. Two types of plastic sheet
were used to simulate defects: vacuum bag (nylon, 0.06 mm thick – defects #1, #2, and #3) and
prepreg backing film (polyethylene, 0.12 mm thick – defects #4, #5, and #6). Three square defects
(25.4 mm × 25.4 mm) of each material were placed in the laminate, one at each of the following
depths: shallow (below the first ply – defects #1 and #4), medium (in the middle – defects #2 and
#5), and deep (above the last ply – defects #3 and #6). All six were evenly spaced along the length
of the laminate. The laminate was oven cured according to the manufacturer recommended three-
hour dwell at 121 °C and 1.5 °C/min ramp rate. Vacuum was applied during the cure to remove
trapped air and consolidate the part.

88

3. Results and Discussion
3.1 Representative Scan
Figure 5-2 shows a representative scan of an acetal block containing two air-filled holes,
which yielded signal deflections corresponding to sample and defect boundaries. Circuitry design
was such that constant magnetic flux in the RC (receive coil) resulted in zero voltage readings at
the probes, and changes in magnetic flux in the RC due to electromagnetic property variations in
the sample caused positive or negative voltage deflections. Direct current produced by a full wave
rectifier was fed to an operational amplifier, and sample-dependent voltage changes in the RC
induced transient signals. The scan displayed in Figure 5-2 reveals six signal deflections
corresponding to spatial variations in the inspected region. Signal deflections correspond to
transitions from air to acetal and the reverse as the sensor rasters over the block. Signal deflections

Figure 5-2. Receive coil output voltage as a function of traverse time. The dotted
line indicates sensor path with signal time points corresponding to traverse
distances to air/acetal and acetal/air interfaces.
89

and amplitudes can be partly attributed to variations in lift-off distance caused by irregular
surfaces. The effect of both sample-dependent and operating parameters on the output voltage of
the RC circuit are discussed in the following sections.
3.2 Effect of Operating Parameters on Detectability
3.2.1 Transmit Coil Frequency
Figure 5-3a shows the general relationship between the frequency of the alternating current
in the TC (𝑓 ) and the output voltage of the RC (𝑉 ), with output voltage maximal at resonant
frequency. Resonant frequency is altered by the electromagnetic properties of materials near the
sensor, thus resulting in a horizontal shift in the curve. With 𝑓 held constant, sample-dependent
shifts in 𝑓 R
result in a change in the output voltage, 𝑉 , with signal deflections, as seen in Figure 5-
2. As shown in Figure 5-3a, the slope of the curve approaches zero when 𝑓 approaches 𝑓 R
, and the
slope increases as the operating frequency moves further from resonance. Therefore, for a given
shift in 𝑓 R
due to sample properties, the observed signal deflection is larger when the value of 𝑓 is
further from the initial resonance value. Figure 5-3b shows the signal deflection amplitude as a
function of frequency that results from increasing the TC frequency in one-kHz increments. Signal
deflections are positive when 𝑓 < 𝑓 R
, and negative when 𝑓 > 𝑓 R
. Thus, resonant frequency can
be determined by the point of inflection. For each sample, resonant frequency was determined by
locating the frequency at which the signal deflections changed from positive to negative.
The resonant frequency of the small-diameter sensor over the acetal was determined to be
951 kHz ± 1 kHz. Signal deflection amplitude as a function of 𝑓 at the acetal-air interface is shown
to be maximum at 960 kHz (Figure 5-3c) or approximately 10 kHz above resonant frequency. The
operating frequency for each scan was selected to maximize signal accordingly.
90

Figure 5-3. (a) General relationship between output voltage of the receive coil
and frequency of the alternating current going through the transmit coil. (b) Signal
deflection amplitude as a function of transmit coil frequency. (c) Amplitude of
signal deflection as a function of frequency for an acetal/air interface at a lift-off
of 0.762 mm.
3.2.2 Lift-Off
For a coil with windings parallel to the surface of a sample, magnetic field strength varies
inversely with 𝑧 , the distance along the axis perpendicular to the coil radius and through the center
of the coil. An 𝑁 -turn coil produces a magnetic field according to Eqn. 5-1, where 𝑅 is the radius
of the coil, and 𝜇 0
is the permeability of free space. For 𝑧 = 0, the magnetic field at the center of
the coil is reduced to Eqn. 5-2.
𝐻 =
𝜇 0
𝑁 𝑅 2
𝐼 2(𝑧 2
+𝑅 2
)
3 2 ⁄
(5-1)
𝐻 0
=
𝜇 0
𝑁𝐼
2𝑅 (5-2)
When 𝑧 = 𝑅 , 𝐻 ≈ 0.35𝐻 0
. By Maxwell’s equations, (Eqns. 5-3 through 5-6) the induced electric
field is proportional to the magnitude of the magnetic field from the coil. Thus, increased lift-off
91

decreases the induced electric field, and for constant permeability, the secondary induced magnetic
field. This in turn reduces the output voltage from the RC.
∇ ∙ 𝑬 =
𝜌 𝜀 0
(5-3)
∇ ∙ 𝑯 = 0 (5-4)
∇ × 𝑬 = −𝜇 𝜕 𝑯 𝜕𝑡
(5-5)
∇ × 𝑯 = 𝜀 𝜕 𝑬 𝜕𝑡
+ 𝜎 𝑬 (5-6)
Signal dependence on lift-off for the acetal block with two air-filled holes is shown in
Figure 5-4. Values given represent the signal deflection amplitude at the leading acetal-air
interface. The lowest value of lift-off tested (0.254 mm) resulted in a peak voltage of 0.208 V.
Increasing the lift-off to 1.27 mm decreased the signal by 57.7% to 0.088 V. At a lift-off of 2.032
mm, the signal dropped 73.1% to 0.056 V. Signal-to-noise was sufficient to detect the interface at
all lift-off values tested, resulting in a signal amplitude of 0.024 V at the largest lift-off of 3.81
mm.

