on the fabrication, characterization and …...naguib’s lab to do some of the polymer...
TRANSCRIPT
On the Fabrication, Characterization and Mechanical Properties
of Melt-Stretched Stochastic Honeycombs
by
Megan Hostetter
A thesis submitted in conformity with the requirements for the degree of
Doctor of Philosophy
Department of Materials Science and Engineering
University of Toronto
© Copyright by Megan Hostetter (2015)
ii
On the Fabrication, Characterization and Mechanical Properties of
Melt-Stretched Stochastic Honeycombs
Megan Hostetter
Doctor of Philosophy 2015
Department of Materials Science and Engineering
University of Toronto
Abstract
This thesis presents a new type of polypropylene (PP) cellular material fabricated through a simple melt-stretching process.
Stochastic honeycombs have an open cell, random honeycomb structure, with webs oriented perpendicular to built-in skins.
This process has the advantage that, for example, PP pellets can be turned into a sandwich panel in one step. It was
demonstrated that despite the randomness in the web structure, the out-of-plane compressive strength of stochastic
honeycombs was repeatable, and exceeded that of commercial PP foams and was comparable to commercial PP
honeycombs.
The key material properties required to produce an this architecture were shown to be a high melt strength and a high
viscosity, branched polymer. The viscosity was shown to affect the total length of the webs in cross-section and the relative
partitioning of material through the skin, transition region and webs. Web thickness was affected by the areal density of the
polymer during fabrication.
Mechanical testing methods were adapted from ASTM standards for honeycombs, and the fabrication method was
advanced from a manual to a machine controlled process. Stochastic honeycombs were shown to buckle elastically,
plastically, and fracture after the peak strength. Elastic and plastic buckling were dominant at lower densities, and plastic
buckling and fracture at higher densities.
A thin-plate buckling model for the strength of stochastic honeycombs was developed and verified experimentally.
The crystallinity of the polymer affected the tensile strength and stiffness, having a linear effect on the buckling strength.
The architecture was composed of webs bound on both sides and webs bound on one side and free on the other. A greater
fraction of bound webs increased the strength of the structure in the buckling model. A fabrication study showed that melt-
stretching the polymer at higher strain rates increased the connectivity and fraction of bound webs. Additionally, higher
density led to a less connected structure. Energy absorption could be improved if the drop in stress from the peak to valley
strength was minimized. Higher melt-stretching rates during fabrication increased the energy absorbed, and higher viscosity
polymers minimized the drop from peak to valley, also increasing the absorbed energy.
iii
Acknowledgements
I would like to extend my sincerest thanks to Prof. Glenn Hibbard, for his support and guidance throughout this
project. Your insights have been invaluable, and you have consistently pushed me to look deeper and think more
closely about the project, and to constantly re-think my assumptions.
I would also like to thank my committee members, Prof. Mansoor Barati, Prof. Keryn Lian and Prof. Hani
Naguib, for their valuable comments and discussions over the years. Additionally, I was granted access to Prof.
Naguib’s lab to do some of the polymer characterization which would not have otherwise been possible. My
thanks also go out to Dr. Dan Grozea, for his patience and assistance when I used his labs for fabrication and
testing, and to Sal Boccia, for his assistance with the scanning electron microscope and polarized light
microscopy.
My thanks also go out to past and current members of the Cellular Hybrid Materials Research Group, for
their valuable discussions and company over the years: Dr. Brandon Bouwhuis, Dr. Evelyn Ng, Dr. Eral Bele,
Bosco Yu, Dr. Balaji Devatha Venkatesh, Karen Chien, Adam Bird, Adam Yaremko, Ante Laušić and Khaled
Abu Samk. Additionally, I would like to thank the summer students who contributed over the years, especially
Vinson Truong who came back for a second year.
Finally, Vajira, thank you for your never-ending faith in me, as well as your understanding and patience
through this process. Thank you to my brothers, who have always and continue to inspire me to do more and go
farther, just to keep up with both of you. And last, but definitely not least, I thank my parents for teaching me
that with enough hard work, nothing is impossible, and to always keep reaching for more.
iv
Table of Contents
Abstract ......................................................................................................................................................... ii
Acknowledgements .......................................................................................................................................... iii
Table of Contents .............................................................................................................................................. iv
List of Figures.................................................................................................................................................. vii
List of Tables .................................................................................................................................................... xv
List of Symbols and Abbreviations ............................................................................................................... xvii
Chapter 1 Introduction ........................................................................................................................................ 1
1.1 Cellular Materials ............................................................................................................................... 1
1.2 Melt-Stretched Stochastic Honeycombs ............................................................................................ 4
1.3 Overview ............................................................................................................................................ 6
References .............................................................................................................................................. 10
Chapter 2 Background ...................................................................................................................................... 12
2.1 Honeycomb Sandwich Panels .......................................................................................................... 12
2.2 Melt Rheology and Structure of Polypropylene ............................................................................... 15
2.3 Buckling of Thin Plates as it Relates to Honeycombs ..................................................................... 18
2.4 Summary .......................................................................................................................................... 23
References .............................................................................................................................................. 24
Chapter 3 Melt-Stretched Honeycombs ........................................................................................................... 31
Iteration 1: Manual Press ........................................................................................................................ 31
Iteration 2: Fully-Instrumented Process ................................................................................................. 33
References .............................................................................................................................................. 36
Chapter 4 Architecture – Process Relationships in Stochastic Honeycombs ................................................... 37
4.1 Introduction ...................................................................................................................................... 37
4.2 Experimental Methods ..................................................................................................................... 37
4.3 Results and Discussion ..................................................................................................................... 38
4.3.1 Architecture ......................................................................................................................... 38
4.3.2 Fabrication .......................................................................................................................... 42
4.4 Conclusions ...................................................................................................................................... 48
References .............................................................................................................................................. 49
v
Chapter 5 Architectural Characteristics of Stochastic Honeycombs Fabricated from Varying Melt
Strength Polypropylenes ................................................................................................................. 51
5.1 Introduction ...................................................................................................................................... 51
5.2 Experimental Procedures .................................................................................................................. 52
5.3 Results and Discussion ..................................................................................................................... 53
5.3.1 Polymer Characterization .................................................................................................... 53
5.3.2 Architectural Characterization ............................................................................................ 56
5.4 Conclusions ...................................................................................................................................... 62
References .............................................................................................................................................. 63
Chapter 6 Stochastic Honeycomb Sandwich Cores.......................................................................................... 65
6.1 Introduction ...................................................................................................................................... 65
6.2 Experimental .................................................................................................................................... 66
6.3 Cellular Architecture ........................................................................................................................ 67
6.4 Mechanical Properties ...................................................................................................................... 69
6.4.1 Uniaxial Compression ......................................................................................................... 69
6.4.2 Three-Point Bend Testing ................................................................................................... 71
6.5 Conclusions ...................................................................................................................................... 75
References .............................................................................................................................................. 77
Chapter 7 Modeling the Buckling Strength of Polypropylene Stochastic Honeycombs .................................. 78
7.1 Introduction ...................................................................................................................................... 78
7.2 Experimental Details ........................................................................................................................ 78
7.3 Results .............................................................................................................................................. 79
7.3.1 Polymer Characterization .................................................................................................... 79
7.3.2 Honeycomb Architecture .................................................................................................... 81
7.3.3 Mechanical Properties of Stochastic Honeycombs ............................................................. 83
7.4 Discussion ........................................................................................................................................ 84
7.5 Conclusions ...................................................................................................................................... 88
References .............................................................................................................................................. 89
Chapter 8 Post-Peak Collapse and Energy Absorption in Stochastic Honeycombs ......................................... 92
8.1 Introduction ...................................................................................................................................... 92
8.2 Experimental Details ........................................................................................................................ 93
vi
8.3 Results .............................................................................................................................................. 94
8.4 Discussion ...................................................................................................................................... 102
8.5 Conclusions .................................................................................................................................... 107
References ............................................................................................................................................ 108
Chapter 9 Summary and Future Work ............................................................................................................ 110
9.1 Fabrication ...................................................................................................................................... 111
9.2 Internal Architecture ...................................................................................................................... 111
9.3 Mechanical Properties .................................................................................................................... 113
9.4 Future Work ................................................................................................................................... 114
Appendix A Polymer Properties, PP-1 to PP-5 .............................................................................................. 116
References ............................................................................................................................................ 117
Appendix B Melt-Stretched Honeycombs with Various Polymers ................................................................ 118
Appendix C Characterization of Stochastic Honeycomb Sandwich Failure .................................................. 121
C.1 Introduction ................................................................................................................................... 121
C.2 Experimental Details ..................................................................................................................... 122
C.3 Results and Discussion .................................................................................................................. 123
C.4 Conclusions ................................................................................................................................... 133
References ............................................................................................................................................ 134
vii
List of Figures
Figure 1.1 Strength–density property map [4]. .....................................................................................2
Figure 1.2 Honeycombs with varying shape: hexagonal [5] (a), hexagonal and triangular [5] (b), triangular
[5] (c), square [5] (d), circular [6] (e) and Voronoi [7]. .......................................................3
Figure 1.3 A high melt strength PP stochastic honeycomb seen from the side in an unreconstructed X-ray
image. The image shows the as-fabricated skins and the core. ............................................4
Figure 1.4 3D reconstructions of a melt-stretched stochastic honeycomb fabricated with high melt strength
PP (a) and linear PP (b), with the as-fabricated skins edited out. The scale bar in (b) is 5 mm and
applies to both images. .........................................................................................................5
Figure 1.5 Two space-filling model plots of the high melt strength PP sample shown in Figure 1.4, from the
top (a) and a sectioned piece from the centre (b). An example of the buttressing structure is
circled in (a), and a hole in the web indicating an archway is shown with the arrow in (b). The
scale bar in (b) is 5 mm and applies to both images. ...........................................................6
Figure 1.6 Flow chart detailing the four key topics in this thesis and the contributions made in each. The
chapter that these contributions are discussed is indicated through the numbers on the left.
............................................................................................................................................... 9
Figure 2.1 Cellular material sandwich panels, with foam, truss and honeycomb core materials, top to
bottom [8]. .........................................................................................................................13
Figure 2.2 Sandwich panel failure mechanisms [23] ..........................................................................13
Figure 2.3 Closed [34] (a) and open [35] (b) cell foams. ....................................................................14
Figure 2.4 In-plane (x-y) and out-of-plane (z) directions on a honeycomb sandwich panel [10]. ......14
Figure 2.5 2D Voronoi honeycomb with a random but even spacing of points [36] (a) and a completely
random spacing of points [37] (b). .....................................................................................15
Figure 2.6 Schematic of the effect of branching and cross-linking on the tensile strength of PP [100].
...........................................................................................................................................17
Figure 2.7 Schematic of a cross-link [72] (a) and long chain branching [100] (b) in a polymer chain, where
Mbp is the average molecular weight between branch points and MLCB is the average molecular
weight of the branches. ......................................................................................................18
Figure 2.8 Strain hardening (a), strain softening (b) and perfectly plastic (c) stress-strain curves [113].
...........................................................................................................................................19
Figure 2.9 Buckling modes of a square honeycomb in out-of-plane compression [114]. ...................19
Figure 2.10 A schematic of a thin plate, with height a and width b, and thickness t in the x-direction.
...........................................................................................................................................20
Figure 2.11 A thin plate under compression with m = 5 half-waves [121]. ..........................................21
viii
Figure 3.1 The inside of the oven used for the manual press, showing the upper and lower platens, the
polymer on the lower platen, and the thermocouple leads (highlighted to increase visibility) (a),
and the manual press locked with the bars on either side in the stretched position as the sample
cools, with the upper platen visible (b). .............................................................................32
Figure 3.2 A stochastic honeycomb fabricated with a high melt strength PP (approximately 20 cm x 22 cm).
The circles indicate areas with flaws in the skin, and the arrow points to a corner where there
was not enough polymer and an interconnected structure was not formed. ......................33
Figure 3.3 A Shimadzu AG-I Universal Testing Machine, and the furnace, a Shimadzu TCE-N300
Thermostatic Chamber (a), with the aluminium platens and thermocouple leads within the
furnace (b). Once the thermocouple leads read 180 oC, the upper platen was lowered and the
furnace moved away (c). ....................................................................................................34
Figure 3.4 Stress (σ) as a function of time (t) during fabrication from the Shimadzu AG-I UTM. ....34
Figure 4.1 Examples of the widely different architectures formed over the fabrication space considered,
with melt-stretch rate = 10 mm min-1
and areal density ρA = 0.09 g cm-2
(a) and ρA = 0.23 g
cm-2
(b), and ρA = 0.15 g cm-2
and = 1 mm min-1
(c), and 100 mm min-1
(d). Scale bar indicates
a length of 20 mm. .............................................................................................................39
Figure 4.2 Process Map illustrating the range of fabrication space (melt-stretch rate, , and areal density,
ρA) considered. The points labelled a-d correspond to the 2D projections seen in Figure 4.1.
...........................................................................................................................................39
Figure 4.3 The 2D projections of the webs for five samples fabricated at = 10 mm min-1
and with ρA =
0.10 g cm-2
(a), 0.13 g cm-2
(b), 0.16 g cm-2
(c), 0.19 g cm-2
(d), and 0.23 g cm-2
(e). Scale bar
indicates a length of 20 mm. ..............................................................................................40
Figure 4.4 The 2D projections of the webs for five samples fabricated with ρA = 0.13 g cm-2
and = 1 mm
min-1
(a), 10 mm min-1
(b), 30 mm min-1
(c), 60 mm min-1
(d), and 100 mm min-1
(e). Scale bar
indicates a length of 20 mm. ..............................................................................................40
Figure 4.5 The number of nodes, N (a), and the total length of webs, L (b), plotted against ρA for = 10 mm
min-1
and 100 mm min-1
. ....................................................................................................40
Figure 4.6 The number of nodes, N, plotted against the total web length, L for melt-stretch rates of 10 mm
min-1
and 100 mm min-1
. ....................................................................................................41
Figure 4.7 2D projections for samples having the same areal density and N/L values (ρA = 0.21 g cm-2
and
N/L = 1.18 cm-1
) but with χ2D = -5 (a) and -25 (b). These samples were fabricated at = 10 mm
min-1
and = 100 mm min-1
respectively. Scale bar indicates a length of 20 mm. ...........41
Figure 4.8 Process Map of melt-stretch rate and areal density ρA, with all fabricated samples labelled by
the corresponding Euler number (χ2D), and overlaid with contour lines illustrating the trend in
N/L. ....................................................................................................................................42
ix
Figure 4.9 Full fabrication curve (σ, stress, and S, extension) showing all four stages, with compressive
stress expressed as negative and tensile stress as positive, with permanent plastic extension
beginning at the origin. The locations marked (i – iv) correspond to the images on the right. At
the beginning of fabrication, the melt (with thickness tblank, and diameter dblank) (i) is compressed
between Al platens in the furnace. In air, the melt (with tmelt, dfinal) is stretched uniaxially (ii), and
the stress increases to a maximum (iii). At the maximum, air ingress begins as the melt fractures,
creating the webs of the stochastic honeycomb (iv). The melt in the webs stretches and some of
the webs fracture, creating partial webs and archways until the sample reaches its final height (h,
dfinal) and cools in air. .........................................................................................................43
Figure 4.10 Stages II, III and IV on the fabrication curve for two samples fabricated with the same
elongation rate ( = 10 mm min-1
) but different densities, ρA = 0.09 g cm-2
and 0.23 g cm-2
(a),
and two samples fabricated with the same density (ρA = 0.14 g cm-2
) and two elongation rates,
= 1 mm min-1
and 100 mm min-1
(b). These curves correspond to the images in Figure 4.1.
...........................................................................................................................................44
Figure 4.11 Peak fabrication stress (σpk) is plotted as a function of areal density (ρA) for = 10 mm min-1
and
100 mm min-1
. ....................................................................................................................44
Figure 4.12 Peak fabrication stress, σpk, and apparent melt stiffness, Emelt (a), and extensional stress, σext (b),
pare plotted as a function of strain rate ( ) for samples with ρA = 0.13 g cm-2
. .................45
Figure 4.13 Apparent elongational viscosity, , is plotted against time, t, for five samples with ρA = 0.13 g
cm-2
and increasing elongation rate, (a), and the peak of the elongational viscosity with time,
, is plotted as a function of strain rate, , on a log–log scale for all samples. ...........46
Figure 4.14 The Process Map plotting elongation rate, , against areal density, ρA, showing the samples
fabricated, with variation in N/L shown in the data points. Contour lines illustrating the trends in
σpk (a), Emelt (b) and (c) are superimposed on top................................................47
Figure 5.1 Parallel-plate rheometry results, giving the viscosity (η) as a function of frequency (ω).
...........................................................................................................................................53
Figure 5.2 Storage modulus (G’) and loss modulus (G") plotted as a function of frequency (ω) for PP-2 and
PP-5. ...................................................................................................................................54
Figure 5.3 The two components of complex viscosity (η' and η") from parallel plate rheology tests: the
curves for PP-1, PP-2, PP-3 and PP-4 each exhibit a linear portion followed by an increase in
slope. This implies that long-chain branching is present in the polymers. PP-5 exhibits a semi-
circular curve, indicating a linear polymer [1, 17]. ............................................................55
Figure 5.4 Mid-height cross-sections for PP-1 (a), PP-2 (b), PP-3 (c), PP-4 (d) and PP-5 (e), each having a
nominal density of ~ 20%. The scale bar in part (e) represents a length of 5 mm and applies to
each figure. ........................................................................................................................56
Figure 5.5 Cross-sectional area fraction (Ac) as a function of position (P) in the sample (from bottom to top)
corresponding to the samples in Figure 5.4. ......................................................................57
x
Figure 5.6 Cross-sectional web length per unit area, lc, (at a nominal relative density of ~ 20%) plotted as a
function of zero-shear viscosity, ηo (a) and crossover frequency, ωc (b). ..........................58
Figure 5.7 Mid-height cross-sections of PP-3 at = 10% (a), 13% (b), 16% (c), and 18% (d). The scale bar
in (d) is 5 mm and applies to all images. ...........................................................................58
Figure 5.8 Mid-height cross-sectional area fraction (Ac) of PP-2 and PP-3 as a function of areal density (ρA).
...........................................................................................................................................59
Figure 5.9 Cross-sectional area fraction (Ac) plotted as a function of position, P, for PP-3 at = 13%. The
central third of the sample was used to determine Ac and lc. .............................................59
Figure 5.10 Cross-sectional area slices from positions a to e in Figure 5.9. The scale bar in (e) is 5 mm and
applies to all figures. ..........................................................................................................59
Figure 5.11 Archway defect in PP-3 ( = 13%) shown in a 3D compilation image (a), with horizontal lines
marking the position of the cross-sections. The cross-section indicated by the top line is shown
in (b), the centre line is (c) and the lowest line is (d). The scale bar is 2 mm and applies to (b), (c)
and (d). ...............................................................................................................................61
Figure 5.12 Buttress defect in PP-1 ( = 11%) shown in 3D compilation image (a), with horizontal lines
marking the position of the cross-sections. The cross-section indicated by the top line is shown
in (b), the centre line is (c) and the lowest line is (d). The scale bar is 2 mm and applies to (b), (c)
and (d). ...............................................................................................................................61
Figure 5.13 Void defects in PP-3 ( = 13%) shown in a 3D compilation image (a), and in three separate
cross-sectional slices (b), (c) and (d). The scale bar is 2 mm and applies to (b), (c) and (d).
...........................................................................................................................................61
Figure 6.1 Top view of an as-fabricated stochastic honeycomb core showing the cellular web structure (a).
Also shown is a three point bend coupon 45 mm x 200 mm, 20 mm core thickness, reinforced
with conventional polypropylene face sheets and an as-fabricated stochastic honeycomb core
compression 30 mm x 30 mm test coupon, 20 mm core thickness (b). .............................65
Figure 6.2 Variation in web thickness, tweb (a) and web length, b (b) over all densities. ....................68
Figure 6.3 SEM images showing the internal structure of the stochastic honeycomb architecture. An
incomplete web with a gap extending to the skin is shown (a), along with a mid-height gap
indicated by the arrow (b), and the complex buttressing of the webs (c). .........................68
Figure 6.4 Typical uniaxial compression curves for stochastic honeycomb cores. ............................70
Figure 6.5 SEM images of the web structure in a ρcore = 11.8% sample (a) and after preloading to strains of
ε ≈ 0.05 (b), ε ≈ 0.25 (c) and ε ≈ 0.35 (d). .........................................................................71
Figure 6.6 Stochastic honeycomb compressive strength, σp (a) and stiffness, E (b) over a range of core
densities. ............................................................................................................................71
Figure 6.7 Typical load-deflection curves for as-prepared stochastic honeycomb in three point bending.
...........................................................................................................................................72
xi
Figure 6.8 Flexural strength, σfl (a) and flexural stiffness, D (b) of stochastic honeycomb cores in three-
point bending. ....................................................................................................................72
Figure 6.9 Failure map plotting regions of dominant sandwich panel failure mechanism. The dominant
failure mechanism at low sheet thickness was face sheet wrinkling, core shearing dominated at
intermediate thickness, and face sheet yielding and fracture dominated at the greatest thickness.
...........................................................................................................................................73
Figure 6.10 Typical three-point bending load–deflection curves illustrating face sheet yield and fracture (tf =
1.75 mm, a), core shearing (tf = 0.94 mm, b) and face sheet wrinkling, (tf = 0.73 mm, c).
...........................................................................................................................................74
Figure 7.1 Heat flow (ΔH/m) plotted as a function of temperature (T) for the cool–heat cycle for PP-1, PP-
2, PP-3 and PP-4 (a), and the two heating curves (for the as-fabricated webs and the base resin)
for PP-1 and PP-4 (b). ........................................................................................................80
Figure 7.2 Stress (σ) – strain (ε) curves for tensile tests on PP-1, PP-2, PP-3 and PP-4, pulled at a rate of 1
mm min-1
. ...........................................................................................................................81
Figure 7.3 Skin thickness as a percentage of height ( ) plotted as a function of total relative density ( ).
...........................................................................................................................................82
Figure 7.4 3D reconstructions of a PP-2 stochastic honeycomb at 9.5% core density. Archways are webs
that do not extend through the entire height of the sample (shown with an arrow), and buttresses
form at the top and bottom the sample where they would join the skin (circled). Each scale bar
represents 5 mm. The images show a top view of the core (a), a side view (b), and views from
the mid-height of the complex structure (c,d). ...................................................................82
Figure 7.5 Stress (σ) – strain (ε) curves for the out-of-plane compression of stochastic honeycombs
fabricated with PP-1, PP-2, PP-3 and PP-4 with core densities (ρcore) of 11% (a), and the peak
strength (σp) of the stochastic honeycombs as a function of ρcore (b). ................................83
Figure 7.6 Surface area over volume (AS/V) as a function of core density (ρcore) for PP-1, PP-2, PP-3 and
PP-4. ...................................................................................................................................85
Figure 7.7 Predicted strength (σcr) from the composite buckling model plotted as a function of web
thickness (tweb) for PP-1 and PP-4, overlaid with the experimental data (σp) plotted as a function
of effective web thickness (teff). .........................................................................................87
Figure 7.8 Normalized web thickness (t/a) plotted as a function of the fraction of bound webs (f) for various
specific strengths (σ/ρ). The core density (ρcore) fraction for a given web thickness is also
displayed. ...........................................................................................................................87
Figure 8.1 3D reconstructions of samples A, B, C and D (a–d, respectively) with core relative densities as
shown in Table 8.1. ............................................................................................................94
Figure 8.2 Successive compressive stress–strain curves for samples A–D with core relative densities as
shown in Table 8.1. The samples were characterized through X-ray tomography in the as-
xii
fabricated condition (εo = 0), after loading to the peak (εp), the beginning of the valley (εv) and
just before the onset of densification (εd). The samples were unloaded and scanned, and then re-
loaded. The stress–strain curves were then stitched together. ...........................................95
Figure 8.3 The mid-height cross-section for sample C (a) and schematics of bound–bound, B (b) and
bound–free, F (c) webs with the length, b and the height, a as shown ..............................96
Figure 8.4 Web CB-1, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve
(εp), at the beginning of the valley (εv) and the end of the valley before densification (εd). The
first two images at each strain show the web from either side, and the third is a through-thickness
cross-section taken in the centre of the web. .....................................................................97
Figure 8.5 Web CB-2, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve
(εp), at the beginning of the valley (εv) and the end of the valley before densification (εd). The
first two images at each strain show the web from either side, and the third is a through-thickness
cross-section taken in the centre of the web. .....................................................................97
Figure 8.6 Web CB-3, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve
(εp), at the beginning of the valley (εv) and the end of the valley before densification (εd). The
first two images at each strain show the web from either side, and the third is a through-thickness
cross-section taken in thirds along the lower image of the web. .......................................99
Figure 8.7 Web CF-1, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve
(εp), at the beginning of the valley (εv) and the end of the valley before densification (εd). The
first two images at each strain show the web from either side, and the third is a through-thickness
cross-section taken in the centre of the web. .....................................................................99
Figure 8.8 Web CF-2, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve
(εp), at the beginning of the valley (εv) and the end of the valley before densification (εd). The
first two images at each strain show the web from either side, and the third is a through-thickness
cross-section taken in the centre of the web. ...................................................................100
Figure 8.9 Web CF-3, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve
(εp), at the beginning of the valley (εv) and the end of the valley before densification (εd). The
first two images at each strain show the web from either side, and the third is a through-thickness
cross-section taken in the centre of the web. ...................................................................100
Figure 8.10 Webs AB-1 (a), BB-1 (b), AF-1 (c), DF-1 (d), and DB-2 (e), in the as-fabricated state (εo) and at the
valley (εv). These webs demonstrate the three collapse mechanisms: elastic buckling, plastic
buckling near the skin, plastic buckling in the central portion of the web, plastic buckling with
horizontal tears, and plastic buckling with horizontal and vertical tears, from a – e respectively.
.........................................................................................................................................101
Figure 8.11 Energy absorbed, J, calculated up to εf = 0.4 (a), εf = (b), εf = εd (c), and εf = εη (d), plotted as
a function of ρcore..............................................................................................................104
Figure 8.12 Percentage drop (D%) from σp to σv a and energy absorbed, J, up until εf = 0.4 b as a function of
ρcore, for PP-1, PP-2, PP-3 and PP-4. ...............................................................................106
xiii
Figure 8.13 Specific strength (σp/ρ) plotted against energy absorbed up until εf = 0.4 for PP-1, PP-2, PP-3 and
PP-4. .................................................................................................................................107
Figure B.1 Zero-shear viscosity (ηo) plotted as a function of melt flow index (MFI) for a number of PP, PE,
and copolymers of PP and PE (CP), along with side images of representative samples.
.........................................................................................................................................118
Figure B.2 Stochastic honeycombs fabricated with HMS PP (a), linear iPP (b) and LDPE (c). The scale bar
on (c) is 3 cm and applies to all three images. .................................................................119
Figure B.3 SEM images of stochastic honeycombs fabricated with HMS PP (a,b), L-iPP (c,d) and LDPE
(e,f), perpendicular to the webs (top row), and at a tilt (bottom row). The HMS PP has buttresses
that support the webs at the transition between the webs and the skin, while the L-iPP and the
LDPE have smooth transition regions with no buttressing. .............................................120
Figure C.1 3D image reconstructed from a micro-CT scan, with a sample size of 40 mm 40 mm 15 mm
(a) and the web structure as seen from the side (b), with the as-fabricated face sheets removed
digitally in order to more clearly reveal the internal web structure. The scale bars are 5 mm each.
.........................................................................................................................................122
Figure C.2 Micro-CT scans of the mid-height cross-sectional slice (sample size 40 mm 40 mm 15 mm)
for the LD sample (a) and the HD sample (b). The scale bar is 5 mm and applies to both figures.
The corresponding area fraction, Ac, as a function of normalized position in the sample, P, from
bottom to top is given in (c). ............................................................................................123
Figure C.3 Partial web in the LD as-fabricated structure. The scale bar is 2 mm. ............................124
Figure C.4 Schematic of a web bound on one side and free on the other. The length, b, height, a, and
thickness, tweb, are indicated. ............................................................................................125
Figure C.5 Uniaxial compression stress–strain curve for a ρcore = 13% stochastic honeycomb. .......125
Figure C.6 Uniaxial compression stress-strain curve for LD (ρcore = 11%) and HD (ρcore = 18%) stochastic
honeycombs to the valley (ε = 0.20). ...............................................................................125
Figure C.7 As-fabricated LD web bounded on each side (a), and the vertical tear created by the horizontal
tensile strain from the compression of the bounding webs (b). The scale bar is 2 mm and applies
to both figures. .................................................................................................................126
Figure C.8 LD-1 as-fabricated from the side (a), straight on (b), in vertical cross-section (c), and partially
compressed from the side (d), straight on (e), and in vertical cross-section (f). The scale bar is
2.5 mm and applies to all the figures. ..............................................................................127
Figure C.9 Thickness profile of LD-1 and the vertical cross-section for comparison. The dashed line marks
the point of initiation of instability. .................................................................................127
xiv
Figure C.10 LD-2 as-fabricated from the side (a), straight on (b), in vertical cross-section (c), and partially
compressed from the side (d), straight on (e), and in vertical cross-section (f). The scale bar is
2.5 mm and applies to all the figures. ..............................................................................129
Figure C.11 Thickness profile of LD-2 and the vertical cross-section for comparison. The dashed lines mark
the locations of initiation of instability. ...........................................................................129
Figure C.12 HD-1 as-fabricated from the side (a), straight on (b), in vertical cross-section (c), and partially
compressed from the side (d), straight on (e), and in vertical cross-section (f). The scale bar is
2.5 mm and applies to all the figures. ..............................................................................130
Figure C.13 Thickness profile of HD-1 and the vertical cross-section for comparison. The dashed line marks
the location of initiation of instability. .............................................................................130
Figure C.14 HD-2 as-fabricated from the side (a), from the opposite side (b), straight on (c), in vertical cross-
section (d), and partially compressed from the side (e), from the opposite side (f), straight on (g),
and in vertical cross-section (h). The scale bar is 2.5 mm and applies to all the figures.
.........................................................................................................................................131
Figure C.15 HD-2 buttress as-fabricated (a), in vertical cross-section (b), and partially compressed (c), in
vertical cross-section (d). The scale bar is 1 mm and applies to all the figures. ..............131
Figure C.16 Thickness profile of HD-2 and the buttress (inset) with vertical cross-sections for comparison.
