development of lightweight materials by meso- and
TRANSCRIPT
Scholars' Mine Scholars' Mine
Doctoral Dissertations Student Theses and Dissertations
Fall 2020
Development of lightweight materials by meso- and Development of lightweight materials by meso- and
microstructure control microstructure control
Myranda Shea Spratt
Follow this and additional works at: https://scholarsmine.mst.edu/doctoral_dissertations
Part of the Materials Science and Engineering Commons
Department: Materials Science and Engineering Department: Materials Science and Engineering
Recommended Citation Recommended Citation Spratt, Myranda Shea, "Development of lightweight materials by meso- and microstructure control" (2020). Doctoral Dissertations. 2955. https://scholarsmine.mst.edu/doctoral_dissertations/2955
This thesis is brought to you by Scholars' Mine, a service of the Missouri S&T Library and Learning Resources. This work is protected by U. S. Copyright Law. Unauthorized use including reproduction for redistribution requires the permission of the copyright holder. For more information, please contact [email protected].
DEVELOPMENT OF LIGHTWEIGHT MATERIALS BY MESO- AND MICRO
STRUCTURE CONTROL
by
MYRANDA SHEA SPRATT
A DISSERTATION
Presented to the Graduate Faculty of the
MISSOURI UNIVERSITY OF SCIENCE AND TECHNOLOGY
In Partial Fulfillment of the Requirements for the Degree
DOCTOR OF PHILOSOPHY
in
MATERIALS SCIENCE AND ENGINEERING
2020
Approved by:
Joseph W. Newkirk, Advisor K. Chandrashekhara
Jeremy Watts Yijia Gu
Ronald O’Malley
© 2020
Myranda Shea Spratt
All Rights Reserved
PUBLICATION DISSERTATION OPTION
iii
This dissertation consists of the following four articles, formatted in the style used
by the Missouri University of Science and Technology:
Paper I, found on pages 13-33, has been published in the Rapid Prototyping
Journal.
Paper II, found on pages 34-63, has been submitted to Metallurgical and
Materials Transactions A.
Paper III, found on pages 64-97, has been accepted for publication by S/N
Applied Sciences.
Paper IV, found on pages 98 -125, has been submitted to Metallurgical and
Materials Transactions B.
iv
ABSTRACT
In this work, two lightweight structures - lattice structures and metal matrix
syntactic foams (MMSF) - were studied. Honeycomb lattices were manufactured by
powder bed selective laser melting (SLM) from 304L stainless steel. The wall thicknesses
of these structures ranged from 0.2 to 0.5 mm. Surface roughness was the primary cause
of dimensional mismatch between the expected and as-built structures with an average
wall thickness increase of 0.12 mm. The strength of the honeycombs increased with
increasing wall thickness. A feature of the SLM microstructure, the melt pool boundary,
was also studied as a part of this work. 3D models of the melt pool boundary network
were developed. The goal of this study was to determine if the melt pool boundary
significantly affects elastic or plastic properties in SLM materials. The study found a
negative correlation between the concentration of the melt pool boundary network and
the yield strength. The melt pool boundaries also seem to contribute to anisotropic plastic
deformation. The copper matrix syntactic foams studied in this work were manufactured
using low-pressure binder injection molding. The binder was a water-agar-glycerin gel
optimized to achieve the highest sintered density for metal matrix syntactic foams.
Syntactic foams use hollow or porous particles as their filler material. The two fillers
used here were porous silica and bubble alumina, selected for their extreme differences in
size, material, and porosity distribution. Three samples for each filler with increasing
filler volume fraction were manufactured. The specific strength of the copper-porous
silica samples increased with increasing volume fraction. The copper-alumina bubble
sample greatly improved the impact resistance of the foam at low filler volume fraction.
v
ACKNOWLEDGMENTS
My work has been greatly helped by the support and assistance of many people,
without which I would not have finished this dissertation.
I would first like to thank my advisor, Dr. Joseph W. Newkirk, for the opportunity
to pursue this research at Missouri S&T. His expertise, guidance, and feedback have been
invaluable over the course of my dissertation.
I would like to acknowledge the Missouri University of Science and Technology
for the funding I received as a part of the Chancellor’s Distinguished Fellowship. The
financial support allowed me the freedom to research my passion and the opportunity to
meet people from both my own and other fields.
I would like to thank Dr. KC for the support and advice he gave over our
collaborative projects, which have enriched my experience during my graduate studies.
I would also like to thank fellow graduate students, both in my own McNutt Hall
and in Toomey Hall, for many years of discussion and spirited debate over our research
topics.
Finally, I would like to thank my family and friends for their unending
encouragement and faith. A special thanks to my dear husband, Cody, whose wise
counsel and encouragement kept me going.
Vi
TABLE OF CONTENTS
PUBLICATION DISSERTATION OPTION...................................................................iii
ABSTRACT....................................................................................................................... iv
ACKNOWLEDGMENTS.................................................................................................. v
LIST OF ILLUSTRATIONS............................................................................................. ix
SECTION
1. INTRODUCTION.................................................................................................... 1
1.1. INTRODUCTION AND RESEARCH OBJECTIVES...................................... 1
1.2. LATTICE STRUCTURES.................................................................................4
1.3. METAL MATRIX SYNTACTIC FOAM.......................................................... 8
1.4. ORGANIZATION OF DISSERTATION........................................................ 11
PAPER
I. BUILD ACCURACY AND COMPRESSION PROPERTIES OF ADDITIVELY MANUFACTURED 304L HONEYCOMB............................................................ 13
ABSTRACT................................................................................................................. 13
1. INTRODUCTION.................................................................................................... 14
2. METHODOLOGY................................................................................................... 16
3. RESULTS AND DISCUSSION..............................................................................20
4. CONCLUSIONS...................................................................................................... 30
REFERENCES ............................................................................................................ 32
II. EFFECT OF THE MELT POOL BOUNDARY NETWORK ON THE ANISOTROPIC MECHANICAL PROPERTIES OF SELECTIVE LASER MELTED 304L........................................................................................................ 34
Page
ABSTRACT.................................................................................................................34
1. INTRODUCTION.................................................................................................... 35
2. MODELING............................................................................................................. 37
3. PROCEDURE ......................................................................................................... 44
4. RESULTS AND DISCUSSION..............................................................................47
4.1. EXPERIMENTAL WORK.............................................................................. 47
4.2. MODEL WORK...............................................................................................49
5. CONCLUSIONS...................................................................................................... 58
REFERENCES ............................................................................................................ 60
III. OPTIMIZATION AND CHARACTERIZATION OF NOVEL INJECTIONMOLDING PROCESS FOR METAL MATRIX SYNTACTIC FOAMS...........64
ABSTRACT ................................................................................................................ 64
1. INTRODUCTION.................................................................................................... 65
2. MATERIALS AND METHODS............................................................................. 69
2.1. POWDER CHARACTERIZATION................................................................ 69
2.2. BINDER DEVELOPMENT............................................................................. 70
2.3. INJECTION MOLDING.................................................................................. 71
2.4. SAMPLE POST-PROCESSING AND CHARACTERIZATION................... 73
3. CALCULATION..................................................................................................... 74
4. RESULTS AND DISCUSSION.............................................................................. 77
4.1. POWDER CHARACTERIZATION................................................................ 77
4.2. BINDER OPTIMIZATION.............................................................................. 80
4.3. INJECTION MOLDING.................................................................................. 84
vii
4.4. BINDER BURNOUT AND SINTERING 87
4.5. MECHANICAL TESTING............................................................................ 91
5. CONCLUSIONS.................................................................................................... 91
CONFLICT OF INTEREST........................................................................................ 92
APPENDIX.................................................................................................................. 93
REFERENCES ............................................................................................................ 94
IV. INFLUENCE OF FILLER MATERIAL AND VOLUME FRACTION ONCOPPER MATRIX SYNTACTIC FOAMS.......................................................... 98
ABSTRACT.................................................................................................................98
1. INTRODUCTION.................................................................................................... 99
2. EXPERIMENTAL PROCEDURE......................................................................... 102
3. RESULTS AND DISCUSSION............................................................................ 106
3.1. POWDER CHARACTERIZATION.............................................................. 106
3.2. SAMPLE CHARACTERIZATION............................................................... 108
3.3. MECHANICAL TESTING............................................................................ 114
4. CONCLUSIONS.................................................................................................... 119
ACKNOWLEDGEMENTS........................................................................................121
REFERENCES........................................................................................................... 121
SECTION
2. CONCLUSIONS AND FUTURE WORK...........................................................126
BIBLIOGRAPHY ........................................................................................................... 130
viii