Figure 5-4. Signal deflection amplitude as a function of lift off for an air/acetal
interface.
92

3.3 Effect of Permeability on Detectability
From Eqn. 5-5, permeability (𝜇 ) affects the electric field induced by an alternating
magnetic field, and hence spatial variations in 𝜇 affect output in the RC. Variability in permeability
was simulated with ferrofluid-filled holes in an acetal block. Scans performed with the small-
diameter coil operating at a frequency of 961 kHz and a lift-off of 0.762 mm are shown in Figure
5-5. The signal obtained increases with increasing permeability values of the fluid. A non-linear
effect would be expected, as we have shown signal-to-noise depends on the difference between
operating frequency and resonance frequency, which in turn is affected by the large differences in
permeability values of the samples.

Figure 5-5. Signal deflection amplitude as a function of magnetic permeability
for a ferrofluid/acetal interface.
3.4 Applicability to CFRP
In addition to applications for the inspection of insulating materials, this technique is
potentially applicable to NDT of CFRPs. The composite defect sample described above was
scanned using the small-diameter coil at a constant lift-off of 0.762 mm along two trajectories –
93

one directly over the line of defects, and one along a control path that excluded defects from the
sensor field of view. The resonant frequency for the composite sample was 943 kHz ± 1 kHz, and
the operational frequency was set to 953 kHz. Scan results are displayed in Figure 5-6a. With this
sensor, maximum signals reached approximately 15 V, much stronger than the values obtained on
insulator samples, which did not exceed 1.5 V. As expected, stronger signals were obtained in the
setting of conductive fibers than in the setting of permittivity or permeability variability alone.
131–
136

Figure 5-6. Output voltage of the receive coil as a function of time, resulting from
a scan across a CFRP sample containing defects labeled 1 through 6. Scans of the
defect side and the control (no defects) side are shown for the (a) small diameter
and (b) large diameter sensor. A scaled outline of the part and defects is placed
above the scans in (b).
Unlike a monolithic metal, CFRP is characterized by anisotropic conductivity, since the
conductive fibers in CFRP laminates can be aligned in any direction. The composite defect sample
contained fibers aligned in two (orthogonal) directions, and fibers were non-uniformly distributed
in three dimensions, thus making flaw detection difficult. As shown in Figure 5-6a, defects in the
CFRP were indistinguishable from the noise from the carbon fibers with the small diameter coil.
94

Modifying coil diameter, and thus the effective region over which the signal is averaged,
enhanced flaw detection. At the expense of spatial resolution, a larger-diameter coil increased
signal-to-noise.
138,139
Figure 5-6b demonstrates improved detectability of the defects in a CFRP
composite. With constant raster velocity, scan distance varies linearly with time. A scaled outline
of the defect side of the composite part is shown in Figure 5-6b; the front and back edges of the
outlined part line up with the base of the corresponding signal deflections, as shown by the dotted
lines. The leading signal deflection was both large in amplitude and broad, obscuring any signals
that might have resulted from defects labeled #1 and #2. The three signal deflections observed in
the middle of the scan correspond to the leading composite/defect interface for defects labeled #3,
#4, and #5. No signal corresponding to defect #6 was observable, indicating that it was either
obscured by the large signal deflection from the back edge, or that the material type at that depth
was undetectable.
4. Conclusions
The results in this study demonstrated expanded capabilities of ECT as it applies to
insulators and CFRP composites. The use of ECT on insulators has been experimentally
demonstrated at high frequencies, because the contribution from permittivity is greater with
increasing frequency. Here we investigated a new approach for designing an ECT sensor in which
the resonant frequency of the transmit and receive coils are specifically tuned to maximize signal
output from the RC, resulting in a highly sensitive sensor capable of detecting defects in insulators
at lower frequencies than previously reported. In addition, the dependence on magnetic
permeability was demonstrated and confirmed to be a contributing factor in ECT.
As the composites industry continues to grow, new NDT methods will be required to meet
evolving needs that are not addressable by current methods. Improvements in ECT techniques can
95

extend applications to complex geometries, to those situations which prohibit the use of coupling
agents, and to materials with variable electromagnetic parameters, including conductivity,
permittivity and permeability. Not only can ECT be applied to cured CFRP composites with
complex geometries, but in principle, it can be used to evaluate uncured CFRP prepreg, a specific
application that requires the absence of a coupling agent. Valuable time and cost savings could be
realized if uncured prepreg laminates could be evaluated for defects prior to cure. The
understanding of ECT as it applies to non-uniformly conductive materials was expanded by the
results of this study, demonstrating increased versatility of this type of NDT method.

*This study is under review with the journal NDT&E.
96

CONCLUDING REMARKS

In the presented work, several research goals were accomplished. Key results of
each study are discussed along with broader implications. Efficient strategies to mitigate
surface voids on composite parts were developed based on the confirmed source of surface
porosity, entrapped air. These strategies were efficient in the sense that parasitic weight on
the final part was avoided and costly added time and materials were not required within
the manufacturing process. As weight is a critical factor for aerospace applications, current
solutions involving the incorporation of a resin-rich surfacing ply were not practical, thus
motivating the work in chapter 1. A systematic study of relevant parameters revealed that
by using prepreg with just a few days of out-time and maintaining as close to full vacuum
as possible will enable the evacuation of air trapped at the composite-tool interface and
result in zero surface porosity. These results enable manufacturers to make small edits to
their practices to reduce the number of steps to produce a surface void-free part.
Hypervelocity impact tests and empirical BLEs indicated that BMG composites
outperform traditional metals such as titanium, motivating the shift toward using BMG
materials in MMOD shielding systems. FML structures using BMG as the metal layer and
CFRP as the composite layer are a potential candidate for MMOD shielding; the BMG
provides high strength and hardness while the CFRP provides structural rigidity, thereby
enabling the use of thin BMG sheets in a structural system. BMG-CFRP FMLs were
evaluated for their interlaminar adhesion, galvanic corrosion, and mechanical properties.
Results indicated that surface treatments commonly used to promote adhesion for
aluminum resulted in equally strong if not stronger interlaminar bonds when applied to
97