The dashed lines mark the locations of initiation of instability. ......................................132
xv
List of Tables
Table 3.1 The program for the automated process for fabrication of stochastic honeycombs, where Fapp is the
chosen applied force, is the chosen melt stretch rate, and h is the chosen final height of the
sample. ....................................................................................................................................34
Table 4.1 Temperature of the platens during melt elongation at the beginning and end of each stage of
fabrication. ..............................................................................................................................44
Table 5.1 Polymer properties for four high melt strength PP (PP-1 to PP-4) and one conventional (linear) PP
(PP-5). .....................................................................................................................................52
Table 5.2 Cross-sectional area fraction (Ac) and web length per unit area (lc) over the central third of the
stochastic honeycomb samples ( ~ 20%). .............................................................................57
Table 6.1 Average web thickness (tweb) and length (b) for four ranges of core relative densities (ρcore).67
Table 7.1 Zero-shear viscosity (ηo) as measured using parallel-plate rheology tests at 1.1Tm [1]. .........79
Table 7.2 Melting and crystallization temperatures (Tm, Tc) and the crystallinity (xc) of the base polymer and
the webs for all four PPs from differential scanning calorimetry. ..........................................80
Table 7.3 Tensile strength (σTS) and strain at break (εu) for all four PPs. ................................................81
Table 7.4 Buckling model parameters determined from the X-ray images of the webs, and the tangent
modulus, Et, as determined from the tensile testing. ...............................................................87
Table 8.1 Relative density (ρcore), cross-sectional area fraction (Ac), peak strength (σp) and valley strength (σv)
for samples A–D. ....................................................................................................................94
Table 8.2 Dimensions of various webs at mid-height from sample C. ...................................................96
Table 8.3 The percentage of webs from each sample type that exhibited each mechanism of deformation at the
valley. Reported values are based on the characterization of at least 35 webs in each sample.102
Table 8.4 Final strain values as a function of density, for each definition of strain. The references indicate the
study or studies where this integration range was selected. ..................................................103
Table 8.5 Zero-shear viscosity (ηo), melting temperature (Tm), crystallinity (xc) and tensile strength (σTS) of
PP-1, PP-2, PP-3 and PP-4 [8, 9], and peak and valley strength (σp, σv), and the energy absorbed (J)
up until εf = 0.4 at ρcore = 10%. ..............................................................................................106
Table C.1 Compressive properties for the low density and high density stochastic honeycombs.........125
xvi
Table C.2 LD-1 dimensions at failure initiation point. ..........................................................................127
Table C.3 LD-2 dimensions at failure initiation points. ........................................................................129
Table C.4 HD-1 dimensions at failure initiation point. .........................................................................130
Table C.5 HD-2 dimensions at failure initiation points. ........................................................................132
Table C.6 Thickness and aspect ratio correlated with failure mechanism .............................................132
xvii
List of Symbols and Abbreviations
A, B, C Sample designations
Ac Average normalized cross-sectional area over the central third of the
honeycomb height
Acells mm2 Area of apparently closed cells in a 2D image of a stochastic honeycomb
As mm2 Surface area
Atotal mm2 Total area of a stochastic honeycomb sample
Aw mm2 Area of a single web in a given cross-section
Awebs mm2 Area of the webs in a 2D image of stochastic honeycombs
a mm Height of the core of the stochastic honeycomb
B mm Mean breadth
B Boundary condition: web is bound on either side
b mm Length of web at a given cross-section
bB mm Average length of bound webs
bF mm Average length of bound-free webs
CP Copolymer of polypropylene and polyethylene
D N mm2 Flexural stiffness
D% % Percentage drop from peak to valley in compression
d mm Diameter
dblank mm Diameter of a polypropylene blank
dfinal mm Diameter of the final stochastic honeycomb
DSC Differential Scanning Calorimetry/Calorimeter
E MPa Elastic modulus
Emelt MPa Melt stiffness
Es MPa Elastic modulus of the solid material
Et MPa Tangent modulus
F N Applied load
F Boundary condition: web is bound on one side and free on the other
Fapp N Applied compressive force during fabrication
FEA Finite Element Analysis
f Length fraction of bound webs
G* Pa Complex shear modulus
G' Pa Storage modulus
xviii
G" Pa Loss modulus
Gc Pa Crossover modulus
HD High density
HDPE High density polyethylene
HMS High Melt Strength
ΔH W Enthalpy
Δhm J g-1
Specific enthalpy of fusion
Δ J g
-1 Specific enthalpy of fusion of a crystalline polymer
h mm Height of stochastic honeycomb (total thickness)
J J cm-3
Energy absorbed during compression
K Buckling coefficient
KB Buckling coefficient for bound webs
KF Buckling coefficient for bound-free webs
L cm Total length of webs at a given cross-section
l mm Length
lc cm-1
Average length of webs normalized by the cross-sectional area over the central
third of the honeycomb height
LCB Long-Chain Branch
LD Low density
LDPE Low density polyethylene
Ls mm Span length
m Number of half-waves in a thin plate under compression
Mbp g mol-1
Average molecular weight between branch points
MFI g/10 min Melt Flow Index
MLCB g mol-1
Average molecular weight of branches
mSHC g Mass of the stochastic honeycomb
Mw g mol-1
Weight-average molecular weight
MWD Molecular Weight Distribution
N Number of nodes
P mm mm-1
Normalized vertical position in the web
PDI Polydispersity index
PE Polyethylene
PP Polypropylene
PP-1 Daploy WB135 HMS polypropylene
xix
PP-2 Daploy WB140 HMS polypropylene
PP-3 Daploy WB180 HMS polypropylene
PP-4 Daploy WB260 HMS polypropylene
PP-5 Accucomp HP0306L polypropylene
PS Polystyrene
PU Polyurethane
Pw mm Perimeter of a single web at a given cross-section
Pwebs % Percentage of webs
S mm Elongation during fabrication
mm min-1
Elongation or melt-stretch rate during fabrication
s mm Deflection
SEM
Scanning Electron Microscopy/Microscope
T oC Temperature
Tc oC Crystallization temperature
Tm oC Melting temperature
t s Time
tblank mm Thickness of a polypropylene blank
teff mm Effective web thickness
tf mm Thickness of commercial polypropylene sheets
tmelt mm Thickness of a polypropylene blank after force has been applied
tplate Mm Thickness of a plate
tskin mm Average skin thickness
% Skin thickness as a percentage of height
tweb mm Thickness of webs
UTM Universal Testing Machine
V cm3 Volume
w mm Width of square sample
xc % Percent crystallinity
δ mm Deflection
ε mm mm-1
Strain
ε s-1
Strain rate
ε0 mm mm-1
Zero strain
ε mm mm
-1 Strain where stress is equal to twice the peak stress
xx
εd mm mm-1
Strain at the onset of densification
εF mm mm-1
Strain at failure in uniaxial extension of a polymer melt
εf mm mm-1
Arbitrary final strain
εp mm mm-1
Strain at the peak of the stress-strain curve
εU mm mm-1
Strain at break during tensile test
εv mm mm-1
Strain at the beginning of the valley on the stress-strain curve
εη mm mm-1
Strain at the maximum efficiency
η Pa s Viscosity
η' Pa s Real component of complex viscosity
η" Pa s Imaginary component of complex viscosity
ηo Pa s Zero-shear viscosity
ηE Pa s Elongational viscosity
η Pa s Apparent elongational viscosity
η Pa s Maximum apparent elongational viscosity
η(ε) Efficiency of energy absorption
λ s Characteristic relaxation time
ν Poisson's ratio
ξ Area fraction of apparently closed cells
ρ kg m-3
or g cm-3
Density
ρ % Relative density of the stochastic honeycomb
ρA g cm-2
Areal density of a stochastic honeycomb
ρA,blank g cm-2
Areal density of a polypropylene blank
ρ
% Relative density of the stochastic honeycomb core
ρPP g cm-3
Density of polypropylene
MPa Stress
cr MPa Critical buckling strength from thin-plate model
ext kPa Plateau extension stress during fabrication
fl MPa Flexural strength
p MPa Peak strength during compression
pk MPa Peak tensile stress during fabrication
pl MPa Plateau strength of a foam under compression
TS MPa Tensile strength
v MPa Valley strength during compression
xxi
τrelax kPa s-1
Effective relaxation rate of polypropylene in the melt
τmelt kPa s-1
Relaxation rate of polypropylene melt during melt stretching in fabrication
φ Aspect ratio of wall height to wall width, a/b
χ, χ2D, χ3D Euler number, in 2D and 3D
ω rad s-1
Angular frequency
ωc rad s-1
Crossover angular frequency
1
Chapter 1
Introduction
1.1 Cellular Materials
Cellular materials are hybrid structures consisting of a solid material and open space. The solid material can be
partitioned into a wide variety of architectures, from a random foam to a two-dimensional honeycomb to an
ordered three-dimensional truss. Using various geometrical and topological arrangements of the solid material,
spaces on a strength–density property map can be reached that were inaccessible otherwise (Figure 1.1). The
goal is to create materials with higher strength and lower density, filling in the upper left side of the property
map. This gap has an upper bound that is defined by the maximum theoretical strength attainable. In the diagram
shown in Figure 1.1, this limit is based on the strength required to break covalent bonds, which in turn is related
to the bond stiffness [1, 2].
Foams are considered 3D cellular materials because the structure varies in all 3 dimensions. In contrast,
honeycombs can be defined by their 2D projection, since along the perpendicular axis the structure is constant.
Variations in honeycomb topology are shown in Figure 1.2, where it can be seen that honeycombs can be
composed of regular shapes (be it hexagonal, triangular, square or rectangular, circular, or a combination of
these), or irregular shapes, as in the Voronoi honeycomb.
The relative density of cellular materials composed of one material and air is defined as the density of the
cellular material normalized by the density of the parent material ( ). This produces a relative density that is
generally less than 0.20. The mechanical properties, specifically the stiffness and strength, of cellular materials
are thus often reported as a function of this relative density. Similarly, the strength and stiffness under
compression can be normalized by the yield strength (σy) and elastic modulus (Es) of the parent material,
respectively, to yield the specific strength and stiffness of the cellular material. Although foams are 3-
dimensional cellular materials, they are generally considered isotropic in each dimension, unlike honeycombs,
which have very different properties in-plane and out-of-plane. The specific plastic collapse strength and specific
stiffness of foams has been found to scale with relative density as [3]:
(1.1)
2
(1.2)
In contrast, the out-of-plane mechanical properties of honeycombs scale with relative density as [3]:
(1.3)
(1.4)
For foams, these are based on idealized unit cells, while for regular honeycombs, the relationships can be
calculated exactly.
Figure 1.1 Strength–density property map [4].
3
Figure 1.2 Honeycombs with varying shape: hexagonal [5] (a), hexagonal and triangular [5] (b), triangular [5] (c), square [5] (d),
circular [6] (e) and Voronoi [7].
Polymeric cellular materials have enabled important advances in the aerospace, automotive and packaging
industries [8, 9], and have become increasingly important for high-strength low-weight applications over the last
few decades [3]. The most common uses of foams and honeycombs are as the core in sandwich panels, where
two relatively thin and stiff sheets are separated by a thicker low-density cellular core [10]. Historically, polymer
core sandwich panels have primarily had metal face sheets [3], but over the past 20 years it has become more
and more common for the face sheets to be a polymer matrix composite adhesively bonded to the core [11, 12].
There are many types of adhesives, the most common being epoxies or other thermosets [11, 13].
Thermoplastic foams are typically produced by using blowing agents such as carbon dioxide or nitrogen to
create bubbles in the melt [3, 14], or by introduction of volatile compounds (either low melting point liquids or
additives which react and/or decompose) that will form vapour bubbles upon heating [3], and then injection
moulding the final product [14]. Thermoplastic honeycombs can be fabricated through a number of methods,
including extruding sheets of plastic and cutting the sheets into blocks which are then thermally welded or
bonded in narrow strips and expanded into a honeycomb shape [15], or extruded sheets cut into strips and slotted
together [15], or through extrusion into a complex mould [16]. Each of these processes require multiple steps to
create a sandwich panel.
Various methods have been used to improve the mechanical properties of polymer foams and honeycombs
as the core for sandwich panels. Fibre addition has been used to improve the properties of bulk polymers [17]
and has recently been used to improve the properties of honeycomb sandwich structures as well [18]. Clay
nanocomposites are also being used to improve the properties of polymer foams [19]. While these methods can
4
improve the mechanical properties of the cellular material, they typically do so at the expense of recyclability
[20]. The greatest potential for recycling sandwich materials is if a single material could be used for the core,
face sheets and adhesive layers [21, 22]. This would allow the entire structure to be recycled without the need for
additional (and costly) steps to separate the different materials.
1.2 Melt-Stretched Stochastic Honeycombs
Melt-stretched stochastic honeycombs are a new type of cellular material, and are so named due to the
fabrication method and the resulting structure. They are fabricated through uniaxially stretching a molten
polymer between metal platens. This creates a complete sandwich structure, shown in an X-ray image from the
side in Figure 1.3. The as-fabricated skins, which form on the metal platens, create in-situ face sheets for
sandwich panels. Additionally, as the melt is stretched, the polymer is redistributed and air enters the structure,
creating an interconnected set of webs. These webs are analogous to the walls in a honeycomb, and are parallel
to the direction of melt-stretching. The structure is relatively consistent through the cross-section, and as such is
almost a 2D cellular material, like a honeycomb. However, while the webs are interconnected, there are no
closed cells, making this structure more like a Voronoi honeycomb with some missing walls (Figure 1.4).
Finally, the term "stochastic" refers to an element of randomness in the structure, as the melt-stretching produces
a different web pattern in every sample.
The structure does not consist only of webs perpendicular to the skin. Figure 1.4 presents 3D
reconstructions of two stochastic honeycombs, one fabricated with high melt strength (HMS) polypropylene (PP)
and one with linear PP, with the built-in skin layers digitally removed so that the core structure is visible. As can
be seen in Figure 1.4a, near the skin, buttresses appear on either side of some HMS PP webs, but are not present
in the linear PP structure (Figure 1.4b). Also, a significant fraction of the webs do not extend entirely through the
Figure 1.3 A high melt strength PP stochastic honeycomb seen from the side in an unreconstructed X-ray image. The image shows
the as-fabricated skins and the core.
5
Figure 1.4 3D reconstructions of a melt-stretched stochastic honeycomb fabricated with high melt strength PP (a) and linear PP (b),
with the as-fabricated skins edited out. The scale bar in (b) is 5 mm and applies to both images.
structure, but form archways. These are also not present in the linear PP structure. As will be seen in Chapters 4
and 5, the processing conditions and polymer type have a significant effect on the resultant architecture.
Other architectural features are more clearly seen by plotting the open space instead of the solid material.
Figure 1.5 shows part of the high melt strength PP stochastic honeycomb in an inverted space-filling model (the
open area is shown in white and the polymer in the webs is empty space). The buttressing can be seen clearly in
the lower left corner of Figure 1.5a (circled), where the space-filled pore has two indents. The space-filling
models clearly illustrate the range of cell geometries formed. Additionally, an opening in the webs is indicated
with an arrow in Figure 1.5b, illustrating the connectivity of the pores. During fabrication, air enters the structure
from the outer perimeter, creating a continuous pathway through the structure. This leads to an open-cell
honeycomb architecture.
In this thesis, stochastic honeycombs were fabricated using five PPs with variations in molecular structure
and melt and bulk properties to determine the effect of polymer characteristics on the internal architecture and
the mechanical properties of the cellular material. There was a factor of six difference in the melt flow index
from 2.1 to 12 g/10 min (a detailed summary of the PP properties are given in Appendix A). Four branched
(HMS) PPs and one isotactic linear PP were chosen to provide this variation, with the branched PPs having
branch densities on the order of ~ 0.2/1000 C atoms [23]). Other polymers were also initially investigated, and
the early fabrication results for these polymers are summarized in Appendix B.
6
Figure 1.5 Two space-filling model plots of the high melt strength PP sample shown in Figure 1.4, from the top (a) and a sectioned
piece from the centre (b). An example of the buttressing structure is circled in (a), and a hole in the web indicating an
archway is shown with the arrow in (b). The scale bar in (b) is 5 mm and applies to both images.
As with foams and honeycombs, the density of the structure affects the mechanical properties, and as is
demonstrated in Figure 1.4, the molecular structure (and thus the rheological properties) of the polymer will
affect the architecture and the mechanical properties as well, as do the fabrication conditions. The objectives of
this thesis were to develop, quantify and explore the fabrication, architecture, and mechanical properties of
stochastic honeycombs, as well as the determining the effect of material properties and fabrication parameters on
the internal architecture, and the effect of material properties and architecture on the compressive strength and
energy absorption capability of melt-stretched stochastic honeycombs.
1.3 Overview
This thesis presents the first five articles on melt-stretched stochastic honeycombs [24-28], with minor edits for
flow and clarity. The objectives of this document are to demonstrate a compelling motivation for this project and
to determine polymer property–stochastic honeycomb architecture–fabrication variable–mechanical property
relationships. Because this is a new hybrid material, it was important to determine the separate effects of
polymer rheology and fabrication variables on the architecture of stochastic honeycombs, as well as the effect of
architecture on the mechanical properties. The development of a buckling model for the out-of-plane
compressive properties of stochastic honeycombs was desired, as was a determination of collapse mechanisms of
the stochastic honeycomb under compression, along with the energy absorption capabilities. Chapter 2 provides
an overview of key background concepts, namely honeycomb sandwich panels, melt rheology of polypropylene,
and thin-plate buckling.
7
Chapter 3 details the fabrication methods for melt-stretching stochastic honeycombs. It includes the
manual press that was used to fabricate samples for the papers presented in Chapters 5 – 8, and the fully
instrumented press developed to study the polymer response during fabrication, presented in Chapter 4.
Chapter 4 presents a study on the effect of fabrication variables on the architecture for one PP resin (M.
Hostetter, G.D. Hibbard. Architecture-process relationships in stochastic honeycombs, Journal of Applied
Polymer Science 132 (2015) 42174 [27]). Stochastic honeycombs are fabricated by melting PP between
aluminium platens and then uniaxially stretching the polymer melt to a specified height. The areal density of PP
and the strain rate during uniaxial extension both had significant effects on the architecture. Higher strain rates
activated more entanglements leading to a greater number of node points and a more interconnected geometry.
Following this, another study compared the 5 PP resins (4 HMS PPs and one linear PP) to determine the
effect of rheological properties on the resulting stochastic honeycomb architecture. The rheological properties of
the melt were found to have a significant effect on the architecture. The internal architecture changes with the
properties of the melt, such as the melt flow index and the zero-shear viscosity. The presence of long-chain
branches increases the entanglement density with branches acting as entanglement points with primary bonds as
opposed to secondary bonds; branched PPs formed stochastic honeycombs with many more webs and web
branch nodes than linear PP, implying a greater number of activated entanglement points. This study is presented
in Chapter 5 (M. Hostetter, G.D. Hibbard. Architectural characteristics of stochastic honeycombs fabricated from
varying melt strength polypropylenes, Journal of Applied Polymer Science 131 (2014) 40074 [25]).
Chapter 6 (M. Hostetter, B. Cordner, G.D. Hibbard. Stochastic honeycomb sandwich cores, Composites
Part B: Engineering 43 (2012) 1024 [24]) presents the first paper published on melt-stretched stochastic
honeycombs, and demonstrates the excellent mechanical properties of stochastic honeycombs. Stochastic
honeycombs were fabricated over a range of densities and tested mechanically in out-of-plane compression and
three-point bending. A subset of bending samples were reinforced by thermally welding conventional PP sheets
to the skins to create a rigid sandwich structure composed entirely of PP. It was found that despite the simple
fabrication method, stochastic honeycombs had a mechanical strength comparable to commercial PP
honeycombs and greater than commercial PP foams.
Chapter 7 (M. Hostetter, G.D. Hibbard. Modeling the buckling strength of polypropylene stochastic
honeycombs, Journal of Materials Science 49 (2014) 8365 [26]) takes the four HMS PPs and, with the aid of X-
ray tomography, presents a thin-plate buckling model to predict the strength of stochastic honeycombs. The four
HMS PPs were analyzed mechanically over a range of densities to compare the strength across the varying
architectures. Since stochastic honeycombs combine the stochastic character of conventional foams with the
8
aligned webs of a honeycomb, the internal architecture is quite complex. The core structure of the stochastic
honeycomb is similar to an irregular honeycomb with some percentage of missing walls. A buckling model
based on architectural data from X-ray tomography was created and verified for the differing architectures of the
stochastic honeycombs created with each HMS PP.
A final study investigated the post-peak collapse behaviour of stochastic honeycombs under uniaxial
compression. The dominant failure mechanisms over a range of densities were identified, and it was found that
stochastic honeycombs collapse through elastic buckling, plastic buckling, and plastic buckling with fracture.
The effect of different definitions of densification strain on the absorbed energy was presented, thus completing
the study of the compressive mechanical properties of stochastic honeycombs. Additionally, it was shown that
the secondary defect structure in stochastic honeycombs compensates for the lack of the closed-cell structure
common to traditional honeycombs. This work is presented in Chapter 8 (M. Hostetter, G.D. Hibbard. Post-peak
collapse and energy absorption in stochastic honeycombs (2015) In preparation for submission to the Journal of
Materials Science [28]).
Appendix A presents the material properties for the four HMS PPs and the linear isotactic PP, while
Appendix B includes fabrication details of other polyethylenes and polypropylenes. Appendix C presents a
conference paper on the characterization of stochastic honeycomb failure (M. Hostetter, G.D. Hibbard.
Characterization of stochastic honeycomb sandwich failure (2013) In: Proceedings of the International
Conference on Composite Materials 19, Montreal, Canada, 28 July – 2 August, pp. 312–319 [29]).
Figure 1.6 presents a flow chart detailing the key topics covered in this thesis: fabrication, material
properties, internal architecture and mechanical properties, and the contributions made in each category. The
three major contributions are characterization of the stochastic honeycomb internal architecture, creation of the
buckling model, and determination of fabrication windows. As seen in Figure 1.4 and Figure 1.5, the architecture
could be quantified using many different measures, including the angle between webs, the length and thickness
of webs, and the web end constraints, among many others. Additionally, the architectural variables change not
only from web-to-web within one sample, but also in a single web with vertical position. I determined the key
architectural variables that will be shown to vary with material properties (zero-shear viscosity), density of the
structure, melt-stretch rate during fabrication, and vertical position in the test sample. This characterization
allowed the creation of a buckling model. I modified the thin-plate buckling model to include additional
architectural information, such as variation in end constraints between types of webs and variation in
dimensions. This modified model predicted the critical collapse strength quite well. Finally, I developed a fully
instrumented fabrication process, and was able to determine three key fabrication parameters that were affected
by the density of the test sample and the melt-stretch rate during fabrication. Through the fabrication parameters,
9
I was able to determine how the web structure is formed during fabrication, and link melt-stretch rate and density
to the fabrication parameters and the resulting stochastic honeycomb architecture.
A summary of the completed work is presented in Chapter 9, along with suggestions for future work.
Figure 1.6 Flow chart detailing the four key topics in this thesis and the contributions made in each. The chapter that these
contributions are discussed is indicated through the numbers on the left.
Fabrication
Developed methods;
parameters
Determined key material properties
Effect of melt-stretch
rate on architecture
Effect of density on
architecture
Material Properties
Effect of melt material
properties on architecture
Effect of bulk material
properties on compressive
strength
Effect of bulk material
properties on post-peak
collapse and energy
absorption
Internal Architecture
Developed characterization
methods
Quantified the internal
architecture
Effect of internal
architecture on compressive
strength
Effect of internal
archtiecture on post-peak
collapse and energy
absorption
Mechanical Properties
Demonstrated strength is repeatable
Strength comparable to commercial PP honeycombs
Developed buckling model
Post-peak failure mechanisms
Energy absorption
Chapter 4. Architecture–process relationships in stochastic honeycombs
Chapter 5. Architectural characteristics of stochastic honeycombs fabricated from varying melt strength
polypropylenes
Chapter 6. Stochastic honeycomb sandwich panels
Chapter 7. Modeling the buckling strength of polypropylene stochastic honeycombs
Chapter 8. Post-peak collapse and energy absorption in stochastic honeycombs
8
7
8
8
4 – 8
4,6,
6,8
6
6,7
7
7
7
5,7
4
4
5
4,6
10
References
[1] M.F. Ashby. Overview No. 80: On the engineering properties of materials, Acta Materialia 37 (1989) 1273–1293.
[2] M.F. Ashby, D.R.H. Jones. Engineering Materials 1, Pergamon Press, Oxford, England, 1980.
[3] L.J. Gibson, M.F. Ashby. Cellular Solids: Structures and Properties. 2 ed., Cambridge University Press,
Cambridge, New York, 1997.
[4] K. Finnegan, G. Kooistra, H.N.G. Wadley, V.S. Deshpande. The compressive response of carbon fiber composite
pyramidal truss sandwich cores, International Journal of Materials Research 98 (2007) 1264-1272.
[5] F. Lipperman. Nucleation of cracks in two-dimensional periodic cellular materials, Computational Mechanics 39
127-139.
[6] J. Chung, A.M. Waas. Compressive response of circular cell polycarbonate honeycombs under inplane biaxial
static and dynamic loading. Part I: experiments, International Journal of Solids and Structures 27 (2002) 729–754.
[7] M.J. Silva, W.C. Hayes, L.J. Gibson. The effects of non-periodic microstructure on the elastic properties of two-
dimensional cellular solids, International Journal of Mechanical Sciences 37 (1995) 1161-1177.
[8] J. Zhang, M.F. Ashby. Mechanical selection of foams and honeycombs used for packaging and energy absorption,
Journal of Materials Science 29 (1994) 157-163.
[9] M.F. Ashby, Y.J.M. Bréchet. Designing hybrid materials, Acta Materialia 51 (2003) 5801-5821.
[10] H.G. Allen. Analysis and design of structural sandwich panels. 1st ed., Pergamon Press, (Oxford, New York),
1969.
[11] N.O. Cabrera, B. Alcock, T. Peijs. Design and manufacture of all-PP sandwich panels based on co-extruded
polypropylene tapes, Composites Part B: Engineering 39 (2008) 1183-1195.
[12] P. Corvaglia, A. Passaro, O. Manni, L. Barone, A. Maffezzoli. Recycling of PP-based Sandwich Panels with
Continuous Fiber Composite Skins, Journal of Thermoplastic Composite Materials 19 (2006) 731-745.
[13] W.S. Burton, A.K. Noor. Structural analysis of the adhesive bond in a honeycomb core sandwich panel, Finite
Elements in Analysis and Design 26 (1997) 213-227.
[14] M. Heckele, W.K. Schomburg. Review on micro molding of thermoplastic polymers, Journal of Micromechanics
and Microengineering 14 (2004) R1-R14.
[15] H.N.G. Wadley. Multifunctional Periodic Cellular Metals, Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences 364 (2006) 31-68.
[16] A. Ziebig, H. Brigasky, W. Lewand. Extrusion device for the production of honeycomb structures. Google Patents,
1987.
[17] C.A. Steeves, N.A. Fleck. Material selection in sandwich beam construction, Scripta Materialia 50 (2004) 1335-
1339.
[18] Y.-M. Jen, L.-Y. Chang. Effect of thickness of face sheet on the bending fatigue strength of aluminum honeycomb
sandwich beams, Engineering Failure Analysis 16 (2009) 1282-1293.
[19] P.H. Nam, P. Maiti, M. Okamoto, T. Kotaka, T. Nakayama, M. Takada, M. Ohshima, A. Usuki, N. Hasegawa, H.
Okamoto. Foam processing and cellular structure of polypropylene/clay nanocomposites, Polymer Engineering &
Science 42 (2002) 1907-1918.
[20] R.E. Smallman, R.J. Bishop. Plastics and Composites. Modern Physical Metallurgy and Materials Engineering.
Oxford: Butterworth-Heinemann, 1999.
11
[21] K.F. Karlsson, T.B. Åström. Manufacturing and applications of structural sandwich components, Composites Part
A: Applied Science and Manufacturing 28 (1997) 97-111.
[22] R. Gendron, ed. Thermoplastic foam processing : principles and applications, CRC Press, Boca Raton, Fla., 2005.
[23] H. Naguib. Personal Communication, 18 Mar 2015, to M. Hostetter.
[24] M. Hostetter, B. Cordner, G.D. Hibbard. Stochastic honeycomb sandwich cores, Composites Part B: Engineering
43 (2012) 1024-1029.
[25] M. Hostetter, G.D. Hibbard. Architectural characteristics of stochastic honeycombs fabricated from varying melt
strength polypropylenes, Journal of Applied Polymer Science 131 (2014) 40074.
[26] M. Hostetter, G.D. Hibbard. Modeling the buckling strength of polypropylene stochastic honeycombs, Journal of
Materials Science 49 (2014) 8365-8372.
[27] M. Hostetter, G.D. Hibbard. Architecture-process relationships in stochastic honeycombs, Journal of Applied
Polymer Science 132 (2015) 42174.
[28] M. Hostetter, G.D. Hibbard. Post-Peak Collapse and Energy Absorption of Stochastic Honeycombs, In preparation
for submission to Journal of Materials Science (2015).
[29] M. Hostetter, G.D. Hibbard. Characterization of stochastic honeycomb failure. International Conference on
Composite Materials 19. Montreal, Canada, 2013. p.312-319.
12
Chapter 2
Background
It will be shown that polypropylene stochastic honeycombs have similar mechanical properties to conventional
PP honeycombs and superior mechanical properties to PP foams. Polypropylene was chosen as the material for
stochastic honeycombs because it has been considered as a substitute for other closed cell foams such as
polyurethane, polyethylene, and polystyrene, due to its good mechanical strength and relatively high melt
temperature [1, 2]. Equally important is its recyclability [3], as polyurethane is a thermoset [4, 5], many
polyethylene foams are crosslinked [5], and polystyrene foam is quite difficult to recycle [6, 7].
Stochastic honeycombs were conceptualized and characterized based on foundational work in the field of
cellular solids, while the fabrication is heavily dependent on the rheology, and mechanical testing of stochastic
honeycombs was built on work done on foams and honeycombs. Significant results in cellular materials,
particularly sandwich panels with foam or honeycomb cores, are presented here. Additionally, the melt
properties of polypropylene are presented, along with methods to study and improve the rheological properties
for a given application. Finally, previous work on the mechanics of foams and honeycombs is presented, with the
focus on the thin plate buckling model and its applicability to honeycombs.
2.1 Honeycomb Sandwich Panels
Sandwich panels consist of a low density core positioned between two thin, stiff, strong face sheets (Figure 2.1)
[8-11]. The sandwich panels can be fabricated from any material, such as metal foams [12], balsa wood [12],
three-dimensional truss structures [10], or thermoplastic cores [13-17]. Thermoplastic sandwich panels are of
interest for their ease of fabrication [15], and the potential, especially for polypropylene, to fully recycle the end
product [17-19].
In bending, sandwich panels exhibit one or more of five failure mechanisms [12, 20-23] (Figure 2.2):
i. Delamination: the adhesion between the face sheet and the core fails, usually near the ends of
the sandwich panel
13
ii. Face Sheet Failure: the face sheet fails under tension or compression at the bottom or top of
the panel, in the centre
iii. Face Sheet Wrinkling: the face sheet fails on the top panel, and wrinkles under compression
near the centre of the panel, due to a local minimum in core stiffness or the face sheet
properties
iv. Core Crushing: at the supports or in the centre of the beam, the core locally crushes
v. Core Failure: core fails in shear
Figure 2.1 Cellular material sandwich panels, with foam, truss and honeycomb core materials, top to bottom [8].
Figure 2.2 Sandwich panel failure mechanisms [23]
14
A failure mode map is created by plotting the core relative density as a function of the face sheet thickness
normalized by the panel span, and at each point the dominant failure mechanism is identified. Boundaries
between dominant modes of failure can thus be identified as a function of the sandwich panel geometry [12, 20-
22]. It has been shown that the strength depends very little on core thickness [24], but the properties of the panel
can be tuned by adjusting the thickness of the core and face sheets [17].
The core of the sandwich panels are usually foams (Figure 2.3) or honeycombs (Figure 2.4). Foams are
generally considered isotropic materials, while honeycombs are isotropic in-plane, and have a different response
out-of-plane. Closed cell foams are common in automotive parts (polyurethane, PU [4, 12, 25]) and packaging
(polyethylene, PE [12, 26-28], and polystyrene, PS [12, 29, 30]). Foams have many geometric irregularities,
such as cell wall thickness variations, wavy distortions of cell walls [31, 32] for closed cell foams, and variation
in strut-cross-sectional area [32] for open cell foams. Closed and open cell foams exhibit non-uniform shape of
cells [31, 32]. Due to this geometrical and topological variation, foams have scatter in the mechanical properties
[33].
Figure 2.3 Closed [34] (a) and open [35] (b) cell foams.
Figure 2.4 In-plane (x-y) and out-of-plane (z) directions on a honeycomb sandwich panel [10].
15
Figure 2.5 2D Voronoi honeycomb with a random but even spacing of points [36] (a) and a completely random spacing of points
[37] (b).
Voronoi honeycombs are often used to model foams in 2D [36-41]. This model is created by laying points
in 2D, and introducing a line bisecting at 90o a non-existent line between each point, creating a random
honeycomb with various sizes and shapes of cells and with a dispersion in the number of walls per cell (Figure
2.5). Foams have also been modelled as a 3D Voronoi tessellation [32].
Much work has been done on honeycombs, both in-plane [33, 36, 37, 39-50] and out-of-plane [51-67]
(Figure 2.4), analytically, using finite element analysis (FEA) [33, 37-39, 42-45, 57-61], and experimentally [43,
47, 48, 51, 52, 57, 61-64]. A number of studies have looked at the effect of defects [33, 37-44, 46], but primarily
under in-plane compression. In-plane, it was shown that the separation distance between defects had little effect
on the stiffness or strength of the honeycomb [42], and that the stiffness decreased as the fraction of missing
walls increased [43].
A single unit cell model for honeycombs, in out-of-plane compression, cannot be used to study defects
such as missing or fractured walls [39], and as a result this has not been studied extensively. Similarly to foams,
curved walls in honeycombs decrease the strength and stiffness, under both in-plane and out-of-plane
compression [68].
2.2 Melt Rheology and Structure of Polypropylene
Polypropylene is an attractive polymer for use in cellular materials because of its low cost, ease of processing,
environmental stability, non-toxicity and recyclability [17, 69-71]. Additionally, it has high stiffness, high
resistance to corrosive chemicals, and when incinerated does not produce harmful by-products [70-72]. PP is
being considered to replace polyvinyl chloride (PVC), PU and PE in many products, but needs improved melt
strength [70, 71, 73] and greater strain hardening capabilities [70, 71]. The low melt strength of PP results in
16
local instabilities in thermoforming, blow moulding, extrusion coating and foaming [70, 71]. The production of
closed cell foams with linear PP is difficult becasue the melt does not exhibit strain hardening under extensional
flow, resulting in holes in cell walls and a semi-open cell structure [71, 74, 75]. Many of the processing
difficulties of PP come about because it has a sharp melting point [73]. Introducing long chain branches (LCBs)
increases the processability and melt strength [69, 71], and this is termed high melt strength (HMS) PP. Melt
strength is related to the limit of extensibility of the polymer melt [25], and a higher melt strength allows greater
extension before rupture [3]. Strain-hardening is a phenomenon that is linked to melt strength, but occurs when
polymer chains align during melt stretching, increasing the stress required to further extend the polymer [76].
Thus, during foaming, if the polymer strain-hardens it increases the membrane stability and stabilizes cell walls,
preventing them from expanding too quickly and rupturing. Poor melt strength in PP can also lead to melt
fracture in extrusion, which, along with ruptured cell walls, can lead to an open cell foam structure [25, 71, 74,
77, 78]. Increased melt strength also limits shear thinning, which occurs when a polymer is drawn and it thins
and elongates from a single point, with decreasing resistance to elongation [79].
Melt strength and the ability to strain harden are key variables in the foamability of PP. However, melt
strength is not a quantity that is easy or convenient to measure, so often the melt flow index (MFI) is used
instead, especially industrially. Also important to the processability of PP are the extensional properties, as
determined through the extensional viscosity.
The melt flow index (MFI) is determined by how much polymer can be pushed through a capillary of set
diameter and length at a given temperature and pressure, in a set amount of time. It is measured in g/10 min as
per ASTM D1238 [80]. It is a crude, empirical way of determining the relative molecular weight [81, 82], and is
related to the shear viscosity [83]. Polymers with low MFI values have longer molecular chains and higher
average molecular weights which form more entanglements in the melt [81]. The MFI is inversely proportional
to zero-shear viscosity of a given polymer melt [84, 85], and once the MFI is determined, it is possible to
calculate the zero-shear viscosity for most linear polymers [84]. It has been shown that the zero-shear viscosity
obtained from MFI data is quite close to zero-shear viscosity measured directly [84]. However, the zero-shear
viscosity is very sensitive to molecular weight distribution, while the MFI is not [85].
Extensional viscosity (ηE), on the other hand, is a well-defined rheological property that is measured by
drawing a polymer filament in a controlled-temperature medium (usually an oil bath) and measuring the force as
a function of time [83, 86, 87]. This is limited in practice because it must be determined under isothermal,
uniform, low-strain conditions [83] that are difficult to achieve because the melt will often rupture before it
reaches a steady state [86, 88-90]. As a result, the extensional properties of a melt are often measured
isothermally but not at steady extensional flow, using an apparent elongational viscosity ( ) [90, 91]. Under
17
, failure is defined as the maximum on the force–extension curve [86], although melt rupture occurs some
time after this point. Melt rupture is preceded by either ductile failure (necking) or cohesive fracture [87, 89].
High density polyethylene (HDPE) and PP were found to exhibit ductile failure in melt extension [92].
Melt strength, when quantified, is a rough measure of the extensional viscosity of the polymer [76], and is
used exclusively in polymer engineering, specifically extrusion and formation of thin films and fibres [82]. In
this context, it is the resistance of a melt to extension or sag [83]. In contrast to the steady state conditions of
extensional viscosity and the isothermal apparent elongational viscosity, melt strength is measured non-
isothermally and under high strain rates, closer to actual processing conditions [83], and is used to compare the
drawability of polymers [81]. Many methods are used to measure melt strength, the most popular being to use a
capillary rheometer equipped with a haul off device and a force transducer to measure the strength of the
polymer melt [112]. The Gottfert Rheotens test for melt tension (measured in cN) and velocity at break
(measured in mm s-1
) [79] is a form of this test.
On the molecular scale, polymers with more entanglements tend to have higher resistance to extensional
deformation yielding, and so a higher melt strength [81]. Melt strength is improved by the presence of a high
molecular weight tail (where a small fraction of the chains have very high molecular weight, creating a tail on
the MWD), or long chain branches [112]. Figure 2.6 demonstrates the difference between melt extension of a
Figure 2.6 Schematic of the effect of branching and cross-linking on the tensile strength of PP [100].
18
linear PP and a cross-linked or highly entangled PP. In order to extend the cross-linked or entangled melt,
covalent bonds have to be broken, increasing the melt strength.
Long chain branches and cross-links (depicted schematically in Figure 2.7) have a very large effect on the
melt strength and rheological properties of polymer melts [71, 93-100]. LCBs increase the number of
entanglements [71], and even low amounts of LCBs can increase the zero-shear viscosity and reduce shear
thinning [69, 101]. Long chain branching is induced in PP through electron beam irradiation [70, 71, 95-97, 102-
105], the introduction of multi-functional monomers [70, 71, 93, 94, 98, 99], and melt grafting in the presence of
peroxides [69-71, 94, 98, 99, 102, 106, 107]. Many of these methods lead to some measure of cross-linking [72].