VITA 138
ix
LIST OF ILLUSTRATIONS
Figure 1.1: Three unit cell designs are shown...................................................................5
Figure 1.2: An optical micrograph of the as-built SLM microstructure of 304L stainlesssteel....................................................................................................................7
Figure 1.3: A SEM image of a copper and alumina bubble syntactic foam...................... 9
PAPER I
Figure 1: The CAD model for the H2 honeycomb specimens.........................................17
Figure 2: The surface of a honeycomb wall.....................................................................21
Figure 3: The width of a wall can be indirectly calculated from the number of melt pool tracks across it, the hatch spacing, and the average width of a melt pool..................................................................................................................23
Figure 4: Etched micrographs of honeycomb specimens................................................. 25
Figure 5: Columnar grains that cross the melt pools were observed.................................26
Figure 6: A schematic of the locations from which hardness measurements were takenis shown............................................................................................................ 26
Figure 7: An example of the stress- displacement curves generated by the compressiontesting................................................................................................................27
Figure 8: A buckled wall after compression testing. T..................................................... 28
Figure 9: Yield and ultimate stress and specific yield stress of the honeycomb areshown............................................................................................................... 31
Figure 10: The yield stress and ultimate stress of the solid material are shown..............31
PAPER II
SECTION Page
Figure 1: The stripes scan pattern showing the input parameters used by the RenishawAM250............................................................................................................. 39
Figure 2: Optical images of SLM 304L in two magnifications.......................................40
x
Figure 3: The model of a single melt pool compared to the actual solidified melt poolin the 304L ss...................................................................................................41
Figure 4: The model of a single layer with two colors to indicate the change in laserdirection as the melt pools were created..........................................................42
Figure 5: The scan pattern rotation is shown beside the model for that scan patternrotation..............................................................................................................42
Figure 6: The single melt pool mesh for the 0° hatch angle model, before and afterBooleans are applied........................................................................................43
Figure 7: The three RVE models after the boundaries were removed.............................43
Figure 8: The 67° hatch angle model rotated around the [100] direction by 0°, 45°,and 90°.............................................................................................................44
Figure 9: The four build orientations of the cylinders are shown....................................46
Figure 10: Anisotropy in the 0°-[100]-0° specimens after compression testing..............50
Figure 11: The 0°-[010]-45° specimens sheared as the other 0° hatch angle specimensdid, but the (001) anisotropy was not elliptical............................................... 51
Figure 12: A hardness map of the 0°-[100]-90° specimen after compression testing.....52
Figure 13: There were two types of lines produced when the MPBN concentration wasplotted over the length of the model - flat and pseudo-sinusoidal.................53
Figure 14: The yield strength vs concentration of the MPNB in the compressiondirection are correlated.................................................................................... 54
Figure 15: The edges of the cylinders rippled after compression testing........................ 55
Figure 16: A close view of the melt pool boundary and cellular structure before (A)and after (B) compression testing....................................................................56
Figure 17: A 0°-[100]-0° specimen after compression testing shown from the side andthe top of the cylinder......................................................................................57
Figure 18: A 0°-[010]-90° specimen after compression testing shown from the sideand the top of the cylinder............................................................................... 58
PAPER III
Figure 1: The particle size distribution by count and volume is shown..........................79
Figure 2: SEM images of each powder are shown...........................................................80
Xi
Figure 3: The response surfaces for the a) wet density, b) dry density, c) bulk sintereddensity, d) open porosity after sintering........................................................82
Figure 4: The percent mass loss in each binder-only injection molded specimen overtim e................................................................................................................85
Figure 5: The percentage of binder in each specimen versus specimen number.............88
Figure 6: The percentage of porous silica in each specimen versus specimen number, ... 88
Figure 7: A element map of a 0% P-S specimen with the carbon, copper, and oxygenelements identified...........................................................................................89
Figure 8: A micrograph of a 15% P-S specimen showing the dispersion of the poroussilica particles and the morphology of the particles after sintering...............90
Figure 9: The strength and density of several MMSF materials from this work and theliterature...........................................................................................................92
PAPER IV
Figure 1: The particle size distribution by count percent and cumulative count percent for each powder where a) is copper, b) is alumina bubble, and c) is porous silica...............................................................................................................107
Figure 2: SEM images of the three powders - a) copper, b) silica, c) and d) alumina -show the morphology and relative size of the powders.................................108
Figure 3: The cross-sections of AB powder particles were used to determine thedensity of the AB powder.............................................................................. 109
Figure 4: Some of the AB particles in the copper matrix filled with copper powder....112
Figure 5: Elemental mapping of a 100% copper specimen shows the distribution ofcopper and oxygen in the sample...................................................................114
Figure 6: EDS map of the elements present in a P-S and copper specimen; elementsare indicated on the figure............................................................................. 115
Figure 7: EDS map of the elements present in an AB and copper specimen; elementsare indicated on the figure............................................................................. 116
Figure 8: The specific modulus and 0.2% offset yield strength for each sample.......... 117
Figure 9: The specific modulus of the porous silica and alumina bubble samplescontrasted with a model for metal foams by Zhu et al [42]......................... 118
Figure 10: The absorbed energy of each sample...........................................................120
LIST OF TABLES
Table 1: The nominal and measured dimensions for the honeycomb samples................. 18
Table 2:The measured wall thickness and cell size values, adjusted to correct forparticles sintered to the walls.............................................................................. 21
Table 3: The calculated width of the honeycomb walls using the number of melt pooltracks, width of an average melt pool, and hatch spacing.................................. 23
PAPER II
Table 1: The test matrix for each sample, where the orientation listed shows the planenormal to the compression direction.................................................................. 45
Table 2: The yield strength, aspect ratio horizonal plane of each cylinder after testingfor each specimen are shown.............................................................................48
Table 3: The maximum, minimum, and difference between the pseudo-sinusoidal|peaks for the three directions of interest.......................................................... 59
PAPER III
Table 1: Details provided by the manufacturer on each metal and ceramic powder usedin this experiment............................................................................................... 70
Table 2: The composition of each slurry in the injection molding reservoir beforeinjection molding is listed.................................................................................. 72
Table 3: The general composition of each sample for the model calculation.................. 75
Table 4: The composition of each binder for the model calculation, all units in volumepercent................................................................................................................ 75
Table 5: The 10th, 50th, 90th percentiles and average particle size in microns for eachpowde................................................................................................................. 80
Table 6: The cubic model b-values calculated for each response and the predictedvalue based on the optimized composition........................................................ 83
Table 7: The error for each response and relevant variance calculations are tabulated. .. 84
xii
PAPER I Page
Table 8: The average volume percent of binder and porous silica in each sample afterinjection molding............................................................................................... 87
Table 9: The relative density of each sample after each stage of sample postprocessing shows the approximate amount of porosity in each sample...........89
PAPER IV
Table 1: The solids loading, filler, and copper volume percent of each sample............103
Table 2: The density and relative density for each stage and sample are shown withtheir standard deviation.....................................................................................110
Table 3: The volume percent of binder in each sample was used to calculate a moreaccurate relative density for each sample......................................................... 111
xiii
SECTION
1. INTRODUCTION
1.1. INTRODUCTION AND RESEARCH OBJECTIVES
Porosity is often considered a defect in a material, but with careful engineering, it
can be turned into a strength. Generally, structural properties such as strength and
stiffness increase as material density increases [1]-[6]. Material design charts produced
by Ashby [7] plot this relationship for many materials. The goal of lightweight structural
material design is to create materials have higher structural property to density ratios
compared to other materials. By controlling porosity or other density-decreasing
additives, such materials can be designed. [1], [2], [7]-[10]. This work investigates two
different structure-controlled materials - lattice structures and metal matrix syntactic
foams.