BMG. This study was the first to evaluate the adhesion between BMG and CFRP, paving
the way for BMG-CFRP FMLs, a structure that presents an opportunity to leverage the
high strength and hardness of BMGs and low density of CFRP.
Galvanic corrosion, which currently prohibits the use of carbon fiber in aluminum-
based FMLs, was shown to be negligible between BMG and carbon fiber. Interlaminar
shear strength and flexural strength were similar for Al-based and BMG-based FMLs, as
the failure primarily occurred within the composite layers. However, tensile strength and
modulus were improved with BMG-based FMLs, which was expected based on the greater
tensile properties of BMG when compared to aluminum. The results presented in this
section confirmed that BMG-based FMLs are feasible and perform as well as if not better
than Al-based FMLs. The absence of galvanic corrosion makes BMGs an especially
promising candidate for CFRP-based FMLs, opening the design space for FMLs to
applications that desire the superior mechanical properties of carbon fibers.
A new ECT-based NDT method for the inspection of insulators and CFRPs was
considered, and the resonance-tuned approach to sensor design was described. Adding to
previous reporting that ECT-based NDT methods detect changes in permittivity, results of
the present work indicate that detection is also based on differences in permeability. By
modifying coil diameter, signal-to-noise was increased and detection of defects was
possible in a CFRP laminate. Not only does this add to the suite of NDT options available
for CFRP inspection, it satisfies currently unaddressed needs. For example, inspection
methods for uncured composite prepreg are unavailable because of the required absence of
a coupling agent; in theory, the NDT method presented here should be applicable to such
materials.
98

RECOMMENDATIONS FOR FUTURE WORK

Though the projects presented here cover a diverse range of topics, each one
addresses some form of a lacking efficiency, bet it inefficient processing methods, the
development of higher performing materials, or inadequate inspection methods. In regard
to surface porosity, strategies were identified that are applicable to commercial prepregs.
Future work should focus on fabricating prepregs with specifically tailored resin
distribution that increases out of plane permeability, thereby making it easier for trapped
air to escape. Such work is already underway at USC with the development of USCPreg, a
prepreg design that employs resin strips instead of a continuous film.
140

Considerable preliminary efforts were made to evaluate the possibility for BMGs
to be used with CFRP in a FML. Results indicated that BMG-CFRP FMLs are possible
and perform equally well if not better than aluminum-based counterparts. Because the
costly expense of BMGs may preclude its use in commercial aircraft, future work should
focus on the application of spacecraft shielding. The hypervelocity impact resistance of
BMG-CFRP FMLs should be evaluated and compared to existing shielding structures used
to protect the International Space Station.
Finally, the detectability of defects in a CFRP laminate supports the likely
applicability of the presented ECT-based NDT method to uncured prepreg. However,
experimentation is required to confirm this hypothesis. In addition, future work should
investigate the effect of coil diameter on signal-to-noise ratio in greater detail. This is
essential if this NDT method is to be used by the composites industry.
99

REFERENCES

1. Campbell, F. C. Manufacturing Technology for Aerospace Structural Materials. (Elsevier, 2006).
2. Centea, T., Grunenfelder, L. K. & Nutt, S. R. A review of out-of-autoclave prepregs – Material
properties, process phenomena, and manufacturing considerations. Compos. Part A Appl. Sci.
Manuf. 70, 132–154 (2015).
3. Grunenfelder, L. K. & Nutt, S. R. Void formation in composite prepregs – Effect of dissolved
moisture. Compos. Sci. Technol. (2010). doi:10.1016/j.compscitech.2010.09.009
4. Grunenfelder, L. K. & Nutt, S. R. Moisture and Pressure Effects on Void Formation in Prepreg
Processed Composites. in Proceedings of the ASC 2010 Conference (2010).
5. Gangloff, J. J., Cender, T. a., Eskizeybek, V., Simacek, P. & Advani, S. G. In situ flow
visualization of void migration during out-of-autoclave thermoset prepreg processing. Proc. 12th
Int. Conf. Flow Process. Compos. Mater. Twente, Netherlands 1–2 (2014).
6. Committee on Space Debris. Orbital Debris: A Technical Assessment. (1995).
7. Scientific and Technical Subcommittee of the United States Nations Committee on the Peaceful
Uses of Outer Space. Technical Report on Space Debris. (1999).
8. Christiansen, E. L., Nagy, K., Lear, D. M. & Prior, T. G. Space station MMOD shielding. Acta
Astronaut. 65, 921–929 (2009).
9. Christiansen, E. L. in Handbook for Designing MMOD Protection (2009).
10. Asundi, A. & Choi, A. Y. N. Fiber Metal Laminates: An Advanced Material for Future Aircraft. J.
Mater. Process. Technol. 63, 384–394 (1997).
11. Vermeeren, C. A. J. R. Application of Carbon Fibers in ARALL Laminates - Report LR-658.
(1991).
12. Vogelesang, L. B. & Vlot, A. Development of fibre metal laminates for advanced aerospace
structures. J. Mater. Process. Technol. 103, 1–5 (2000).
13. Kolesnikov, B., Herbeck, L. & Fink, A. CFRP/titanium hybrid material for improving composite
bolted joints. Compos. Struct. 83, 368–380 (2008).
14. Sinmazçelik, T., Avcu, E., Bora, M. Ö. & Çoban, O. A review: Fibre metal laminates, background,
bonding types and applied test methods. Mater. Des. 32, 3671–3685 (2011).
15. Lin, C. T., Kao, P. W. & Yang, F. S. Fatigue behaviour of carbon fibre-reinforced aluminium
laminates. Composites 22, 135–141 (1991).
16. Cortés, P. & Cantwell, W. J. Structure–properties relations in titanium-based thermoplastic fiber–
metal laminates. Polym. Compos. 27, 264–270 (2006).
17. Cantwell, W. J. & Cortes, P. The fracture properties of a fibre – metal laminate based on
magnesium alloy. Compos. Part B 37, 163–170 (2006).
18. Abdullah, M. R. & Cantwell, W. J. The impact resistance of fiber–metal laminates based on glass
fiber reinforced polypropylene. Polym. Compos. 27, 700–708 (2006).
19. Langdon, G. S., Nurick, G. N., Karagiozova, D. & Cantwell, W. J. in Dynamic Failure of
Materials and Structures 269–296 (2010). doi:10.1007/978-1-4419-0446-1
20. Khan, S. U., Alderliesten, R. C. & Benedictus, R. Delamination in Fiber Metal Laminates
(GLARE) during fatigue crack growth under variable amplitude loading. Int. J. Fatigue 33, 1292–
100