Additionally, the melt strength of PP can be increased without long chain branching by the addition very high
molecular weight PE chains [103], organo-clay nanocomposites [108, 109], carbon nanotubes [110, 111], and
the addition of inorganic nano-fillers [69, 107-111].
Figure 2.7 Schematic of a cross-link [72] (a) and long chain branching [100] (b) in a polymer chain, where Mbp is the average
molecular weight between branch points and MLCB is the average molecular weight of the branches.
2.3 Buckling of Thin Plates as it Relates to Honeycombs
There are three types of mechanical responses of a cellular material under uniaxial compression (see Figure 2.8):
[12, 47, 113]
i. Strain hardening: the stress increases steeply in the elastic region, and during plastic
deformation the rate of increase slows but does not stop. At densification, the slope increases
again. This is found in some foams and in-plane compression of honeycombs.
19
ii. Strain softening: the stress increases steeply in the elastic region to a peak, after which the
stress decreases to a valley, and then increases again once densification begins. This is found
in out-of-plane compression of honeycombs.
iii. Perfectly plastic: the stress increases steeply in the elastic region, and then levels off, with a
slope of zero, at the plateau stress. This continues until densification begins. This is found in
ideal foams.
For out-of-plane compression of honeycombs, it is assumed that all of the cell walls are equally loaded, and the
peak compressive strength is determined by competition between elastic buckling and plastic microbuckling
[64]. Out-of-plane buckling of a square honeycomb is demonstrated in Figure 2.9.
Euler determined the buckling strength of a column under compression, and in 1890 Bryan [115] extended
his theory to buckling of a thin plate restrained on one or more unloaded edges. The relationship was put in its
current form by Timoshenko and Gere in 1961 [116]. The critical buckling strength of a thin plate is [117-119]:
Figure 2.8 Strain hardening (a), strain softening (b) and perfectly plastic (c) stress-strain curves [113].
Figure 2.9 Buckling modes of a square honeycomb in out-of-plane compression [114].
20
(2.1)
where: σcr: the critical buckling stress
K: constant for a given system that incorporates the edge constraints
Et : the tangent modulus from the tensile σ–ε curves
ν: Poisson’s Ratio
tplate: thickness of the plate
b: width of the plate
A schematic of a thin plate with the dimensions and axes is shown in Figure 2.10.
The elastic thin plate buckling model assumes that: [117]
i. Deflections are very small (less than the thickness of the plate).
ii. The middle plane of the plate remains a neutral surface.
iii. Plane sections rotate during bending to remain normal to the neutral surface.
iv. Shearing forces are neglected; loads are resisted by the bending and twisting of plate
elements.
v. The thickness of the plate is small compared to the other dimensions.
These assumptions are fairly good in the elastic region, before the peak in the stress-strain curve.
There are three common edge constraint conditions, termed free, simply supported, and built-in (also
termed clamped). Free means that the edge is not restrained in any way; however, if both unloaded edges are
free, the thin plate is considered a column [120]. A simply supported edge condition requires that there is no
curvature or deflection at the edges [117]:
Figure 2.10 A schematic of a thin plate, with height a and width b, and thickness t in the x-direction.
21
Figure 2.11 A thin plate under compression with m = 5 half-waves [121].
(2.2)
where y and z are defined in Figure 2.10, and δ is the deflection in x. The built-in edge conditions state that there
is no deflection or slope at either boundary [117]:
(2.3)
Most models of out-of-plane compression assume simply supported edge conditions (e.g., [52]), although one
showed that the actual edge constraints were between those of simply supported and built-in, but closer to
simply-supported [120].
A simply supported rectangular plate loaded in compression will develop sinusoidal buckles [117, 118]
(see Figure 2.11), and the number of half-waves, m, increases as the aspect ratio a/b increases [117, 118]. It has
also been shown that when a/b >2, K varies very little with aspect ratio [119].
For the simply supported condition, a web bound on both unloaded edges, K, is determined by [117, 118]:
(2.4)
where = a/b. If the web is simply supported on one side and free on the other, K is found by solving the
following relation [117, 118]:
(2.5)
where r, s, p and q are placeholder variables to simplify the expression, and are defined as:
22
(2.6a)
(2.6b)
It has been shown that there is no interaction between neighbouring vertical edges in out-of-plane
compression [21], implying that the application of a thin-plate model to an extended structure like a honeycomb
is valid.
In a sandwich panel, the loaded edges are also restrained by a face sheet, and it was shown that a single
face sheet increases the critical stress supported by the honeycomb over that of the core alone, and that two face
sheets further increases the critical stress [51].
Elastic buckling is followed by inelastic collapse, leading to localized failure [57]. Fracture begins at the
drop from peak to valley [65], and it has been shown that even for low-velocity and low energy impact of a
sandwich panel with carbon fibre face sheets, the residual compressive strength (i.e. strength after buckling has
begun) can be 50% lower than the peak strength [39, 64].
Deformation of the cell walls was shown to be uniform for ε < 0.003 [64], which is in the elastic region.
Wrinkling of the cell walls was observed at larger strains [64], and the overall collapse mechanism varies with
density [39, 64]. At low density, failure is primarily elastic buckling [39, 64], and plastic collapse at high density
[39]. These two mechanism interact at intermediate densities [39].
The energy absorbed during compression is the integral of the stress-strain curve up to a final strain,
usually the densification strain:
(2.7)
where J is the energy absorbed per unit volume and εd is the densification strain. Densification occurs when the
cellular material has collapsed to enough of a degree that the cell walls begin impacting each other, hindering
further compression [12]. For foams, the densification strain is defined as the strain at which the slope of the
stress–strain curve approaches the stiffness of the solid material, Es [122]. Densification strain has also been
defined as the strain at which the stress is equal to twice the peak stress (for out-of-plane compression of
honeycombs) [123], and the strain at the peak efficiency for energy absorption [113, 124, 125]:
23
(2.8)
where η(ε) is the efficiency of the energy absorption as a function of strain. Which densification value is used
will change the quantification of the energy absorbed during compression or impact testing.
2.4 Summary
This work on melt-stretched stochastic honeycombs has built on the diverse bodies of knowledge in cellular
materials, polymer rheological properties and the buckling and collapse of thin plates. Specifically, previous
work on foams, honeycombs, and sandwich panels that pertain to the work done on stochastic honeycombs have
been summarized. Additionally, the melt flow index and melt strength properties of polypropylene, as well as
specific rheological properties and elongational viscosity of molten polymers, as it pertains to the fabricated
stochastic honeycomb and during fabrication are detailed. Finally, previous work on buckling of thin plates and
post-peak buckling of honeycombs and foams was presented.
24
References
[1] P. Spitael, C.W. Macosko, A. Sahnoune. Extensional rheology of polypropylene and its effect on foaming. Proc.
SPE ANTEC 2002, 2002.
[2] J. Zhang, N. Kikuchi, V. Li, A. Yee, G. Nusholtz. Constitutive modeling of polymeric foam material subjected to
dynamic crash loading, International Journal of Impact Engineering 21 (1998) 369-386.
[3] R. Gendron, ed. Thermoplastic foam processing : principles and applications, CRC Press, Boca Raton, Fla., 2005.
[4] M. Szycher. Handbook of polyurethanes, CRC Press, Boca Raton, FL (US), 1999.
[5] A.H. Landrock. Handbook of Plastic Foams: Types, Properties, Manufacture and Applications, Noyes
Publications, Mill Road, Park Ridge, New Jersey, 1995.
[6] H. Ukei, T. Hirose, S. Horikawa, Y. Takai, M. Taka, N. Azuma, A. Ueno. Catalytic degradation of polystyrene
into styrene and a design of recyclable polystyrene with dispersed catalysts, Catalysis Today 62 (2000) 67-75.
[7] R.S. Stein. Polymer recycling: opportunities and limitations, Proceedings of the National Academy of Sciences
89 (1992) 835-838.
[8] M.F. Ashby, Y.J.M. Bréchet. Designing hybrid materials, Acta Materialia 51 (2003) 5801-5821.
[9] M. He, W. Hu. A study on composite honeycomb sandwich panel structure, Materials & Design 29 (2008) 709-
713.
[10] H.N.G. Wadley. Multifunctional periodic cellular metals, Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences 364 (2006) 31-68.
[11] A.F. Johnson, G.D. Sims. Mechanical properties and design of sandwich materials, Composites 17 (1986) 321-
328.
[12] L.J. Gibson, M.F. Ashby. Cellular Solids: Structures and Properties. 2 ed., Cambridge University Press,
Cambridge, New York, 1997.
[13] H. Ning, G.M. Janowski, U.K. Vaidya, G. Husman. Thermoplastic sandwich structure design and manufacturing
for the body panel of mass transit vehicle, Composite Structures 80 (2007) 82–91.
[14] M.M. Sommer, M. Päβler, R. Schledjewski, M. Stack. Thermoplastic Composite Sandwich Structure for Sportive
Applications, Sandwich Structures 7: Advancing with Sandwich Structures and Materials 7 (2005) 957-966.
[15] A. Passaro, P. Corvaglia, O. Manni, L. Barone, A. Maffezzoli. Processing‐properties relationship of sandwich
panels with polypropylene‐core and polypropylene‐matrix composite skins, Polymer Composites 25 (2004) 307-
318.
[16] R. Ferri, B.V. Sankar. A comparative study on the impact resistance of composite laminates and sandwich panels,
Journal of Thermoplastic Composite Materials 10 (1997) 304-315.
[17] N.O. Cabrera, B. Alcock, T. Peijs. Design and manufacture of all-PP sandwich panels based on co-extruded
polypropylene tapes, Composites Part B: Engineering 39 (2008) 1183-1195.
[18] B. Alcock, N.O. Cabrera, N.M. Barkoula, A.B. Spoelstra, J. Loos, T. Peijs. The mechanical properties of woven
tape all-polypropylene composites, Composites Part A: Applied Science and Manufacturing 38 (2007) 147-161.
[19] P. Corvaglia, A. Passaro, O. Manni, L. Barone, A. Maffezzoli. Recycling of PP-based sandwich panels with
continuous fiber composite skins, Journal of Thermoplastic Composite Materials 19 (2006) 731-745.
[20] T.C. Triantafillou, L.J. Gibson. Failure mode maps for foam core sandwich beams, Materials Science and
Engineering 95 (1987) 37-53.
25
[21] E.W. Andrews, N.A. Moussa. Failure mode maps for composite sandwich panels subjected to air blast loading,
International Journal of Impact Engineering 36 (2009) 418-425.
[22] R.P. Ley, W. Lin, U. Mbanefo. Face sheet wrinkling in sandwich structures. NASA, 1999.
[23] A.G. Evans, J.W. Hutchinson, N.A. Fleck, M.F. Ashby, H.N.G. Wadley. The topological design of
multifunctional cellular metals, Progress in Materials Science 46 (2001) 309-327.
[24] M.K. Khan. Compressive and lamination strength of honeycomb sandwich panels with strain energy calculation
from ASTM standards, Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace
Engineering 220 (2006) 375-386.
[25] J.L. Throne. Thermoplastic Foam Extrusion: An Introduction, Hanser Gardner Publications, Cincinnati, Ohio,
2004.
[26] N.J. Mills, Y. Masso-Moreu. Finite element analysis (FEA) applied to polyethylene foam cushions in package
drop tests, Packaging Technology and Science 18 (2005) 29-38.
[27] E.C.L. Cardoso, A.B. Lugão, L.G. Andrade E. Silva. Crosslinked polyethylene foams, via EB radiation,
Radiation Physics and Chemistry 52 (1998) 197-200.
[28] B. Brandel, R.S. Lakes. Negative Poisson's ratio polyethylene foams, Journal of Materials Science 36 (2001)
5885-5893.
[29] M.B. Hocking. Relative merits of polystyrene foam and paper in hot drink cups: Implications for packaging,
Environmental Management 15 (1991) 731-747.
[30] B. Song, W.W. Chen, S. Dou, N.A. Winfree, J.H. Kang. Strain-rate effects on elastic and early cell-collapse
responses of a polystyrene foam, International Journal of Impact Engineering 31 (2005) 509-521.
[31] J.L. Grenestedt, K. Tanaka. Influence of cell shape variations on elastic stiffness of closed cell cellular solids,
Scripta Materialia 40 (1998) 71-77.
[32] K. Li, X.L. Gao, G. Subhash. Effects of cell shape and strut cross-sectional area variations on the elastic
properties of three-dimensional open-cell foams, Journal of the Mechanics and Physics of Solids 54 (2006) 783-
806.
[33] J. Hohe, C. Beckmann. Probabilistic homogenization of hexagonal honeycombs with perturbed microstructure,
Mechanics of Materials 49 (2012) 13-29.
[34] L.J.M. Jacobs, K.C.H. Danen, M.F. Kemmere, J.T.F. Keurentjes. A parametric study into the morphology of
polystyrene-co-methyl methacrylate foams using supercritical carbon dioxide as a blowing agent, Polymer 48
(2007) 3771–3780.
[35] G. Bitetti, S. Mileti, M. Marchetti. Characterization of low-density open-cell foams for space inflatable
applications, Journal of Spacecraft and Rockets 46 (2009) 210-217.
[36] M.J. Silva, W.C. Hayes, L.J. Gibson. The effects of non-periodic microstructure on the elastic properties of two-
dimensional cellular solids, International Journal of Mechanical Sciences 37 (1995) 1161-1177.
[37] H.X. Zhu, J.R. Hobdell, A.H. Windle. Effects of cell irregularity on the elastic properties of 2D Voronoi
honeycombs, Journal of the Mechanics and Physics of Solids 49 (2001) 857-870.
[38] V. Lucarini. From symmetry breaking to Poisson point process in 2D Voronoi tessellations: the generic nature of
hexagons, Journal of Statistical Physics 130 (2013) 1047-1062.
[39] M.J. Silva, L.J. Gibson. The effects of non-periodic microstructure and defects on the compressive strength of
two-dimensional cellular solids, International Journal of Mechanical Sciences 39 (1997) 549-563.
26
[40] H.X. Zhu, S.M. Thorpe, A.H. Windle. The effect of cell irregularity on the high strain compression of 2D
Voronoi honeycombs, International Journal of Solids and Structures 43 (2006) 1061-1078.
[41] C. Chen, T.J. Lu, N.A. Fleck. Effect of imperfections on the yielding of two-dimensional foams, Journal of the
Mechanics and Physics of Solids 47 (1999) 2235-2272.
[42] X.E. Guo, L.J. Gibson. Behavior of intact and damaged honeycombs: a finite element study, International Journal
of Mechanical Sciences 41 (1999) 85–105.
[43] J.M. Albuquerque, M. Fátima Vaz, M.A. Fortes. Effect of missing walls on the compression behaviour of
honeycombs, Scripta Materialia 41 (1999) 167-174.
[44] F. Lipperman. Nucleation of cracks in two-dimensional periodic cellular materials, Computational Mechanics 39
127-139.
[45] J. Lee, J.B. Choi, K. Choi. Application of homogenization FEM analysis to regular and re-entrant honeycomb
structures, Journal of Materials Science 31 (1996) 4105-4110.
[46] J. Sjölund, A. Karakoç, J. Freund. A new approach for determining the hidden material parameters of a
honeycomb, Journal of Materials Science 49 (2014) 4998-5005.
[47] E.A. Flores-Johnson, Q.M. Li, R.A.W. Mines. Degradation of Elastic Modulus of Progressively Crushable Foams
in Uniaxial Compression, Journal of Cellular Plastics 44 (2008) 415-434.
[48] J.W. Klintworth, W.J. Stronge. Elasto-plastic yield limits and deformation laws for transversely crushed
honeycombs, International Journal of Mechanical Sciences 30 (1988) 273-292.
[49] P. Qiao, J. Wang. Mechanics of Composite Sinusoidal Honeycomb Cores, Journal of Aerospace Engineering 18
(2005) 42-50.
[50] L.J. Gibson, M.F. Ashby, G.S. Schajer, C.I. Roberston. The Mechanics of Two-Dimensional Cellular Materials,
Proceedings of the Royal Society A 382 (1982) 25-42.
[51] Y. Aminanda, B. Castanié, J. Barrau, P. Thevenet. Experimental analysis and modeling of the crushing of
honeycomb cores, Applied Composite Materials 12 (2005) 213-227.
[52] F. Côté, V.S. Deshpande, N.A. Fleck, A.G. Evans. The out-of-plane compressive behavior of metallic
honeycombs, Materials Science and Engineering: A 380 (2004) 272-280.
[53] D. Mohr, M. Doyoyo. Large plastic deformation of metallic honeycomb: orthotropic rate-independent
constitutive model, International Journal of Solids and Structures 41 (2004) 4435–4456.
[54] J.R. Hohe, W. Becker. Effective stress-strain relations for two-dimensional cellular sandwich cores:
homogenization, material models, and properties, Applied Mechanics Reviews 55 (2002) 61-87.
[55] K.W. Kwan, Y. Lin, A.H.W. Ngan. Cascading collapse of honeycomb columns under compression along the long
axis, HKIE Transactions 20 (2013) 148-156.
[56] E. Wu, W.-S. Jiang. Axial crush of metallic honeycombs, International Journal of Impact Engineering 19 (1997)
439–456.
[57] A. Wilbert, W.-Y. Jang, S. Kyriakides, J.F. Floccari. Buckling and progressive crushing of laterally loaded
honeycomb, International Journal of Solids and Structures 48 (2011) 803–816.
[58] X. Fan, I. Verpoest, D. Vandepitte. Finite element analysis of out-of-plane compressive properties of
thermoplastic honeycomb, Journal of Sandwich Structures and Materials 8 (2006) 437-458.
[59] W. Miller, C.W. Smith, K.E. Evans. Honeycomb cores with enhanced buckling strength, Composite Structures
93 (2011) 1072–1077.
27
[60] G. Kress, M. Winkler. Honeycomb sandwich residual stress deformation pattern, Composite Structures 89 (2009)
294–302.
[61] D. Asprone, F. Auricchio, C. Menna, S. Morganti, A. Prota, A. Reali. Statistical finite element analysis of the
buckling behavior of honeycomb structures, Composite Structures 105 (2013) 240-255.
[62] S. Xu, J.H. Beynon, D. Ruan, G. Lu. Experimental study of the out-of-plane dynamic compression of hexagonal
honeycombs, Composite Structures 94 (2012) 2326–2336.
[63] D.M. Wang, Z.W. Wang. Experimental investigation into the cushioning properties of honeycomb paperboard,
Packaging Technology and Science 21 (2008) 309-316.
[64] B.P. Russell, V.S. Deshpande, H.N.G. Wadley. Quasistatic deformation and failure modes of composite square
honeycombs, Journal of Mechanics of Materials and Structures 3 (2008) 1315-1340.
[65] J. Zhang, M.F. Ashby. The out-of-plane properties of honeycombs, International Journal of Mechanical Sciences
34 (1992) 475-489.
[66] T. von Kármán, H.-S. Tsien. The buckling of thin cylindrical shells under axial compression, Journal of The
Aeronautical Sciences 8 (1941) 303-312.
[67] S. Liang, H.L. Chen. Investigation on the square cell honeycomb structures under axial loading, Composite
Structures 72 (2006) 446-454.
[68] A.E. Simone, L.J. Gibson. The effects of cell face curvature and corrugations on the stiffness and strength of
metallic foams, Acta Materialia 46 (1998) 3929–3935.
[69] S. Li, M. Xiao, D. Wei, H. Xiao, F. Hu, A. Zheng. The melt grafting preparation and rheological characterization
of long chain branching polypropylene, Polymer 50 (2009) 6121-6128.
[70] M. Sugimoto, T. Tanaka, Y. Masubuchi, J.I. Takimoto, K. Koyama. Effect of chain structure on the melt
rheology of modified polypropylene, Journal of Applied Polymer Science 73 (1999) 1493-1500.
[71] F.-H. Su, H.-X. Huang. Rheology and melt strength of long chain branching polypropylene prepared by reactive
extrusion with various peroxides, Polymer Engineering and Science 50 (2010) 342-351.
[72] C. He, S. Costeux, P. Wood-Adams, J.M. Dealy. Molecular structure of high melt strength polypropylene and its
application to polymer design, Polymer 44 (2003) 7181-7188.
[73] H.C. Lau, S.N. Bhattacharya, G.J. Field. Melt strength of polypropylene: its relevance to thermoforming,
Polymer Engineering & Science 38 (1998) 1915-1923.
[74] K. Jayaraman, A. Chaudhary. Entangling additves enhance polypropylene foam quality, Plastics Research Online
(2011).
[75] C. Gabriel, H. Munstedt. Strain hardening of various polyolefins in uniaxial elongational flow, Journal of
Rheology 47 (2003) 619-630.
[76] J.R. Fried. Polymer Science and Technology. 2nd ed., Prentice Hall PTR, Upper Saddle River, NJ, 2003.
[77] C.H. Lee, K.J. Lee, H.G. Jeong, S.W. Kim. Growth of gas bubbles in the foam extrusion process, Advances in
Polymer Technology 19 (2000) 97-112.
[78] C.B. Park, L.K. Cheung. A study of cell nucleation in the extrusion of polypropylene foams, Polymer
Engineering & Science 37 (1997) 1-10.
[79] E.M. Phillips. Melt Strength Technical Bulletin. Swedesboro, New Jersey: Sterigenics Advanced Applications,
2005.
28
[80] ASTM D1238 Standard Test Method for Melt Flow Rates of Thermoplastics by Extrusion Plastometer, ASTM
International (2004).
[81] S. Muke, I. Ivanov, N. Kao, S N. Bhattacharya. The melt extensibility of polypropylene, Polymer International 50
(2001) 515-523.
[82] N.G. McCrum, C.P. Buckley, C.B. Bucknall. Principles of Polymer Engineering. Second Edition ed., Oxford
University Press, 1997.
[83] S.K. Goyal. The influence of polymer structure on melt strength behavior of PE resins, 51 (1995) 25-28.
[84] A. Dutta. On viscosity - melt flow index relationship, Rheologica Acta 23 (1984) 565-569.
[85] A. Shenoy, D.R. Saini. Thermoplastic Melt Rheology and Processing, Marcel Dekker, Inc., New York, New
York, 1996.
[86] A.Y. Malkin, C.J.S. Petrie. Some conditions for rupture of polymer liquids in extension, Journal of Rheology 41
(1997) 1-25.
[87] Y. Ide, J.L. White. Investigation of failure during elongational flow of polymer melts, Journal of Non-Newtonian
Fluid Mechanics 2 (1977) 281–298.
[88] A. Considère. Memoire sur l'emploi du fer et de l'acier dans les constructions, Annales des Ponts et Chaussées 9
(1885).
[89] J. Lee, S.E. Solovyov, T.L. Virkler, C.E. Scott. On modes and criteria of ABS melt failure in extension,
Rheologica Acta 41 (2002) 567-576.
[90] J.M. Dealy, R.G. Larson. Structure and Rheology of Molten Polymers: From Structure To Flow Behavior and
Back Again, Hanser Verlag, 2006.
[91] F.H. Su, J.H. Yan, H.X. Huang. Structure and melt rheology of long‐chain branching polypropylene/clay
nanocomposites, Journal of Applied Polymer Science 119 (2011) 1230-1238.
[92] Y. Ide, J.L. White. Experimental study of elongational flow and failure of polymer melts, Journal of Applied
Polymer Science 22 (1978) 1061-1079.
[93] R.P. Lagendijk, A.H. Hogt, A. Buijtenhuijs, A.D. Gotsis. Peroxydicarbonate modification of polypropylene and
extensional flow properties, Polymer 42 (2001) 10035-10043.
[94] B. Wong, W.E. Baker. Melt rheology of graft modified polypropylene, Polymer 38 (1997) 2781-2789.
[95] D. Auhl, J. Stange, H. Münstedt, B. Krause, D. Voigt, A. Lederer, U. Lappan, K. Lunkwitz. Long-chain branched
polypropylenes by electron beam irradiation and their rheological properties, Macromolecules 37 (2004) 9465-
9472.
[96] B. Krause, M. Stephen, S. Volkland. Long‐chain branching of polypropylene by electron‐beam irradiation in the
molten state, Journal of Applied Polymer Science 99 (2006) 260-265.
[97] Z.I. Ali, H.A. Youssef, H.M. Said, H.H. Saleh. Influence of electron beam irradiation and polyfunctional
monomer loading on the physico-chemical properties of polyethylene/polypropylene blends, Advances in
Polymer Technology 25 (2006) 208-217.
[98] J. Tian, W. Yu, C. Zhou. The preparation and rheology characterization of long chain branching polypropylene,
Polymer 47 (2006) 7962-7969.
[99] D. Graebling. Synthesis of branched polypropylene by a reactive extrusion process, Macromolecules 35 (2002)
4602-4610.
29
[100] W. Zhao, Y. Huang, X. Liao, Q. Yang. The molecular structure characteristics of long chain branched
polypropylene and its effects on non-isothermal crystallization and mechanical properties, Polymer 54 (2013)
1455-1462.
[101] A. Malmberg, C. Gabriel, T. Steffl, H. Münstedt, B. Löfgren. Long-chain branching in metallocene-catalyzed
polyethylenes investigated by low oscillatory shear and uniaxial extensional rheometry, Macromolecules 35
(2002) 1038-1048.
[102] F. Yoshii, K. Makuuchi, S. Kikukawa, T. Tanaka, J. Saitoh, K. Koyama. High‐melt‐strength polypropylene with
electron beam irradiation in the presence of polyfunctional monomers, Journal of Applied Polymer Science 60
(1996) 617-623.
[103] M. Sugimoto, Y. Masabuchi, J. Takimoto, K. Koyama. Melt rheology of polypropylene containing small amounts
of high molecular weight chain. I. Shear flow, Journal of Polymer Science Part B: Polymer Physics 39 (2001)
2692-2704.
[104] S. urzbeck, . ster, H. M nstedt, T.Q. Nguyen, R. Gensler. Rheological properties of two polypropylenes with
different molecular structure, Journal of Rheology 43 (1999).
[105] M. Rätzsch. Reaction mechanism to long-chain branched PP, Journal of Macromolecular Science, Part A 36
(1999) 1759-1769.
[106] Z. Zhang, D. Wan, H. Xing, Z. Zhang, H. Tan, L. Wang, J. Zheng, Y. An, T. Tang. A new grafting monomer for
synthesizing long chain branched polypropylene through melt radical reaction, Polymer 53 (2012) 121-129.
[107] A. Fina, D. Tabuani, G. Camino. POSS grafting on PPgMA by one-step reactive blending, Polymer 50 (2009)
218-226.
[108] P. Cassagnau. Melt rheology of organoclay and fumed silica nanocomposites, Polymer 49 (2008) 2183-2196.
[109] M. Bailly, M. Kontopoulou. Preparation and characterization of thermoplastic olefin/nanosilica composites using
a silane-grafted polypropylene matrix, Polymer 50 (2009) 2472-2480.
[110] G.W. Lee, S. Jagannathan, M.L. Minus, S. Kumar. Carbon nanotube dispersion and exfoliation in polypropylene
and structure and properties of the resulting composites, Polymer 49 (2008) 1831-1840.
[111] D. Xu, Z. Wang. Role of multi-wall carbon nanotube network in composites to crystallization of isotactic
polypropylene matrix, Polymer 49 (2008) 330-338.
[112] J. Zwynenburg. Predicting polyolefin foamability using melt rheology. Associated Polymer Labs, 2008.
[113] Q.M. Li, I. Magkiriadis, J.J. Harrigan. Compressive strain at the onset of densification of cellular solids, Journal
of Cellular Plastics 42 (2006) 371-392.
[114] Z. Xue, J.W. Hutchinson. Crush dynamics of square honeycomb sandwich cores, International Journal for
Numerical Methods in Engineering 65 (2006) 2221-2245.
[115] G.H. Bryan. On the stability of a plane plate under thrusts in its own plane, with applications to the "buckling" of
the sides of a ship, Proceedings of the London Mathematical Society s1-22 (1890) 54-67.
[116] J.M. Gere, S.P. Timoshenko. Theory of Elastic Stability. 2nd ed., McGraw-Hill, New York, 1961.
[117] H.G. Allen, P.S. Bulson. Background to Buckling, McGraw-Hill Book Co., New York, 1980.
[118] F. Bloom, D. Coffin. Handbook of Thin Plate Buckling and Postbuckling, Chapman & Hall/CRC, Boca Raton,
Florida, 2000.
[119] O. Vilnay, K.C. Rockey. A generalised effective width method for plates loaded in compression, Journal of
Constructional Steel Research 1 (1981) 3-12.
30
[120] E. Lundquist, E. Stowell. Critical compressive stress for flat rectangular plates supported along all edges and
elastically restrained against rotation along the unloaded edges. 1942.
[121] C.D. Moen, B.W. Schafer. Elastic buckling of thin plates with holes in compression or bending, Thin-Walled
Structures 47 (2009) 1597-1607.
[122] B. Wang, Z. Peng, Y. Zhang, Y. Zhang. Compressive response and energy absorption of foam EPDM, Journal of
Applied Polymer Science 105 (2007) 3462-3469.
[123] O.B. Olurin, N.A. Fleck, M.F. Ashby. Deformation and fracture of aluminium foams, Materials Science and
Engineering: A 291 (2000) 136–146.
[124] M. Avalle, G. Belingardi, R. Montanini. Characterization of polymeric structural foams under compressive
impact loading by means of energy-absorption diagram, International Journal of Impact Engineering 25 (2001)
455–472.
[125] P.J. Tan, J.J. Harrigan, S.R. Reid. Inertia effects in uniaxial dynamic compression of a closed cell aluminium
alloy foam, Journal of Materials Science & Technology 18 (2013) 480-488.
31
Chapter 3
Melt-Stretched Honeycombs
Melt-stretched honeycombs are fabricated by uniaxially stretching molten polymer between two platens to the
desired thickness of the panel, and letting the polymer cool. The resulting structure consists of two skins, parallel
and initially in contact with the platens, separated by a network of interconnected webs oriented parallel to the
direction of melt-stretching. Hot tack adhesion between the polymer and the platens is strong enough to
overcome the cohesiveness of the melt and create the webbed structure, and then as the sample cools and
contracts, it separates spontaneously from the platens. Initially, the honeycombs were fabricated using a manual
press, and then a fully instrumented process was developed. These are detailed below.
Iteration 1: Manual Press
The manual press consisted of two aluminium platens, one with handles and one without, and a wooden
framework for the melt-stretching process. The Al platens were pre-heated in the oven, set at 180 oC (Figure
3.1a), and the temperature was monitored with thermocouples and an IR thermometer. The polymer was placed
on the lower platen for approximately 20 minutes, until it reached the molten state. The lower platen was then
placed on a wooden base, and the framework for the press was placed on top (Figure 3.1b). The upper platen was
compressed on top of the lower platen, and then raised to the desired height using wooden levers through the
handles, and locked in place as the sample cooled. Consistently fabricating successful samples was difficult, as a
number of challenges presented themselves: temperature, adhesion, and controlling the force applied and the
melt-stretch rate.
The temperature in the oven was monitored, but it was known to have (relatively) hot and cold spots, and
since the platens had to be removed from the oven to place in the press, it was not possible to physically attach
the thermocouple leads to the platens. Instead, they rested on the platens or hung in the air in the oven. Over
large areas, portions of the polymer would reach the molten state sooner than other portions.
32
Figure 3.1 The inside of the oven used for the manual press, showing the upper and lower platens, the polymer on the lower platen,
and the thermocouple leads (highlighted to increase visibility) (a), and the manual press locked with the bars on either side
in the stretched position as the sample cools, with the upper platen visible (b).
In the manual press, the force applied to the polymer melt through compression of the upper platen was
not controllable or even. The upper platen was manually pushed down onto the lower platen, resulting in uneven
applied force and variation in the applied force from sample-to-sample. Similarly, the melt was stretched by
applying a downward force on the levers shown in Figure 3.1b and then locking the upper platen in place with
the levers. This led to little control and a large amount of variability in the melt stretch rate sample-to-sample.
On average, the force applied and melt stretch rate were estimated to be approximately 500 N and 100 mm min-1
,
respectively.
Hot tack adhesion between the molten polymer and the platens is necessary in order to stretch the polymer
uniaxially. Generally, PP was found to have good hot tack adhesion to the aluminium, but occasionally polymer
residue was left on the platens, which could be cleaned with ethanol. PP was also found to adhere well to
galvanized steel in initial fabrication trials, but since steel has lower thermal conductivity than aluminium (< 80
W m-1
oC
-1 and ~ 237 W m
-1 oC
-1 [1], respectively), the steel platens took longer to cool and so aluminium was a
better choice. To increase hot tack adhesion, water was poured on heated Al platens to introduce a hydroxyl
layer. This did increase the hot tack adhesion, but it also increased the adhesion after the fabricated sample had
cooled, making it difficult to remove. As a result, this method was not used.
Figure 3.2 shows a HMS PP sample fabricated in the manual press. Circled are some holes that appeared
in the as-fabricated skin. It was unclear what caused a localized loss of adhesion, but when the platens were
properly cleaned with ethanol and there was no polymer residue on the plates, the skin was much more complete.
Additionally, if the polymer itself was unevenly spread across the platen, and the applied force was not enough
33
to even it out, in some areas the polymer would not form continuous webs. Indeed, as indicated with the arrow,
individual polymer beads would form disconnected columns of polymer during melt-stretching.
Note that despite the above-mentioned inherent uncertainty in fabrication consistency from the manual
press, once conditions were developed, samples could be fabricated with controlled relative densities and
compressive strength that were on the order of ± 10% [2].
Figure 3.2 A stochastic honeycomb fabricated with a high melt strength PP (approximately 20 cm x 22 cm). The circles indicate
areas with flaws in the skin, and the arrow points to a corner where there was not enough polymer and an interconnected
structure was not formed.
Iteration 2: Fully-Instrumented Process
While it was possible to produce reliable samples with the manual press, it was desirable to have more control
over the process, especially the temperature, applied force, and melt stretch rate. A set of platens were fabricated
with attachments such that they could be inserted into a Shimadzu AG-I Universal Testing Machine (UTM). A
Shimadzu TCE-N300 Thermostatic Chamber was wheeled forward to encompass the platens, and then wheeled
back once the polymer was molten. Figure 3.3a shows the thermostatic chamber encompassing the platens, and
Figure 3.3b shows the platens inside the chamber. The thermostatic chamber could control the temperature
within ± 1 oC, and a thermocouple was used to monitor the temperature of the platens in real time. The
aluminium platens were attached with four screws at the corners to steel plates with small spacers leaving an air
34
Figure 3.3 A Shimadzu AG-I Universal Testing Machine, and the furnace, a Shimadzu TCE-N300 Thermostatic Chamber (a), with
the aluminium platens and thermocouple leads within the furnace (b). Once the thermocouple leads read 180 oC, the upper
platen was lowered and the furnace moved away (c).