Lattice structures are porous foams with engineered geometry. [1], [10].
Periodicity is beneficial for these structures in that the properties are more predictable. A
unit cell, a shape translated and repeated in 2 or 3 dimensions where each cell is on the
micro to meso length scale, is one of the simplest lattice structure designs. This design is
often inspired by crystal lattice structures and trusses [1], [11]. Lattice structure designs
also include graded cellular structures, unit cells with curved surfaces, auxetic unit cells,
and more [2], [10]-[13]. Lattice structures are typically manufactured using either
investment casting or an additive manufacturing (AM) method. Selective laser melting
(SLM), a powder bed AM process, is often used to make complex lattice structures that
are not otherwise possible to fabricate. The powder bed process is ideal for these
structures as the powder can support the lattice while it is built. The literature includes
much research on the process-microstructure-property relationship, but there are aspects
of that relationship that are poorly defined [14]—[16]. One aspect involves a
microstructure feature known as the melt pool boundary. The fish-scale shape of a
solidified melt pool is a well-known feature of the SLM process, and it appears in nearly
every as-built, etched micrograph of SLM materials. The early literature on the melt pool
boundary indicates that may be a failure point in the structure [16]—[19]. The effect of the
melt pool boundary may be magnified in the thin features of the SLM lattice structures.
SLM-manufactured lattice structures are also vulnerable to dimensional mismatch and
surface roughness, demanding rigorous optimization and analysis be performed on the
structures [20].
Metal matrix syntactic foams are three-phase particulate composites. The MMSF
consists of a metal matrix embedded with hollow or porous filler particles — usually
ceramic or glass. Porosity in these composites is controlled by the particle size of the
filler. The filler surrounds the porosity, shielding the metal matrix from the pores. The
filler also reinforces the matrix as in solid metal matrix composites [8]. These materials
have a wide range of potential applications through the primary research vector has been
lightweight structural applications [21]-[25]. Heat sinks, electrical and vibrational
shielding, ballistic armor, packaging, and other applications are other potential uses for
these materials [4], [5], [26], [27]. The design of MMSF materials is limited by sources
of porous or hollow filler material, material compatibility between the phases, and
manufacturing the composites [9], [28], [29]. Porosity-containing filler powders are not
2
3
available in a wide variety of sizes, materials, or densities. The ones that are available are
not compatible with every matrix material or process. For good material compatibility,
both the filler and the matrix must survive the melting or sintering process without
undesirable diffusion, corrosion, or interfacial reactions. There are three main processing
techniques used for MMSF manufacturing. These are stir casting, pressure infiltration,
and binder injection molding [5]. Injection molding is the method used in this work. In
this technique, the metal and filler powders are combined with a binding agent to form a
slurry. This slurry is then injected into a mold. Parts are densified by sintering. Slurry
segregation, matrix porosity, and material compatibility during sintering are the key
challenges to overcome with this processing technique.
The knowledge gaps targeted in this research included the process-microstructure-
property relationship in SLM lattice materials and injection molded metal matrix
syntactic foams. This work expanded the body of research involving structure-controlled
lightweight structural materials. Four papers were written that each address a specific
objective, as follows:
1. Identify key build quality control issues in SLM lattice structures and study
their effects on the properties of the lattice.
2. Study an aspect of SLM microstructure that has received little attention in the
past, the melt pool boundary, and determine the effect of the melt pool boundary on
anisotropic material properties.
3. Develop a water-based binder for use in a low-pressure injection molding
processing, specifically optimized to decrease matrix porosity in a metal matrix syntactic
foam.
4
4. Determine the effect of filler material and volume percent on the properties and
manufacturing of metal matrix syntactic foams.
1.2. LATTICE STRUCTURES
Lattice structures are designed by manipulating the meso-scale geometry of the
material. Many unit cell lattice structures are trusses derived from crystal lattice
structures. Honeycombs are common 2D structures based on the HCP structure [30].
Cubic truss structures, based on BCC or FCC structures, are common 3D lattice unit cells
[10], [11], [31]-[33]. Other designs include curved surfaces such as gyroids and non
crystal structure inspired truss designs like the octet truss [2], [10]—[12], [34], [35].
Figure 1.1 shows a honeycomb, a BCC truss, and a gyroid unit cell. The unit cell design
can be optimized by grading the unit cell density if, for example, certain areas of a part
need to take more load than others or if successive failure of the part is desired. Density
grading can be done by changing the strut size or the unit cell size in the target area [36].
A critical part of the lattice structure design process is modeling the mechanical
properties. The mechanical properties of lattice structures are geometry-dominated [30],
[34], [37]. For example, honeycomb structures are stronger vertically than in-plane with
the honeycomb octagons, regardless of the material used [38]. The unit cell properties
can be modeled with mathematics, but modern tools - such as finite element analysis
(FEA) - are more common [1], [10], [20], [30], [31], [34], [35], [37], [38]. A
representative volume element can be used to model the structure, rather than an entire
part. Unit cell designs are particularly useful for research and development if a part is not
yet envisioned for a specific application or if the part shape is confidential.
5
Figure 1.1: Three unit cell designs are shown- A) a 2D honeycomb, B) a BCC truss, andC) a gyroid.
Lattice structures can be manufactured using investment casting or an additive
manufacturing technique [10], [39]. Some 2D lattice structures can be made using
extrusion as well [39], [40]. Investment casting of lattice structures is time-consuming
and costly. The molds used for this method are made by coating a sacrificial polymer in
ceramic slurry. After the polymer is removed, the mold is filled with a suitably fluid
metal [4]. This method is well-understood in the literature, but the strut size and
complexity are limited by the flowability of the molten metal. Sheet folding and welding
can be used to make some lattice structure designs. In this method, sheets of metal are
deformed into the desired shapes then welded together. Powder bed selective laser
melting (SLM) is an excellent method of manufacturing lattice structures on a small scale
[10], [14]-[16]. The powder bed supports the lattice as it is built, and it can also be easily
removed from an open-celled lattice structure after building. The laser used as a heat
source is precisely controlled and capable of printing thin features on the order of a
hundred microns. The laser melts powder in a pre-determined path, layer by layer, with
fresh powder spread over the bed after each layer. This is known as the laser scan pattern.
The melt pool that forms from the melted powder solidifies rapidly as it cools. The size
and shape of the melt pool depends on the laser energy, residual heat from the
surrounding powder and dense part, and the properties of the material.
Research into the process-microstructure-property relationships in SLM parts is
active, and the majority is focused on the process parameter influence on properties [15],
[16], [18], [41]-[46]. The distinctive fish-scale microstructure of SLM materials is a
result of the laser scan pattern used to make the part [15], [16], [44], [45], [47], [48]. In
steel alloys, the solidified melt pools are outlined by a thin feature, called the melt pool
boundary, with a cellular sub-grain structure inside the melt pool. [16]—[19]. Figure 1.2
shows these microstructural features. Literature on the melt pool boundary seems to
indicate that it is a weak feature in the material. For example, cracks in fractured SLM
material may follow the melt pool boundary [19], [49]. Grains in SLM materials are often
columnar and grow in the build direction, often through several melt pools. These grains
are often oriented in the same direction [15], [16], [18], [48]. Porosity defects are usually
related to the powder, assuming other variables are optimized [45], [50], [51]. Gases and
water trapped between or inside particles can cause inclusions and gas porosity [11],
[51]-[53] All of these and more variables affect the final SLM material [47].