1303 (2011).
21. Wang, W. X., Takao, Y., Kashima, I. & Matsubara, T. CFRP/Al-FRML Based on Nano-composite
Coating and Its Mechanical Properties. in 48th AIAA/ASME/ASCE/AHS/ASC Structures, Structural
Dynamics, and Materials Conference – 15th – April 23-26, 2007 – Honoloulu, HI (2007).
doi:10.2514/6.2007-2268
22. Xue, J., Wang, W., Takao, Y. & Matsubara, T. Reduction of thermal residual stress in carbon fiber
aluminum laminates using a thermal expansion clamp. Compos. Part A 42, 986–992 (2011).
23. Wang, W. et al. Development of galvanic corrosion-resistant CFML based on composite coating.
J. Soc. Mater. Sci. 56, 420–425 (2007).
24. Non Destructive Testing. (A. A. Balkema, Rotterdam, 1996).
25. Prakash, R. in Non-Destructive Testing Techniques 1–9 (New Academic Science, 2012).
26. Herring, M. L. & Fox, B. L. The effect of a rapid curing process on the surface finish of a carbon
fibre epoxy composite. Compos. Part B Eng. 42, 1035–1043 (2011).
27. Darrow, D. C. Elimination of Mold surface Porosity on Composite Parts. J. Adv. Mater. (1995).
28. Ridgard, C. Out of Autoclave Composite Technology for Aerospace, Defense and Space
Structures. Proc. SAMPE 2009 Conf. (2009).
29. Herring, M. L., Mardel, J. I. & Fox, B. L. The effect of material selection and manufacturing
process on the surface finish of carbon fibre composites. J. Mater. Process. Technol. 210, 926–940
(2010).
30. Hamill, L., Centea, T., Nilakantan, G. & Nutt, S. SURFACE POROSITY IN OUT-OF-
AUTOCLAVE PREPREG PROCESSING : CAUSES AND REDUCTION STRATEGIES.
(2014).
31. Grigoriev, M. M., Han, J., Woo, R. & Cheng, D. Effect of Atmospheric Pressure Plasma
Treatment on Surface Porosity of OOA Composites. SAMPE 2270–2283 (2013).
32. Grigoriev, M. M., Hogikyan, A., Cheng, D. & Blvd, M. B. POROSITY-FREE MOLDED
SURFACES FOR OUT-OF- AUTOCLAVE COMPOSITES. (2014).
33. Cender, T. a., Simacek, P. & Advani, S. G. Resin film impregnation in fabric prepregs with dual
length scale permeability. Compos. Part A Appl. Sci. Manuf. 53, 118–128 (2013).
34. Gangloff, J. J., Daniel, C. & Advani, S. G. A model of two-phase resin and void flow during
composites processing. Int. J. Multiph. Flow 65, 51–60 (2014).
35. Bloom, L. D., Napper, M. a., Ward, C. & Potter, K. On the evolution of the distribution of
entrapped air at the tool/first ply interf‘ace during lay-up and debulk. Adv. Manuf. Polym. Compos.
Sci. 1, 36–43 (2015).
36. Kardos, J. L. & Dave, R. Void Growth and Resin Transport During Processing of Thermosetting -
Matrix Composites. (1986).
37. Grunenfelder, L. K. & Nutt, S. R. Out Time Effects on VBO Prepreg and Laminate Properties.
Proceedings of the SAMPE 2011 Conference (2011).
38. Ahn, K. J., Peterson, L., Seferis, J. C., Nowacki, D. & Zachman, H. G. Prepreg Aging in Relation
to Tack. J. Appl. Polym. Sci. 45, 399–406 (1992).
39. Grunenfelder, L. K., Centea, T., Hubert, P. & Nutt, S. R. Effect of room-temperature out-time on
tow impregnation in an out-of-autoclave prepreg. Compos. Part A Appl. Sci. Manuf. (2013).
40. Centea, T., Kratz, J. & Hubert, P. The effect of out-time and cure cycle on the consolidation of
out-of-autoclave prepregs. in Proceedings of the 11th International Conference on Flow
101