Table 3.1 The program for the automated process for fabrication of stochastic honeycombs, where Fapp is the chosen applied force,
is the chosen melt stretch rate, and h is the chosen final height of the sample.
Wait for start
Direction Down
Stroke rate 10 mm min-1
Stop when F ≥ Fapp N
Direction Up
Stroke rate mm min-1
Hold stroke when S = h mm
Figure 3.4 Stress (σ) as a function of time (t) during fabrication from the Shimadzu AG-I UTM.
35
gap. The steel plates were attached to rods that linked to the UTM. The thermocouple leads were inserted into
the gap between the steel and Al plates, with two on the top and two on the bottom. The leads were placed such
that they were in contact with the Al platens and not the steel. Once the temperature of the platens reached 180
oC, the upper platen was lowered into contact with the molten polymer (F ≈ 150 N), and the thermostatic
chamber was opened and wheeled back, as in Figure 3.3c. The UTM was programmed to follow a set of steps to
fabricate the melt-stretched honeycomb from this point. Table 3.1 lists the steps in this process. The upper platen
was compressed at a rate of 10 mm min-1
until the force reached a pre-set force, Fapp. The upper platen was then
raised, uniaxially, at a given melt stretch rate, , until it reached a chosen stroke limit, for the height of the final
sample (h). These variables, along with temperature, could then be varied systematically. The output from the
UTM could be plotted as shown in Figure 3.4, where the stress has been plotted as a function of time during
fabrication (where tensile stresses are positive).
At the tensile peak of the fabrication curve in Figure 3.4, the molten polymer undergoes cohesive failure,
and the web structure is created as air rushes in to the centre of the sample. The hot tack adhesion of the polymer
to the platens must exceed the cohesive strength of the polymer to allow this to happen, and the web structure
that is formed is governed in large part by the surface tension of the melt.
36
References
[1] Y.A. Çengel. Heat Transfer: A Practical Approach. 2nd Ed. ed., McGraw-Hill Higher Education, New York, NY,
2003.
[2] M. Hostetter, B. Cordner, G.D. Hibbard. Stochastic honeycomb sandwich cores, Composites Part B: Engineering
43 (2012) 1024-1029.
37
Chapter 4
Architecture – Process Relationships in Stochastic Honeycombs
M. Hostetter, G.D. Hibbard. Architecture-process relationships in stochastic honeycombs, Journal of Applied
Polymer Science 132 (2015) 42174
4.1 Introduction
Chapter 3 detailed the fabrication process of melt-stretched stochastic honeycombs. Using the fully instrumented
process, this study exams the mechanical response of the polymer melt during melt stretching. The response is
linked to the fabrication variables, Fapp and , as is the resulting architecture, and used as a means to characterize
differences in web interconnectivity.
4.2 Experimental Methods
PP sheets were fabricated by melting PP-1 beads (MFI = 2.4 g/10 min, ASTM D1238 [1], see Appendix A) at
1.1Tm (180 oC) between Al platens attached to a Shimadzu AG-I UTM. The temperature was controlled using a
Shimadzu TCE-N300 Thermostatic Chamber. A force of 500 N was applied to the PP melt, and the stroke was
held constant as the PP sheet cooled in air. Disc-shaped blanks of PP were cut from the sheets (thickness, tblank =
1.5 – 4.5 mm and diameter, dblank = 8.2 ± 0.3 cm) with areal densities (ρA,blank) ranging from 0.13 – 0.37 g cm-2
.
Stochastic honeycombs were fabricated from the blanks (again at 1.1Tm), by applying a preliminary compressive
force (Fapp = 200 – 5000 N) at 10 mm min-1
, before melt-stretching at varying elongation rates (melt-stretch
rates) of = 1, 10, 30, 60 and 100 mm min-1
(corresponding to strain rates of 0.005, 0.040, 0.125, 0.235 and
0.370 s-1
) in ambient air to a height of h = 10 mm. The temperature of the platens was monitored using a 4-lead
thermocouple with leads attached at opposing corners. Five samples were fabricated in each set of conditions,
and were then illuminated from underneath and photographed. The photographs were edited in ImageJ to
produce binary images of the 2D projection of the webs that could be used as inputs for analysis in ImageJ and
MatLab.
38
4.3 Results and Discussion
4.3.1 Architecture
While the process to fabricate stochastic honeycombs is simple, their internal architecture is not. This study
focuses on the formation of the initial web pattern; the simple back-lit images given in Figure 4.1 are able to
illustrate this formation, and are used as the basis of architectural characterization. Due to the presence of
internal archways and partial webs, the projection of the webs on either side of the samples varied slightly as
only the webs directly in contact with the skin were imaged, but the variation within one sample was inside the
standard deviation of the samples fabricated under identical conditions.
Figure 4.2 presents a process map of the melt-stretch rates, , and areal densities, ρA, used during
fabrication. The areal densities varied by a factor of three and the melt-stretch rate by two orders of magnitude.
Figure 4.3 shows the webs of samples fabricated with = 10 mm min-1
and ρA over the range of 0.10 g cm-2
to
0.23 g cm-2
. Over this density range, the structure changes dramatically. At ρA = 0.10 g cm-2
the structure has a
large number of branch points, or nodes, and a large total length of webs. As ρA increases, the number of nodes,
N, decreases and the total length of webs, L, decreases as well. Figure 4.4 shows the webs of five samples at
constant areal density (ρA = 0.13 g cm-2
) and with the melt-stretch rate changing from = 1 mm min-1
to 100
mm min-1
. Qualitatively, the distance between nodes becomes smaller and the number of nodes increases as
increases. N and L are plotted as a function of ρA for = 10 mm min-1
and 100 mm min-1
in Figure 4.5. N and L
both decrease as ρA increases, and increase as increases. Both variables depend more strongly on ρA than , and
there appears to be a larger difference between = 10 mm min-1
and = 100 mm min-1
at low densities for N. A
greater number of nodes implies a more branched structure and a greater number of connected webs. Figure 4.6
plots N against L for = 10 mm min-1
and 100 mm min-1
; for each , the samples fit along a linear line, in which
ρA decreases as N and L increase. As decreases from 100 mm min-1
to 10 mm min-1
, the slope of the N–L plot
decreases from 4.3 nodes cm-1
to 2.4 nodes cm-1
. N/L can be calculated for each sample, and this value is in
effect the number of nodes per unit length of web.
39
Figure 4.1 Examples of the widely different architectures formed over the fabrication space considered, with melt-stretch rate = 10
mm min-1 and areal density ρA = 0.09 g cm-2 (a) and ρA = 0.23 g cm-2 (b), and ρA = 0.15 g cm-2 and = 1 mm min-1 (c), and
100 mm min-1 (d). Scale bar indicates a length of 20 mm.
Figure 4.2 Process Map illustrating the range of fabrication space (melt-stretch rate, , and areal density, ρA) considered. The points
labelled a-d correspond to the 2D projections seen in Figure 4.1.
40
Figure 4.3 The 2D projections of the webs for five samples fabricated at = 10 mm min-1 and with ρA = 0.10 g cm-2 (a), 0.13 g cm-2
(b), 0.16 g cm-2 (c), 0.19 g cm-2 (d), and 0.23 g cm-2 (e). Scale bar indicates a length of 20 mm.
Figure 4.4 The 2D projections of the webs for five samples fabricated with ρA = 0.13 g cm-2 and = 1 mm min-1 (a), 10 mm min-1
(b), 30 mm min-1 (c), 60 mm min-1 (d), and 100 mm min-1 (e). Scale bar indicates a length of 20 mm.
Figure 4.5 The number of nodes, N (a), and the total length of webs, L (b), plotted against ρA for = 10 mm min-1 and 100 mm min-1.
Two samples with N/L = 1.18 cm-1
(ρA = 0.21 g cm-2
, = 10 and 100 mm min-1
) are shown in Figure 4.7.
In the 2D projection, the sample at lower (Figure 4.7a) has a more open structure, while the sample at higher
(Figure 4.7b) has more (projected) closed cells. This can be seen through the Euler number (χ) which is a
measure used in 2D cellular solids and 3D foams to give a quantitative value for the connectedness of a structure
[2, 3]. Practically, χ2D is determined through image analysis by subtracting the number of "holes" in an area
(shaded in Figure 4.7) from the number of connected areas [4, 5]. Note that over the entire 3-dimensional
structure of the stochastic honeycombs, χ3D = 1, since each pore must be connected back to the sample perimeter.
41
Figure 4.6 The number of nodes, N, plotted against the total web length, L for melt-stretch rates of 10 mm min-1 and 100 mm min-1.
Figure 4.7 2D projections for samples having the same areal density and N/L values (ρA = 0.21 g cm-2 and N/L = 1.18 cm-1) but with
χ2D = -5 (a) and -25 (b). These samples were fabricated at = 10 mm min-1 and = 100 mm min-1 respectively. Scale bar
indicates a length of 20 mm.
In a 2D projection, however, partial webs give the appearance of closed cells, which relates to the tortuosity of
the path from sample perimeter to centre. For the samples in Figure 4.7, the values of χ2D are -5 and -25,
respectively; Figure 4.7b has a more connected structure with a greater number of 'closed' cells (i.e., partial
webs) than Figure 4.7a. Over the entire range of samples, χ2D ranged from 0 (ρA = 0.23 g cm-2
, = 10 mm min-1
)
to -94 (ρA = 0.14 g cm-2
, = 100 mm min-1
).
The connectedness of the structure can also be quantified by examining the area fraction of the
(apparently) closed cells. At the outer perimeter there is a fractal-like structure that extends some (varying)
distance towards the centre of the sample before the webs become interconnected. This length scale is expressed
as an area fraction of the sample in which the webs are not interconnected, and is calculated through 1 minus the
area fraction of closed cells, disregarding the area of the webs:
42
(4.1)
where Acells is the area of the closed cells, Atotal is the area of the stochastic honeycomb sample, Awebs is the area of
the webs, and ξ is the fractal area fraction. ξ represents the outer area fraction of air ingress through which the
fractal nature of the stochastic honeycomb extends. For the image in Figure 4.7a, ξ = 0.75, while for Figure 4.7b,
ξ = 0.48. In general χ2D and ξ are negatively correlated.
Overall, N/L, χ2D, and ξ all change with the areal density and with the melt-stretch rate. To illustrate how
the architecture varies over the range of samples fabricated, contour lines representing the ratio N/L were
superimposed on the Process Map, where the samples are grouped by their values of χ2D (Figure 4.8). The values
of N/L increase as ρA decreases and increases, and the same is generally true for χ2D.
Figure 4.8 Process Map of melt-stretch rate and areal density ρA, with all fabricated samples labelled by the corresponding Euler
number (χ2D), and overlaid with contour lines illustrating the trend in N/L.
4.3.2 Fabrication
In order to understand how these internal architectures form, one would like to have an in-situ process parameter
that can be used to track the state of the system during fabrication. The force required to maintain the fixed melt
stretching rate as a function of displacement is a useful indicator because it provides a value for the overall
response of the molecular chains. Figure 4.9 presents the force response (normalized by the area of the melt-
compressed blank to give an effective stress) over one complete fabrication cycle ( = 10 mm min-1
and ρA =
0.21 g cm-2
), where tension is defined to be positive. This process can be considered as having four stages. In
Stage I, the molten blank is compressed to a pre-selected force at a compression rate of 10 mm min-1
. The melt
flows outward on the platens as tblank decreases to tmelt and dblank increases to dfinal. In Stage II, the compressive
43
Figure 4.9 Full fabrication curve (σ, stress, and S, extension) showing all four stages, with compressive stress expressed as negative
and tensile stress as positive, with permanent plastic extension beginning at the origin. The locations marked (i – iv)
correspond to the images on the right. At the beginning of fabrication, the melt (with thickness tblank, and diameter dblank)
(i) is compressed between Al platens in the furnace. In air, the melt (with tmelt, dfinal) is stretched uniaxially (ii), and the
stress increases to a maximum (iii). At the maximum, air ingress begins as the melt fractures, creating the webs of the
stochastic honeycomb (iv). The melt in the webs stretches and some of the webs fracture, creating partial webs and
archways until the sample reaches its final height (h, dfinal) and cools in air.
load is removed and melt is stretched uniaxially in air at a selected elongation rate between 1 and 100 mm min-1
.
Initially, the polymer experiences elastic recovery, but once σ > 0 it is under tension at a constant melt stiffness
(Emelt) and the stress increases to a peak tensile stress. Stage III begins at the peak stress (σpk) with the creation of
new internal surface area due to a combination of flow of the polymer chains and air infiltration. Air ingress then
inflates the core, creating the webs of the stochastic honeycomb, and the stress decreases quickly. In Stage IV,
the stress reaches a nearly constant value as the final web pattern has been formed and the webs stretch at the
plateau extensional value (σext). Some of the webs undergo ductile failure as the sample is stretched to its final
height, h, and partial webs and archways are formed [6]. The temperature at each stage as a function of the
elongation rate is detailed in Table 4.1.
Figure 4.10 shows the tensile portion of the fabrication curves for four samples. Figure 4.10a compares
samples fabricated with = 10 mm min-1
at ρA = 0.09 and 0.23 g cm-2
. The lower density sample has a higher σpk
and a larger Emelt, while σext does not vary with density. Figure 4.10b compares two samples with ρA = 0.14 g cm-2
and = 1 mm min-1
and 100 mm min-1
. The sample with the higher elongation rate has a larger σpk, a higher
Emelt, and a larger σext. Figure 4.11 plots σpk as a function of ρA for = 10 and 100 mm min-1
. σpk decreases as a
function of density for = 10 mm min-1
, but as increases, σpk becomes less dependent on ρA, and is
independent of ρA by = 100 mm min-1
. σpk, Emelt, and σext are plotted as a function of for ρA = 0.13 g cm-2
in
Figure 4.12. σpk, Emelt and σext all increase as increases, independent of density.
44
Table 4.1 Temperature of the platens during melt elongation at the beginning and end of each stage of fabrication.
Elongation Rate,
(mm/min)
Temperature, T (oC)
Beginning of
Stage I
Beginning of
Stage II
Beginning of
Stage III
Beginning of
Stage IV End of Stage IV
1 180 179.3 ± 0.7 165.7 ± 1.8 164.2 ± 0.2 141.0 ± 1.6
10 180 179.6 ± 0.6 178.1 ± 0.4 177.9 ± 0.2 175.9 ± 0.5
30 180 179.3 ± 1.5 178.8 ± 0.3 178.6 ± 0.1 178.0 ± 0.4
60 180 179.4 ± 0.2 178.5 ± 0.2 178.4 ± 0.1 178.1 ± 0.1
100 180 179.0 ± 0.7 178.6 ± 0.2 178.4 ± 0.3 178.1 ± 0.2
Figure 4.10 Stages II, III and IV on the fabrication curve for two samples fabricated with the same elongation rate ( = 10 mm min-1)
but different densities, ρA = 0.09 g cm-2 and 0.23 g cm-2 (a), and two samples fabricated with the same density (ρA = 0.14 g
cm-2) and two elongation rates, = 1 mm min-1 and 100 mm min-1 (b). These curves correspond to the images in Figure
4.1.
Figure 4.11 Peak fabrication stress (σpk) is plotted as a function of areal density (ρA) for = 10 mm min-1 and 100 mm min-1.
45
Figure 4.12 Peak fabrication stress, σpk, and apparent melt stiffness, Emelt (a), and extensional stress, σext (b), pare plotted as a function
of strain rate ( ) for samples with ρA = 0.13 g cm-2.
It is well known that the tensile strength and stiffness of polymers below their crystallization temperature
are strain rate dependent [7, 8]; this is also true for polymer melts [9-11], although the strength and stiffness are
not often measured. At the beginning of the melt extension (Stage II), the polymer is relaxing while the net
compressive stress is being reduced. An additional set of tests were conducted where the samples were held at
the compressed thickness for 15 s before extension began (i.e. loading to a preset force and then holding at that
displacement). From these tests, an effective relaxation rate of τrelax = 2.1 ± 0.1 kPa s-1
was determined from the
stress–time curve. This value was compared to the slope of the initial portion of Stage II for the different
elongation rates. For = 1 mm min-1
, τmelt = 1.7 ± 0.2 kPa s-1
, and for = 100 mm min-1
, τmelt = 147 ± 12 kPa s-1
(calculated from the σ–time curve at the onset of melt stretching). Therefore, at the lowest elongation rate, and
when σ < 0 in Stage II, the polymer is experiencing relaxation without any added tensile extension, while at =
100 mm min-1
, the polymer is being extended much faster than the relaxation rate.
At the transition from Stage II to Stage III, the melt morphology changes rapidly as air enters from the
perimeter. In rheological studies of the elongation of polymer melts, PP failed through necking [12-15].
However, those studies were done on narrow filaments where the diameter was much smaller than the length, so
that d/l << 1 (i.e., not flat blanks such as in the present study where d/l ≈ 6.0 – 10.0). A large and quick drop in
stress from the peak indicates either rapid ductile necking or cohesive failure [15, 16]. A molten filament
undergoing cohesive failure would break into two pieces [16], and ductile failure would result in an elongating
neck leading to eventual fracture [14]. In the present case, the melt is neither fracturing across the entire cross
section, nor thinning to a point, but undergoes a combination of the two mechanisms that results in the complex
web architecture of stochastic honeycombs. The surface tension of the molten polymer greatly influences this
46
process, as the surface area of the polymer increases rapidly. The webs also undergo necking during web
extension, likely in both the drop in Stage III and the plateau in Stage IV.
Generally, the apparent elongational viscosity, , and the strain at failure, εF, are reported when referring
to the elongational properties of polymer melts [12, 15-22]. is calculated [20] through:
(4.2)
where t is time, σ is stress, and is the strain rate. Since is a function of time as well as strain rate, it is
continuously changing during fabrication; is plotted as a function of time, so the curve exhibits a peak as seen
in the σ–extension curves (Figure 4.9 and Figure 4.10). –t is shown in Figure 4.13a for five samples at the
same density and varying . At any given time, is larger for samples pulled at a faster , as is also seen in
rheological melt extension studies [15, 16, 21, 22]. The peaks of the –t curves occur at higher
as
decreases, and at the peak (
) is plotted as a function of in Figure 4.13b for all samples. However, the
literature shows that should increase as increases [17], which is opposite to the trend observed here. This
can be explained by taking the change in temperature into account. From the beginning to the end of Stage II, the
temperature of the platens dropped by less than half a degree (0.4 ± 0.2 oC) for a melt stretching rate of = 100
mm/min, while for = 1 mm/min, the temperature dropped by more than ten degrees (13.6 ± 1.8 oC) as shown in
Table 4.1. The decrease in temperature causes the larger apparent elongation viscosity at low , as has been
shown in the literature [16].
Figure 4.13 Apparent elongational viscosity,
, is plotted against time, t, for five samples with ρA = 0.13 g cm-2 and increasing
elongation rate, (a), and the peak of the elongational viscosity with time, , is plotted as a function of strain rate, ,
on a log–log scale for all samples.
47
Figure 4.14 The Process Map plotting elongation rate, , against areal density, ρA, showing the samples fabricated, with variation in
N/L shown in the data points. Contour lines illustrating the trends in σpk (a), Emelt (b) and (c) are superimposed on
top.
48
Contour lines for σpk, Emelt, and are overlaid on the Process Maps in Figure 4.14, with the data points
representing the value of N/L for each sample. The peak PP melt stress and apparent melt stiffness both increase
as ρA decreases and as increases. Similarly, the apparent melt viscosity at the peak increases as increases, and
is also slightly affected by ρA. These response parameters are all affected by the molecular structure of the
polymer. At the onset of air ingress, structural changes are occurring in the polymer at two distinct length scales.
On the molecular scale, polymer chains are shifting and sliding past each other, and stretching between
entanglements, while on the macroscale, webs are forming and branching, and extending between the skins.
Immediately preceding the peak of the fabrication curve, the web must begin to fracture at the outer perimeter,
small cracks (cavities) form in the melt and the fine edge structure around the perimeter of the blank is formed
(e.g. Figure 4.3 and Figure 4.4). In Figure 4.3a, with = 10 mm min-1
and ρA = 0.10 g cm-2
, the fractal area
fraction ξ = 0.5, while in Figure 4.3e ( = 10 mm min-1
, ρA = 0.23 g cm-2
), ξ = 1. The low density sample appears
to have activated a higher percentage of entanglements on a molecular scale, creating a larger number of nodes
than the high density sample. Figure 4.4a ( = 1 mm min-1
, ρA = 0.13 g cm-2
) has ξ = 0.7, while Figure 4.4e ( =
100 mm min-1
, ρA = 0.13 g cm-2
) has ξ = 0.3. Here, as increases, more entanglements are activated and the
average number of nodes in the architecture increases from 245 (Figure 4.4a) to 440 (Figure 4.4e). Figure 4.3a
and Figure 4.4e have σpk > 70 kPa, while Figure 4.3e and Figure 4.4a have σpk ≈ 40 kPa. This higher σpk could
indicate the higher stress required to deform the polymer to allow a path for air into the centre of the sample.
4.4 Conclusions
The melt-stretching of PP stochastic honeycombs was studied in detail and sub-divided into four stages based on
the mechanical response of the polymer melt. At the peak stress, cavitation created air pockets that then
propagated as cracks towards the centre of the structure. The low melt-stretching rates were found to be on the
same order of magnitude as the apparent relaxation rate of the PP, while at higher melt stretching rates the PP
was being extended much more quickly than the polymer chains could realign. Lower melt-stretch rates and
higher density both led to lower peak stress (σpk) and stiffness (Emelt) during melt stretching, as well as higher
apparent elongation viscosity ( ). The melt-stretch rate played an important role in determining the
architecture of the webs. A slower rate gave a less negative Euler number, χ2D, while a faster rate exceeded the
ability of the polymer chains to reorient, creating a structure with more apparent (projected) closed cells (smaller
ξ, and more negative χ2D). At lower melt stretch-rates, density also played a significant role in the architecture,
with a smaller number of nodes per length of webs, N/L, at higher density. These results illustrate that it will be
possible to tune the final architecture by varying the fabrication conditions.
49
References
[1] ASTM D1238 - 13 Standard Test Method for Melt Flow Rates of Thermoplastics by Extrusion Plastometer. vol.
D1328. West Conshohocken, PA: ASTM International, 2014.
[2] G.E. Schröder-Turk, W. Mickel, S.C. Kapfer, M.A. Klatt, F.M. Schaller, M.J.F. Hoffmann, N. Kleppmann, P.
Armstrong, A. Inayat, D. Hug, M. Reichelsdorfer, W. Peukert, W. Schwieger, K. Mechke. Minkowski tensor
shape analysis of cellular, granular and porous structures, Advanced Materials 23 (2011) 2535-2553.
[3] L.J. Gibson, M.F. Ashby. Cellular Solids: Structures and Properties. 2 ed., Cambridge University Press,
Cambridge, New York, 1997.
[4] . hser, . M cklich. Statistical Analysis of Microstructures in Materials Science, Wiley, Chichester, West
Sussex, England, 2000.
[5] K. Michielsen, H. De Raedt. Integral-geometry morphological image analysis, Physics Reports 347 (2001) 461–
538.
[6] M. Hostetter, G.D. Hibbard. Architectural characteristics of stochastic honeycombs fabricated from varying melt
strength polypropylenes, Journal of Applied Polymer Science 131 (2014) 40074.
[7] J.R. Fried. Polymer Science and Technology. 2nd ed., Prentice Hall PTR, Upper Saddle River, NJ, 2003.
[8] Y. Zhou, P.K. Mallick. Effects of temperature and strain rate on the tensile behavior of unfilled and talc-filled
polypropylene. Part I: Experiments, Polymer Engineering & Science 42 (2002) 2449-2460.
[9] H. Münstedt. Dependence of the elongational behavior of polystyrene melts on molecular weight and molecular
weight distribution, Journal of Rheology 24 (1980) 847-867.
[10] H.M. Laun. Prediction of elastic strains of polymer melts in shear and elongation, Journal of Rheology 30 (1986)
459-501.
[11] H.M. Laun, H. Münstedt. Elongational behaviour of a low density polyethylene melt I. Strain rate and stress
dependence of viscosity and recoverable strain in the steady-state. Comparison with shear data. Influence of
interfacial tension, Rheologica Acta 17 (1978) 415-425.
[12] A.Y. Malkin, C.J.S. Petrie. Some conditions for rupture of polymer liquids in extension, Journal of Rheology 41
(1997) 1-25.
[13] J.L. White, Y. Ide. Instabilities and failure in elongational flow and melt spinning of fibers, Journal of Applied
Polymer Science 22 (1978) 3057-3074.
[14] Y. Ide, J.L. White. Investigation of failure during elongational flow of polymer melts, Journal of Non-Newtonian
Fluid Mechanics 2 (1977) 281–298.
[15] Y. Ide, J.L. White. Experimental study of elongational flow and failure of polymer melts, Journal of Applied
Polymer Science 22 (1978) 1061-1079.
[16] J. Lee, S.E. Solovyov, T.L. Virkler, C.E. Scott. On modes and criteria of ABS melt failure in extension,
Rheologica Acta 41 (2002) 567-576.
[17] J.M. Dealy, R.G. Larson. Structure and Rheology of Molten Polymers: From Structure To Flow Behavior and
Back Again, Hanser Verlag, 2006.
[18] A. Considère. Memoire sur l'emploi du fer et de l'acier dans les constructions, Annales des Ponts et Chaussées 9
(1885).
50
[19] G.H. McKinley, O. Hassager. The Considère condition and rapid stretching of linear and branched polymer
melts, Journal of Rheology 43 (1999) 1195-1212.
[20] F.H. Su, J.H. Yan, H.X. Huang. Structure and melt rheology of long‐chain branching polypropylene/clay
nanocomposites, Journal of Applied Polymer Science 119 (2011) 1230-1238.
[21] R.P. Lagendijk, A.H. Hogt, A. Buijtenhuijs, A.D. Gotsis. Peroxydicarbonate modification of polypropylene and
extensional flow properties, Polymer 42 (2001) 10035-10043.
[22] W. Minoshima, J.L. White, J.E. Spruiell. Experimental investigation of the influence of molecular weight
distribution on the rheological properties of polypropylene melts, Polymer Engineering & Science 20 (1980)
1166-1176.
51
Chapter 5
Architectural Characteristics of Stochastic Honeycombs Fabricated
from Varying Melt Strength Polypropylenes
M. Hostetter, G.D. Hibbard. Architectural characteristics of stochastic honeycombs fabricated from varying melt
strength polypropylenes, Journal of Applied Polymer Science 131 (2014) 40074
5.1 Introduction
The internal architecture of polymer-based cellular materials is determined in part by the rheological behaviour
of the polymer in question. For example, when a polymer is expanded from the melt state to a closed cell foam,
the stability of the walls is paramount, and two of the most important properties for wall stability are the melt
strength [1-3] and the strain hardening ability of the melt [4-6].
Long chain branching in polypropylene has been shown to increase the processability of PP for industrial
purposes. For instance, Stange and Münstedt [7, 8] studied the rheological properties of LCB PPs and the effect
of the LCBs on foaming of PP. They concluded that the presence of branches led to increased cell density and
cell diameter in the PP foam, as well as an increased expansion ratio. Similarly, Nam et al. [9] determined that
the most important variable in determining foam density was the degree of long chain branching of the PP, and
that lower densities of foam were achievable when using LCB PPs. Additionally, Gotsis et al. [10] modified a
linear PP to an LCB PP, and improved the processability of the PP in foaming and thermoforming.
The mechanical properties of closed cell foams depend on the internal cellular architecture of the polymer
foams, which in turn is controlled by a convolution of processing parameters and the physical properties of the
bulk polymer [11]. This study looks at the effect LCB in PP can have on the internal architecture of stochastic
honeycombs. The influence of polymer rheology and areal density of the melt on the architecture of stochastic
honeycombs was investigated by using four different high melt strength PPs and one conventional (control) PP
to fabricate stochastic honeycombs by melt-stretching. Parallel-plate rheometry tests were used to characterize
52
the viscoelastic properties of the starting polymer, and X-ray tomography was used to characterize the
architecture of the as-fabricated honeycombs.
5.2 Experimental Procedures
The five polypropylenes used in this study were Daploy WB135 HMS, WB140 HMS, WB180 HMS, WB260
HMS (Borealis, Austria) and Accucomp HP0306L (Aclo, Canada), referred to, respectively, as PP-1 to PP-5
(details in Appendix A). To complement the manufacturer supplied melt flow index (ASTM Standard D1238
[12]), each of the five types of PP were characterized by differential scanning calorimetry (DSC) and parallel-
plate rheometry. DSC scans were used to determine the melting temperature and define the processing
conditions for each polymer. Samples were put through a heat-cool-heat cycle at 10 oC min
-1 between room
temperature and 200 oC on a Q20 DSC (TA Instruments). The melting temperatures of the five PP types ranged
from 147 oC to 167
oC and are summarized in Table 5.1. Parallel-plate viscosity tests were performed on an
Advanced Rheometric Expansion System (ARES, TA Instruments) using a dynamic shear test (also called a
dynamic test or a frequency sweep) with sample dimensions of 1 mm through-thickness and 25 mm diameter.
The test sweeps the angular frequency from ω = 0.01 to 100 rad s-1
, with a fixed strain rate of 1% at 1.1 times the
melting temperature (Tm). This temperature allowed direct comparison between the rheological properties of the
polymers at the fabrication temperature. The stochastic honeycombs were fabricated as detailed in Chapter 3 (in
the manual press), by first heating each polymer on an Al platen (at 1.1Tm) until a viscous melt was formed.
Plate and polymer were subsequently removed from the furnace and placed in a custom-built press, where a
second Al platen was compressed on top of the molten PP with a force of ~ 500 N. Subsequently the upper plate
was raised to the desired height (in this case, 15 mm) at a rate of ~ 5 mm s-1
and locked in place. Upon cooling,
the sample spontaneously separated from the platens, resulting in an integrated sandwich structure with upper
and lower skins separated by a network of interconnected webs. Four samples for each polymer, with relative
densities ( ) ranging from 11% to 18% were then characterized in a Skyscan 1172 micro-CT X-ray scanner
Table 5.1 Polymer properties for four high melt strength PP (PP-1 to PP-4) and one conventional (linear) PP (PP-5).
Polymer Tm
(oC)
ηo
( 104 Pa s)
λ (s) Gc
( 104 Pa)
ωc (rad s-1
) MFI
+
(g/10 min)
PP-1 162 2.9 ± 0.6 0.081 ± 0.002 1.5 ± 0.2 12.4 ± 0.3 2.4
PP-2 163 2.9 ± 0.7 0.214 ± 0.066 0.7 ± 0.3 5.0 ± 1.7 2.1
PP-3 163 1.9 ± 0.5 0.039 ± 0.005 1.8 ± 0.5 26.2 ± 3.5 6.0
PP-4 147 3.7 ± 0.5 0.137 ± 0.005 1.5 ± 0.3 7.6 ± 2.0 2.4
PP-5 167 0.4 ± 0.1 0.015 ± 0.001 3.7 ± 1.0 66.2 ± 2.5 12 + provided by the manufacturer
53
(Micro Photonics Inc.) at 44 kV source voltage, 188 μA source current, with an exposure of 158 ms and a
rotational step size of 0.4 degrees. The radiographs were reconstructed into a series of cross-sections, with a
voxel edge length of 35 μm, and binarized to aid the analysis. The cross-sectional slices were then stacked,
giving a three-dimensional model of the internal architecture.
5.3 Results and Discussion
5.3.1 Polymer Characterization
Figure 5.1 plots the results of the parallel-plate rheometry tests, with viscosity (η) given as a function of angular
frequency (ω). Zero-shear viscosity (ηo) was determined by extrapolating the low-shear plateau to a hypothetical
"zero" shear. This parameter is used as a benchmark to compare the viscosity of polymers tested under different
conditions, but at the same temperature [13]. All five polymers exhibited a range of viscous properties under
three identical tests, with the variation given in Table 5.1. Under the frequency range available, only PP-5
exhibited a plateau at low shear, and thus a true zero-shear viscosity. The ηo values given in Table 5.1 are thus
the viscosity at 0.01 rad s-1
, which is a lower bound estimate for ηo.
For an ideally elastic solid, Hooke's Law applies, so that when the stress is changed, the strain changes
immediately, with the modulus acting as a constant of proportionality. For a purely viscous solid, the strain is
time-dependent and perfectly out of phase with the stress, by π/2 rad. A viscoelastic solid, as most polymers are,
has a phase-lag between 0 and π/2 rad. The complex modulus (Eqn. (5.1)) has an elastic response (the storage
modulus, G') that is in phase with the stress, and a viscous response (the loss modulus, G") that is out of phase
with the stress [14, 15]:
Figure 5.1 Parallel-plate rheometry results, giving the viscosity (η) as a function of frequency (ω).
54
(5.1)
Figure 5.2 plots the storage modulus, G', and the loss modulus, G", as a function of ω for two polymers
(PP-2 and PP-5). At low frequency, G" is larger than G', implying that the viscous response dominates at low
frequency [14]. The terminal behaviour of PP-5 at low ω for G' (as indicated by the steep decline as ω decreases)
indicates that this is a linear polymer [1]. This allows the characteristic relaxation time, λ, to be determined,
which is the time-scale separating the predominantly viscous response (t < λ) from the predominantly elastic
response (t > λ) of the polymer melt. λ is determined by [16]:
(5.2)
from which it can be seen that λ = ω-1
where G' and G" intersect. This gives the λ values in Table 5.1.
Figure 5.3 illustrates the relationship between the two components of the complex viscosity, η' and η",
where [16]:
(5.3a)
(5.3b)
From here, it can be seen that the curves for PP-1 to PP-4 exhibit a change in slope after the second derivative
goes through an inflection point at higher η'. This is typically seen for long chain branching in PP, while the
semi-circular curve for PP-5 is indicative of a linear PP [1, 17].
Figure 5.2 Storage modulus (G’) and loss modulus (G") plotted as a function of frequency (ω) for PP-2 and PP-5.
55
Figure 5.3 The two components of complex viscosity (η' and η") from parallel plate rheology tests: the curves for PP-1, PP-2, PP-3
and PP-4 each exhibit a linear portion followed by an increase in slope. This implies that long-chain branching is present
in the polymers. PP-5 exhibits a semi-circular curve, indicating a linear polymer [1, 17].
The crossover point between the storage and loss moduli (ωc,Gc) in Figure 5.2 gives some information
about the breadth of the molecular weight distribution (MWD), and about the weight-average molecular weight
(Mw) for linear PP [9, 18-21]. However, the relationships become more complicated when the polymers are
branched. The polydispersity index (PDI) is inversely correlated with Gc [9, 19, 21], and in turn, Gc is inversely
correlated with either MWD or the degree of branching. For polymers with the same MWD but different Gc, this
measure gives information about the branching density [9], but without specific information about the MWD this
is not a useful relationship. However, a higher G' at lower ω implies more elastic behaviour, and thus a higher
branch density [22], and this can be applied to the four HMS PP. From this measure, PP-1, PP-2 and PP-4 have
very similar branch densities, while PP-3 has less branching. This is corroborated through the relaxation times,
where a longer relaxation time implies a more branched structure [22].