For lattice structures, the SLM process parameters are tuned for different goals
than in bulk parts [11], [54]-[56] Lattice structures have both high surface area and
geometry-dominated properties. The SLM process parameters should be tuned to reduce
dimensional mismatch between the intended and as-built geometry in these materials.
Dimensional mismatch can occur in several ways. Surface roughness can increase the
volume and mass of the structure enough to change the unit cell density. It also causes
6
7
Figure 1.2: An optical micrograph of the as-built SLM microstructure of 304L stainless steel, etched electrolytically in a 60:40 nitric acid: water solution.
local variation is strut width [20], [35]. The surface roughness of SLM parts comes in two
forms. First, un-melted powder welded to the surfaces of the struts is a relatively uniform
addition to the surface of the part. Second, melt pools at the edges of the part can also add
a rippled texture to the surface of the part from surface tension. Models that include a
variable strut diameter more accurately match experimental data, indicating that the
variation has a significant effect on the properties [31]. Dimensional mismatch can also
be caused by misprinting due to design limitations. Overhangs and smooth curved
surfaces can fail or include stair-stepping. Each machine has a lower limit on part width.
The melt pool size may be on the order of the width of a feature, which may mean that
the overlap between melt pools should to be adjusted to hit the target dimension of that
feature or vice versa [55]. Closed internal spaces will be filled with powder. For the
lattice structure, any unit cell should be tested for printability before printing a full part
[31], [33]. Also, it is often appropriate to attempt to model the actual as-built shape of the
printed part if time and cost allow [20], [55].
1.3. METAL MATRIX SYNTACTIC FOAM
Metal matrix syntactic foams (MMSFs) are particulate composites with
lightweight structural applications. The filler particles used in these materials contain
porosity which lightens the structure. Figure 1.3 shows an example of the MMSF
microstructure. MMSFs can be tailored to a specific application by changing the matrix
material, filler material, filler volume fraction, and filler shell thickness. The notable
lightweight structural properties of these materials are their excellent energy absorption,
specific strength, and specific stiffness [9], [24], [25], [28], [57]-[59]. These materials
can also have properties of both closed-celled foams and metal matrix composites as
well, such as wear resistance, vibrational damping, electrical and thermal conductivity,
and more [4], [26], [60]-[62].
Material compatibility between the filler and the matrix material is vital to the
quality production of MMSF parts. Many defects in these materials occur between the
matrix and filler during processing. Defects include unwanted porosity in the matrix,
fractured hollow spheres, unwanted matrix-filler interactions, and density gradients [21],
[25], [63]-[66]. The metal matrix is in contact with the filler during processing, usually at
high temperatures. This can cause chemical reactions between the matrix and filler -
inclusions, diffusion, and corrosion [21], [25], [28], [57], [67]-[69]. Certain metal-filler
combinations are particularly detrimental, such as alloys with very high oxygen affinities
8
9
Figure 1.3: A SEM image of a copper and alumina bubble syntactic foam.
(aluminum, for example) paired with oxide-containing ceramic or glass fillers. Beneficial
grain refinement can occur as well near the surface of the filler [26]. Because of their
relatively low softening temperature, many glass hollow spheres soften at the required
processing temperatures, which can collapse the internal porosity. [70], [71]. Pure silica,
which has a high softening temperature, can dissolve alloying elements from the metal
that lower the softening point of the glass [72].
There are two main types of porosity-containing filler materials available. First,
there are naturally occurring or by-product fillers. These include expanded clay/perlite
and fly ash cenospheres [29], [59], [73], [74]. The fly ash cenospheres in particular are
widely utilized MMSF filler materials because they are otherwise a waste product [9],
[75]. Naturally occurring fillers usually contain impurities; are often an aluminosilicate
with glassy phases; and may contain fractured particles and porous shell walls [22], [23],
10
[26], [28], [57], [60], [61], [65], [69], [76], [77]. Second, hollow filler materials
manufactured in a laboratory are available. Ceramic hollow spheres tend to be larger than
0.1 mm in size which may not be acceptable for all processing techniques [21], [25], [67],
[68]. Glass and metal hollow spheres are also available [58], [66], [70]-[72], [78]. Glass
filler can be available in smaller particle sizes than the ceramic spheres. The available
selection of hollow or porous filler limits the matrix materials and processes that can be
used [79].
The manufacturing of MMSF materials requires balancing several competing
factors. First, the filler particulates should be well-dispersed in the material. The filler
particles have very low density (less than 1 g/cm3) compared the matrix, so they tend to
float or segregate out of mixtures. Second, the particulates should keep their internal
porosity intact. The porosity can collapse if the particles soften or melt during processing.
The particles can fracture from mechanical stresses during processing. The matrix
material can infiltrate the pores if the shell wall is porous or corroded away. As a
corollary to the second factor: the selected manufacturing technique should be a near-net
shape process as subtractive machining will expose the filler porosity. Third, matrix
porosity should be eliminated from the material so that only the (controlled) filler
porosity remains.
With these criteria and restrictions, there are few processing techniques that can
be used to manufacture MMSF materials. Stir casting is one of the more popular methods
of making MMSFs in the literature [26], [57], [60], [61]. In this method, an impellor is
used to mix the filler into molten metal. The filler tends to float due to low density, so the
casting and cooling of the MMSF must be rapid. The impellor can fracture the filler due
to high shear forces. Further, this process only works with castable metals. Another
common processing method is infiltration [21]-[23], [25], [26], [28], [67], [68], [74],
[76], [80]. In this method, a bed of the filler is infiltrated by the molten metal with or
without pressure assistance. Infiltration can leave matrix porosity behind if the wetting
between the filler and the metal is poor or if the metal has a high viscosity. If the
infiltration pressure is too high, the filler can fracture. Powder metallurgy methods such
as binder injection molding are less common in the literature [78], [81]-[83]. In injection
molding, the filler, metal, and binder will tend to segregation while the slurry is non-solid
[66], [84]. Further, the hollow spheres can fracture if the shear forces during
compounding are too high [78]. The sintering temperature can also be an issue. If a
liquid-phase sintering process is used, it can affect the filler similar to the melt processes.
If no liquid phase is used and no pressure, then matrix porosity can be an issue.
1.4. ORGANIZATION OF DISSERTATION
An outline of this dissertation’s research objectives was given in Section 1.1. The
background research on lattice structures and metal matrix syntactic foams was reviewed
in Section 1.2 and 1.3. Four papers, each addressing one of the research objectives, were
included in this dissertation.
The first paper focused on the characterization and compression testing of SLM
honeycomb structures with increasing wall thickness. The dimensional mismatch
between the 3D model of the honeycomb structures and the as-built specimens was
thoroughly characterized. The effect of wall thickness on the mechanical properties of the
honeycomb lattice structures was investigated.
11
12
In the second paper, the melt pool boundary network, a microstructural feature of
the SLM process, was investigated. This feature is not well understood in the literature,
though a growing body of research indicates that it does influence the properties of SLM
parts. The melt pool boundary network was modeled using 3D software. The correlation
between this feature and the elastic and plastic properties of selective laser melted 304L
steel was investigated.
The third paper shifted the focus from lattice structures to metal matrix syntactic
foams. A water-based binder suitable for low-pressure injection molding of MMSF
material was developed. MMSF samples with a porous silica filler and a copper matrix
were produced and characterized.