Processing of Composite Materials (2012).
41. Putnam, J. & Seferis, J. C. Prepreg gas permation as a function of fiber orientation and aging time.
J. Adv. Mater. 26, 36–41 (1995).
42. Centea, T. & Hubert, P. Measuring the impregnation of an out-of-autoclave prepreg by micro-CT.
Compos. Sci. Technol. 71, 593–599 (2011).
43. Tang, J., Lee, W. I. & Springer, G. S. Effects of cure pressure on resin flow, voids, and mechanical
properties. J. Compos. Mater. (1987). doi:10.1177/002199838702100502
44. Liu, L., Zhang, B.-M., Wang, D.-F. & Wu, Z.-J. Effects of cure cycles on void content and
mechanical properties of composite laminates. Compos. Struct. 73, 303–309 (2006).
45. Centea, T. & Hubert, P. Out-of-autoclave prepreg consolidation under deficient pressure
conditions. J. Compos. Mater. 0021998313494101- (2013). doi:10.1177/0021998313494101
46. Boey, F. Y. C. & Lye, S. W. Void reduction in autoclave processing of thermoset composites -
Part 1: High pressure effects on void reduction. Composites 23, 261–265 (1992).
47. Winter, J. C. F. De. Using the Student ’ s t -test with extremely small sample sizes. Pr.
Assessment, Res. Evalutaion 18, 1–12 (2013).
48. Gu, Y., Li, M., Zhang, Z. & Li, Y. Effects of resin storage aging on rheological property and
consolidation of composite laminates. Polym. Compos. 30, 1081–1090 (2009).
49. Hamill, L., Centea, T. & Nutt, S. Surface porosity during vacuum bag-only prepreg processing :
Causes and mitigation strategies. Compos. Part A 75, 1–10 (2015).
50. Ryan, S. & Christiansen, E. L. Micrometeoroid and Orbital Debris ( MMOD ) Shield Ballistic
Limit Analysis Program TM-2009-2147 89. (2010).
51. Davidson, B. M. et al. Investigating Amorphous Metal Composite Architectures as Spacecraft
Shielding **. 1–7 (2012). doi:10.1002/adem.201200313
52. Johnson, W. L. Bulk Glass-Forming Metallic Alloys: Science and Technology. MRS Bull. 24, 42–
56 (1999).
53. Busch, R., Liu, W. & Johnson, W. L. Thermodynamics and kinetics of the Mg65Cu25Y10 bulk
metallic glass forming liquid. J. Appl. Phys. 83, 4134 (1998).
54. Jin, K. & Lo ̈ ffler, J. F. Bulk metallic glass formation in Zr–Cu–Fe–Al alloys. Appl. Phys. Lett. 86,
241909 (2005).
55. Wang, Q., Qiang, J., Wang, Y., Xia, J. & Dong, C. Bulk metallic glass formation in Cu–Zr–Ti
ternary system. J. Non. Cryst. Solids 353, 3425–3428 (2007).
56. Hofmann, D. C. et al. Semi-solid Induction Forging of Metallic Glass Matrix Composites. JOM
61, 11–17 (2009).
57. Peker, A. & Johnson, W. L. A highly processable. Appl. Phys. Lett. 63, 2342–2345 (1993).
58. Hays, C. C., Kim, C. P. & Johnson, W. L. Microstructure Controlled Shear Band Pattern
Formation and Enhanced Plasticity of Bulk Metallic Glasses Containing in situ Formed Ductile
Phase Dendrite Dispersions. 2901–2904 (2000).
59. Hofmann, D. C. et al. Designing metallic glass matrix composites with high toughness and tensile
ductility. Nature 451, 1085–1089 (2008).
60. Hofmann, D. C. et al. Development of tough, low-density titanium-based bulk metallic glass
matrix composites with tensile ductility. Proc Natl Acad Sci U S A 105, 20136–20140 (2008).
61. Wu, Y., Xiao, Y., Chen, G., Liu, C. T. & Lu, Z. Bulk metallic glass composites with
102

transformation-mediated work-hardening and ductility. Adv Mater 22, 2770–2773 (2010).
62. Pauly, S., Gorantla, S., Wang, G., Kuhn, U. & Eckert, J. Transformation-mediated ductility in
CuZr-based bulk metallic glasses. Nat. Mater. 9, 473–477 (2010).
63. Launey, M. E., Hofmann, D. C., Suh, J., Kozachkov, H. & Johnson, W. L. Fracture toughness and
crack-resistance curve behavior in metallic glass- matrix composites glass-matrix composites.
241910, 1–4 (2013).
64. Schramm, J. P., Hofmann, D. C., Demetriou, M. D. & Johnson, W. L. Metallic-glass-matrix
composite structures with benchmark mechanical performance. Appl. Phys. Lett. 97, 1–4 (2010).
65. Yang, C., Liu, R. P., Zhan, Z. J., Sun, L. L. & Wang, W. K. High speed impact on
Zr41Ti14Cu12.5Ni10Be22.5 bulk metallic glass. Mater. Sci. Eng. A 426, 298–304 (2006).
66. Martin, M., Sekine, T., Kobayashi, T., Kecskes, L. & Thadhani, N. N. High-Pressure Equation of
the State of a Zirconium-Based Bulk Metallic Glass. Metall. Mater. Trans. A 38, 2689–2696
(2007).
67. Turneaure, S. J., Winey, J. M. & Gupta, Y. M. Compressive shock wave response of a Zr-based
bulk amorphous alloy. Appl. Phys. Lett. 84, 1692 (2004).
68. Yuan, F. & Prakash, V. Spall strength and Hugoniot elastic limit of a zirconium-based bulk
metallic glass under planar shock compression. (2007). doi:10.1557/JMR.2007.0053
69. Zhuang, S., Lu, J. & Ravichandran, G. Shock wave response of a zirconium-based bulk metallic
glass and its composite. Appl. Phys. Lett. 80, 4522 (2002).
70. Hamill, L. et al. Hypervelocity Impact Phenomenon in Bulk Metallic Glasses and Composites.
Adv. Eng. Mater. 16, 85–93 (2014).
71. Hofmann, D. C., Hamill, L., Christiansen, E. & Nutt, S. Hypervelocity Impact Testing of a
Metallic Glass-Stuffed Whipple Shield. Adv. Eng. Mater. (2015). doi:10.1002/adem.201400518
72. Gebert, A. et al. Investigations on the electrochemical behaviour of Zr-based bulk metallic glasses.
Mater. Sci. Eng. A 267, 294–300 (1999).
73. Salimon, A. I., Ashby, M. F., Bréchet, Y. & Greer, A. L. Bulk metallic glasses: what are they good
for? Mater. Sci. Eng. A 375 –377, 385–388 (2004).
74. Hofmann, D. C., Hamill, L., Christiansen, E. & Nutt, S. Hypervelocity Impact Testing of a
Metallic Glass-Stuffed Whipple Shield. Adv. Eng. Mater. 17, 1313–1322 (2015).
75. Shin, K. C., Lee, J. J. & Lee, D. G. A study on the lap shear strength of a co-cured single lap joint.
J. Adhes. Sci. Technol. 14, 123–139 (2000).
76. Kim, W.-S. & Lee, J.-J. Adhesion strength and fatigue life improvement of co-cured
composite/metal lap joints by silane-based interphase formation. J. Adhes. Sci. Technol. 21, 125–
140 (2007).
77. Shin, K. C. & Lee, J. J. Tensile load-bearing capacity of co-cured double lap joints. J. Adhes. Sci.
Technol. 14, 1539–1556 (2000).
78. Shin, K. C. & Lee, J. J. Fatigue characteristics of a co-cured single lap joint subjected to cyclic
tensile loads. J. Adhes. Sci. Technol. 16, 347–359 (2002).
79. Molitor, P., Barron, V. & Young, T. Surface treatment of titanium for adhesive bonding to
polymer composites: a review. Int. J. Adhes. Adhes. 21, 129–136 (2001).
80. Aktas, A. & Polat, Z. Improving Strength Performance of Adhesively Bonded Single-lap
Composite Joints. J. Compos. Mater. 44, 2919–2928 (2010).
81. Malpass, B. W., Packham, D. E. & Bright, K. A Study of the Adhesion of Polyethylene to Porous
103