The rheometric data can be also compared to the manufacturer supplied melt flow index (MFI). MFI is
defined as the mass of polymer passing through a cylinder of set dimensions under a given load and fixed
temperature (2.16 kg and 230 oC for PP) in ten minutes, and so is reported in g/10 min, after ASTM Standard
D1238 [12]. It is expected that higher viscosity would lead to less polymer flowing through the cylinder, and
vice versa, and indeed it has been found that MFI generally has an inverse relationship with ηo [16]. However, ηo
is sensitive to both branching and molecular weight distribution [16], leading to differences between the
viscosities of PP-1 and PP-4, even though these two polymers have the same MFI (Table 5.1).
56
5.3.2 Architectural Characterization
The varying rheological properties of the five polypropylenes under consideration lead to differences in the self-
assembled internal architecture of the as-fabricated stochastic honeycombs. The initial areal density of polymer
before fabrication (i.e. the mass per unit area before stretching) also affects the architecture. Finally, the structure
also varies in the through-thickness direction, with the formation of various types of web defects. Variations in
local structure due to the starting polymer type, areal density, and position are discussed below.
Figure 5.4 presents mid-height cross-sections for the five different polypropylenes at a nominal relative
density of ~ 20%. The four HMS PP samples appear comparable, while PP-5 has a much less integrated
structure. Similar to the case of melt fracture during conventional foaming, for relative densities below ~ 20%,
the webs for PP-5 (the linear PP) tended to fail, with the molten polymer flowing back to the skins. Although the
branched morphology of the webs of each stochastic honeycomb type was comparable, there were fewer, thicker
webs for the linear PP.
The normalized cross-sectional areas for each sample were plotted as a function of position, from bottom
to top, and are shown in Figure 5.5. PP-1 to PP-4 all had similar curves, with an initially steep transition from
unit area fraction at the skins to the central third of the sample, at which point a relatively constant plateau
developed. In the plateau, an average normalized cross-sectional area (Ac = Awebs/Atotal) could be determined,
where the average was taken over the central third of the sample height and the standard deviation was taken as
the variation over the same. The web length per unit area within the cross-section (lc = L/Atotal) could also be
determined over the plateau, and like the area, the length of the webs was nearly constant over the middle third
of the specimens (Table 5.2). It is important to note that while PP-5 had a similar density and cross-sectional
area in the plateau to the four HMS PPs, the total length of webs was much less (0.07 mm mm-2
compared to
0.17 ± 0.02 mm mm-2
).
Figure 5.4 Mid-height cross-sections for PP-1 (a), PP-2 (b), PP-3 (c), PP-4 (d) and PP-5 (e), each having a nominal density of ~ 20%.
The scale bar in part (e) represents a length of 5 mm and applies to each figure.
57
Figure 5.5 Cross-sectional area fraction (Ac) as a function of position (P) in the sample (from bottom to top) corresponding to the
samples in Figure 5..
Table 5.2 Cross-sectional area fraction (Ac) and web length per unit area (lc) over the central third of the stochastic honeycomb
samples ( ~ 20%).
Polymer Ac (mm2 mm
-2) lc (mm mm
-2)
PP-1 0.113 ± 0.005 0.195 ± 0.007
PP-2 0.118 ± 0.003 0.200 ± 0.003
PP-3 0.074 ± 0.007 0.147 ± 0.004
PP-4 0.086 ± 0.006 0.172 ± 0.013
PP-5 0.095 ± 0.005 0.071 ± 0.001
The cross-sectional area fraction over the central third of the honeycomb is an important parameter when
considering the mechanical performance because bucking tends to occur where the webs are thinnest and
weakest. However, no trend was observed for Ac as a function of the rheological properties: PP-3, PP-4 and PP-5
have similar cross-sectional areas (Table 5.2) but very different viscosities and rheological properties (Table
5.1). In contrast, the total web length per unit area does seem to be determined by the rheological properties of
the polymer. Figure 5.6a shows an increase in the web length with increasing zero-shear viscosity, before a
plateau is reached at higher values of ηo, while the web length was seen to decrease with increasing crossover
frequency (Figure 5.6b). Crossover frequency is governed by the segmental relaxation time, and as such a lower
ωc implies a more branched structure [22], and more branching indicates increasing entanglement density. It may
therefore be the case that a certain local density of entanglement points needs to be activated during melt-
stretching in order to create each new web. If there are more entanglements grouped closer together, the resulting
melt-stretched architecture will have a greater number of relatively short webs, leading to a greater total web
length. In contrast, if there are fewer entanglements, the webs will be longer and fewer in number, as seen with
PP-5.
In addition to the rheological properties, the starting areal density before melt stretching also affects the
58
Figure 5.6 Cross-sectional web length per unit area, lc, (at a nominal relative density of ~ 20%) plotted as a function of zero-shear
viscosity, ηo (a) and crossover frequency, ωc (b).
Figure 5.7 Mid-height cross-sections of PP-3 at = 10% (a), 13% (b), 16% (c), and 18% (d). The scale bar in (d) is 5 mm and
applies to all images.
resultant architecture. Figure 5.7 shows mid-height cross-sections for PP-3 at overall relative densities of =
10%, 13%, 16% and 18% (corresponding to initial areal densities of 0.14, 0.18, 0.21 and 0.25 g cm-2
,
respectively, before melt-stretching to a height of 15 mm). The mid-height web area fraction increases linearly
with the overall relative density, from 0.031 ± 0.005 at 10% to 0.074 ± 0.007 at 18%. This leads to the question
of whether the mid-height area fraction increases with relative density due to an increase in the web length or the
web thickness. Some insight into this question can be obtained by plotting the cross-sectional area (Ac) for PP-2
and PP-3 as a function of areal density in Figure 5.8. For areal density increases of 50% and 80%, the mid-height
web area fractions increased by 50% and 150% for PP-2 and PP-3, respectively. On the other hand, the
normalized lengths were within 10% and 15% of the mean value for PP-2 and PP-3. Thus, the cross-sectional
area fraction increases linearly with density, while the web length is nearly constant. This implies that as the
density increases, the webs generally become thicker, as opposed to becoming more numerous. It was seen
previously that polymer entanglement, as indicated by the viscosity and relative molecular weight, affects the
59
Figure 5.8 Mid-height cross-sectional area fraction (Ac) of PP-2 and PP-3 as a function of areal density (ρA).
Figure 5.9 Cross-sectional area fraction (Ac) plotted as a function of position, P, for PP-3 at = 13%. The central third of the sample
was used to determine Ac and lc.
Figure 5.10 Cross-sectional area slices from positions a to e in Figure 5. The scale bar in (e) is 5 mm and applies to all figures.
60
total length of webs; here it is seen that the mid-specimen cross-sectional area has a nearly linear dependence on
the amount of material present for a given stretching height.
More subtle characteristics of the cellular architecture can be seen by examining a particular sample in
more depth. Figure 5.9 plots the normalized cross-sectional area as a function of position for PP-3 at a relative
density of 13%. The transition from the relatively constant middle plateau (a) to the fully dense outer skins
begins gradually (b) before steeply increasing through a linear area fraction increase (c – e). Sections through the
core at each of these positions are shown in Figure 5.10. By comparing sequential sections through the sample
height it is possible to identify several distinct types of web defects: archways between incomplete webs,
buttresses supporting some webs near the skin, and air pockets, or voids, in the thicker portions of the webs.
Archways are partial webs that do not stretch entirely from one skin to the other. They can be bounded on one
side or on two by adjoining webs, and provide mechanical support to the adjoining webs even though they do not
provide a direct through-thickness path for load transmission. Figure 5.11 shows an example of an archway (in
PP-3, = 13%), with a 3D representation and three horizontal cross-sections through the archway. Buttresses are
small, partial webs that support the complete webs. They are located adjacent, and attached, to the skin. When a
compressive load is added, they act to resist web buckling and rotation. Figure 5.12 shows an example of two
buttresses (from PP-1, = 11%), with a 3D representation and three horizontal cross-sections through the webs.
Figure 5.13 shows an example of three voids in a web from PP-3 ( = 13%), with a 3D representation and
three horizontal cross-sections through the voids. Voids generally occur near the top or bottom of the sample
where the webs are thicker. For PP-3, the total volume of air pockets in the samples increased from 0.25 cm3 to
0.60 cm3 as the density increased from 10% to 18%. PP-1, PP-2 and PP-4 have a higher total volume of voids,
with volumes ranging from 0.60 cm3 to 0.90 cm
3 for densities increasing from = 11% to 18%.
All four HMS PP types, and to a certain extent PP-5, exhibited archways. Arches occurred in gaps
between webs, and were larger near the mid-height of the structure and smaller near the skin. Buttresses and
voids, however, were not distributed uniformly between the stochastic honeycombs fabricated with different
polymer types. Buttresses were almost non-existent in PP-5, and rare in PP-3, the least viscous of the four HMS
PP. They were much more common in PP-1, PP-2 and PP-4. Voids were present for each polymer type, but were
more common in the least viscous polymers (PP-3 and PP-5). Collectively, these web defects represent a
secondary type of structure, superimposed upon the network of continuous webs spanning the distance between
opposing skin surfaces.
61
Figure 5.11 Archway defect in PP-3 ( = 13%) shown in a 3D compilation image (a), with horizontal lines marking the position of the
cross-sections. The cross-section indicated by the top line is shown in (b), the centre line is (c) and the lowest line is (d).
The scale bar is 2 mm and applies to (b), (c) and (d).
Figure 5.12 Buttress defect in PP-1 ( = 11%) shown in 3D compilation image (a), with horizontal lines marking the position of the
cross-sections. The cross-section indicated by the top line is shown in (b), the centre line is (c) and the lowest line is (d).
The scale bar is 2 mm and applies to (b), (c) and (d).
Figure 5.13 Void defects in PP-3 ( = 13%) shown in a 3D compilation image (a), and in three separate cross-sectional slices (b), (c)
and (d). The scale bar is 2 mm and applies to (b), (c) and (d).
62
5.4 Conclusions
Micro-CT characterization of stochastic honeycombs produced over a range of densities and from
polypropylenes having a range of rheological parameters revealed the complex internal architecture of stochastic
honeycombs. Within a given specimen, the structure changed continuously from position to position, but over
the central third of the honeycomb height, the cross-sectional area and the total length of webs were nearly
constant. For a given polypropylene, the mid-height cross-sectional area was seen to increase linearly with the
areal density of the molten polymer before melt-stretching. However, the total length of webs was relatively
constant over the density range examined. This led to the conclusion that as more polymer was available during
melt-stretching, the webs formed tended to be thicker, rather than more numerous. In addition, it was determined
that the total length of webs was largely governed by the rheological properties of the polymer. Finally, all of the
samples, across all densities and polymer types, showed secondary defects such as archways, buttresses and
voids, adding to the complexity of structural characterization.
63
References
[1] F.-H. Su, H.-X. Huang. Rheology and melt strength of long chain branching polypropylene prepared by reactive
extrusion with various peroxides, Polymer Engineering and Science 50 (2010) 342-351.
[2] S.K. Goyal. The influence of polymer structure on melt strength behavior of PE resins, 51 (1995) 25-28.
[3] J.M. Dealy, K.F. Wissbrun. Melt Rheology and Its Role in Plastics Processing: Theory and Applications, Van
Nostrand Reinhold, Dordrecht, The Netherlands, 1999.
[4] M. Sugimoto, T. Tanaka, Y. Masubuchi, J.I. Takimoto, K. Koyama. Effect of chain structure on the melt
rheology of modified polypropylene, Journal of Applied Polymer Science 73 (1999) 1493-1500.
[5] H.C. Lau, S.N. Bhattacharya, G.J. Field. Melt strength of polypropylene: Its relevance to thermoforming,
Polymer Engineering & Science 38 (1998) 1915-1923.
[6] S. Li, M. Xiao, D. Wei, H. Xiao, F. Hu, A. Zheng. The melt grafting preparation and rheological characterization
of long chain branching polypropylene, Polymer 50 (2009) 6121-6128.
[7] J. Stange, H. Münstedt. Effect of long-chain branching on the foaming of polypropylene with azodicarbonamide,
Journal of Cellular Plastics 42 (2006) 445-467.
[8] J. Stange, H. Munstedt. Rheological properties and foaming behavior of polypropylenes with different molecular
structures, Journal of Rheology 50 (2006) 907-923.
[9] G.J. Nam, J.H. Yoo, J.W. Lee. Effect of long-chain branches of polypropylene on rheological properties and
foam-extrusion performances, Journal of Applied Polymer Science 96 (2005) 1793-1800.
[10] A.D. Gotsis, B.L.F. Zeevenhoven, A.H. Hogt. The effect of long chain branching on the processability of
polypropylene in thermoforming, Polymer Engineering and Science 44 (2004) 973-982.
[11] R. Gendron, ed. Thermoplastic foam processing : principles and applications, CRC Press, Boca Raton, Fla., 2005.
[12] ASTM D1238 Standard Test Method for Melt Flow Rates of Thermoplastics by Extrusion Plastometer, ASTM
International (2004).
[13] J.R. Fried. Polymer Science and Technology. 2nd ed., Prentice Hall PTR, Upper Saddle River, NJ, 2003.
[14] A.P. Deshpande, J.M. Krishnan, S. Kumar, eds. Rheology of Complex Fluids, Springer, New York, 2010.
[15] J. Margolis, ed. Engineering plastics handbook: thermoplastics, properties and applications, McGraw-Hill, New
York, 2006.
[16] A. Shenoy, D.R. Saini. Thermoplastic Melt Rheology and Processing, Marcel Dekker, Inc., New York, New
York, 1996.
[17] Z. Zhang, D. Wan, H. Xing, Z. Zhang, H. Tan, L. Wang, J. Zheng, Y. An, T. Tang. A new grafting monomer for
synthesizing long chain branched polypropylene through melt radical reaction, Polymer 53 (2012) 121-129.
[18] J.C. Chadwick, F.P.T.J. van der Burgt, S. Rastogi, V. Busico, R. Cipullo, G. Talarico, J.J.R. Heere. Influence of
Ziegler−Natta catalyst regioselectivity on polypropylene molecular weight distribution and rheological and
crystallization behavior, Macromolecules 37 (2004) 9722-9727.
[19] C. Tzoganakis. A rheological evaluation of linear and branched controlled‐rheology polypropylenes, The
Canadian Journal of Chemical Engineering 72 (1994) 749-754.
[20] . Bernreitner, W. Neiβl, M. Gahleitner. Correlation between molecular structure and rheological behaviour of
polypropylene, Polymer Testing 11 (1992) 89-100.
64
[21] J.M. Dealy, R.G. Larson. Structure and Rheology of Molten Polymers: From Structure To Flow Behavior and
Back Again, Hanser Verlag, 2006.
[22] L. Wang, X. Jing, H. Cheng, X. Hu, L. Yang, Y. Huang. Rheology and crystallization of long-chain branched
poly(l-lactide)s with controlled branch length, Industrial and Engineering Chemistry Research 51 (2012) 10731-
10741.
65
Chapter 6
Stochastic Honeycomb Sandwich Cores
M. Hostetter, B. Cordner, G.D. Hibbard. Stochastic honeycomb sandwich cores, Composites Part B: Engineering
43 (2012) 1024
6.1 Introduction
In the present study, a high melt strength (HMS) PP (PP-1) was used to fabricate stochastic honeycombs through
a simple melt-stretching process. The fabrication process will be described, and it will be seen that this process
creates a random structure that varies from sample to sample. The cellular architecture, seen in Figure 6.1a,
resembles an open cell honeycomb. Compression test coupons and sandwich panels (Figure 6.1b) were
fabricated over a range of densities for out-of-plane compression and three-point bending tests.
Figure 6.1 Top view of an as-fabricated stochastic honeycomb core showing the cellular web structure (a). Also shown is a three
point bend coupon 45 mm x 200 mm, 20 mm core thickness, reinforced with conventional polypropylene face sheets and
an as-fabricated stochastic honeycomb core compression 30 mm x 30 mm test coupon, 20 mm core thickness (b).
66
6.2 Experimental
The fabrication method presented here is a new, simple, low-cost process that requires no blowing agents,
additives or adhesives. The operating temperature in the furnace was 180 oC, and the press consisted of a simple
frame with adjustable height settings to guide the position of twin aluminium platens. The internal temperature
of the furnace and the temperature of the Al platens were measured by thermocouple and an infrared laser
thermometer was used to monitor the temperature of the polymer in the furnace and in the press.
The platens were preheated in the furnace until they reached a temperature between 60 and 100 oC. HMS
PP-1 were then placed on the lower platen and left in the furnace until they formed a viscous melt
(approximately 10 to 12 minutes). At this point, the platens along with the PP were removed from the furnace
and placed in the press. The platens were compressed for approximately 30 seconds to ensure hot tack adhesion
between the upper platen and the PP. The upper platen was then raised vertically in the press, expanding the
polymer uniaxially and spontaneously creating the webbed structure. It was locked in place at the desired sample
height and the press was left to cool. Once the press reached a temperature of approximately 35 oC (after 6 – 7
minutes) the PP stochastic honeycomb separated itself from the Al platens.
The fabrication window spans expansion heights of 10 to 35 mm, with relative densities ranging from 5%
to 25%. For this study, a height of 20 mm was selected with relative densities ranging from 7 – 12%. Given the
relatively large variability in local cell structure, at least five samples were tested for each density range. The
relative density of each core was estimated according to Eqn. (6.1), where mSHC is the sample mass, tskin is the
total average skin thickness (over approximately 50 measurements), w is the width of the square sample, V is the
volume, is the relative density of the core, and ρPP (0.905 g cm-3
) is the density of the bulk PP:
(6.1)
30 mm x 30 mm coupons were cut from the as-fabricated samples and used for compression testing at a
cross-head speed of 1 mm min-1
(after [1]) in a Shimadzu AG-I Universal Testing Machine (UTM). Web
collapse mechanisms during compression testing were investigated by pre-loading samples to characteristic
uniaxial strain values and studying the deformed web structure in a scanning electron microscope (SEM). Three
point bend samples were also made in the same manner, and cut into samples 200 mm long and 45 mm wide
(after [2, 3]). The three-point bend tests were conducted at a cross-head speed of 1 mm min-1
and span length (Ls)
of 100 mm.
67
PP face sheets were purchased commercially, having thicknesses of tf = 0.73 mm, 0.94 mm, 1.19 mm and
1.75 mm. The sheets were cut to size and joined to a subset of the three point bend samples using thermal
welding. The reinforcing sheets were heated on the upper and lower Al platens in the furnace until they formed a
partial melt (i.e., the PP was soft but not flowing). The lower platen was then fixed in the press as before, with
the three-point bend sample placed on top, and the upper Al platen fixed in the press at a height of 20 mm. This
provided sufficient pressure to ensure adhesion between the PP face sheets and the as-fabricated sample while
not affecting the web structure.
6.3 Cellular Architecture
Since the cellular architecture spontaneously forms as the opposing Al platens are separated, there is large
variation in the as-fabricated web thickness, web length and angle between adjacent webs. This variation was
characterized by imaging the cross-section of the samples at mid-height and averaging the thickness over the
web length in order to give a single value per web. Given the relatively large dispersions of structures at a
particular density and the relatively narrow range of densities considered (~ 5%), it was difficult to establish a
clear trend with density. This is illustrated in Table 6.1, where the mean and standard deviation in web thickness
(tweb) and length (b) for four density ranges is presented. There was a large coefficient of variance (standard
deviation normalized by the mean) for both the thickness and length parameters; the coefficient of variance was
between 25% and 30% for the thickness values, and for the length values, the coefficient of variance was found
to be between 55% and 75%, over all densities. Figure 6.2 presents histograms of web thickness and web length
as global histograms covering all densities. Measurements from approximately 500 webs are included here. The
thickness histogram can be seen to be almost Gaussian, while the length histogram is positively skewed. The
inset on Figure 6.2b illustrates this variation in web length.
There is a complex network of partial webs that connect the slender webs of the stochastic core to the
built-in skin. This can be seen in Figure 6.3, which presents a series of SEM images of stochastic honeycombs
cut at mid-height. While much of the mass of the structure is contained in either the skin or the complete webs
Table 6.1 Average web thickness (tweb) and length (b) for four ranges of core relative densities (ρcore).
(%) tweb (mm) b (mm)
7.9 – 8.0 0.547 ± 0.138 4.20 ± 2.29
8.0 – 8.5 0.483 ± 0.120 4.27 ± 2.58
9.4 – 9.7 0.477 ± 0.130 4.21 ± 3.14
10.5 – 11.0 0.544 ± 0.156 4.47 ± 2.46
68
Figure 6.2 Variation in web thickness, tweb (a) and web length, b (b) over all densities.
Figure 6.3 SEM images showing the internal structure of the stochastic honeycomb architecture. An incomplete web with a gap
extending to the skin is shown (a), along with a mid-height gap indicated by the arrow (b), and the complex buttressing of
the webs (c).
(i.e. those webs extending continuously from lower to upper skin), there are also numerous incomplete webs.
These webs may have gaps which extend to the built-in skin (Figure 6.3a) or be fully contained within the web
(indicated by an arrow in Figure 6.3b).
There are three key properties of the PP-1 used in this study that allow the architecture to be formed: hot
tack adhesion, a low melt flow index, and the low surface energy of PP. The first allows the PP to form an
immediate bond with the Al when the platen is hot [4]. The PP adheres strongly to the Al in the melt state,
allowing it to be expanded uniaxially in the press. Secondly, the PP has a low melt flow index (MFI) of 2.4 g/10
min (as measured by ASTM D1238 [5]) which limits the flow of PP in the melt state [6]; the high melt strength
of the PP used in this study (0.31 N following [7]), was higher than that of conventional PP, which is typically
between 0.01 N to 0.20 N [8]. Lastly, while PP has good hot tack adhesion to Al, it has very low surface energy,
making it a poor adhesive. Special surface treatments are required, or polar compounds need to be grafted on to
69
the polymer backbone, to increase the surface energy enough for it to function as an adhesive [4]. The low
surface energy is an asset in this fabrication process as it causes the sample to spontaneously separate from the
Al platens when it has cooled.
Finally, because the sandwich core is composed of a single material, the recyclability of stochastic
honeycombs is high; it can be ground up and pelletized without need for separating the face sheet from the core.
In certain cases, it may also be possible to refabricate the samples without the need for repelletization. For
example, preliminary studies have indicated that it is possible to fabricate a stochastic honeycomb core, crush it
in uniaxial compression, and then re-expand the crushed core without any significant loss in the refabricated
mechanical properties. This characteristic of stochastic honeycombs may become important in energy absorption
applications.
6.4 Mechanical Properties
6.4.1 Uniaxial Compression
Figure 6.4 shows typical uniaxial compression stress–strain curves for three different densities of stochastic
honeycombs. The curves consist of an initial peak stress, followed by a broad valley in which the stress is
relatively constant until final densification. The overall form of the curves resembles that seen for out-of-plane
compression in conventional honeycombs (e.g. [9]). To determine how the webs collapsed during compression,
deformation in the same mid-height location of a ρcore = 11.5% sample was tracked from the as-prepared state
through to final densification; Figure 6.5 presents SEM images of the as-fabricated structure (Figure 6.5a) and
after pre-loading to strains of ε ≈ 0.05 (Figure 6.5b), ε ≈ 0.25 (Figure 6.5c) and ε ≈ 0.35 (Figure 6.5d). The
slenderness of the web structure can be seen in the as-prepared state (Figure 6.5a). Local plastic buckling
(indicated by arrows) can be seen in samples pre-loaded to the peak stress (Figure 6.5b). Large scale global
buckling had occurred by the time the valley stress had been reached (Figure 6.5c) and continued through
densification (Figure 6.5d).
Figure 6.6 shows the peak strength (σp) and the stiffness (E) plotted against the relative density. While the
strength increased from 1.0 MPa at 7.2% relative density to 2.4 MPa at 11.5% relative density, the specific
strength values were approximately constant at 21.8 ± 2.0 kPa m3 kg
-1. Similarly the stiffness varied from 60
MPa to 130 MPa over the same density range, and the specific stiffness was nearly constant at 1.18 ± 0.09 MPa
m3 kg
-1. For out-of-plane compression of honeycombs, the following relationships have been found [9]:
70
(6.2)
(6.3)
These relationships are determined geometrically, based on regular hexagons [9]. When the graphs in Figure 6.6
are fit, for stochastic honeycombs, the following relationships are apparent:
(6.4)
(6.5)
These results match the theoretical honeycomb relationships very closely.
Comparing these specific strength and stiffness values to those seen for commercial foams and
honeycombs over a similar density range yields encouraging results. The specific strength for stochastic
honeycombs (21.8 kPa m3 kg
-1) is approximately 3 to 5 times higher than that of conventional, commercial PP
foams (e.g. 3 – 4 kPa m3 kg
-1 [10] and 3 – 8 kPa m
3 kg
-1 [11]); likewise the specific stiffness (1.18 MPa m
3 kg
-1)
was approximately 4 to 10 times higher (e.g. 0.10 – 0.30 MPa m3 kg
-1 [12]) than that of commercial PP foam.
When compared to PP honeycombs of comparable densities, the specific strength was on par with those found
for commercial honeycombs (16 – 28 kPa m3 kg
-1 [13] and 10 – 22 kPa m
3 kg
-1 [14]). Similarly, the specific
stiffness of the stochastic honeycombs was comparable to that of one commercial honeycomb (0.78 – 1.21 MPa
m3 kg
-1 [13]) and slightly higher than that of another commercial PP honeycomb (0.19 – 0.46 MPa m
3 kg
-1 [14]).
Figure 6.4 Typical uniaxial compression curves for stochastic honeycomb cores.
71
Figure 6.5 SEM images of the web structure in a ρcore = 11.8% sample (a) and after preloading to strains of ε ≈ 0.05 (b), ε ≈ 0.25 (c)
and ε ≈ 0.35 (d).
Figure 6.6 Stochastic honeycomb compressive strength, σp (a) and stiffness, E (b) over a range of core densities.
6.4.2 Three-Point Bend Testing
The stochastic honeycombs were first tested in three-point bending in the as-fabricated condition. Figure 6.7
shows typical load–deflection (F–δ) curves for samples having relative densities of 8.8%, 9.8%, and 10.8%. As
the density increased, the samples failed at a lower deflection and a higher load. The built-in face sheets meant
that there was a continuous transition from core to outer skin, making it difficult to distinguish between the
nearly simultaneous face sheet and core fracture mechanisms. Figure 6.8 plots the flexural strength (σfl) and
flexural stiffness (D) of the stochastic honeycomb cores as a function of relative density. The flexural strength
72
Figure 6.7 Typical load-deflection curves for as-prepared stochastic honeycomb in three-point bending.
Figure 6.8 Flexural strength, σfl (a) and flexural stiffness, D (b) of stochastic honeycomb cores in three-point bending.
and flexural stiffness increased from 0.53 MPa to 0.95 MPa and from 125.5 kN mm2 to 328.2 kN mm
2
respectively as the relative density increased from 8.4% to 10.6%. The significantly greater flexural strength and
stiffness of the higher relative density stochastic honeycomb cores is partly explained by differences in the built-
in skin thickness. For example, the average skin thickness on a ρcore = 11.0% sample was ~ 0.24 mm while the
average skin thickness on a ρcore = 8.7% sample was ~ 0.14 mm.
73
The stochastic honeycomb cores were also reinforced with polypropylene sheets having thicknesses
varying from 0.73 mm to 1.75 mm. Sandwich panels under three-point bend testing typically fail by one of four
mechanisms: delamination of the face sheet, wrinkling of the face sheet, core shear failure, and face sheet yield
and fracture [9]. All four of these failure mechanisms were observed in the reinforced sandwich structures of the
present study. However, it was found that through careful sample preparation, delamination of the face sheet
could essentially be eliminated. Figure 6.9 summarizes the failure mechanisms in the form of a failure map for
these samples, where the dashed lines show the transition between dominant failure modes. It is apparent that at
the lower face sheet thicknesses and lower densities, wrinkling was the dominant failure mechanism. As the face
sheet thickness increased, along with the density, core shearing and then face sheet yield and fracture
consecutively became the dominant failure mechanisms. Typical load–displacement curves for each of these
failure mechanisms are shown in Figure 6.10.
The samples that failed primarily through face sheet yield and fracture failed at the highest applied loads
(Figure 6.10a). These samples generally deformed to between 4 and 6 mm deflection and then snapped, failing
abruptly. For tf = 1.75 mm, the flexural strength and stiffness varied from 5.46 MPa and 1140 kN mm2 at 7.6%
relative density to 6.12 MPa and 1300 kN mm2 at a relative density of 8.8%, respectively.
Core shearing (Figure 6.10b) occurred mostly in the samples with face sheet thicknesses of 0.94 mm and
1.19 mm. In both cases, cracking noises could be heard as the core progressively sheared, corresponding to
periodic load drops in the load-displacement curves. For both face sheet thicknesses, the flexural strength and
stiffness increased with core density. When tf = 0.94 mm, the flexural strength and stiffness increased from 4.32
Figure 6.9 Failure map plotting regions of dominant sandwich panel failure mechanism. The dominant failure mechanism at low
sheet thickness was face sheet wrinkling, core shearing dominated at intermediate thickness, and face sheet yielding and
fracture dominated at the greatest thickness.
74
Figure 6.10 Typical three-point bending load–deflection curves illustrating face sheet yield and fracture (tf = 1.75 mm, a), core
shearing (tf = 0.94 mm, b) and face sheet wrinkling, (tf = 0.73 mm, c).
MPa and 808 kN mm2 at 7.4% relative density to 5.24 MPa and 982 kN mm
2 at 9.1% relative density,
respectively. Similarly, when tf = 1.19 mm, the flexural strength and stiffness increased from 4.48 MPa and 982
kN mm2 at 7.5% relative density to 4.65 MPa and 1198 kN mm
2 at 9.4% relative density, respectively.
75
Wrinkling of the face sheet (Figure 6.10c) caused the samples to fold in on themselves, and most did not
fracture but rather slid off of the supports after reaching displacements of more than 65 mm. Samples that failed
by face sheet wrinkling had the thinnest PP sheet reinforcement and failed at the lowest peak loads. When tf =
0.73 mm, the flexural strength and stiffness increased from 1.88 MPa and 558 kN mm2 at 7.0% relative density
to 2.73 MPa and 596 kN mm2 at 8.9% relative density, respectively.
In the reinforced sandwich panels, failure was either largely controlled by the face sheet (in face sheet
yield and fracture) or by the core (in core shearing and wrinkling). When sample failure was controlled by the
face sheet (for tf = 1.75 mm), the specific flexural strength varied by less than 2% and the stiffness varied by
less than 10% over all the samples, at 78.3 ± 0.9 kPa m3 kg
-1 and 16.0 ± 1.7 N m
5 kg
-1 respectively.
Alternatively, when sample failure occurred primarily in the core, much larger variation in the specific flexural
strength and stiffness was seen. For tf = 1.19 mm (failure by core shearing), the specific flexural strength was
60.8 ± 5.5 kPa m3 kg
-1, a variation of 9%, and the specific flexural stiffness was 13.0 ± 2.3 N m
5 kg
-1, a variation
of 17%. For tf = 0.73 mm (failure by wrinkling), the specific flexural strength and stiffness were 30.9 ± 3.1 kPa
m3 kg
-1 and 7.4 ± 2.1 N m
5 kg
-1, variations of 10% and 28% respectively.
At constant core density, the peak load increased with increasing face sheet thickness. Likewise, at
constant face sheet thickness, the peak load increased with increasing core density. It should be noted that
relatively larger sample-to-sample variability was seen in the flexural properties of the as-prepared samples than
the reinforced samples. For example, at a core relative density of 8.8%, the flexural strength values for the as-
prepared samples were within 15%, while at the same density, the reinforced samples varied by 5% on average.
In flexure, fracture occurs at the outer surface and is more sensitive to changes in the local built-in or reinforced
face sheet thicknesses. The reinforced sheets were not only much thicker, but also more uniform than the built-in
skins, accounting for the reduction in variability in the reinforced samples. Further development of stochastic
honeycomb sandwich panels could include using alternative materials for the face sheets. Glass-fibre reinforced
PP sheets, for example, would have higher tensile strength and stiffness, increasing the flexural properties of the
sandwich panels, but could still be thermally welded in place.
6.5 Conclusions
This study has shown that with minimal equipment, and using a simple, one step process, it is possible to
fabricate a PP stochastic honeycomb core that has comparable compressive performance to conventional PP
honeycomb cores and can exceed the performance of conventional PP foams. Mechanical testing methods were
76
developed, adapted from the ASTM Standards for honeycombs (see ASTM C365, C393 and D790 [1, 2, 15]) for
sample size and strain rate. It was demonstrated that the strength of stochastic honeycombs at a given density
was repeatable despite the random structure and sample-to-sample variation. The built-in skin allows external
face sheets to be joined to the core without using an adhesive. Employing a single thermoplastic material allows
the stochastic honeycombs to be recycled without any special preparation, and allows for the possibility of
refabrication (melting down the samples and re-expanding them uniaxially in the press) without repelletization.
When the as-fabricated stochastic honeycomb cores were tested in three-point bending, the panels failed
by face sheet yield and fracture, and the flexural strength and stiffness varied with the core relative density.
When the cores were reinforced with external PP sheets, all four of the typically observed sandwich panel failure
mechanisms were observed: face sheet yield and fracture, core shearing, face sheet wrinkling and delamination.
With careful sandwich preparation it was possible to eliminate delamination in the reinforced samples. For all of
the sandwich panels, the flexural strength and stiffness increased with core relative density at a constant face
sheet thickness, and also increased with increasing face sheet thickness at constant core relative density.
77
References
[1] ASTM C365 Standard Test Method for Flatwise Compressive Properties of Sandwich Cores. ASTM
International, 2005.
[2] ASTM C393 Standard Test Method for Core Shear Properties of Sandwich Constructions by Beam Flexure,
ASTM International (2006).
[3] ASTM D7250 Standard Practice for Determining Sandwich Beam Flexural and Shear Stiffness, ASTM
International (2006).
[4] I. Skeist. Handbook of Adhesives. 2d ed., Van Nostrand Reinhold Co, New York, 1977.
[5] ASTM D1238 Standard Test Method for Melt Flow Rates of Thermoplastics by Extrusion Plastometer, ASTM
International (2004).
[6] Ides.com -. Typical Properties of Polypropylene (PP). <http://www.ides.com/generics/PP/PP_typical_properties.
htm> Accessed 4 Nov 2011.
[7] J. Zwynenburg. Predicting Polyolefin Foamability Using Melt Rheology. Associated Polymer Labs, 2008.
[8] H.C. Lau, S.N. Bhattacharya, G.J. Field. Melt strength of polypropylene: Its relevance to thermoforming,
Polymer Engineering & Science 38 (1998) 1915-1923.