The fourth paper used the process developed in the third paper to test the impact
of two different fillers on the foam properties. The porous silica and alumina bubble
powders were selected for their differences in size, material, and porosity distribution.
The contrast strongly showed the strengths and weaknesses of the two types of filler. The
effect of filler material and volume fraction on the yield strength and impact energy of
the MMSF materials was studied.
13
PAPER
I. BUILD ACCURACY AND COMPRESSION PROPERTIES OF ADDITIVELY MANUFACTURED 304L HONEYCOMB
ABSTRACT
Purpose: Thin walled metal honeycomb structures made with a selective laser
melting process were analyzed for their build quality and compression properties in this
study. The honeycombs were tested over increasing densities. Methodology: The density
of the honeycombs was changed by increasing wall thickness of each sample. The
honeycombs were tested under compression. Differences between the CAD model and
the as-built structure were quantified by measuring physical dimensions. The
microstructure was evaluated by optical microscopy, density measurements, and
microhardness. Findings: The Vickers Hardness (HV) of the honeycomb structures was
209±14 at 50g load. The compression ultimate and yield strength of the honeycomb
material were shown to increase as the wall thickness of the honeycomb samples
increased. The specific ultimate strength also increased with wall thickness, while the
specific yield stress of the honeycomb remained stable at 42 ± 2.7 MPa/g/cm3. The
specific ultimate strength minimized near 0.45 mm wall thickness at 82 ± 5 MPa/g/cm3,
and increased to 134 ± 3 MPa/g/cm3 at 0.6 mm wall thickness. Value: This work
highlights a single lightweight metal structure, the honeycomb, built by additive
manufacturing. The honeycomb is an interesting structure because it is a well-known
building material in the lightweight structural composites field, but is still considered a
relatively complex geometric shape to fabricate. As shown here, AM techniques can be
used to make complex geometric shapes with strong materials to increase the design
flexibility of the lightweight structural component industry.
1. INTRODUCTION
14
Additive manufacturing (AM) research continues to be popular as interest in
metal AM grows. However, the expensive trial and error adjustment of process
parameters to tune properties prohibits many industries from utilizing AM. [1], [2] Trial
and error procedures are usually done when there is not enough information about a
process to accurately model it. Accurate models require investigation into how process
parameters affect the as-built geometry and microstructure of a part and how those
features, in turn, affect the properties.[3] As accurate models of the AM processes are
built and validated, the cost to design and fabricate new components is reduced. This
paper explores how the as-built geometry of a lattice structure differs from the CAD
model, and how changing the geometry of the part affected the compression properties.
This structure was chosen due to increasing need for lighter transportation equipment -
especially in aerospace - which requires high specific stiffness and specific strength
materials. AM-built metal lattice structures are of particular interest due to the design
flexibility of AM.
Additive manufacturing is commonly used to make complex geometries -
sometimes with internal features or convoluted patterns. A gyroid is a good example of
one such shape. [4] One issue with making these parts is that many of them are not
machine-able after building. While this may be less of a problem in wire-fed systems,
many AM processes involve the use of metal powder. Powder particles can weld to the
surfaces of the part and become trapped in internal features during printing. Both of these
issues can affect the properties and performance of the part. Further, AM is a layer-by
layer process. The layers create a ‘stair-step’ pattern on the surface of a part, contributing
to the surface roughness. Another issue is anisotropy due to the layering of the melt pools
- which adds the build direction to the list of variables that alter the mechanical
properties. All of these features can be controlled by process parameters, and powder (or
wire) quality to some extent. [5], [6] Models of melt pool formation and the resulting
effect on the surface quality have been done with some success. [7], [8]
Beyond the surface features, the internal features such as the solidification
structure, the dislocation and porosity distribution in the part, and the residual stress must
also be identified. [9] produced an in-depth characterization of the AM stainless steel
microstructure. These internal features are clearly tied to the process parameters (such as
laser power, scan speed, scan pattern, etc.), but how they are correlated is a complex
subject. For example, [10] found that remelting the layer increased density and reduced
surface roughness, and they determined the microstructural change associated with this
process. [11] found that residual stress can be controlled partially through part
orientation, energy density, and part spacing.
There are several particular challenges involved in additively manufacturing
lattices. For example, [12] found that there is a mechanical property drop when
transitioning from a thick to thin walled structures. This drop can be altered by improving
process parameters for thin walled parts. Their work suggests that surface quality is
15
increasingly important as wall thickness decreases. A design challenge in AM lattice
structures is that they must be self-supporting. A schoen gyroid and the schwatz diamond
are both examples of self-supporting lattices used in the literature. [13] Some lattice
geometries must be built in a particular orientation, or avoid a particular orientation.
Honeycombs, for example, cannot be built with the hexagons perpendicular to the build
plate without overhang problems. During mechanical testing, the strain can be difficult to
measure due to the lack of flat surfaces to which an extensometer can be attached. One
method of accurate strain measurement differential image correlation (DIC). [14] These
and other issues can make design and development of AM lattices difficult.
This study involves investigation of a component with a complex, but relatively
well-understood geometry. Two-dimensional honeycomb samples were printed in a
selective laser melting process. 304L stainless steel powder was used as the printing
medium. The build quality and microstructure were evaluated using a variety of methods
including hardness, microstructure, and dimensional differences between the as-built and
CAD model. The samples were then tested under compression.
2. METHODOLOGY
16
2D honeycombs were built with a Renishaw AM 250 powder bed system. 304L
stainless steel powder was used - a detailed analysis of which was published by [15]. The
powder was sieved before use to remove any particles above 63 microns in diameter. The
Renishaw uses a 1070 nm NdYAG laser. The following build parameters were kept
constant throughout the experiment: laser power of 200W, spot size of 70 p,m, layer
thickness of 50 p,m, hatch spacing of 85 microns, point distance of 60 microns, scan
speed of 800 mm/s, and flowing argon gas. It should be noted also that the Renishaw
AM250 uses a point-to-point laser pattern, not a continuous laser pattern.
An example of the CAD model for a honeycomb specimen can be seen in Figure
1. The figure shows labels for the cell size (C), wall thickness (t), height (Z), length (L),
and width (W) of the specimen. Four different samples were printed, with 15 specimens
per sample. The cell size remained constant at 3.97 mm, and the wall thickness was
varied from 0.1 to 0.5 mm. All the relevant CAD model dimensions are shown in Table
1. The specimens with 0.1 mm wall thickness were visibly porous, and were not used for
further experimentation.
17
WLT
C
Z
Figure 1: The CAD model for the H2 honeycomb specimens, where the relevantdimensions are labeled.
Calipers were used to measure the dimensions listed in Table 1. Each
measurement was taken three times in random positions across each of the 15 specimens
per sample set. Comparison between the as-built geometry and CAD model was done
using t-tests at 0.05 significance level. Two measurements of density were done. First,
Archimedes’ method was used, with isopropyl alcohol as the immersion liquid. Second,
the density of the honeycomb was found via a geometric measurement. This calculation
assumes the volume of the honeycomb as a cube.
18
Table 1: The nominal and measured dimensions for the honeycomb samples. Nominaldimensions are in bold.