Alumina Films Using the Scanning Electron Microscope. J. Appl. Polym. Sci. 18, 3249–3258
(1974).
82. Packham, D. E., Bright, K. & Malpass, B. W. Polyethylene to Aluminium. J. Appl. Phys. 18,
3237–3247 (1974).
83. Tanoglu, M., Mcknight, S. H., Palmese, G. R. & Gillespie, J. W. Use of silane coupling agents to
enhance the performance of adhesively bonded alumina to resin hybrid composites. Int. J. Adhes.
Adhes. 18, 431–434 (1998).
84. Jang, J. & Kim, E. U. I. K. Corrosion Protection of Epoxy-Coated Steel Using Different Silane
Coupling Agents. J. Appl. Polym. Sci. 71, 585–593 (1999).
85. Yang, J. & Dauskardt, R. H. Hybrid coupling layers for bulk metallic glass adhesion. J. Mater.
Res. 28, 3164–3169 (2013).
86. Molitor, P. & Young, T. Adhesives bonding of a titanium alloy to a glass fibre reinforced
composite material. Int. J. Adhes. Adhes. 22, 101–107 (2002).
87. Kinloch, A. J. Review: The science of adhesion. J. Mater. Sci. 15, 2141–2166 (1980).
88. Lim, A. S., Melrose, Z. R., Thostenson, E. T. & Chou, T. Damage sensing of adhesively-bonded
hybrid composite/steel joints using carbon nanotubes. Compos. Sci. Technol. 71, 1183–1189
(2011).
89. Wang, X. & Gupta, V. Construction and characterization of chemically joined stainless steel/E-
glass composite sections. Mech. Mater. 37, 1198–1209 (2005).
90. Ochoa-Putman, C. & Vaidya, U. K. Mechanisms of interfacial adhesion in metal – polymer
composites – Effect of chemical treatment. Compos. Part A 42, 906–915 (2011).
91. Zhang, X., Sun, B., Feng, W., Zhang, Q. & Li, Q. Wetting Behavior of Polymer Melts on Bulk
Metallic Glasses. Appl. Mech. Mater. 404, 25–31 (2013).
92. Hofmann, D. C., Roberts, S. & Johnson, W. L. Twin Roll Sheet Casting of Bulk Metallic Glasses
and Composites in an Inert Environment. 18 (2012).
93. Lin, C. T. & Kao, P. W. Fatigue delamination growth in carbon aluminium laminates. Compos.
Part A 27, 9–15 (1996).
94. Lawcock, G. D., Ye, L., Mai, Y. W. & Sun, C. T. Effects of Fibre/Matrix Adhesion on Carbon-
Fibre-Reinforced Metal Laminates - 1. Residual Strength. Compos. Sci. Technol. 57, 1609–1619
(1997).
95. Lawcock, G., Ye, L., Mai, Y.-W. & Sun, C.-T. The effect of adhesive bonding between aluminum
and composite prepreg on the mechanical properties of carbon-fiber-reinforced metal laminates.
Compos. Sci. Technol. 57, 35–45 (1997).
96. Xia, Y., Wang, Y., Zhou, Y. & Jeelani, S. Effect of strain rate on tensile behavior of carbon fiber
reinforced aluminum laminates. Mater. Lett. 61, 213–215 (2007).
97. Botelho, E. C., Pardini, L. C. & Rezende, M. C. Evaluation of hygrothermal effects on the shear
properties of Carall composites. Mater. Sci. Eng. A 452, 292–301 (2007).
98. Tavakkolizadeh, M. & Saadatmanesh, H. Galvanic Corrosion of Carbon and Steel in Aggressive
Environments. J. Compos. Constr. 5, 200–210 (2001).
99. Ireland, R., Arronche, L. & Saponara, V. La. Electrochemical investigation of galvanic corrosion
between aluminum 7075 and glass fiber/epoxy composites modified with carbon nanotubes.
Compos. Part B 43, 183–194 (2012).
100. Tucker, W. C., Brown, R. & Russell, L. Corrosion Between a Graphite/Polymer Composite and
104