[9] L.J. Gibson, M.F. Ashby. Cellular Solids: Structures and Properties. 2 ed., Cambridge University Press,
Cambridge, New York, 1997.
[10] JSP. <http://techdocs.jsp.com> JSP Docs. Accessed 4 Nov 2011.
[11] F.P.C.F.P. Corporation. Arplank Technical information - FPC Foam Products Corporation - Expanded
polystyrene, Bead foam. <http://www.fpcfoam.com/eperan-polypro-tech.html> Accessed 4 Nov 2011.
[12] I. Beverte. Deformation of polypropylene foam Neopolen® P in compression, Journal of Cellular Plastics 40
(2004) 191-204.
[13] Hexacor Ltd. Hexacor - Honeycomb Core Materials. <http://www.hexacor.com> Accessed 4 Nov 2011.
[14] Nida-Core Corporation. Polypropylene Honeycomb Core, Composite Reinforcements - Nida-Core Technical
Data. <http://nida-core.com/english/> Accessed 4 Nov 2011.
[15] ASTM D790 Standard Test Methods for Flexural Properties of Unreinforced and Reinforced Plastics and
Electrical Insulating Materials.
78
Chapter 7
Modeling the Buckling Strength of Polypropylene Stochastic
Honeycombs
M. Hostetter, G.D. Hibbard. Modeling the buckling strength of polypropylene stochastic honeycombs, Journal of
Materials Science 49 (2014) 8365 with kind permission from Springer Science and Business Media
7.1 Introduction
Chapter 5 detailed the effect of rheological properties on the architecture of stochastic honeycombs, and Chapter
6 surveyed the out-of-plane compressive and flexural properties. However, the mechanical properties will
depend on both the polymer material properties (i.e. crystallinity, tensile strength) and the architecture (i.e. cross-
sectional area, arrangement of the webs). In order to build a predictive model for their strength, these effects
need to be separated. To accomplish this, the present study used four HMS PPs with different viscosities and
branching structure to fabricate stochastic honeycombs over a range of densities. The samples were
characterized by X-ray tomography and tested in out-of-plane compression in order to develop a composite
buckling model as a first step towards predicting the strength of stochastic honeycombs.
7.2 Experimental Details
The stochastic honeycombs characterized in this study were fabricated from PP-1, PP-2, PP-3 and PP-4. The
rheological properties of these PPs were characterized in Chapter 5 [1] and Appendix A, and their zero-shear
viscosities (ηo) are presented again in Table 7.1. Each of these four PPs contain long chain branches, which
increases their viscosity [1]. To see if melt-stretching during fabrication affected the crystallinity, differential
scanning calorimetry (DSC) was performed using a Q20 DSC (TA Instruments) on the as-received resin as well
as on portions of webs cut from the as-fabricated stochastic honeycombs. The samples in the DSC were put
through heat–cool–heat cycles between 40 oC and 200
oC at 10
oC min
-1. Tensile coupons were moulded with a
79
Table 7.1 Zero-shear viscosity (ηo) as measured using parallel-plate rheology tests at 1.1Tm [1].
Polymer ηo (x104 Pa s)
PP-1 3.24 ± 0.24
PP-2 3.31 ± 0.39
PP-3 2.15 ± 0.21
PP-4 3.94 ± 0.18
gauge length of 18 mm, width of 8 mm and thickness of 2 mm. The tensile tests were performed on a Shimadzu
EZ-L UTM at a stroke rate of 1 mm min-1
for four samples per polymer. The stochastic honeycombs were
fabricated as before by melting the polymer resins at 1.1Tm (as determined by DSC), compressing the melt
between Al platens and then melt-stretching at ~ 5 mm min-1
to a height of 14 mm. The samples were cut into 40
mm 40 mm 14 mm blocks after cooling. The as-fabricated skin thickness (tskin) was measured by taking ~
200 measurements with a point micrometer around the outside edge of each sample. A minimum of four samples
from each polymer, of increasing density, were scanned by X-ray tomography (Skyscan 1172 micro-CT scanner)
with an 181 μA and 44 kV source. The result, after reconstruction using NRecon software, is a digital 3D
representation of the structure in which each voxel has an edge length of 35 μm, allowing the internal
architecture to be analyzed mathematically. A minimum of three samples at six to seven different densities for
each PP were compressed uniaxially in out-of-plane compression at 1 mm min-1
in a Shimadzu AG-I UTM.
7.3 Results
7.3.1 Polymer Characterization
Figure 7.1a gives the DSC cool–heat thermal cycle for all four starting PP resins, representing a calorimetric
baseline for the polymers with no thermal history. The melt temperature (Tm), crystallization temperature (Tc),
and percent crystallinity (xc) were determined from the baseline. Crystallinity is defined as:
(7.1)
where Δhm is the specific enthalpy of fusion, determined from the area under the melting curve in the DSC, and
Δ is the specific enthalpy of fusion of a crystalline polymer [2] (previously reported values for PP have been
on the order of 210 J g-1
[3-6]). PP-1, PP-2 and PP-3 exhibit similar calorimetric profiles while PP-4 has a lower
Tm, Tc, and crystallinity (Table 7.2). Lower Tm and lower crystallinity are linked, as the crystallinity influences
the melting temperature [7]. The crystallinity of all four PPs is also lower than that of isotactic PP (xc ~ 50 –
60%) [8] as expected, since the HMS PPs are substantially branched [1]. Figure 7.1b shows the melting curves
of PP-1 and PP-4 in more detail. It can be seen that PP-1 has a smooth curve, while the melting curve for PP-4
80
Figure 7.1 Heat flow (ΔH/m) plotted as a function of temperature (T) for the cool–heat cycle for PP-1, PP-2, PP-3 and PP-4 (a), and
the two heating curves (for the as-fabricated webs and the base resin) for PP-1 and PP-4 (b).
Table 7.2 Melting and crystallization temperatures (Tm, Tc) and the crystallinity (xc) of the base polymer and the webs for all four
PPs from differential scanning calorimetry.
Polymer Tm (oC) base polymer Tm (
oC) as-fabricated Tc (
oC) xc (%) base xc (%) webs
PP-1 162.1 ± 1.0 165.3 ± 0.7 125.4 ± 2.2 42.9 ± 1.8 44.3 ± 2.7
PP-2 161.2 ± 0.9 163.9 ± 1.3 125.3 ± 1.1 41.5 ± 1.7 42.5 ± 2.5
PP-3 163.0 ± 0.4 165.3 ± 0.4 126.5 ± 0.2 43.2 ± 1.0 43.8 ± 1.5
PP-4 147.7 ± 1.0 150.8 ± 0.6 110.3 ± 1.1 28.5 ± 2.1 28.2 ± 2.5
has a shoulder at lower temperature. The shoulder indicates that this PP likely has at least some syndiotactic
character, leading to two crystalline phases [9].
Additional samples cut from the webs of as-fabricated stochastic honeycombs were also characterized by
DSC. A direct comparison between the as-fabricated webs and the base resin was obtained from the first and
second heating cycles, respectively, on the as-fabricated webs, with the initial heat cycle giving the thermal
history from the melt-stretching fabrication process. The initial and second heat curves for PP-1 and PP-4 are
shown in Figure 7.1b to demonstrate the effect of this thermal history. Note that PP-2 and PP-3 have very similar
curves to PP-1 and so are not shown here. Since fabrication involves melt stretching, it is reasonable to assume
that there may be some degree of strain-induced crystallinity in the webs. In fact, a small change can be seen
through the melting temperature of the as-fabricated webs, which were ~ 2.8 oC higher than that of the base
properties (Table 7.2). A greater percent crystallinity and larger crystal size are both expected to increase Tm, as
determined from the peak of the melting curve. This peak shifts to higher temperatures in the DSC because more
energy is required to melt the crystals [7, 10-14]. However, the larger percent error associated with enthalpy
based crystallinity measurements made it difficult to unambiguously determine an increase in xc.
81
Figure 7.2 Stress (σ) – strain (ε) curves for tensile tests on PP-1, PP-2, PP-3 and PP-4, pulled at a rate of 1 mm min-1.
Table 7.3 Tensile strength (σTS) and strain at break (εu) for all four PPs.
Polymer σTS (MPa) εu (mm mm-1
)
PP-1 32.1 ± 0.8 0.14 ± 0.02
PP-2 30.9 ± 0.9 0.11 ± 0.02
PP-3 29.4 ± 0.5 0.10 ± 0.01
PP-4 18.2 ± 1.6 > 1.0
The tensile stress–strain curves for PP-1 to PP-4 are shown in Figure 7.2, and the maximum strength as
well as the strain at break are given in Table 7.3. It is evident that PP-1, PP-2 and PP-3 have very similar
strengths, stiffnesses and strains at fracture. PP-4, however, has a much lower strength and stiffness, and a much
larger strain at fracture. A typical fracture strain for PP is 4.0, with a strain of ~ 0.12 at yield [15, 16]. PP-4 was
the only PP that approached this amount of deformation, while PP-1, PP-2 and PP-3, with similar strengths, all
fractured just beyond the yielding point, at a strain of ~ 0.1. PP-4 is quite ductile in tension, while PP-1, PP-2
and PP-3 are comparatively brittle; the significantly lower crystallinity of PP-4 likely plays a role in the
increased ductility [17, 18]. However, this difference in ductility is not expected to affect the overall
compressive strength of the stochastic honeycombs, as out-of-plane compressive failure first occurs through
elastic instability followed by plastic buckling [19]. As a result, the tensile properties outside of the elastic region
should have comparatively little role in determining the peak compressive strength. Note that this difference in
ductility may affect the post-peak collapse mechanisms, but this was not the subject of the present study.
7.3.2 Honeycomb Architecture
Conventional honeycomb sandwich panels have a separately bonded face sheet on either side of a low density
core. In contrast, stochastic honeycombs have a facing skin that is formed during the melt-stretching fabrication
step. As a result, the skin itself is a part of the overall cellular architecture. The skin thickness, tskin, was
normalized by the height of the samples to produce a dimensionless face sheet parameter ( ) and plotted as a
82
Figure 7.3 Skin thickness as a percentage of height ( ) plotted as a function of total relative density ( ).
Figure 7.4 3D reconstructions of a PP-2 stochastic honeycomb at 9.5% core density. Archways are webs that do not extend through
the entire height of the sample (shown with an arrow), and buttresses form at the top and bottom the sample where they
would join the skin (circled). Each scale bar represents 5 mm. The images show a top view of the core (a), a side view (b),
and views from the mid-height of the complex structure (c,d).
function of the relative density, , in Figure 7.3. increases with at the same rate for all four resins, but the
line for PP-3 is shifted upwards towards thicker skins from the other three PPs. This PP had the lowest viscosity
(Table 7.1), suggesting that the relative partitioning of melt between skin and core may be largely controlled by
the resistance to flow in the melt state. The skin thickness can also be used to define a density parameter for the
core, as shown in Chapter 6:
83
(7.2)
where ρPP is the density of polypropylene (0.905 g cm-3
), m is the mass of the sample, w2 is the area of the
sample, and h is the total height.
The distribution of material within the honeycomb core was studied by X-ray tomography. Figure 7.4
shows 3D reconstructions of the X-ray scans for a PP-2 sample with = 9.5%. Figure 7.4a shows the web
structure for the central 90% of the core. It can be seen that many of the webs have a connectivity of 3 (defined
as the number of webs that meet at a node), but also that many webs are unbound on one vertical edge. Figure
7.4b shows a side view of a vertical slice of the upper 50% of the structure, where the upper as-fabricated skin
can be seen, along with the edges of the webs. Figure 7.4c and d show different areas in the structure,
highlighting web buttresses, arches and missing walls.
7.3.3 Mechanical Properties of Stochastic Honeycombs
Mechanical properties for the four grades of polypropylene, with core densities ranging from 7 – 14%, were
measured in uniaxial out-of-plane compression (Figure 7.5). Across all samples, the structure first underwent
elastic compression to a peak load (σp). After the peak, plastic post-buckling occurs and the strength decreases to
a valley stress after which densification begins as the webs collapse into each other. Figure 7.5a shows
representative stress–strain curves for each of the four different PPs at a core density of 11%. Even though all
four of these samples have the same core density, PP-1 and PP-2 have peak strengths nearly double that of PP-3
and PP-4. Figure 7.5b plots the peak strength as a function of core relative density for each of the four PPs. Over
Figure 7.5 Stress (σ) – strain (ε) curves for the out-of-plane compression of stochastic honeycombs fabricated with PP-1, PP-2, PP-3
and PP-4 with core densities (ρcore) of 11% (a), and the peak strength (σp) of the stochastic honeycombs as a function of
ρcore (b).
84
the range of densities, PP-1and PP-2 are very similar, as are PP-3 and PP-4. PP-1 and PP-2 show a three-fold
increase in strength over a core density range of 8 – 14%, while PP-3 shows a two-fold increase in strength over
a core density range of 7 – 11%, and PP-4 shows a two-fold increase in strength over a core density range of 9 –
14%. The following section develops a mechanical model to can explain these ranges of properties.
7.4 Discussion
Significant work has been done on the out-of-plane properties of regular honeycombs through experimentation
[20-24], buckling models [21-25] and finite element analysis (FEA) [25-27]. Irregular honeycombs have been
studied in in-plane compression through FEA [28-33], as have regular and irregular honeycombs with defects
(missing walls) [33-36]. Given the complex internal architecture of stochastic honeycombs, the first step is to
provide quantitative relationships for the web geometry and connectivity. The relative significance of material
and architectural effects can then be seen by developing a buckling based model for the compressive strength of
stochastic honeycombs.
The key architectural factors would be the geometry of the webs (length, thickness) and their connectivity.
In order to determine these properties, the structure was analyzed using Minkowski functionals. These
relationships come from integral geometry, and can be used to quantify the connectivity and geometry of
complex shapes [37, 38]. They are often used in the analysis of open cell foams and other porous materials [37,
39], and can be used to extract quantifiable parameters from 3D reconstructions or 2D images [40].
In 3D, the Minkowski functionals are volume, V, surface area, AS, mean breadth, B, and the Euler number,
χ3D [41, 42]. Mean breadth applies only to convex surfaces [42], so it is neglected in the present analysis. V, AS
and χ3D were computed by running binary stacks of images from the X-ray tomography data through a Matlab
program developed by Legland et al. [43]. χ3D is a measure of the connectivity of the structure, and is defined as
the number of disconnected volumes minus the number of fully enclosed pockets [41, 42]. For the as-fabricated
stochastic honeycomb, χ3D = 1 for all PPs, because the structure is fully connected and there are only minor
enclosed air pockets that can be neglected. The volume and surface area were computed over the central third of
the height of the structures, which was shown to have a constant cross-section in Chapter 5 [1]. The V and AS
functionals can be expressed as the ratio AS/V, which increases as the number of webs goes up and decreases as
the webs become thicker. Surface area-to-volume ratios are often used to characterize foams, and specifically as
a measure of porosity [44-47]. This variable can also be seen as the inverse of an effective web thickness:
85
(7.4)
Figure 7.6 plots AS/V as a function of core density; AS/V decreases as core density increases, and at
approximately the same rate for each PP. This implies that as the core becomes denser, the webs generally
become thicker rather than increasing significantly in total length, matching the conclusions drawn in Chapter 5.
Note that PP-3 generally has a higher AS/V for a given than the other polymers (e.g., at 9.5% core density,
PP-3 has an AS/V value of 8 mm-1
, while PP-1 has an AS/V value of 5 mm-1
) which suggests that in addition to
affecting the partitioning between core and skins, the shear viscosity may also affect the total length of webs in
the cross-section.
Buckling of a thin plate is an extension of Euler’s buckling of a column, developed by Bryan in 1890 [48]
and extended by Timoshenko and Gere in 1961 [49]:
(7.5)
where Et (tangent modulus) and ν (Poisson’s ratio) are material property values, K is related to the end
constraints of the thin plate, and tweb (thickness) and b (length) are geometric parameters. The thickness of the
webs, tweb, was chosen as the independent variable, as this would allow relatively simple comparison to the
experimental strength values through use of the Minkowski functionals (in the form of (AS/V)-1
= teff). Since
failure in stochastic honeycombs is initially elastic, the tangent modulus is taken as the slope of the initial linear
elastic portion of the tensile stress–strain curves for each PP. Note that a mean value of ν = 0.4 was taken for the
Poisson’s ratio in the present study [16, 50-53].
Figure 7.6 Surface area over volume (AS/V) as a function of core density (ρcore) for PP-1, PP-2, PP-3 and PP-4.
86
End constraints divide the webs in the stochastic honeycomb into two broad groups: bound on each side
(denoted with a subscript B) and bound on one side, free on the other (denoted with a subscript F). The value of
K will thus be different for each type of web, as will the length of the webs. Mid-height X-ray images were used
to classify each web as either B- or F-type. For each stochastic honeycomb, the length fraction of bound (f) and
bound–free (1 – f) webs were determined, as were the average length of the webs (bB and bF). To determine K,
the type of constraint must first be chosen. In this first study, a conservative simply supported boundary
condition was selected. This choice minimizes the K values, and implies that the webs have no curvature or
deflection at the edges parallel to the loading axis [54]. Given the buckling and folding observed in the study
presented in Chapter 6 [19], this is reasonable since the model applies to the initial peak stress, on the cusp of
buckling failure. In addition, K is a function of the axial aspect ratio of the webs, or a/b, where a is the height of
the sample without the skin. Values for KB and KF are thus obtained using the standard methods [54, 55], as
detailed in Chapter 2.
The thin plate buckling equation can be split into a material properties portion and an architectural portion,
as seen in Eqn. (7.6):
(7.6)
Since tweb is the independent variable and there is a fraction of B and F webs, this can be re-written as Eqn. (7.7):
(7.7)
σcr changes linearly with Et, while f has a smaller effect. Buckling model parameters (KB, KF, bB and bF) varied
by less than 10% and are summarized in Table 7.4 along with the tangent modulus, Et. Overall, there was good
agreement between the composite buckling model and the experimentally measured peak strengths (model and
experimental data for the weakest and strongest honeycombs are shown in Figure 7.7).
Some insight into the role of the geometric parameters in determining the mechanical performance of the
stochastic honeycombs can be seen by plotting contour lines of density and specific strength over the
architectural variables of web thickness (here normalized by the sandwich core height, a) and the fraction of
bound webs, f in Figure 7.8 (Et = 600 MPa and bB, bF, KB and KF taken as the average of the values given in
Table 7.4). These relationships help to explain the experimental observations of the present study.
87
Table 7.4 Buckling model parameters determined from the X-ray images of the webs, and the tangent modulus, Et, as determined
from the tensile testing.
Variable PP-1 PP-2 PP-3 PP-4
KB 4.04 4.05 4.01 4.00
bB (mm) 6.08 7.26 6.38 6.57
f 0.550 0.636 0.489 0.508
KF 0.465 0.506 0.446 0.458
bF (mm) 4.35 5.00 3.96 4.23
Et (MPa) 784 704 691 427
Figure 7.7 Predicted strength (σcr) from the composite buckling model plotted as a function of web thickness (tweb) for PP-1 and PP-
4, overlaid with the experimental data (σp) plotted as a function of effective web thickness (teff).
Figure 7.8 Normalized web thickness (t/a) plotted as a function of the fraction of bound webs (f) for various specific strengths (σ/ρ).
The core density (ρcore) fraction for a given web thickness is also displayed.
88
PP-1 and PP-2 have the highest strengths of the four polymers, with a similar Et and greater fraction of
bound webs, f, than the other two PPs. PP-3 has a similar Et to PP-1 and PP-2, but a lower f. Perhaps more
significantly, however, is the range of web thicknesses in this architecture. PP-1, 2 and 4 have teff ranging from
0.18 mm to 0.25 mm over a core density range of 9 – 11%. Over the same density range, teff for PP-3 ranged
from 0.13 mm to 0.18 mm. Thus, when plotted as a function of core density, PP-3 appears to be much weaker
than PP-1 and PP-2, but when strength is plotted as a function of web thickness, much of the difference
disappears. Finally, PP-4 has a significantly lower Et than the other PPs, likely a factor of the significantly lower
crystallinity observed in PP-4, and accounting for its lower strength.
7.5 Conclusions
Differential scanning calorimetry and tensile tests were performed on four PPs with varying rheological
properties. Stochastic honeycombs with relative densities ranging from 10 – 17% were then fabricated for each
PP, and the as-fabricated skin thicknesses and core density were determined for each sample. A subset of
samples were characterized by X-ray tomography to determine the internal architecture, and the samples were
subsequently crushed in out-of-plane compression. Minkowski functionals were used in conjunction with the
architectural parameters from the X-ray scan data to develop a composite thin plate buckling model as a first step
towards modeling the strength of stochastic honeycombs.
From the Minkowski functionals, a teff (= (AS/V)-1
) value was calculated and it was shown that this
parameter can be more valuable than ρcore when determining the relative strengths of stochastic honeycombs
fabricated from different PPs. PP-3 has a lower teff for a similar ρcore to PP-1 and PP-2, leading to a lower
strength. A comparison of PP-1 and PP-2 with PP-4 demonstrates the importance of the stiffness of the base
polymer. Through the thin plate buckling model, it was shown that the PP-4 stochastic honeycomb is weaker
than the PP-1 and PP-2 stochastic honeycombs due to the stiffness of the polymer and not to a difference in the
architecture.
The composite buckling model separates the effects of material properties from architecture, giving
insight into the mechanisms that strengthen stochastic honeycombs. This also demonstrates that as a first step, a
simple buckling model can be used to study the effect of missing or broken walls in a regular or irregular
honeycomb in out-of-plane compression.
89
References
[1] M. Hostetter, G.D. Hibbard. Architectural characteristics of stochastic honeycombs fabricated from varying melt
strength polypropylenes, Journal of Applied Polymer Science 131 (2014) 8.
[2] D.W. Van Krevelen. Properties of Polymers, Their Estimation and Correlation with Chemical Structure. 4 ed.,
Elsevier, New York, 2009.
[3] A. Galeski. Crystallization. in: Karger-Kocsis J, (Ed.). Polypropylene: An A-Z Guide, vol. 2. Kluwer Publishers,
Dordrecht, 1999. pp. 135-141.
[4] C. Maier, T. Calafut. Polypropylene: The Definitive User's Guide and Databook, Plastics Design Library,
Norwich, NY, 1998.
[5] F. Ide, A. Hasegawa. Studies on polymer blend of nylon 6 and polypropylene or nylon 6 and polystyrene using
the reaction of polymer, Journal of Applied Polymer Science 18 (1974) 963-974.
[6] P. Svoboda, C. Zeng, H. Wang, L.J. Lee, D.L. Tomasko. Morphology and mechanical properties of
polypropylene/organoclay nanocomposites, Journal of Applied Polymer Science 85 (2002) 1562-1570.
[7] G. Natta, P. Corradini. Structure and properties of isotactic polypropylene, Il Nuovo Cimento Series 10 15 (1960)
40-51.
[8] A.S. Nielsen, D.N. Batchelder, R. Pyrz. Estimation of crystallinity of isotactic polypropylene using Raman
spectroscopy, Polymer 43 (2002) 2671-2676.
[9] S. Haftka, K. Könnecke. Physical properties of syndiotactic polypropylene, Journal of Macromolecular Science,
Part B: Physics 30 (1991) 319-334.
[10] H. Asakawa, K. Nishida, T. Kanaya, M. Tosaka. Giant single crystal of isotactic polypropylene showing near-
equilibrium melting temperature, Polymer Journal 45 (2013) 287-292.
[11] K. Yamada, M. Hikosaka, A. Toda, S. Yamazaki, K. Tagashira. Equilibrium melting temperature of isotactic
polypropylene with high tacticity. 2. Determination by optical microscopy, Macromolecules 36 (2003) 4802-
4812.
[12] K. Yamada, M. Hikosaka, A. Toda, S. Yamazaki, K. Tagashira. Equilibrium melting temperature of isotactic
polypropylene with high tacticity: 1. Determination by differential scanning calorimetry, Macromolecules 36
(2003) 4790-4801.
[13] S.Z.D. Cheng, J.J. Janimak, A. Zhang, E.T. Hsieh. Isotacticity effect on crystallization and melting in
polypropylene fractions: 1. Crystalline structures and thermodynamic property changes, Polymer 32 (1991) 648–
655.
[14] K. Nakamura, S. Shimizu, S. Umemoto, A. Thierry, B. Lotz, N. Okui. Temperature dependence of crystal growth
rate for α and β forms of isotactic polypropylene, Polymer Journal 40 (2008) 915-922.
[15] J.R. Fried. Polymer Science and Technology. 2nd ed., Prentice Hall PTR, Upper Saddle River, NJ, 2003.
[16] CES Polymer Selector. Cambridge, UK: Granta Material Intelligence, 2013.
[17] G. Perego, G.D. Cella, C. Bastioli. Effect of molecular weight and crystallinity on poly(lactic acid) mechanical
properties, Journal of Applied Polymer Science 59 (1996) 37-43.
[18] A. van der Wal, J.J. Mulder, R.J. Gaymans. Fracture of polypropylene: The effect of crystallinity, Polymer 39
(1998) 5477–5481.
90
[19] M. Hostetter, B. Cordner, G.D. Hibbard. Stochastic honeycomb sandwich cores, Composites Part B: Engineering
43 (2012) 1024-1029.
[20] S. Xu, J.H. Beynon, D. Ruan, G. Lu. Experimental study of the out-of-plane dynamic compression of hexagonal
honeycombs, Composite Structures 94 (2012) 2326–2336.
[21] J. Zhang, M.F. Ashby. The out-of-plane properties of honeycombs, International Journal of Mechanical Sciences
34 (1992) 475-489.
[22] F. Côté, V.S. Deshpande, N.A. Fleck, A.G. Evans. The out-of-plane compressive behavior of metallic
honeycombs, Materials Science and Engineering: A 380 (2004) 272-280.
[23] Y. Aminanda, B. Castanié, J. Barrau, P. Thevenet. Experimental analysis and modeling of the crushing of
honeycomb cores, Applied Composite Materials 12 (2005) 213-227.
[24] H.S. Lee, S.H. Hong, J.R. Lee, Y.K. Kim. Mechanical behavior and failure process during compressive and shear
deformation of honeycomb composite at elevated temperatures, Journal of Materials Science 37 (2002) 1265-
1272.
[25] S. Liang, H.L. Chen. Investigation on the square cell honeycomb structures under axial loading, Composite
Structures 72 (2006) 446-454.
[26] A. Domenico, A. Ferdinando, M. Costantino, M. Simone, P. Andrea, R. Alessandro. Statistical finite element
analysis of the buckling behavior of honeycomb structures, Composite Structures 105 (2013).
[27] X. Fan, I. Verpoest, D. Vandepitte. Finite element analysis of out-of-plane compressive properties of
thermoplastic honeycomb, Journal of Sandwich Structures and Materials 8 (2006) 437-458.
[28] J. Hohe, C. Beckmann. Probabilistic homogenization of hexagonal honeycombs with perturbed microstructure,
Mechanics of Materials 49 (2012) 13-29.
[29] M.J. Silva, W.C. Hayes, L.J. Gibson. The effects of non-periodic microstructure on the elastic properties of two-
dimensional cellular solids, International Journal of Mechanical Sciences 37 (1995) 1161-1177.
[30] H.X. Zhu, J.R. Hobdell, A.H. Windle. Effects of cell irregularity on the elastic properties of 2D Voronoi
honeycombs, Journal of the Mechanics and Physics of Solids 49 (2001) 857-870.
[31] H.X. Zhu, S.M. Thorpe, A.H. Windle. The effect of cell irregularity on the high strain compression of 2D
Voronoi honeycombs, International Journal of Solids and Structures 43 (2006) 1061-1078.
[32] J. Lee, J.B. Choi, K. Choi. Application of homogenization FEM analysis to regular and re-entrant honeycomb
structures, Journal of Materials Science 31 (1996) 4105-4110.
[33] M.J. Silva, L.J. Gibson. The effects of non-periodic microstructure and defects on the compressive strength of
two-dimensional cellular solids, International Journal of Mechanical Sciences 39 (1997) 549-563.
[34] J.M. Albuquerque, M. Fátima Vaz, M.A. Fortes. Effect of missing walls on the compression behaviour of
honeycombs, Scripta Materialia 41 (1999) 167-174.
[35] X.E. Guo, L.J. Gibson. Behavior of intact and damaged honeycombs: a finite element study, International Journal
of Mechanical Sciences 41 85-105.
[36] A.-J. Wang, D.L. McDowell. Effects of defects on in-plane properties of periodic metal honeycombs,
International Journal of Mechanical Sciences 45 (2003) 1799–1813.
[37] C. Lautensack, T. Sych. 3D image analysis of open foams using random tesselations, Image Analysis and
Stereology 25 (2006) 87-93.
91
[38] K.R. Mecke. Additivity, Convexity, and Beyond: Applications of Minkowski Functionals in Statistical Physics.
in: Mecke KR, Stoyan D, (Eds.). Lecture Notes in Physics, vol. 554. Springer Berlin Heidelberg, 2000. pp. 111-
184.
[39] G. Frenkel, R. Blumenfeld, P.R. King, M.J. Blunt. Topological analysis of foams and tetrahedral structures,
Advanced Engineering Materials 11 (2009) 169-176.
[40] G.E. Schröder-Turk, W. Mickel, S.C. Kapfer, M.A. Klatt, F.M. Schaller, M.J.F. Hoffmann, N. Kleppmann, P.
Armstrong, A. Inayat, D. Hug, M. Reichelsdorfer, W. Peukert, W. Schwieger, K. Mechke. Minkowski tensor
shape analysis of cellular, granular and porous structures, Advanced Materials 23 (2011) 2535-2553.
[41] K. Michielsen, H. De Raedt. Integral-geometry morphological image analysis, 347 (2001) 461–538.
[42] . hser, . M cklich. Statistical Analysis of Microstructures in Materials Science, Wiley & Sons, West Sussex,
England, 2000.
[43] D. Legland, K. Kiêu, M.-F. Devaux. Computation of Minkowski measures on 2D and 3D binary images, Image
Analysis & Stereology 26 (2007) 83-92.
[44] A.G. Mikos, A.J. Thorsen, L.A. Czerwonka, Y. Bao, R. Langer, D.N. Winslow, J.P. Vacanti. Preparation and
characterization of poly(l-lactic acid) foams, Polymer 35 (1994) 1068-1077.
[45] A.M. Kraynik, D.A. Reinelt, F. van Swol. Structure of random foam, Physical Review Letters 93 (2004).
[46] L. Lu, S.J. Peter, M.D. Lyman, H.-L. Lai, S.M. Leite, J.A. Tamada, J.P. Vacanti, R. Langer, A.G. Mikos. In vitro
degradation of porous poly(l-lactic acid) foams, Biomaterials 21 (2000) 1595–1605.
[47] J. Zeschky, T. Hofner, C. Arnold, R. Weibmann, D. Bahloul-Hourlier, M. Scheffler, P. Greil. Polysilsesquioxane
derived ceramic foams with gradient porosity, Acta Materialia 53 (2005) 927–937.
[48] G.H. Bryan. On the stability of a plane plate under thrusts in its own plane, with applications to the "buckling" of
the sides of a ship, Proceedings of the London Mathematical Society s1-22 (1890) 54-67.
[49] J.M. Gere, S.P. Timoshenko. Theory of Elastic Stability. 2nd ed., McGraw-Hill, New York, New York, 1961.
[50] J.A. Rinde. Poisson's ratio for rigid plastic foams, Journal of Applied Polymer Science 14 (1970) 1913-1926.
[51] W.D. Callister, D.G. Rethwisch. Fundamentals of Materials Science and Engineering: An Integrated Approach. 3
ed., John Wiley & Sons, Inc., Hoboken, New Jersey, 2008.
[52] A.M. Zihlif, R.A. Duckett, I.M. Ward. The Poisson's ratio of ultra-drawn polyethylene and polypropylene fibres
using Michelson interferometry, Journal of Materials Science 13 (1978) 1837-1840.
[53] D.W. Van Krevelen, K. Te Nijenhuis. Properties of Polymers. Their Correlation with Chemical Structure; Their
Numerical Estimation and Prediction from Additive Group Contributions. Amsterdam: Elsevier, 2009.
[54] H.G. Allen, P.S. Bulson. Background to buckling, McGraw-Hill Book Co., New York, New York, 1980.
[55] F. Bloom, D. Coffin. Handbook of thin plate buckling and postbuckling, Chapman & Hall/CRC, Boca Raton,
Florida, 2000.
92
Chapter 8
Post-Peak Collapse and Energy Absorption in Stochastic Honeycombs
M. Hostetter, G.D. Hibbard. Post-peak collapse and energy absorption in stochastic honeycombs (2015) In
preparation for submission to the Journal of Materials Science
8.1 Introduction
Light-weight cellular materials are used in many applications where in addition to high strength– and stiffness–
to–weight ratios, the energy absorption upon collapse is important. In these materials the internal architecture
generally collapses under a combination of deformation mechanisms that result in the overall stress–strain
response after failure has been initiated. Foams, considered ideal energy absorbers [1], have a nearly constant
(plateau) collapse strength up to densification. In contrast, honeycombs in out-of-plane compression have a
much higher initial peak strength, followed by a generally steeper descent to a valley strength, where the stress
eventually rises again during densification [2]. For a given material and density, honeycombs tend to absorb
more total energy in out-of-plane compression than foams, but the stress decrease can be a drawback in certain
applications [1].
Thermoplastic foams are typically produced by using blowing agents such as carbon dioxide or nitrogen to
create bubbles in the melt [2, 3], or by introduction of volatile compounds (either low melting point liquids or
additives which react and/or decompose) that will form vapour bubbles upon heating [2], and then injection
moulding the final product [3]. Thermoplastic honeycombs can be fabricated through a number of methods,
including extruding sheets of plastic and cutting the sheets into blocks which are then thermally welded or
bonded in narrow strips, and expanded into a honeycomb shape [4], or extruded sheets cuts into strips and slotted
together [4], or through extrusion into a complex mold [5]. Each of these processes require multiple steps to
create a sandwich panel.
In contrast, melt-stretched stochastic honeycomb sandwiches can be fabricated in a single step from the
polymer melt, and have comparable mechanical properties to commercially available honeycombs [6]. Not only
does this simple melt-stretching process create a material that can be easily recycled, but it also allows for the
93
possibility of refabrication. For example, in an initial study, samples were fabricated by melt-stretching, tested
mechanically to collapse, re-melted and stretched to form a new honeycomb, and tested mechanically once more
[6]; the strength of the refabricated samples fit within the standard deviation of as-fabricated samples at the same
densities. This means that one need not go through the normal steps of recycling (e.g. separating, regrinding and
pelletizing [7]); one can simply re-melt the collapsed honeycomb, and melt-stretch it once again into a sandwich
panel. This processing characteristic may be particularly attractive for energy absorption applications.