H1 H2 H3 H4
Nominal Wall Thickness (mm) 0.2 0.3 0.4 0.5
Measured Wall Thickness 0 .3 ± 0 .0 1 0 .4 4 ± 0 .0 2 0 .5 2 ± 0 .0 2 0 .5 9 ± 0 .0 2
Nominal Cell Size (mm) 3.97 3.97 3.97 3.97
Measured Cell Size 3 .8 5 ± 0 .0 2 3 .8 2 ± 0 .0 3 3 .8 4 ± 0 .0 2 3 .8 3 ± 0 .0 2
Nominal Height (Z) 25.4 25.4 25.4 25.4
Measured Z (mm) 2 6 .7 3 ± 0 .0 9 2 6 .3 1 ± 0 .7 2 6 .8 6 ± 0 .0 7 2 6 .1 8 ± 0 .6 9
Nominal Width (W) 25.15 25.73 24.38 20.83
Measured W (mm) 2 5 .2 9 ± 0 .0 4 2 5 .8 4 ± 0 .1 5 2 4 .2 7 ± 0 .0 4 2 0 .9 9 ± 0 .0 4
Nominal Length (L) 26.67 27.28 26.67 22.86
Measured L (mm) 2 6 .6 5 ± 0 .0 5 2 7 .3 7 ± 0 .0 6 2 6 .7 3 ± 0 .0 5 2 2 .9 4 ± 0 .1 4
Microhardness was measured with ASTM E384-17. A Vicker’s indenter at 50g
load was applied to polished and etched specimens. The microhardness was found across
several specimens in multiple locations. Optical microstructure images of select
specimens were taken before and after compression testing. Specimens were
electrolytically etched in a 60:40 nitric acid: water solution.
Compression in the z-direction was done according to ASTM C365. Five
specimens from each sample set were tested. Specimens were loaded between two
parallel platens at a rate of 3 mm/minute. Strain was not directly measured, due to sample
19
constraints, but was instead measured from the LVDT of the machine and normalized to
the starting height of the specimen. Because of this, accurate measurement of the
modulus could not be done using this data. The compression stress was calculated using
Equation 1, noted as the stress of the honeycomb, and also by using Equation 2, noted as
the stress of the solid material. The specific stress was calculated using Equation 3. The
ultimate stress is defined as the maximum stress the honeycomb AM produced structure
withstood prior to plastic failure. The 2% offset yield stress was calculated using
MATLAB code.
FOi j —M W*L Equation 1
_ F*psas — W*L*pp Equation 2
X — P̂H Equation 3
where oh is the honeycomb stress, F is the load applied, Wis the width of the specimen, L
is the length, is the os is the stress in the cell walls, ps is the density of the solid material,
Ph is the density of the honeycomb, and X is the specific stress.
As stated, Equation 2 determines the stress in the solid material, or the stress in
the cell walls. The derivation of Equation 2 is as follows:
pos — — Equation 4
where As is the cross-sectional area of the solid portion of the honeycomb. As can be
derived from the density of the material as shown in Equation 5.
mPs H*A Equation 5
20
where H is the height of the honeycomb and m is the mass. The density of the honeycomb
(Equation 6) was used to bring Equation 2 into a similar form as Equation 1. Equation 7
shows the final form of As.
pH = ——— Equation 6h*w*l n
.Ac = ———— Equation 7 ̂ W*L*ps n
3. RESULTS AND DISCUSSION
Table 1 shows the CAD model dimensions of the honeycombs (notated as
nominal) and the average and standard deviation of the as-built dimensions. All measured
values were found to be significantly different from the CAD model dimensions (t-test,
a=0.05. The overall width and length of the honeycombs were less than 1% different than
the nominal dimensions. The cell size, height, and wall thickness measured dimensions
were all larger than the nominal dimensions by more than 1%. The most likely reason for
the height mismatch is a post-processing error. An EDM was used to remove the
specimens from the build plate, and care was not taken to perfectly match the CAD
model’s height. For the cell size and wall thickness dimensions, the presence of unmelted
particles sintered to the surface of the walls is likely the majority of the discrepancy.
These particles can be seen with the naked eye. Figure 2 is an example of the surface of a
honeycomb wall, as shown by an optical microscope with the ability to capture depth.
Because the maximum particle size of the powder is 63 microns in diameter, the cell size
and wall thickness measurements can be adjusted to account for the particles. These
adjustments, shown in Table 2, significantly reduced the differences between the nominal
21
and as-built dimensions. This is especially true of the cell size, where the difference was
reduced below 1%.
Figure 2: The surface of a honeycomb wall, shown to highlight the presence of particlessintered to the walls.
Table 2:The measured wall thickness and cell size values, adjusted to correct for particlessintered to the walls
H1 H2 H3 H4
Nom inal W all Th ickness (mm) 0.2 0.3 0.4 0.5M easured W all Thickness 0.3±0.01 0.44±0.02 0.52±0.02 0.59±0.02% Difference from Nom inal 51.0% 46.7% 29.8% 18.2%
Particle Size Adjustm ent 0.18±0.01 0.31±0.02 0.39±0.02 0.46±0.02
% Difference from Nom inal -12.0% 4.7% -1.7% -7.0%Nom inal Cell Size (mm) 3.97 3.97 3.97 3.97M easured Cell Size 3.85±0.02 3.82±0.03 3.84±0.02 3.83±0.02% Difference from Nom inal 3.1% 3.8% 3.3% 3.4%
Particle Size Adjustm ent 3.97±0.02 3.95±0.03 3.96±0.02 3.96±0.02
% Difference from Nom inal 0.1% -0.6% -0.2% -0.2%
22
There are potentially two other mechanisms that could contribute to the mismatch
between the as-built and nominal wall thickness. First, the natural surface roughness of
the wall, not including particles sintered to the wall. An accurate accounting of this
surface roughness could not be done due to the many particles clinging to the walls.
Second, the melt pool size and hatch spacing of the process are set prior to printing, and
were not adjusted to account for the thin wall specimens. It is likely that these melt pool
widths did not perfectly match the sizes intended by the CAD model. To check this, the
width of the honeycomb wall was calculated from the maximum width of the melt pool,
the number of laser tracks in each wall thickness, and the hatch spacing of the laser
tracks. The number of tracks is known from the microstructure, and the hatch spacing is a
set build parameter. The maximum width of the melt pool, however, is not easily
determined. The ‘top’ surface of the melt pool is the area of greatest width, and this area
is remelted in subsequent layers. Thus, direct measurement of the maximum width of the
melt pool is not possible using microstructure images. An approximation of the
maximum melt pool width can be made by measuring the width of the visible melt pool
structure. This was done with imageJ software and resulted in an approximate melt pool
width of 0.126±0.020 mm. The width of the wall can be calculated as shown in Figure 3
The melt pools are offset by the hatch spacing between laser tracks from the midpoint of
the melt pool. Table 3 shows the results of the calculation. The calculated widths are
within a standard deviation of the nominal, with the exception of the 5 mm specimen.
This data does indicate that for thin walled specimens, the wall thickness of the as-built
structure can differ from the CAD model simply due to the melt pool track width. Out of
hatch spacing, melt pool size, and number of laser tracks, the hatch spacing could be
altered to accommodate the thin walls the most easily. Any change to the build
parameters could significantly alter the density of the part, however, so care should be
taken.
23
Half melt pool width
Figure 3: The width of a wall can be indirectly calculated from the number of melt pool tracks across it, the hatch spacing, and the average width of a melt pool as shown in the
figure.
Table 3: The calculated width of the honeycomb walls using the number of melt pool tracks, width of an average melt pool, and hatch spacing.
H1 H2 H3 H4
Number of tracks 3 4 5 7
Nominal width (mm) 0.2 0.3 0.4 0.5
Calculated width (mm) 0.21 ± 0.2 0.30 ± 0.2 0.38 ± 0.2 0.55 ± 0.2
The density of the cell wall material - measured by Archimedes’ method - was
between 95% and 99% dense across all specimens. There was some evidence of lack of
fusion porosity in the walls. The density of the honeycomb, including the void space
within the cells, is listed in Table III. As the wall thickness increases, the density also
increases.