Metals. J. Compos. Mater. 24, 92–102 (1990).
101. Bellucci, F. Galvanic Corrosion Between Nonmetallic Composites and Metals II. Effect of Area
Ratio and Environmental Degradation. Corrosion 48, 281–291 (1992).
102. Wang, W. H., Dong, C. & Shek, C. H. Bulk metallic glasses. Mater. Sci. Eng. R 44, 45–89 (2004).
103. Telford, M. The case for bulk metallic glass. Materials Today 36–43 (2004).
104. Peter, W. H. et al. Localized corrosion behavior of a zirconium-based bulk metallic glass relative
to its crystalline state. Intermetallics 10, 1157–1162 (2002).
105. Pang, S. J., Zhang, T., Asami, K. & Inoue, A. Synthesis of Fe-Cr-Mo-C-B-P bulk metallic glasses
with high corrosion resistance. Acta Mater. 50, 489–497 (2002).
106. Scully, J. R., Gerbert, A. & Payer, J. H. Corrosion and related mechanical properties of bulk
metallic glasses. J. Mater. Res. 22, 302–313 (2007).
107. Sun, B. A., Cheung, K. P., Fan, J. T., Lu, J. & Wang, W. H. Fiber metallic glass laminates. J.
Mater. Res. 25, 2287–2291 (2010).
108. Wu, G. & Yang, J. The Mechanical Behavior of GLARE Laminates for Aircraft Structures. JOM
57, 72–79 (2005).
109. Ye, L., Afaghi-Khatibi, A., Lawcock, G. & Mai, Y. Effect of fibre/matrix adhesion on residual
strength of notched composite laminates. Compos. Part A 29, 1525–1533 (1998).
110. Park, S. Y., Choi, W. J. & Choi, H. S. The effects of void contents on the long-term hygrothermal
behaviors of glass/epoxy and GLARE laminates. Compos. Struct. 92, 18–24 (2010).
111. Khalili, S. M. R., Mittal, R. K. & Kalibar, S. G. A study of the mechanical properties of
steel/aluminium/GRP laminates. Mater. Sci. Eng. A 412, 137–140 (2005).
112. Carrillo, J. G. & Cantwell, W. J. Mechanical properties of a novel fiber – metal laminate based on
a polypropylene composite. Mech. Mater. 41, 828–838 (2009).
113. de Vries, T. J., Vlot, A. & Hashagen, F. Delamination behavior of spliced Fiber Metal Laminates.
Part 1. Experimental results. Compos. Struct. 46, 131–145 (1999).
114. Cortes, P. & Cantwell, W. J. The prediction of tensile failure in titanium-based thermoplastic fibre
– metal laminates. Compos. Sci. Technol. 66, 2306–2316 (2006).
115. Botelho, E. C., Almeida, R. S., Pardini, L. C. & Rezende, M. C. Elastic properties of
hygrothermally conditioned glare laminate. Int. J. Eng. Sci. 45, 163–172 (2007).
116. Carrillo, J. G. & Cantwell, W. J. Scaling effects in the tensile behavior of fiber-metal laminates.
Compos. Sci. Technol. 67, 1684–1693 (2007).
117. Hamill, L. & Nutt, S. R. Adhesion of metallic glass and epoxy in composite-metal bonding (under
review). Compos. Part B Eng.
118. Zhang, X. G. in Uhlig’s Corrosion Handbook 123–143 (2011).
119. in Metals Handbook, Ninth Edition, Volume 13 675 (1987).
120. Cytec Industries Inc. CYCOM ® 5320-1 Epoxy Resin System. 324, 1–11 (2015).
121. Louzguine-Luzgin, D. V, Louzguina-Luzgina, L. V & Churyumov, A. Y. Mechanical Properties
and Deformation Behavior of Bulk Metallic Glasses. Metals (Basel). 3, 1–22 (2013).
122. in ASM Handbook 2, 62–122 (1990).
123. Stoica, M. et al. Strain distribution in Zr64.13Cu15.75Ni10.12Al10 bulk metallic glass
investigated by in situ tensile tests under synchrotron radiation. J. Appl. Phys. 104, 1–7 (2008).
105

124. Hexcel Corporation. HexTow ® IM7. 1–2 (2016).
125. Sophian, A. & Tian, G. Y. Electromagnetic and eddy current NDT : A review Electromagnetic and
eddy current NDT : a review. Insight 43, (2001).
126. Auld, B. A. & Moulder, J. C. Review of Advances in Quantitative Eddy Current Nondestructive
Evaluation. J. Nondestruct. Eval. 18, 3–36 (1999).
127. Garcia-Martin, J., Gomez-Gil, J. & Vazquez-Sanchez, E. Non-Destructive Techniques Based on
Eddy Current Testing. Sensors 11, 2525–2565 (2011).
128. Yi, J. & Lee, S. Analytical Solution for Impedance Change Due to Flaws in Eddy Current Testing.
J. Nondestruct. Eval. 4, 197–202 (1984).
129. Gäbler, S., Heuer, H. & Heinrich, G. Measuring and Imaging Permittivity of Insulators Using
High-Frequency Eddy-Current Devices. in IEEE Transactions on Instrumentation and
Measurement 64, 2227–2238 (2015).
130. Mizukami, K., Mizutani, Y., Todoroki, A. & Suzuki, Y. Design of eddy current-based dielectric
constant meter for defect detection in glass fiber reinforced plastics. NDT E Int. 74, 24–32 (2015).
131. Goeje, M. P. D. E. & Wapenaar, K. E. D. Non-destructive inspection of carbon fibre-reinforced
plastics using eddy current methods. Composites 23, 147–157 (1992).
132. Lange, R. & Mook, G. Structural analysis of CFRP using eddy current methods. NDT E Int. 27,
241–248 (1994).
133. Schulze, M. H., Heuer, H. & Ku, M. High-resolution eddy current sensor system for quality
assessment of carbon fiber materials. Microsyst. Technol. 16, 791–797 (2010).
134. Heuer, H., Schulze, M. H. & Meyendorf, N. in Non-Destructive Evaluation (NDE) of Polymer
Matrix Composites 33–55 (2013).
135. Cheng, J. et al. Role of interlaminar interface on bulk conductivity and electrical anisotropy of
CFRP laminates measured by eddy current method. NDT E Int. 68, 1–12 (2014).
136. Mizukami, K., Mizutani, Y., Todoroki, A. & Suzuki, Y. Detection of delamination in
thermoplastic CFRP welded zones using induction heating assisted eddy current testing. NDT E
Int. 74, 106–111 (2015).
137. McGushion, K. M-Wave Imaging: Using Dual Tuned Resonant Coils to Image Metal, Graphite
Composite and Plastic. (2012).
138. Cherry, M. R., Welterr, S. S. & Blodgett, R. P. Development of high resolution eddy current
imaging using an electro-mechanical sensor. in AIP Conference Proceedings 1430, 324–331
(2012).
139. Nalladega, V., Sathish, S., Jata, K., Blodgett, M. & Knopp, J. in Electromagnetic Nondestructive
Evaluation (XII) 115–122 (2009).
140. Grunenfelder, L. K., Dills, A., Centea, T. & Nutt, S. Effect of prepreg format on defect control in
out-of-autoclave processing. Compos. Part A 93, 88–99 (2017).