Stochastic honeycombs have a cross-sectional area fraction of unity at the skins which decreases to a
cross-sectional area fraction (Ac) that is related to the core density and is nearly constant over the central third of
the honeycomb height [8]. Between the skin and the plateau, as the cross-sectional area is changing rapidly,
buttresses and archways between neighbouring webs provide added support to resist rotation at the top and
bottom of the stochastic honeycomb. These architectural features are not present in conventional honeycombs
(where the web cross-section is constant over the sample height) and would therefore be expected to affect the
post-peak collapse mechanisms and energy absorption capacity.
Given this architectural complexity and the potential usefulness of melt-stretched honeycombs as energy
absorbers, the following study investigates the post-peak collapse mechanisms by X-ray tomography, and the
results are compared to the compression curves from previously analyzed coupons (Chapter 7, [9]).
8.2 Experimental Details
The stochastic honeycombs were fabricated by melting high melt strength polypropylene (PP-1) at 1.1 times the
melting temperature, compressing the melt, and stretching it uniaxially to the desired height of the sample, h =
14 mm, in the manual press (details in Chapter 3). Square samples with an edge length of w = 40 mm were cut
from fabricated sheets. Samples were fabricated over the core relative density (ρcore) range of 7 – 18%. The skin
thickness for all the samples were measured, and varied from 2% to 5% of the height of the sample. The
stochastic honeycombs were tested in out-of-plane compression at a stroke rate of 1 mm min-1
. The internal
structure of four samples (ρcore = 10.6, 13.0, 15.8 and 17.7%), called A, B, C and D respectively, were
characterized through X-ray tomography using a Skyscan 1172 micro-CT scanner. To investigate the post-peak
collapse mechanisms, samples were pre-loaded to characteristic strain values and examined once again by X-ray
tomography. Samples A and D were compressed to the beginning of the valley (εv = 0.17 and 0.20 respectively),
and unloaded, releasing the stored elastic energy. Samples B and C underwent the same process, but were
94
characterized after pre-loading to the peak (εp= 0.05 and 0.06 respectively), the beginning of the valley (εv = 0.13
and 0.18), and just before the onset of densification (εd = 0.33 and 0.36).
8.3 Results
Figure 8.1 presents reconstructed 3-dimensional X-ray tomography views of the 4 sample types. Previous work
has shown that the dimensions of the webs vary within any given sample, in terms of the length and thickness of
the webs, as well as two classes of webs with different end constraints. Additionally, sample-to-sample, the
number and arrangement of the webs differs, and as density increases, the webs become thicker, even as the total
length of webs remains the same [8, 9]. The stress–strain curves for each sample, indicating where the samples
were unloaded and scanned, are presented in Figure 8.2, with Table 8.1 summarizing ρcore, Ac, the peak
compressive stress, σp, and the valley stress, σv, for all 4 sample types. Note that the stress upon re-loading was
lower than the stress upon unloading. In effect, the load–unload–reload testing procedure of the present study is
a type of interrupted cyclic loading. During the first few cycles of a cyclic fatigue test, polypropylene exhibits a
strength decrease on the order of 10 – 20% [11-13]; during stress relaxation, voids are created, beginning the
crack formation process [14]. The lower stress upon re-loading of the compression coupon after X-ray
tomography at εp and εv (Figure 8.2) is likely due to the fact that a fast decay process begins immediately upon
unloading in the portions of the stochastic honeycomb that have deformed the most.
Figure 8.1 3D reconstructions of samples A, B, C and D (a–d, respectively) with core relative densities as shown in Table 8..
Table 8.1 Relative density (ρcore), cross-sectional area fraction (Ac), peak strength (σp) and valley strength (σv) for samples A–D.
Sample ρcore (%) Ac (mm2 mm
-2) σp (MPa) σv (MPa)
A 10.6 0.085 ± 0.004 2.23 0.88 ± 0.01
B 13.0 0.079 ± 0.003 2.90 0.99 ± 0.06
C 15.8 0.099 ± 0.004 3.52 1.53 ± 0.04
D 17.7 0.113 ± 0.005 4.61 2.25 ± 0.11
95
Figure 8.2 Successive compressive stress–strain curves for samples A–D with core relative densities as shown in Table 8.. The
samples were characterized through X-ray tomography in the as-fabricated condition (εo = 0), after loading to the peak
(εp), the beginning of the valley (εv) and just before the onset of densification (εd). The samples were unloaded and
scanned, and then re-loaded. The stress–strain curves were then stitched together.
The post-peak regions of the stress–strain curves in Figure 8.2 represent a complex sequence of failure
mechanisms. In what follows, we will first present the failure of sample type C in order to illustrate the
progression of collapse for a given architecture. Then we will move to compare failure in samples of varying
core density.
Figure 8.3a shows the mid-height cross-section (P = 0.5, where P is the fractional vertical position in the
sample, from 0 at the bottom to 1 at the top) of sample C, with ρcore = 15.8%. Six specific webs, covering two
broad categories, are highlighted. The classification scheme of the webs is based on those that are bound at both
ends (denoted with a subscript B) and those that are free on one side and bound on the other (denoted with a
subscript F) (Figure 8.3b,c). Additionally, anywhere two or more webs meet is called a node.
The 45 interior webs in sample C were characterized after excluding exterior webs that were cut when the
as-fabricated sheets were sectioned into samples. Of the 45 webs, 26 (58%) were classified as bound (CB), and
19 (42%) were bound–free (CF). The six webs marked on Figure 8.3a will be discussed in detail. Table 8.2 gives
the mid-height length (b), area (Aw), perimeter (Pw) and average thickness (tweb) for these six webs, as well as the
height/length ratio, a/b. tweb was calculated by averaging 5 measurements of the web thickness along the length
of the web at mid-height.
96
Figure 8.3 The mid-height cross-section for sample C (a) and schematics of bound–bound, B (b) and bound–free, F (c) webs with
the length, b and the height, a as shown
Table 8.2 Dimensions of various webs at mid-height from sample C.
Web b (mm) Aw (mm2) Pw (mm) tweb (mm) a/b
CB-1 8.7 5.20 23.5 0.63 ± 0.35 1.55
CB-2 5.6 2.12 13.3 0.37 ± 0.07 2.40
CB-3 8.3 3.48 8.9 0.41 ± 0.16 1.64
CF-1 1.8 0.83 5.1 0.39 ± 0.18 7.34
CF-2 1.4 0.84 4.9 0.56 ± 0.41 9.57
CF-3 3.1 2.11 7.8 0.61 ± 0.16 4.43
Figure 8.4 and Figure 8.5 show 3D reconstructions from two different perspectives on CB-1 and CB-2,
taken at strains corresponding to the points labelled εo, εp, εv and εd in Figure 8.2, and a through-thickness cross-
section from the centre of the webs (bmid) extending from one skin to the other. These two webs demonstrate
typical collapse mechanisms for bound webs in stochastic honeycombs. The web CB-1 has a larger length, b,
than CB-2, and has a much lower aspect ratio (1.55 compared to 2.40). Both webs show a slight initial curvature,
which is more pronounced at the peak (εp). This increase in curvature is due to plastic deformation, but
pronounced buckling has not yet begun. By the beginning of the valley, at εv, CB-2 has a significant fold with
some small cracks that do not extend through the thickness of the web, while CB-1 shows two horizontal
fractures perpendicular to the loading axis at the fold. The ends of the fold and horizontal fractures terminate in
vertical fractures near the nodes. At the end of the valley, in CB-1 the edges of the fold have completely
fractured, while the curvature of CB-2 has increased without fracturing.
97
Figure 8.4 Web CB-1, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve (εp), at the beginning of
the valley (εv) and the end of the valley before densification (εd). The first two images at each strain show the web from
either side, and the third is a through-thickness cross-section taken in the centre of the web.
Figure 8.5 Web CB-2, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve (εp), at the beginning of
the valley (εv) and the end of the valley before densification (εd). The first two images at each strain show the web from
either side, and the third is a through-thickness cross-section taken in the centre of the web.
98
In addition to dimensional variance and curvature, other defects, such as the hole in the centre of CB-3
(Figure 8.6), were present. The through-thickness cross-sections for CB-3 are from the left, centre and right of the
images directly above the cross-sections. CB-3 has a similar length, b, to CB-1, as seen in Table 8.2, and collapses
in much the same way. At the peak, the curvature of the webs has increased, and at the valley, horizontal cracks
are beginning to form, but have not yet fractured through the web thickness. By the end of the valley, at εd, a
vertical tear where the web meets the node has formed, and horizontal fractures have appeared around the hole.
Overall, the mechanism of collapse in this web is the same as in CB-1. The hole may affect the strength of the
web, but it does not affect the collapse mechanism.
Webs that are bound on one side and free on the other had more variation in dimension and in collapse
mechanism. Three webs, termed CF-1, CF-2 and CF-3, are presented in Figure 8.7 – Figure 8.9, along with their
through-thickness profiles. Figure 8.7 presents one web that has a noticeable change in thickness through the
height of the web. At a vertical position, P (the fractional position in the core, from 0 at the bottom to 1 at the
top) of 0.1, the thickness is 0.53 ± 0.10 mm, decreasing to a minimum of 0.33 ± 0.17 mm at P = 0.7, and
increasing to 0.68 ± 0.23 mm at P = 0.9. At the peak strain, the curvature has increased, and by the valley strain,
a large fold and small horizontal fractures have appeared. The upper fold occurs near the minimum thickness at
P = 0.7, and fractures in the thicker portion closer to the bottom. The fractures grow and the curvature at the fold
increases as densification approaches (Figure 8.7, εd). CF-2, in Figure 8.8, has a much more dramatic change in
thickness, as seen in the thickness profiles. The thickness decreases from 1.72 ± 1.10 mm at P = 0.9 to 0.29 ±
0.21 mm at P = 0.3, and increases slightly to 0.47 ± 0.41 mm at P = 0.1. The thickness on any horizontal plane
through this web decreases sharply from the axial node to the free edge, as seen in the large standard deviations,
as well as varying from top to bottom. All of the deformation occurs in the thinner portions (P = 0.1 to 0.4), with
a large kink at the valley (εv) and a horizontal fracture above. CF-3 (Figure 8.9), in contrast, is thicker overall
(maximum tweb = 1.01 ± 0.78 mm at P = 0.9) and has a more uniform thickness profile. This web, however,
contains a small archway, and this causes the web to collapse by localized buckling near the bottom (P = 0.2, tweb
= 0.48 ± 0.14 mm), as opposed to fracturing or buckling near the centre of the web. The bend in the web at
densification occurs near P = 0.4, with tweb = 0.56 ± 0.17 mm.
99
Figure 8.6 Web CB-3, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve (εp), at the beginning of
the valley (εv) and the end of the valley before densification (εd). The first two images at each strain show the web from
either side, and the third is a through-thickness cross-section taken in thirds along the lower image of the web.
Figure 8.7 Web CF-1, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve (εp), at the beginning of
the valley (εv) and the end of the valley before densification (εd). The first two images at each strain show the web from
either side, and the third is a through-thickness cross-section taken in the centre of the web.
100
Figure 8.8 Web CF-2, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve (εp), at the beginning of
the valley (εv) and the end of the valley before densification (εd). The first two images at each strain show the web from
either side, and the third is a through-thickness cross-section taken in the centre of the web.
Figure 8.9 Web CF-3, in the as-fabricated condition (εo), after loading to the peak of the stress–strain curve (εp), at the beginning of
the valley (εv) and the end of the valley before densification (εd). The first two images at each strain show the web from
either side, and the third is a through-thickness cross-section taken in the centre of the web.
101
The detailed analysis that was performed on sample C was extended to samples A, B, and D. In these
samples the interior webs were analyzed as before (the number of webs characterized was as follows: AB – 30,
AF – 14; BB – 18, BF – 17; DB – 28, DF – 10). Select webs demonstrating the range of collapse mechanisms are
shown in Figure 8.10, with these images taken as-fabricated (εo) and at the valley (εv). Some webs showed no
plastic deformation at the valley, indicating that these webs had buckled elastically and rebounded when the load
was removed (Figure 8.10a). Webs were also observed to buckle plastically, where some buckled near the skin
(Figure 8.10b), and some in the central portions of the web (Figure 8.10c). In addition to plastic buckling, many
webs fractured. Some only fractured perpendicular to the loading direction, termed horizontally (Figure 8.10d),
and some also fractured parallel to the loading direction, termed vertically (Figure 8.10e).
The post-peak collapse mechanisms can be divided into three main categories: elastic buckling, plastic
buckling, and plastic buckling with fracture. These mechanisms vary from web to web within one sample, from
sample to sample, and with density, but the mechanism evident at εv is the one that will be discussed. Table 8.3
presents the overall fraction of webs in each sample that failed in each manner, averaged between the bound and
bound–free webs. Generally, the lower density sample, A, had thinner webs, and it was only here where a
significant percentage (~ 10%) of webs showed elastic buckling. It is likely that these webs absorbed mainly
elastic energy and released that deformation upon unloading, as seen in Figure 8.10a. 6% of B and 4% of C webs
failed in the same manner, and none of the D webs exhibited only elastic buckling. Plastic buckling can be
broken down into webs that buckled near the skin and those that buckled in the central portion of the core.
Buckling near the skin occurred where there was a significant change in the profile of the web near the upper or
Figure 8.10 Webs AB-1 (a), BB-1 (b), AF-1 (c), DF-1 (d), and DB-2 (e), in the as-fabricated state (εo) and at the valley (εv). These webs
demonstrate the three collapse mechanisms: elastic buckling, plastic buckling near the skin, plastic buckling in the central
portion of the web, plastic buckling with horizontal tears, and plastic buckling with horizontal and vertical tears, from a –
e respectively.
102
Table 8.3 The percentage of webs from each sample type that exhibited each mechanism of deformation at the valley. Reported
values are based on the characterization of at least 35 webs in each sample.
Collapse Mechanism Sample A
ρcore = 10.6%
Sample B
ρcore = 13.0%
Sample C
ρcore = 15.8%
Sample D
ρcore = 17.7%
Elastic Buckling 10% 6% 4% 0%
Plastic Buckling 43% 38% 32% 17%
Plastic Buckling and
Fracture 47% 56% 64% 83%
lower skin. Across B and F webs, as the cross-sectional area fraction increased, the likelihood of fracture
occurring in the web increased as well. This leads to the conclusion that the distribution of material through the
core is important for predicting the core collapse mechanisms.
8.4 Discussion
A hollow cylinder under out-of-plane compression plastically buckles in a series of folds [15, 16] and this
analysis has been extended analytically and through finite element analysis to square honeycomb unit cells [17,
18] and other honeycomb geometries [19-22]. Stochastic honeycombs do not have closed cells as most
honeycombs have, but the collapse mechanisms of two or more webs joined at a node could be expected to be
linked. It has been shown that thin walls joining at a corner, and under compression, have a linked collapse
mechanism. These corners were termed "super-folding elements" [23, 24] that undergo coordinated plastic
buckling across more than one web. However, due to the irregular dimensions of the webs in stochastic
honeycombs, this does not appear to occur. The specific collapse mechanism, i.e., the number of folds as well as
the vertical position of each fold, of every web in sample C were compared across all webs that joined at a given
axial node. No significant link between the vertical location or number of folds was found. The nodes tend to be
thicker than the webs, and in many cases where a thinner web had buckled and met a thicker node, a vertical
fracture developed (see Figure 8.4a, εv).
In a previous study of melt-stretched stochastic honeycombs, we developed a thin-plate buckling model to
examine the peak strength in out-of-plane compression (Chapter 7 [9]). That study looked at four PPs over a
range of core densities (28 sample types in total) but did not examine the post-peak portions of the stress–strain
curves or the energy absorption. The energy absorption capacity of these sample types has now been analyzed
and those results are reported here for the first time.
The energy absorbed per unit volume can be found by integrating under the σ–ε curve to an arbitrary final
strain, εf:
103
(8.1)
Various choices of strain for εf appear in the literature, and the choice of εf can greatly affect J. Four particular
choices for εf will be compared. They are defined as:
i. εf is a fixed strain across all samples and densities, εf = 0.4 [25, 26]
ii. εf is the strain at which σ = 2σp, [27]
iii. εf is the onset of densification, εd, where εd is the strain at the intersection of the tangent to the
σ–ε curve in the valley and the tangent at σ = 2σp [28, 29]
iv. εf is the strain at maximum efficiency (εη), as defined by: [30, 31]
(8.2a)
(8.2b)
The εf values for PP-1 (corresponding to the X-ray tomography samples) are given in Table 8.4, as a function of
density. The total energy absorbed for all PP-1 samples calculated using each of these definitions for εf is plotted
in Figure 8.11. All four definitions show an initially linear increase in J with density, and the change in slope at
the highest density samples is most pronounced when using the εf = εd and εf = εη indices (Figure 8.11c,d).
Around 11 – 12% core density, the percentage drop from peak strength to valley strength exhibits an
increase in sample-to-sample variation, and the average value appears to level off or begin to decrease. This
transition could be a change in dominant collapse mechanism, from primarily elastic and plastic buckling to
Table 8.4 Final strain values as a function of density, for each definition of strain. The references indicate the study or
studies where this integration range was selected.
ρcore (%) εf = 0.40 [25, 26] εf = [27] εf = εd [28, 29] εf = εη [30, 31]
7.9 0.40 0.51 ± 0.07 0.27 ± 0.04 0.42 ± 0.08
8.6 0.40 0.66 ± 0.09 0.39 ± 0.09 0.41 ± 0.09
9.6 0.40 0.78 ± 0.00 0.65 ± 0.00 0.48 ± 0.07
10.6 0.40 0.78 ± 0.00 0.61 ± 0.01 0.52 ± 0.03
11.5 0.40 0.73 ± 0.05 0.55 ± 0.10 0.39 ± 0.10
12.3 0.40 0.73 ± 0.02 0.54 ± 0.02 0.33 ± 0.04
13.0 0.40 0.72 ± 0.00 0.51 ± 0.02 0.46 ± 0.18
104
Figure 8.11 Energy absorbed, J, calculated up to εf = 0.4 (a), εf =
(b), εf = εd (c), and εf = εη (d), plotted as a function of ρcore.
primarily plastic buckling and fracture. The energy absorption plots, when measured up to εd or εη, demonstrate
this transition, while J measured up to ε = 0.4 or does not capture the change. By the time the stochastic
honeycomb reaches densification, both high and low density samples have fractured. Correspondingly, the
energy absorbed for all samples when densification is included is linear, as seen for εf = 0.40 and εf = . When
measuring J with εf = εd or εη, only the higher density samples have fractured, while the lower density samples
have elastically or plastically buckled.
105
Fracturing of the webs requires greater initial energy input than buckling alone, but once a web has
fractured that loading path is no longer able to support further load. Sample A has the least percentage of
fracturing overall, although sample B had a greater percentage of BB webs that did not fracture. Up to εv, these
samples absorbed the same amount of energy, = 0.19 J cm
-3. Samples C and D absorbed
= 0.34 and 0.57 J
cm-3
. From sample A to sample D, density increased by 60% and the energy absorption capacity almost tripled.
However, at εv, buckling and fracture of the webs has just begun, so the effect of fracture on the energy
absorption may not have been felt. Reducing the frequency of fracturing at the valley would improve the energy
absorption per unit volume of material used in the form of stochastic honeycombs. In absolute terms, σp and σv
increase with density [6, 9], but as density increases the percentage drop from the peak strength to the valley
strength stays relatively constant (± 15%).
Foams are considered ideal energy absorbers because there is a nearly constant plateau strength as
opposed to a peak strength decreasing to a valley. As a result, comparing the plateau strength (σpl) of a foam to
the peak strength of a honeycomb is misleading, but we can compare σv to σpl for two commercial PP foams
measured at ε = 0.25 [32, 33]. At ρcore = 8%, the valley strength of the stochastic honeycomb (0.52 ± 0.21 MPa)
is comparable to the plateau strength of the foams (0.53 – 0.55 MPa). Additionally, as ρcore increases, σv
increases more quickly than σpl for the same PP foams (i.e., at = 11.4%, σv = 1.20 ± 0.10 MPa and σpl =
0.75 MPa [32], and at ρcore = 13.0%, σv = 1.39 ± 0.04 MPa and σpl = 0.93 MPa [33]). Thus the stochastic
honeycombs are absorbing more energy than the PP foams, even when the energy absorbed up to the peak stress
is neglected. This demonstrates that stochastic honeycombs have good potential for energy absorbing
applications.
Figure 8.12 plots the percentage drop (D%) from σp to σv and the energy absorbed up to εf = 0.4 as a
function of core relative density for the four PPs from the previous buckling model study [9], where PP-1 is the
polymer presented in the Results section of this article. Table 8.5 presents the material properties of these
polymers detailed in previous studies (zero-shear viscosity, ηo, melting temperature, Tm, crystallinity, xc, and
tensile strength, σTS [8, 9]), as well as the peak and valley strengths, and J for ρcore = 10%. As seen in Figure 8.12
and Table 8.5, there is no statistically significant difference between PP-1 and PP-2. PP-3 is a lower viscosity
PP, and as a result, the internal architecture has fewer buttresses and archways to provide a secondary support
structure [8]. Indeed, while PP-1 and PP-2 each show a 55% decrease from σp to σv (with a coefficient of
variation of 15%), PP-3 shows a 78% decrease, with a coefficient of variation of 7%. Corresponding to this large
decrease, PP-3 absorbed less energy than PP-1 and PP-2. PP-4, on the other hand, has the highest viscosity, and
D% actually decreases from 50% at ρcore = 9.1% to 30% at ρcore = 13.5%. At ρcore = 11%, the cross-sectional area
fraction over the central third of the honeycomb height (Ac) is the same for PP-1, PP-2 and PP-4. However, as
106
ρcore increases, Ac increases by 33% and 48% for PP-1 and PP-2 receptively, while it remains constant for PP-4
with a coefficient of variation of 6%. Since PP-4 was shown to have a similar skin thickness to PP-1 and PP-2
[9], the extra material must be in buttressing and archways, providing a greater secondary support structure, and
leading to a decrease in D% as ρcore increases. However, since PP-4 has a lower crystallinity than PP-1 and PP-2
(Table 8.5), it has lower σp and σv, and thus less energy absorbed overall.
Comparing this data for D% to the literature, it was found that over a variety of materials (Aluminium,
steel, thermosets and fibre-reinforced epoxy) fabricated into regular honeycombs [21, 26, 34-37], D% varied
from ~ 40% – 85%. Within a given material, D% did not vary significantly with density, and the mean, median
and mode for the data set are, respectively, 60%, 50% and 61.3%. Thus, it appears that the secondary defect
structure consisting of buttresses and archways compensates for the lack of closed cells in stochastic
honeycombs, and can make them superior to the traditional closed-cell honeycombs.
Figure 8.12 Percentage drop (D%) from σp to σv a and energy absorbed, J, up until εf = 0.4 b as a function of ρcore, for PP-1, PP-2, PP-3
and PP-4.
Table 8.5 Zero-shear viscosity (ηo), melting temperature (Tm), crystallinity (xc) and tensile strength (σTS) of PP-1, PP-2, PP-3 and PP-
4 [8, 9], and peak and valley strength (σp, σv), and the energy absorbed (J) up until εf = 0.4 at ρcore = 10%.
Polymer ηo
(x104 Pa s)
Tm (oC) xc (%) σTS (MPa)
σp (MPa) at
ρcore = 10%
σv (MPa) at
ρcore = 10%
J (J cm-3
) at
ρcore = 10%,
εf = 0.4
PP-1 3.24 ± 0.24 162.1 ± 1.0 44.3 ± 2.7 32.1 ± 0.8 2.50 ± 0.08 1.09 ± 0.01 0.47 ± 0.01
PP-2 3.21 ± 0.39 161.2 ± 0.9 42.5 ± 2.5 30.9 ± 0.9 2.21 ± 0.40 1.02 ± 0.12 0.50 ± 0.04
PP-3 2.15 ± 0.21 163.0 ± 0.4 43.8 ± 1.5 29.4 ± 0.5 1.29 ± 0.04 0.27 ± 0.09 0.19 ± 0.05
PP-4 3.94 ± 0.18 147.7 ± 1.0 28.2 ± 2.5 18.2 ± 1.6 1.01 ± 0.15 0.55 ± 0.12 0.27 ± 0.05
107
Figure 8.13 Specific strength (σp/ρ) plotted against energy absorbed up until εf = 0.4 for PP-1, PP-2, PP-3 and PP-4.
Finally, Figure 8.13 plots the specific strength (σp/ρ) as a function of energy absorbed (up to εf = 0.4) for
PP-1 to PP-4. For a given specific strength, PP-3 absorbs the least energy and PP-4 absorbs the most. However,
PP-1 and PP-2 have a greater total energy absorption capacity due to their significantly higher strengths
compared to PP-4.
8.5 Conclusions
The post-peak collapse mechanisms were shown to depend on the distribution of material through the stochastic
honeycomb as well as the geometry of the webs themselves. Unlike honeycombs, each stochastic honeycomb
sample will have some webs that exhibit elastic buckling, some plastic buckling and some plastic buckling with
fracture. However, the frequency of each mechanism varies with density. At lower densities, elastic buckling is
more common, and at higher densities, plastic buckling and plastic buckling with fracture dominate. The change
in dominant failure mechanisms occurs between 10% and 12% core density. This is demonstrated by a change in
slope in the energy absorption curves when energy is calculated up to the densification strain, εd, or the strain at
maximum efficiency of energy absorption, εη. Additionally, it was shown that the secondary defect structure,
consisting of buttresses and archways, increases as the viscosity of the polymer increases. That leads to a
decrease in the percentage drop from the peak to the valley during out-of-plane compression (D%), increasing the
energy absorption capacity of the stochastic honeycomb structure. Finally, it was shown that stochastic
honeycombs have a greater energy absorption capacity than commercial PP foams through comparing the valley
strength, σv, to the plateau strength of PP foams.
108
References
[1] M.F. Ashby. Metal foams, Butterworth-Heinemann, Boston, MA, 2000.
[2] L.J. Gibson, M.F. Ashby. Cellular Solids: Structures and Properties. 2 ed., Cambridge University Press,
Cambridge, New York, 1997.
[3] M. Heckele, W.K. Schomburg. Review on micro molding of thermoplastic polymers, Journal of Micromechanics
and Microengineering 14 (2004) R1-R14.
[4] H.N.G. Wadley. Multifunctional periodic cellular metals, Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences 364 (2006) 31-68.
[5] A. Ziebig, H. Brigasky, W. Lewand. Extrusion device for the production of honeycomb structures. Google
Patents, 1987.
[6] M. Hostetter, B. Cordner, G.D. Hibbard. Stochastic honeycomb sandwich cores, Composites Part B: Engineering
43 (2012) 1024-1029.
[7] I. Hamerton, A. Azapagic, A. Emsley. Polymers: The Environment and Sustainable Development, John Wiley &
Sons, Hoboken, New Jersey, 2003.
[8] M. Hostetter, G.D. Hibbard. Architectural characteristics of stochastic honeycombs fabricated from varying melt
strength polypropylenes, Journal of Applied Polymer Science 131 (2014) 40074.
[9] M. Hostetter, G.D. Hibbard. Modeling the buckling strength of polypropylene stochastic honeycombs, Journal of
Materials Science 49 (2014) 8365-8372.
[10] ASTM D1238 Standard Test Method for Melt Flow Rates of Thermoplastics by Extrusion Plastometer, ASTM
International (2004).
[11] S. Rabinowitz, P. Beardmore. Cyclic deformation and fracture of polymers, Journal of Materials Science 9 (1974)
81-99.
[12] T. Ariyama. Cyclic deformation and relaxation characteristics in polypropylene, Polymer Engineering and
Science 33 (1993) 18-25.
[13] T. Ariyama. Stress relaxation behavior after cyclic preloading in polypropylene, Polymer Engineering and
Science 33 (1993) 1494-1501.
[14] R.P. Wool, W.O. Statton. Dynamic polarized infrared studies of stress relaxation and creep in polypropylene,
Journal of Polymer Science: Polymer Physics Edition 12 (1974) 1575-1586.
[15] T. von Kármán. The buckling of thin cylindrical shells under axial compression, Journal of The Aeronautical
Sciences 8 (1941) 303-312.
[16] W. Horton, S. Bailey, A. Edwards. Nonsymmetric buckle patterns in progressive plastic buckling, Experimental
Mechanics 6 (1966) 433-444.
[17] J. Rhodes. Buckling of thin plates and members -- and early work on rectangular tubes, Thin-Walled Structures
40 (2002) 87-108.
[18] S. Liang, H.L. Chen. Investigation on the square cell honeycomb structures under axial loading, Composite
Structures 72 (2006) 446-454.
[19] J. Zhang, M.F. Ashby. The out-of-plane properties of honeycombs, International Journal of Mechanical Sciences
34 (1992) 475-489.
109
[20] W. Miller, C.W. Smith, K.E. Evans. Honeycomb cores with enhanced buckling strength, Composite Structures
93 (2011) 1072–1077.
[21] A. Wilbert, W.-Y. Jang, S. Kyriakides, J.F. Floccari. Buckling and progressive crushing of laterally loaded
honeycomb, International Journal of Solids and Structures 48 (2011) 803–816.
[22] F. Scarpa, S. Blain, D. Perrott, M. Ruzzene, J.R. Yates. Elastic buckling of hexagonal chiral cell honeycombs,
Composites Part A: Applied Science and Manufacturing 38 (2007) 280–289.
[23] T. Wierzbicki. Crushing analysis of metal honeycombs, International Journal of Impact Engineering 1 (1983)
157-174.
[24] S. Santosa, T. Wierzbicki. On the modeling of crush behavior of a closed-cell aluminum foam structure, Journal
of Mechanics and Physics of Solids 46 (1998) 645–669.
[25] B.I. Chaudhary, R.P. Barry, M.H. Tusim. Foams made from blends of ethylene styrene interpolymers with
polyethylene, polypropylene and polystyrene, Journal of Cellular Plastics 36 (2000) 397-421.
[26] B.P. Russell, V.S. Deshpande, H.N.G. Wadley. Quasistatic deformation and failure modes of composite square
honeycombs, Journal of Mechanics of Materials and Structures 3 (2008) 1315-1340.
[27] O.B. Olurin, N.A. Fleck, M.F. Ashby. Deformation and fracture of aluminium foams, Materials Science and
Engineering: A 291 (2000) 136–146.
[28] Q.M. Li, I. Magkiriadis, J.J. Harrigan. Compressive strain at the onset of densification of cellular solids, Journal
of Cellular Plastics 42 (2006) 371-392.
[29] M.F. Ashby. The properties of foams and lattices, Philosophical Transactions of the Royal Society A:
Mathematical, Physical and Engineering Sciences 364 (2006) 15-30.
[30] M. Avalle, G. Belingardi, R. Montanini. Characterization of polymeric structural foams under compressive
impact loading by means of energy-absorption diagram, International Journal of Impact Engineering 25 (2001)
455–472.
[31] P.J. Tan, J.J. Harrigan, S.R. Reid. Inertia effects in uniaxial dynamic compression of a closed cell aluminium
alloy foam, Journal of Materials Science & Technology 18 (2013) 480-488.
[32] BASF PlasticsPortal - Neopolen P. <http://www.plasticsportal.net/wa/plasticsEU~en_GB/portal/show/
content/products/foams/neopolen_p> Accessed 3 Mar 2015.
[33] JSP. ARPRO Technical Properties, Engineering Expertise, High Performance Molding Materials.
<www.arpro.com/tech-docs/europe> Accessed 3 Mar 2015.
[34] S. Xu, J.H. Beynon, D. Ruan, G. Lu. Experimental study of the out-of-plane dynamic compression of hexagonal
honeycombs, Composite Structures 94 (2012) 2326–2336.
[35] D. Asprone, F. Auricchio, C. Menna, S. Morganti, A. Prota, A. Reali. Statistical finite element analysis of the
buckling behavior of honeycomb structures, Composite Structures 105 (2013) 240-255.
[36] F. Côté, V.S. Deshpande, N.A. Fleck, A.G. Evans. The out-of-plane compressive behavior of metallic
honeycombs, Materials Science and Engineering: A 380 (2004) 272-280.
[37] E. Wu, W.-S. Jiang. Axial crush of metallic honeycombs, International Journal of Impact Engineering 19 (1997)
439–456.
110
Chapter 9
Summary and Future Work
This thesis presented the first studies on melt-stretched stochastic honeycombs. The mechanical properties
provided a motivation for the further study of stochastic honeycombs, as well as a baseline to improve upon. The
fabrication variables affect the architecture of the melt-stretched stochastic honeycomb, as do the rheological
properties of the polymer melt. The internal architecture and the bulk material properties of the polymer then
determine the mechanical properties of this hybrid material.
Four high melt strength PPs and one linear PP were used to fabricate stochastic honeycombs. The key
material properties to ensure an interconnected web structure were high viscosity and the ability to strain-harden.
Also important to creating a mechanically strong structure was the presence of secondary defects, such as
buttressing and archways. These were more prevalent in PPs with ηo > 2.5 x 104 Pa s and MFI < 3 g/10 min.
Additionally, polymers with good hot-tack adhesion to metals and low surface energy increase the ease of
fabrication. If a polymer is foam-able, it should be able to be successfully melt-stretched.
From a technological standpoint, the fabrication method was advanced from a manual press to a fully
instrumented process. With the original set-up, there was no fine control over the melt-stretch rate, applied force,
or temperature. The controlled process has improved temperature control (within ± 1 oC), and machine control
over the applied force and melt-stretching rate. An understanding of the fabrication curves was developed such
that the mechanical response of the polymer melt was tied to physical changes in the internal architecture after
fabrication. The force and the melt-stretch rate were identified as key process control variables. A method to
thermally weld PP face sheets to the melt-stretched stochastic honeycomb was also developed, involving
applying enough force to form the thermal weld without compromising the integrity of the core. This led to the
creation of all-PP sandwich panels. Characterization methods were developed to analyze the 3D internal
architecture (determined from X-ray tomography) and the 2D projections of the webs. Topological parameters
(in the form of Minkowski functionals) were adapted to be applicable to stochastic honeycombs, as was the
Euler number for connectivity of a cellular structure. Other characterization variables (such as N, N/L, ξ) had to
be developed to distinguish between architectures that were visually different but had similar architectural
parameters. Mechanical testing methods were adapted from the ASTM Standards for honeycombs and sandwich
panels pertaining to sample size, number of repetitions and strain rate. Through the mechanical testing it was
111
shown that despite the random, open-cell structure of stochastic honeycombs, the mechanical properties were
repeatable. An analytical buckling model for stochastic honeycombs was also developed based on the buckling
of a thin plate. Specific results will be discussed further below.
9.1 Fabrication
The fabrication process involves melting PP at 1.1Tm and applying some preliminary force to ensure adhesion
between the upper Al platen and the polymer. When the force is applied, the polymer melt expands radially. The
melt is then stretched uniaxially, and the stress applied increases to a peak, at which point the melt undergoes a
combination of cohesive fracture and ductile failure as air rushes in to the centre. The tunable fabrication
variables are areal density, ρA, and melt-stretch rate, . The fabrication parameters, corresponding to the response
of the polymer melt, were defined to be peak stress, σpk, melt stiffness, Emelt, and the apparent elongational
viscosity at the peak, . These parameters were chosen for the insight they gave into the deformation of the
polymer during melt-stretching.