Images of the microstructure of the honeycombs can be seen in Figure 4. The
white arrow indicates the build direction and the black arrows indicate melt pool
boundaries. The dark phase visible in the etched micrographs is a cellular microstructure,
a result of fast solidification. It should be noted that two laser parameters were used to
build these structures. The first is optimized to reduce surface roughness, known as the
‘border scan’. This is done only on the outer surface of the component. The second set of
parameters is optimized to improve density and is done inside the component. Due to the
feature size of these parts, and the width of the melt pools, the border scan dominates the
majority of these structures. This could be detrimental to the properties of the honeycomb
samples. Columnar grains that crossed melt pools parallel to the build direction were
also observed as shown in Figure 5 shows some of these features.
The microhardness was evaluated for samples H1 and H2 in four different
locations- the border scan along the L-Z plane, the midpoint between two walls in the W-
L plane, the midpoint in the L-Z plane, and across the L-Z plane. A schematic of each
location is shown in Figure 6. The location of measurement did not appear to have a
significant effect on the test results (F-test, 95% confidence interval). The most notable
aspect of this test is that any difference between the border scan and the internal portion
of the cell wall could not be resolved using this test. This could indicate that the two
24
25
different scans did not notably affect the properties of the honeycombs. The 304L
honeycomb had a Vicker’s hardness (HV) of 209±14 at 50g load. Using the same set-up,
wrought 304 was also measured, with a HV of 237±9 at 50 g load.
BA
DC
20 pm
Figure 4: Etched micrographs of honeycomb specimens where build direction is indicated with white arrow and melt pool edges are pointed out with black arrows. A and B) 0.3
mm nominal thickness, image of the triple point where three struts meet. C and D) Crossection of a 0.4 mm nominal thickness honeycomb.
An example of the honeycomb stress-strain curves is shown in Figure 7. There are
six regions of interest indicated on the graph. Region 1 is the toe region, which is an
artifact of the compression testing equipment. Region 2 is the elastic region, from which
the modulus values were found. Region 3 is the yield stress of the honeycomb. Region 4
26
Figure 5: Columnar grains that cross the melt pools were observed, and two of these features are marked by black lines. The build direction is indicated by a white arrow.
Figure 6: A schematic of the locations from which hardness measurements were taken is shown. A) is the top of three honeycomb cells. B) is the cell wall of one honeycomb cell.
Straight black lines indicate where the hardness was measured.
27
is a plastic deformation region. Region 5 is the ultimate strength, and region 6 is the
typical cycling behavior seen in most honeycomb compression testing. Further work in
this area will involve finite element modeling to predict the properties of other wall
thicknesses and geometries.
Figure 7: An example of the stress- displacement curves generated by the compression testing. Six regions of interest are labelled.
Figure 8 shows an etched optical micrograph of a specimen after compression
testing. The failure of the honeycomb began at the mid-point of the cell walls by
plastically buckling. There was no evidence of cracking in the failure of these. An
interesting feature of this microstructure occurs near the compression side of the buckled
wall. The solidified melt pools along this side appeared to have been compressed and
elongated. This could indicate that there is a microstructure impact on these structures,
which typically have geometry-dominated properties. Work done by Wang et. al. [9] has
28
shown there is evidence that a large number of dislocations exist within the melt pools of
a similar alloy after laser additive manufacturing, but little to no dislocations in the melt
pool boundaries. This microstructure behavior, along with the unique crystallographic
grain orientation and crystallographic texture present in SLM material may have
influenced the failure of the honeycombs. [16], [17] A thorough review of the
compression properties of 304L after SLM processing is future work for this project.
Figure 8: A buckled wall after compression testing. The build direction is indicated by anarrow.
Figure 9 includes the stress data for the honeycomb samples, using Equation 1 to
calculate the stress in the sample. R-squared values on the figure indicate the fit to either
a 2nd order polynomial line or a linear line. The yield stress in the honeycomb appears to
increase linearly with wall thickness. The specific yield strength appears to stay relatively
level in the range tested. This indicates that the yield is noticeably influenced by the
29
geometry of the specimen, but that the density increase with increasing wall thickness is
approximately equal to the yield increase. The ultimate strength of the honeycomb
appears to increase quadratically with the wall thickness. The specific ultimate strength
appears to do the same. The specific strength appears to reach a minimum between H1
and H2, but further testing is needed to confirm this. It is unknown, but expected, that a
maximum specific strength will occur outside the tested range. This would indicate an at
least 3rd order dependence on the wall thickness. More testing outside the range tested
here is needed to confirm this suspicion. Regardless of the trends, the specific properties
remained higher than fully dense 304L stainless steel, which has a specific yield of 26
MPa/g/cm3 and a specific ultimate strength of 70 MPa/ g/cm3 (both values in tension)
according to the ASM aerospace specification. [18]
Figure 10 includes the stress data for the honeycomb samples, using Equation 2
to calculate the stress. This gives the stress experienced by the cell walls during testing.
The yield stress within the cell walls remains relatively stable across the specimens. This
is indicating the yield stress of the AM 304L, which should remain constant. The tensile
yield strength of 304L with 0% cold working is 210 MPa. [18] The yield strength of the
AM material is higher than the conventional material at around 320 MPa, possibly due to
the large number of dislocations that typically form in AM stainless steels. [9] The
ultimate strength in the cell walls increases quadratically as the wall thickness increases,
however there appears to be a minimum between H1 and H2, similar to the specific
strength of the honeycomb. This data indicates that there is a relationship between the
geometry of the specimen and the ultimate strength. The failure and failure prediction of
these honeycombs was investigated in depth by Anandan et. al. [19].
30
4. CONCLUSIONS
The thin-walled honeycombs manufactured in this study showed evidence of
deviation from the intended structure. This deviation exists both in the overall dimensions
of several samples and in the wall thickness and cell size of all the samples. Much of this
deviation can be attributed to particles sintered to the outer surfaces of the honeycombs.
The microstructure of the honeycombs was a typical AM microstructure. The
compression strength of the honeycomb and the solid material both increased from 92 to
267 MPa and 687 to 1052 MPa, respectively, with increasing wall thickness. The specific
compression strength of the honeycombs appears to reach a minimum of 82 ± 5
MPa/g/cm3 at 0.44 mm wall thickness with an increase to 134 ± 3 MPa/g/cm3 at 0.6 mm
wall thickness. The yield stress of the cell wall material was stable at 319 ± 16 MPa and
the yield stress of the honeycomb structure increased from 45 MPa to 83 MPa over the
wall thicknesses tested. The specific yield stress of the honeycomb structure was stable at
42 ± 3 MPa/g/cm3. Microstructural evaluation of the compression tested samples
identified the failure as plastic buckling of the cell walls.
31
140
120
100
80
60
40
20
0
160
Figure 9: Yield and ultimate stress and specific yield stress of the honeycomb are shown (Equation 1 was used to calculate the stress).
c3
xfixfiCD
1200
1000
800
600
400
200
0
Yield Stress
• Ultimate Stress
: * ▼
0.9476 * ^
R2 = 0.3541 ■*----------------- ♦
T-----1-----1-----1-----1-----1-----1-----1-----1
0.3 0.35 0.4 0.45 0.5 0.55 0.6Measured Wall Thickness (mm)
Figure 10: The yield stress and ultimate stress of the solid material are shown (Equation 2was used to calculate stress.)
Spec
ific
Stre
ss (M
Pa/d
ensit
y)
32
REFERENCES
[1] P. Hanzl, M. Zetek, T. Baksa, and T. Kroupa, “The Influence of Processing Parameters on the Mechanical Properties of SLM Parts,” ProcediaEng., vol. 100, pp. 1405-1413,Jan. 2015.
[2] Y. Huang, M. C. Leu, J. Mazumder, and A. Donmez, “Additive Manufacturing: Current State, Future Potential, Gaps and Needs, and Recommendations,” J.Manuf. Sci. Eng., vol. 137, no. 1, p. 014001, Feb. 2015.