Abstract (if available)

Abstract The increase in demand for commercial and space travel has motivated the development of more efficient materials selection and manufacturing practices. More and more people are engaging in air travel, and space exploration is constantly pushing the boundaries of current capabilities. Both efforts require materials that maximize mechanical performance and minimize weight. In addition, shorter processing times, cheaper manufacturing methods, and more versatile evaluation techniques are desired. In the following chapters, five projects are described in which such efforts are addressed. ❧ In chapter 1, a particular kind of defect found in composites produced from out-of-autoclave (OoA) prepregs, surface porosity, is investigated. Currently, the composites industry employs several strategies to eliminate surface porosity including the use of a release film in between the composite part and tool plate, using a resin-rich ply at the composite-tool interface, or painting the cured composite part to fill in the voids. Each solution involves added time, cost, and/or weight. With a goal of developing manufacturing methods that more efficiently eliminate surface porosity, experimental efforts were made to identify the source of surface porosity and possible mitigation methods. Results confirmed that surface porosity was primarily associated with air that was trapped at the tool-prepreg interface during layup. The magnitude and distribution of surface porosity was affected by multiple parameters, including vacuum hold time, freezer storage and out-time, and material and process modifications that affect air evacuation. The results indicated that prepreg out-time (and thus tack) and vacuum quality were the primary drivers of surface porosity. ❧ Chapters 2 – 4 explores new, high-performance materials for the application of hypervelocity impact shielding for spacecraft. High-velocity impacts with debris are a major cause of concern for spacecraft and satellites. Developing new materials that can protect against these threats, while still providing low-density and sufficient toughness to survive launch loads, is important for future spacecraft design. In chapter 2, hypervelocity impacts are used to estimate the ballistic limit for bulk metallic glass (BMG) and their composites and to investigate spalling behavior. The composites are shown to have excellent combinations of hardness and toughness for use as shields. Chapters 3 and 4 expand on the potential for BMG to be used in high performance, low density structures by pairing it with carbon fiber reinforced polymer (CFRP) composites in a fiber metal laminate (FML) structure. ❧ A first step in this effort is determining best practices to promote adhesion between BMG and CFRP

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Fabrication and analysis of prepregs with discontinuous resin patterning for robust out-of-autoclave manufacturing

PDF

In situ process monitoring for modeling and defect mitigation in composites processing

PDF

Efficient manufacturing and repair of out-of-autoclave prepreg composites

PDF

Material and process development and optimization for efficient manufacturing of polymer composites

PDF

Sustainable manufacturing of out-of-autoclave (OoA) prepreg: processing challenges

PDF

Characterization, process analysis, and recycling of a benzoxazine-epoxy resin for structural composites

PDF

Defect control in vacuum bag only processing of composite prepregs

PDF

Chemical recycling of amine/epoxy composites at atmospheric pressure

PDF

In situ process analysis for defect control during composites manufacturing

PDF

Shock wave response of in situ iron-based metallic glass matrix composites

PDF

Development of composite oriented strand board and structures

PDF

Metallic syntactic foams synthesis, characterization and mechnical properties

PDF

Void evolution in vacuum bag-only prepregs

PDF

In-situ characterization of nanoscale opto-electronic devices through optical spectroscopy and electron microscopy

PDF

Slurry based stereolithography: a solid freeform fabrication method of ceramics and composites

PDF

Printed electronics based on carbon nanotubes and two-dimensional transition metal dichalcogenides

PDF

Design and characterization of flow batteries for large-scale energy storage

PDF

Integration of cost effective luminescent materials into organic light emitting diodes (OLEDs) and molecular approaches to improve the device efficiency

PDF

Plasmonic enhancement of catalysis and solar energy conversion

PDF

The effect of lattice structure and porosity on thermal conductivity of additively-manufactured porous materials

Asset Metadata

Creator Hamill, Lee (author)

Core Title Developing efficient methods for the manufacture and analysis of composites structures

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Materials Science

Publication Date 09/15/2017

Defense Date 08/14/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag adhesion,bulk metallic glass,carbon fiber,composites,eddy current,fiber metal laminates,hypervelocity impact,non destructive testing,OAI-PMH Harvest,prepreg,spacecraft shielding

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Nutt, Steven (committee chair), Centea, Timotei (committee member), Emerson, Jane (committee member), Ravichandran, Jayakanth (committee member)

Creator Email leehamill1@gmail.com,lhamill@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-429325

Unique identifier UC11264284

Identifier etd-HamillLee-5724.pdf (filename),usctheses-c40-429325 (legacy record id)

Legacy Identifier etd-HamillLee-5724.pdf

Dmrecord 429325

Document Type Dissertation

Rights Hamill, Lee

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