The peak stress and melt stiffness both increased as density decreased. This is believed to be due to the
higher fraction of polymer chains that must be adhered to the Al platens where there is less mass present. This
would provide anchoring points for the chains, increasing the stress required to move them and increase σpk and
Emelt. Additionally, as density increased the apparent elongational viscosity at the peak increased as well, albeit
to a much lesser degree. Elongational viscosity (ηE) should be independent of density, but in the non-steady state
environment of this fabrication process, some small effect was seen on , likely due to differences in strain
rate.
As the melt-stretch rate increases, the peak stress becomes almost independent of ρA. Increasing
increases σpk and Emelt. At any specific time, is larger for a higher , but because the melt-stretching process is
not temperature controlled, decreases as increases. Additionally, at low , the melt-stretching rate is
lower than the relaxation rate, while at higher , the melt is stretched faster than it can relax.
9.2 Internal Architecture
The internal architecture of stochastic honeycombs is complex, with buttresses, voids and partial webs
superimposed on an open-cell honeycomb structure with bound and bound–free webs. The internal web structure
112
is then sandwiched between two parallel skins. Methods to quantify the architecture from the X-ray scans had to
be developed, to extract meaningful parameters from the countless possible variables in the stochastic
architecture. The cross-sectional area fraction, Ac, decreases from 1 at the skins to a minimum plateau value over
the central third of the honeycomb height. Across that central third, the total length of the webs per area, lc, is
also constant. The webs themselves extend from one skin to the other, and can be classified as bound on both
sides (bound), or bound on one side and free on the other (bound–free). Partial webs, such as archways and
buttresses, serve to support the complete webs and do not extend from skin to skin.
The melt-stretch rate during fabrication has a large effect on the architecture of stochastic honeycombs.
Higher stretch rates increase the number of nodes per length of webs (N/L), implying shorter webs (i.e., lower b)
overall. The total number of nodes increases as increases, and to a lesser extent the total length of webs also
increases as increases. The Euler number in 2D, χ2D, is defined as the number of connected areas minus the
number of apparently closed cells, where the webs consist of 1 connected area. A higher stretch rate leads to a
more negative Euler number, and thus a more tortuous pathway for the air to flow through. ξ was defined as the
outer area fraction of air ingress before the webs become interconnected and form apparently closed cells. ξ thus
increases as χ2D approaches 1, since a larger fraction of air ingress means a less interconnected structure. A
higher stretch rate leads to smaller ξ.
The melt rheology has a large effect on the architecture of stochastic honeycombs. Increased viscosity
leads to a greater total length of webs, and more uniform partitioning of material between the skin and the core.
The PPs with lower viscosity had more material in the skins and the transition to the plateau, and less in the webs
themselves. If the viscosity was too low, like the linear PP-5, many of the webs failed during melt-stretching and
no interconnected structure was formed. On the opposite end, higher viscosity and melt strength led to an
increase in partial webs, such as buttresses, in the structure. Total web length increased as the viscosity
increased, but reached a plateau at the highest viscosity. This implies that an increase in entanglement density
leads to an increase in web length and the conclusion that activation of a certain local density of entanglement
points is required to form a new web.
The applied force during fabrication affects the areal density during melt stretching, and thus the final
density, so Fapp and ρA can be collapsed to one variable. Higher density structures have a less negative Euler
number and as such a larger outer area fraction of air ingress (greater ξ). At lower density, the architecture has
more apparent closed cells, and lower ξ. Lower density structures also have a higher numbers of nodes per length
of webs, N/L. Higher density leads to thicker skin and increased web thickness. The increase in web thickness
can be seen through the increase in Ac as density increased, while lc remained the same. Additionally, over the
113
central third of the honeycomb height, as density increased the surface area-to-volume ratio decreased, and thus
the effective web thickness, teff, increased.
9.3 Mechanical Properties
Under out-of-plane compression, stochastic honeycombs, like regular honeycombs, experience elastic
compression to a peak load, a decrease in stress to a valley, and then an increase in stress at densification. The
structure changes sample-to-sample, but for a given density the strength is consistent. Stochastic honeycombs
collapse through a combination of elastic buckling, plastic buckling, and plastic buckling with fracture, with
elastic buckling more common at lower density and plastic buckling with fracture more prevalent at higher
density. The total energy absorbed depends on the final strain chosen, but the peak strength is comparable to out-
of-plane compressive strength of commercial PP honeycombs and the valley strength is comparable to the
plateau strength of commercial PP foams. The sample-to-sample variation under three-point bending was
reduced when face sheets were thermally welded to the skin creating sandwich panels, and the sandwich panel
failure mechanisms depend on face sheet thickness, core density, and core thickness. The exponents relating
specific strength and stiffness to relative density are 1.59 and 1.01, respectively, and are very close to the
theoretical hexagonal honeycomb exponents of 1.67 (or 5/3) and 1.
The material properties of the PP used to fabricate stochastic honeycombs affect the mechanical properties
of the honeycomb. The crystallinity of the polymer affects the tensile strength and stiffness of the webs, and thus
the buckling strength. The tangent modulus is a key input to the material property portion of the buckling model,
and during buckling the portion of the web at the fold experiences a tensile stress. Increased crystallinity allows
the webs to resist buckling under greater applied stresses.
Thicker webs increases the strength of the stochastic honeycomb in out-of-plane compression, as the
buckling strength is related to the web thickness, squared. Shorter webs (in the in-plane directions, i.e., lower b)
should also increase the strength of the stochastic honeycomb. Additionally, a greater fraction of webs bound on
both sides as opposed to bound on one side and free on the other will increase the buckling strength, as these
webs can support more load. This can be seen through the buckling coefficient, K, that is related to the end
constraints of the webs. KB is approximately an order of magnitude larger than KF, so the bound webs have a
much greater effect on the strength of the stochastic honeycomb than the bound–free webs. An interconnected
structure is also believed to absorb more energy than a less connected structure. Finally, some architectures have
114
more buttresses and archways, and it was shown that these features contribute to an increase in the valley
strength compared to the peak strength (i.e., decreased D%), which increases the energy absorption as well.
9.4 Future Work
The original motivation for this project came from the idea that a sandwich structure could be created with less
environmental impact than what is currently available commercially. Using the simple and low-cost, low-impact
fabrication method, the carbon footprint of the stochastic honeycombs would be expected to be lower than that
of commercial foams and honeycombs. A detailed life-cycle analysis of the impact of foams and honeycombs, as
well as stochastic honeycomb cores, would give the best overview of the relative and absolute environmental
impact. In order to quantify this, the energy requirements to make foams, honeycombs and stochastic
honeycombs must be determined.
Taking this idea a step further, we would like to create stochastic honeycombs using biodegradable
polymers. This would eliminate the need for recycling at end-of-life, and reduce the environmental impact of the
product. From Chapter 5, the effect of melt properties of the polymer on the architecture, it is possible to define a
range of parameters, such as melt flow index and melt strength, from which it can be predicted whether or not
the polymer will be able to form an interconnected stochastic honeycomb structure. Creep tests on the polymers
would allow a better measurement of the zero-shear viscosity than is obtainable from the parallel-plate
rheometry tests, and determining the surface tension of the molten polymers at the fabrication temperature
should give added insight into the mechanism of web formation.
For PP stochastic honeycombs, some questions about the relationships between fabrication, architecture
and mechanical properties remain. It has been shown that a lower viscosity polymer stochastic honeycomb has
more material in the skins and transition region than in the central third of the height of the honeycomb. Also,
higher density stochastic honeycombs have thicker skin as well. The partitioning of material through the
honeycomb is thus affected by the density and the rheology of the polymer; it could also be affected by the melt-
stretching rate during fabrication. If the relative partitioning of material is affected, then the prevalence of
secondary defects in the transition region could be affected as well. Additionally, the varying structures created
at a single density with varying melt-stretching rates may have different strengths, so it should be possible to
determine a structure that is the most efficient, i.e. the highest strength per mass of material. This could be
related to the number of nodes per length of webs, or another characterization variable, and linking those
variables to the strength of the honeycomb would increase the tunability of the stochastic honeycomb properties
115
for a given application. Compression testing structures at the same density fabricated at varying melt-stretching
rates would begin to solve this question.
Mechanically, it is important to expand our existing knowledge of the mechanical properties to include the
impact resistance, and shear and fatigue properties. Depending on the application, any of these properties could
be very important. In order to get reliable data for the shear strength and shear modulus, a series of three- and
four-point bend tests need to be performed.
It has also been shown that during fabrication, air ingress can be directed to come from the top or bottom
(or both) of the structure as opposed to solely from the outer perimeter of the melt through the introduction of
platens with holes. A square or other grid of holes can be used to create an almost regular honeycomb in which
the air enters each cell from the top or bottom, and thus the cells themselves are closed, and the honeycomb
structure has a series of holes in the as-fabricated skin. This closed cell stochastic honeycomb then more
accurately reflects the mathematical construct of a Voronoi honeycomb. With a structure of almost entirely
bound webs, the out-of-plane compressive strength should increase. Additionally, there should be an optimal
hole size and hole spacing to achieve the most efficient architecture.
Finally, all of this work has been towards fabricating flat sheets of a PP stochastic honeycomb. It is
possible to create the open-cell honeycomb structure using two non-parallel platens, or with a more complex
profile, such as corrugated Al platens. The melt would still be stretched uniaxially, so the properties in any
direction other than axially would be related to the shear properties as opposed to the out-of-plane compressive
properties. However, being able to form complex shapes is an advantage that foams have traditionally had over
honeycombs; melt-stretched honeycombs may be able to fill this gap.
116
Appendix A
Polymer Properties, PP-1 to PP-5
The known material properties for PP-1 to PP-5 are presented here. The melting and crystallization temperatures
(Tm, Tc) and crystallinity (xc) were determined from differential scanning calorimetry, and the tensile strength
(σTS), tangent modulus (Et) and ultimate strain (εu) from tensile tests (detailed in Chapter 7). Elongational
viscosity tests were performed on PP-1 and PP-5 (ηE and the temperature of the test, ). Additionally, parallel
plate rheometry tests (ηo, Gc, ωc and λ) were carried out on all five polymers (detailed in Chapter 5).
Polymer PP-1 PP-2 PP-3 PP-4 PP-5
Technical Name WB135HMS+ WB140HMS
+ WB180HMS
+ WB260HMS
+ HP0306L
*
ρ (g cm-3
)‡ 0.905 0.905 0.905 0.905 0.905
ρmelt (g cm-3
) ‡ 0.714 0.714 0.714 0.714
LCB (n/1000 C) ‡ 0.21 0.21
MFI (g/10 min) ‡
2.4 2.1 6 2.4 12
Tensile Modulus
(MPa) ‡
2000 2000 2000 900 1600
Melt Strength (cN) ‡ 32 36 10 27
Melt Extensibility
(mm s-1
) ‡
250 255 245 250
D (MPa) ‡ 1600
fl (MPa) ‡ 46
Tm (oC) 162.1 ± 1.0 161.2 ± 0.9 163.0 ± 0.4 147.7 ± 1.0 166.7
Tm,fab (oC) 165.3 ± 0.7 163.9 ± 1.3 165.3 ± 0.4 150.8 ± 0.6
Tc (oC) 125.4 ± 2.2 125.3 ± 1.1 126.5 ± 0.2 110.3 ± 1.1 124.1
xc (%) 44.3 ± 2.7 42.1 ± 2.5 43.8 ± 1.5 28.2 ± 2.5 43.7
TS (MPa) 32.1 ± 0.8 30.9 ± 0.9 29.4 ± 0.5 18.2 ± 1.6
Et (MPa) 784.4 ± 16.1 703.5 ± 27.7 690.8 ± 25.8 426.6 ± 23.2
εu (mm mm-1
) 0.14 ± 0.02 0.11 ± 0.02 0.10 ± 0.01 > 1
ηE (kPa s) 334.7 120.9
(
oC) 155 165
ηo (x104 Pa s) 3.24 ± 0.24 3.31 ± 0.39 2.15 ± 0.21 3.94 ± 0.18 0.48 ± 0.01
Gc (kPa) 16.2 ± 0.5 7.0 ± 2.7 17.9 ± 5.1 14.7 ± 3.3 37.5 ± 10.0
ωc (rad s-1
) 12.5 ± 0.4 5.0 ± 1.7 26.2 ± 3.5 7.6 ± 2.0 66.2 ± 2.5
λ (s) 0.080 ± 0.002 0.214 ± 0.066 0.039 ± 0.005 0.137 ± 0.031 0.015 ± 0.001 + Borealis AG, Daploy
TM HMS PP [1]
* Aclo Compounders, Accucomp PP [2]
‡ Provided by the manufacturer
117
References
[1] Boralis AG. Daploy(TM) HMS Polypropylene for Foam Extrusion <http://www.borealisgroup.com>
Accessed 26 Nov 2013
[2] Aclo Compounders <http://www.aclocompounders.com/accucomp.htm> Accessed 26 Nov 2013
118
Appendix B
Melt-Stretched Honeycombs with Various Polymers
During fabrication, parallel skins are formed on the Al platens. The skins are separated by webs parallel to the
direction of applied tensile stress. The skin is approximately 40% thicker on the bottom of the sample (in contact
with the lower platen) than on the top of the sample. This is due to a combination of gravitational effects on the
melt and the surface tension of the polymer as the molten polymer adheres to the upper platen and is pulled
upward, redistributing the polymer vertically.
The webs form an interconnected network that is nearly constant in one dimension, and varies in the
others, making in analogous to a 2D cellular material. Figure B.1 plots the zero-shear viscosity (ηo) of a number
of polymers (PP, PE, and copolymers of PE and PP, CP) as a function of the melt flow index (MFI) of these
polymers. Viscosity and MFI have an inverse relationship, as demonstrated here. The fabricated structures, from
the side with the skin visible, of select samples are also shown in Figure B.1. The highest viscosity PE, with MFI
= 0.5 g/10 min, did not expand at all (Image 1). For this polymer, the
Figure B.1 Zero-shear viscosity (ηo) plotted as a function of melt flow index (MFI) for a number of PP, PE, and copolymers of PP
and PE (CP), along with side images of representative samples.
119
cohesive forces in the polymer exceeded the adhesive force to the Al platens. Image 2 is indicative of all four
HMS PP (MFI = 2.1 – 6 g/10 min, ηo = 2 – 4 x 104 Pa s). Images 3 – 6 show the various structures composed of
individual columns or webs, obtained from PE, CP and PP for M I ≥ 5 g/10 min. In Image 5, the upper skin
cooled and separated from the platen and then had no webs or columns to act as supports. For many of these
lower viscosity polymers, it was apparent that the polymer stretched during fabrication and created webs, but
then flowed back to the skin before the structure cooled.
Figure B.2 shows images from the top of melt-stretched honeycombs fabricated with HMS PP, linear
isotactic PP (L-iPP), and low density polyethylene (LDPE). In order to form these L-iPP and LDPE structures, a
greater amount of polymer was used and the platens were cooled quickly. Figure B.3 shows SEM images from
the top and from an angle for the same three polymers. The webs of the HMS PP have buttresses and archways,
which create a secondary defect structure that helps support the webs. In contrast, the L-iPP and LDPE webs
transition smoothly to the skin, with no buttressing at all.
Figure B.2 Stochastic honeycombs fabricated with HMS PP (a), linear iPP (b) and LDPE (c). The scale bar on (c) is 3 cm and applies
to all three images.
120
Figure B.3 SEM images of stochastic honeycombs fabricated with HMS PP (a,b), L-iPP (c,d) and LDPE (e,f), perpendicular to the
webs (top row), and at a tilt (bottom row). The HMS PP has buttresses that support the webs at the transition between the
webs and the skin, while the L-iPP and the LDPE have smooth transition regions with no buttressing.
121
Appendix C
Characterization of Stochastic Honeycomb Sandwich Failure
M. Hostetter, G.D. Hibbard. Characterization of stochastic honeycomb sandwich failure (2013) In: Proceedings
of the International Conference on Composite Materials 19, Montreal, Canada, 28 July – 2 August, pp. 312–319
C.1 Introduction
Foams and honeycombs are composite materials in that they combine the properties of the base material with
empty space through an internal cellular architecture. Stochastic honeycomb structures are a new type of cellular
composite and a variation on the traditional plastic foam or honeycomb sandwich core material. Honeycombs are
essentially two dimensional, with regular, repeating unit cells, while foams have a range of cell size, web
thicknesses and lengths. Stochastic honeycombs are essentially two dimensional, like honeycombs, but have web
lengths and thicknesses that are variable (Figure C.1). Even in the through-thickness direction, the stochastic
honeycomb webs can change in length, thickness, and orientation.
Produced in a simple, low-cost process, the out-of-plane compressive properties of polypropylene (PP)
stochastic honeycombs rival that of PP honeycombs and exceed that of commercial PP foams [1]. For example,
over a relative density range of 7 – 12% the compressive strength of stochastic honeycombs ranged from 1.0 –
2.5 MPa [1], while commercial honeycomb strengths ranged from 1.1 – 2.4 MPa [2-5].
Depending on the architecture and loading configuration, honeycombs have been shown to fail by
buckling, bending, or fracture [6, 7]. Analytical solutions for failure of honeycombs by elastic buckling [6],
plastic collapse (bending) [7], and fracture [6] have been determined and tested experimentally. The failure
mechanism has been found to depend on the density of the honeycomb [6], with elastic buckling typically
occurring at low densities and fracture at higher densities when the local stress exceeds the fracture stress of the
material. Since honeycombs have a regular structure, the geometry of one unit cell is representative of the whole,
and the same failure mechanism is generally seen throughout the structure. This is not the case for stochastic
honeycombs given their complex internal architecture. Multiple failure mechanisms are expected depending on
122
Figure C.1 3D image reconstructed from a micro-CT scan, with a sample size of 40 mm 40 mm 15 mm (a) and the web structure
as seen from the side (b), with the as-fabricated face sheets removed digitally in order to more clearly reveal the internal
web structure. The scale bars are 5 mm each.
the local geometry of the webs. In addition, studies have shown that flaws or defects in a thin-walled structure
(such as the webs in a stochastic honeycomb, or the walls in a conventional honeycomb) can concentrate failure
at that location [8-11]. As such, we would expect that the initiation of these failure mechanisms would be
triggered by local defect structures in the webs. This study is a first look at the detailed failure mechanisms in
stochastic honeycombs.
C.2 Experimental Details
High melt strength polypropylene was heated on a metal platen in an oven at 180 oC until it became a viscous
melt. It was then removed from the oven and placed in a press, where another platen was pressed on top.
Pressure was applied, and the upper platen was raised to the height desired for the sample (in this case, 15 mm)
and locked in place. The sample spontaneously separated from the plates upon cooling, resulting in a sandwich
structure that had integrated upper and lower face sheets separating a network of interconnected webs (Figure
123
C.1). Two samples, with relative densities of ρLD = 11% (low density, LD) and ρHD = 18% (high density, HD),
were scanned in a Skyscan 1172 high resolution micro-CT scanner. The micro-CT scans were reconstructed into
a series of cross-sections in the through-thickness direction, with voxel size of 35 μm 35 μm 35 μm. The
samples were then loaded in out-of-plane compression at a cross-head displacement rate of 1 mm min-1
to a
compressive strain of 0.20, and characterized by micro-CT.
C.3 Results and Discussion
The detailed collapse mechanisms of the LD and HD sandwich cores were investigated by comparing micro-CT
scans of individual webs before and after uniaxial compression. The starting structure of the LD and HD
samples is shown in Figure C.2. Figure C.2a and 2b show mid-height cross-sectional slices. The number of webs
decreased and the thickness of the webs increased as the density increased from 11% to 18%. Figure C.2c plots
the cross-sectional area fraction of the stochastic honeycomb as a function of normalized through-thickness posi-
Figure C.2 Micro-CT scans of the mid-height cross-sectional slice (sample size 40 mm 40 mm 15 mm) for the LD sample (a)
and the HD sample (b). The scale bar is 5 mm and applies to both figures. The corresponding area fraction, Ac, as a
function of normalized position in the sample, P, from bottom to top is given in (c).
124
Figure C.3 Partial web in the LD as-fabricated structure. The scale bar is 2 mm.
tion in the sample (0 indicates the bottom of the sample, 1 indicates the top). Near both the top and bottom, the
cross-sectional area fraction approaches 1 as the webs transform into the built-in skin. From Figure C.2c, it is
also clear that the higher density of the HD sample is due to the greater cross-sectional area in the central 50% of
the overall sandwich thickness. Over this middle half of the sample height, the area fraction of the LD sample
was 0.085 0.006, while the area fraction of the HD sample was 0.114 0.008. Note that even though the
cross-sectional area over the middle half of the sample is effectively constant, the web structure is constantly
changing in the through-thickness direction.
A fraction of the webs in the as-fabricated structure only partially span the distance between opposing face
sheets, as seen in Figure C.3. While not providing a continuous load path of their own, these partial webs do still
provide support to the adjacent complete webs, which are either bounded on both sides, or bounded on one side
and free on the other. The three generally considered restraint conditions for thin plates are termed simply
supported, built-in (or clamped), and free [12, 13]. Simply supported, for a thin plate, is analogous to pin-jointed
for a column or beam, and means that the plate is free to rotate around the hinge, but there is no translational
displacement. In comparison, built-in edge conditions are taken to mean that there is no rotation or translation at
the joint, and free is just that, the edge is unrestrained [12, 13]. When modelling, the edge constraint is often
chosen to be either simply supported or built-in [14]. However, in practice, this boundary is considered to be in
between the classic simply-supported and built-in boundary conditions [6, 15]. A schematic of a web supported
on one side with appropriate dimensions is given in Figure C.4.
The uniaxial compression stress-strain curve of stochastic honeycombs is comparable to that of
conventional honeycombs with an initial peak stress due to plastic buckling, followed by a decrease towards a
valley stress, followed by a final stress increase due to densification (e.g. Figure C.5 [1]). Figure C.6 illustrates
125
Figure C.4 Schematic of a web bound on one side and free on the other. The length, b, height, a, and thickness, tweb, are indicated.
Figure C.5 Uniaxial compression stress–strain curve for a ρcore = 13% stochastic honeycomb.
Figure C.6 Uniaxial compression stress-strain curve for LD (ρcore = 11%) and HD (ρcore = 18%) stochastic honeycombs to the valley
(ε = 0.20).
Table C.1 Compressive properties for the low density and high density stochastic honeycombs.
Sample σp (MPa) σv (MPa) E (MPa)
LD 2.23 0.92 59
HD 4.61 2.23 139
126
stress-strain curves for the LD and HD samples of the present study, which were compressed to the valley stress
beyond the initial peak and then unloaded for subsequent micro-CT characterization. Note that the density
increase from 11% to 18% corresponded to peak strength, valley strength, and elastic modulus increases of
105%, 140%, and 135%, respectively (values summarized in Table C.1). Classification of the web failure
mechanisms was based on the web end constraint conditions. The complete webs that were bounded on both
sides tended to tear vertically, as seen in Figure C.7, from the LD sample. This study focuses on the webs that
are bounded on one side and free on the other. Six webs from each sample were analyzed for their dimensions
and failure mechanisms, termed LD-1 to LD-6, and HD-1 to HD-6. Four specific examples are presented below
(LD-1, LD-2, HD-1, and HD-2), with general results of the overall analysis discussed afterwards.
The web (LD-1) shown in Figure C.8 is from the low density sample. Two views are shown of the as-
fabricated (Figure C.8a and 8b) and partially compressed (Figure C.8d and 8e) webs, while Figure C.8c and 8f
show cross-sections along the height (through-thickness) of the sandwich. Figure C.9 plots the thickness of the
web (x-axis) against the normalized position through the height (y-axis), giving a thickness profile of the web,
also shown in cross-section. The dashed line indicates the point of buckling failure. The web thickness was ~ 0.2
mm and ~ 0.4 mm at the top and bottom (respectively) of the sample and 0.15 mm in the centre. This web did
not buckle at the thinnest point, but rather just above the mid-point along the height dimension. From Figure
C.8c, the cross-section of the as-fabricated web shows that there is already some slight curvature in the area
where the web buckled, so while LD-1 did not buckle at the thinnest point, it did buckle at the area just above the
thinnest point where some curvature was already present. The thickness/length aspect ratio of this web was tweb/b
= 0.10 (Table C.2).
Figure C.7 As-fabricated LD web bounded on each side (a), and the vertical tear created by the horizontal tensile strain from the
compression of the bounding webs (b). The scale bar is 2 mm and applies to both figures.
127
Figure C.8 LD-1 as-fabricated from the side (a), straight on (b), in vertical cross-section (c), and partially compressed from the side
(d), straight on (e), and in vertical cross-section (f). The scale bar is 2.5 mm and applies to all the figures.
Figure C.9 Thickness profile of LD-1 and the vertical cross-section for comparison. The dashed line marks the point of initiation of
instability.
Table C.2 LD-1 dimensions at failure initiation point.
P (mm mm-1
) tweb (mm) tweb/b
0.52 0.15 0.104
128
Another web (LD-2) from the same sample is shown in Figure C.10. This web undergoes a somewhat
more complex failure mechanism, with both web buckling and web fracture being seen in the post-loaded scan.
Figure C.10b and 10c show that the web has some curvature before compression, and afterwards, in addition to
the tear, LD-2 buckled at the bottom (P = 0.11) and the top (P = 0.89). Unlike LD-1, this web was thinner near
the top and bottom (where buckling occurred). Figure C.10a and 10b show that this web was already somewhat
curved before loading, leading to increased bending strains at the thinnest regions, causing the upper and lower
buckles, and then the tear just below the upper buckle. From the thickness profile (Figure C.11), the web tears at
tweb = 0.32 mm, where the aspect ratio is 0.189. The two buckles occur where the web is 0.29 mm and 0.31 mm
thick, and the aspect ratios are 0.103 and 0.152, for the bottom and top, respectively (Table C.3).
The web shown in Figure C.12, in contrast, is from the high-density sample (HD-1). Similarly to LD-1, it
is bound on one side and free on the other. As Figure C.12b and 12c show, this web had some curvature to begin
with, and buckled in the centre in the direction of the initial curvature. At the buckle, this web is at its thinnest
point (tweb = 0.21 mm, Table C.4) as seen in the thickness profile in Figure C.13. The thickness at the top and
bottom of this web is ~ 0.4 mm, so its thinnest point is only 50% of its maximum thickness. This web had a thick
supporting web, as seen straight-on in Figure C.12a. The supporting web buckled at the top and tore, as seen in
Figure C.12d. HD-1 is thicker than LD-1 (0.21 mm compared to 0.15 mm), but has a lower aspect ratio because
it spans a longer distance. This is typical of the HD sample; it had thicker webs, as can be seen from Figure C.2,
but the webs are also longer, leading to lower aspect ratios overall.
HD-2, shown in Figure C.14, is another web from the high-density sample. In addition to being supported
on one side, it is supported by a buttress at the top, which is visible in Figure C.14b and on the opposite side of
the view shown in Figure C.14a. While this web has little initial curvature, it is tilted to one side, as can be seen
in Figure C.14c and 14d. After compression, this web has buckled near the top (tweb = 0.46 mm, Table C.5) and
torn a short distance below the buckle. The buttress supplies extra support for this web, and tears itself, as can be
seen in Figure C.15. The thickness profile for HD-2 (Figure C.16) shows that the buckle occurs at the thinnest
point of the web, and the tear occurs at the upper end of the buttress (Figure C.15c). The thickness profile for the
buttress, shown in the inset for Figure C.16, demonstrates that the buttress also tore near its thinnest point,
stressed by the buckle of the web it was supporting. This web tore near a buckle, as the LD-2 web did. HD-2 had
tweb = 0.50 mm at the tear, while LD-2 had tweb = 0.32 mm. Even though HD-2 was 50% thicker than LD-2, it has
a lower aspect ratio (0.125 for HD-2 and 0.189 for LD-2). Again, this is due to the longer webs in the HD sample
compared to the LD sample.
Table C.6 summarizes the thickness and aspect ratio data at failure for all of the webs considered in this
first study. As some webs had multiple failures, these are each separately recorded in the table. The table is
129
Figure C.10 LD-2 as-fabricated from the side (a), straight on (b), in vertical cross-section (c), and partially compressed from the side
(d), straight on (e), and in vertical cross-section (f). The scale bar is 2.5 mm and applies to all the figures.
Figure C.11 Thickness profile of LD-2 and the vertical cross-section for comparison. The dashed lines mark the locations of initiation
of instability.
Table C.3 LD-2 dimensions at failure initiation points.
P (mm mm-1
) tweb (mm) tweb/b
0.11 0.29 0.103
0.70 0.32 0.189
0.89 0.31 0.152
130
Figure C.12 HD-1 as-fabricated from the side (a), straight on (b), in vertical cross-section (c), and partially compressed from the side
(d), straight on (e), and in vertical cross-section (f). The scale bar is 2.5 mm and applies to all the figures.
Figure C.13 Thickness profile of HD-1 and the vertical cross-section for comparison. The dashed line marks the location of initiation
of instability.
Table C.4 HD-1 dimensions at failure initiation point.
P (mm mm-1
) tweb (mm) tweb/b
0.42 0.21 0.042
131
Figure C.14 HD-2 as-fabricated from the side (a), from the opposite side (b), straight on (c), in vertical cross-section (d), and partially
compressed from the side (e), from the opposite side (f), straight on (g), and in vertical cross-section (h). The scale bar is
2.5 mm and applies to all the figures.
Figure C.15 HD-2 buttress as-fabricated (a), in vertical cross-section (b), and partially compressed (c), in vertical cross-section (d).
The scale bar is 1 mm and applies to all the figures.
132
Figure C.16 Thickness profile of HD-2 and the buttress (inset) with vertical cross-sections for comparison. The dashed lines mark the
locations of initiation of instability.
Table C.5 HD-2 dimensions at failure initiation points.
P (mm mm-1
) tweb (mm) tweb/b
0.72 0.50 0.125
0.82 0.31 0.345
0.90 0.46 0.123
Table C.6 Thickness and aspect ratio correlated with failure mechanism
Web tweb (mm) tweb/b Failure Mechanism
HD-1 0.21 0.042 buckle
LD-3 0.18 0.054 buckle
LD-3 0.19 0.058 buckle
HD-4 0.39 0.076 buckle
HD-4 0.32 0.095 buckle
LD-1 0.15 0.103 buckle
LD-2 0.29 0.103 buckle
HD-4 0.30 0.103 buckle
LD-5 0.35 0.115 tear
HD-3 0.57 0.119 buckle
HD-2 0.46 0.123 buckle
HD-2 0.50 0.125 buckle
LD-4 0.60 0.128 tear
LD-4 0.61 0.128 buckle/tear
HD-5 0.93 0.145 tear
LD-2 0.31 0.152 buckle
LD-5 0.38 0.161 buckle
LD-2 0.32 0.189 tear
LD-5 0.45 0.217 tear
LD-4 0.51 0.250 tear
HD-2 0.31 0.345 tear
133
sorted by aspect ratio (tweb/b), and the failure mechanism at each failure is recorded as well. Critical buckling
stress for a thin wall, first developed by Bryan in 1890 [16] is a function of the aspect ratio, tweb/b, squared. A
slender web, with a low aspect ratio, will have a lower critical buckling stress. A web with a higher aspect ratio
will hold more stress, until reaching either its critical buckling stress or the local fracture strength of the material.
From the table, it is clear that if the aspect ratio is greater than 0.17, the web will fail by tearing. Conversely, if
the aspect ratio is less than 0.11, the web will fail by buckling only. Between these values, the web will either
tear or buckle depending on the local environment and geometry. For example, LD-2 with an aspect ratio of
0.152 buckled above a tear, so the web’s constraints had radically changed before that buckle occurred. Critical
buckling stress is also a function of the end constraints, which depend in part on the vertical aspect ratio of the
webs, a/b.
C.4 Conclusions
The strength and stiffness of stochastic honeycombs are determined by the complex internal architecture of the
webs. Micro-CT scans were used to track the detailed local failure mechanisms within the core of the sandwich
structure for ‘low’ (11%) and ‘high’ (18%) relative density variants. Post failure micro-CT characterization
showed that web buckling and web fracture were seen in both sample types. Web buckling tended to occur at the
thinnest portions of the web, whether this was at mid-height or not. The subset of partially constrained webs (i.e.
those supported on one side and free on the other) could be classified into one of two failure modes: web buckle
only and web buckle plus web fracture. Webs with a thickness-to-length aspect ratio of tweb/b < 0.11 tended to
buckle only, while higher aspect ratio webs would both buckle and fracture (tweb/b > 0.18).
134
References
[1] M. Hostetter, B. Cordner, G.D. Hibbard. Stochastic honeycomb sandwich cores, Composites Part B: Engineering
43 (2012) 1024-1029.
[2] C. Nida-Core. Polypropylene Honeycomb Core, Composite Reinforcements - Nida-Core Technical Data. 2008.
[3] L. Hexacor. Hexacor - Honeycomb Core Materials. 2009.
[4] I. Plascore. Honecyomb Core - Aluminum Honecyomb - Plastic Honeycomb Panels. 2009.
[5] Polycore Composites - Polycore Australia - Polycore Honeycomb.
[6] J. Zhang, M.F. Ashby. The out-of-plane properties of honeycombs, International Journal of Mechanical Sciences
34 (1992) 475-489.
[7] T. Wierzbicki. Crushing analysis of metal honeycombs, International Journal of Impact Engineering 1 (1983)
157-174.
[8] C.D. Moen, B.W. Schafer. Elastic buckling of thin plates with holes in compression or bending, Thin-Walled
Structures 47 (2009) 1597-1607.
[9] D.S. Chehil, S.S. Dua. Buckling of Rectangular Plates With General Variation in Thickness, Journal of Applied
Mechanics 40 (1973) 745-751.
[10] J. Rhodes. Buckling of thin plates and members -- and early work on rectangular tubes, Thin-Walled Structures
40 (2002) 87-108.
[11] R. Brighenti. Buckling of cracked thin-plates under tension or compression, Thin-Walled Structures 43 (2005)
209-224.
[12] H.G. Allen, P.S. Bulson. Background to buckling, McGraw-Hill Book Co., New York, 1980.
[13] F. Bloom, D. Coffin. Handbook of thin plate buckling and postbuckling, Chapman & Hall/CRC, Boca Raton,
Florida, 2000.
[14] A.P. Chattopadhyay, E. Frink, X.J. Xin, K.B. Lease. Elastic buckling characteristics of triangular tubes under
uniaxial loading, Thin-Walled Structures 52 (2012) 29-41.
[15] K.S. Kandil, C.R. Calladine. Classical, local buckling of tubes having rectangular cross-sections, International
Journal of Mechanical Sciences 28 (1986) 789-797.
[16] G.H. Bryan. On the Stability of a Plane Plate under Thrusts in its own Plane, with Applications to the "Buckling"
of the Sides of a Ship, Proceedings of the London Mathematical Society s1-22 (1890) 54-67.