[3] M. M. Francois et al., “Modeling of additive manufacturing processes for metals: Challenges and opportunities,” Curr. Opin. Solid State Mater. Sci., vol. 21, no. 4, pp. 198-206, Aug. 2017.
[4] C. Yan, L. Hao, A. Hussein, and D. Raymont, “Evaluations of cellular lattice structures manufactured using selective laser melting,” Int. J. Mach. Tools Manuf., vol. 62, pp. 32-38, 2012.
[5] S. Karnati, I. Axelsen, F. F. Liou, and J. W. Newkirk, “Investigation of tensile propertie sof bulk and LSM fabricated 304L stainless steel using various gage length specimens,” in Solid Freeform Fabrication Proceedings, 2016, p. 13.
[6] M. Jamshidinia and R. Kovacevic, “The influence of heat accumulation on the surface roughness in powder-bed additive manufacturing,” Surf. Topogr. Metrol. Prop., vol. 3, no. 1, p. 014003, Feb. 2015.
[7] G. Strano, L. Hao, R. M. Everson, and K. E. Evans, “Surface roughness analysis, modelling and prediction in selective laser melting,” J. Mater. Process. Technol., vol. 213, no. 4, pp. 589-597, Apr. 2013.
[8] S. A. Khairallah, A. T. Anderson, A. Rubenchik, and W. E. King, “Laser powder- bed fusion additive manufacturing: Physics of complex melt flow and formation mechanisms of pores, spatter, and denudation zones,” Acta Mater., vol. 108, pp. 36-45, Apr. 2016.
[9] Y. M. Wang et al., “Additively manufactured hierarchical stainless steels with high strength and ductility,” Nat. Mater., vol. 17, no. 1, pp. 63-71, Oct. 2017.
[10] E. Yasa and J.-P. Kruth, “Microstructural investigation of Selective Laser Melting 316L stainless steel parts exposed to laser re-melting,” Procedia Eng., vol. 19, pp. 389-395, Jan. 2011.
33
[11] A. S. Wu, D. W. Brown, M. Kumar, G. F. Gallegos, and W. E. King, “An Experimental Investigation into Additive Manufacturing-Induced Residual Stresses in 316L Stainless Steel,” Metall. Mater. Trans. A, vol. 45, no. 13, pp. 6260-6270, Dec. 2014.
[12] B. Brown, W. Everhart, and J. Dinardo, “Characterization of bulk to thin wall mechanical response transition in powder bed AM,” RapidPrototyp. J., vol. 22, no. 5, pp. 801-809, Aug. 2016.
[13] A. Hussein, L. Hao, C. Yan, R. Everson, and P. Young, “Advanced lattice support structures for metal additive manufacturing,” J. Mater. Process. Tech., vol. 213, pp. 1019-1026, 2013.
[14] V. Br0tan, O. Fergani, K. S0rby, and T. Welo, “Mechanical Properties of Biocompatible 316L Steel Rhombic Dodecahedron Lattice Structures,” in Solid Freedom Fabrication, 2016, pp. 2087-2094.
[15] C. S. Kriewall, A. T. Sutton, M. C. Leu, J. W. Newkirk, and B. Brown, “Investigation of heat-affected 304L ss powder and its effect on built parts in selective laser melting,” in Solid Freeform Fabrication Symposium, 2016, pp. 625639.
[16] Y. Kok et al., “Anisotropy and heterogeneity of microstructure and mechanical properties in metal additive manufacturing: A critical review,” Mater. Des., vol. 139, pp. 565-586, Feb. 2018.
[17] T. DebRoy et al., “Additive manufacturing of metallic components - Process, structure and properties,” Prog. Mater. Sci., vol. 92, pp. 112-224, Mar. 2018.
[18] “ASM Material Data Sheet.” [Online]. Available: http://asm.matweb.com/search/SpecificMaterial.asp?bassnum=MQ304L.[Accessed: 25-May-2018].
[19] S. Anandan et al., “Failure In metal honeycombs manufactured by selective laser melting of 304 L stainless steel under compression,” VirtualPhys. Prototyp., vol. 14, no. 2, pp. 114-122, Apr. 2018.
34
II. EFFECT OF THE MELT POOL BOUNDARY NETWORK ON THE ANISOTROPIC MECHANICAL PROPERTIES OF SELECTIVE LASER
MELTED 304L
Myranda Spratt1; Joseph W. Newkirk1; Okanmisope Fashanu2; K. Chandrashekara2
1Materials Science and Engineering Department, Missouri University of Science andTechnology, Rolla, MO 65409
2Mechanical Engineering Department, Missouri University of Science and Technology,Rolla, MO 65409
ABSTRACT
Anisotropic mechanical properties are a well-known issue in selective laser
melted materials. The microstructure produced by selective laser melting is directional,
including the solidified melt pool structures and the grains. In this work, the melt pool
boundary's effects on 304L stainless steel's compressive properties were investigated. The
melt pool boundary effects were investigated experimentally and numerically. 304L
stainless steel solid cylinders were built in four directions using three hatch angle
rotations - 0°, 67°, and 105°. The twelve samples were compression tested, and the
results were analyzed. Results showed that both the hatch angle and build orientation
influenced the concentration of melt pool boundaries present in the manufactured
samples. A negative correlation of compressive strength to the melt pool boundaries'
concentration was also observed, indicating that the melt pool boundary negatively
affected the material's strength. Local anisotropic plastic deformation was also observed
in the 0° hatch angle samples in that the aspect ratio of the compressed cylinders was
greater than one. In those samples, it was observed that directions that plastically
deformed more contained a higher concentration of the melt pool boundaries than those
that deformed less.
35
1. INTRODUCTION
Additive manufacturing (AM) is a useful method of fabricating components with
complex internal features. AM processes are particularly efficient in fabricating low-
volume, high-value components. [1]-[4] One popular AM technique for making metal
components is selective laser melting (SLM) [4]. In SLM, a powder bed is used to
support the structure as it is being built. The laser melts the powder in a specified pattern
which creates the part layer-by-layer. As the laser melts the powder, a melt pool is
created. The melt pools are overlapped to create the dense part, remelting some of the
previously melted material. This precisely controlled process can be used to create
intricate structures with thin walls, such as lattice structures, as well as bulk parts [5]-[9].
The microstructure of SLM materials has a distinctive fish-scale pattern [6], [10],
[11] . In steel alloys, the fish-scale pattern appears as a thin band around the edge of the
melt pool, known as the melt pool boundary (MPB). The interior of the melt pool often
solidifies into a cellular structure [6], [12]. The different features in the solidified melt
pool may or may not include a detectable chemical difference. Rapid cooling (recorded at
104 K/s for SLM [13]) leaves a large number of dislocations and residual stresses [6],
[12] , [14]. The combination of heat from the laser and the material’s thermal properties
adjusts the size and shape of a melt pool as it forms and the phases that form as it cools
[15], [16]. This melt pool is curved - a result of the Gaussian laser beam distribution and
surface tension [12], [15], [17]. The material solidifies from the edges of the melt pool
towards the interior of the melt pool. Often, grain will grow epitaxially in the build
direction through several melt pools [12], [13], [18]. These grains often have the same
crystallographic orientation [12].
The relationship between the SLM process, the microstructure, and the properties
is complex and subject to much investigation [6], [12], [13]. Anisotropy in particular has
been recorded in strength, fatigue, and other properties of SLM parts, across many alloys
[12], [13], [19]-[24]. The build orientation and the laser scan pattern are notable process
parameters that induce anisotropic properties. This is likely due to the influence these
process parameters have on the microstructure of SLM material [12], [13], [15], [25],
[26]. The laser scan pattern is the path the laser takes in each layer. In many cases, this
path is repeated in each layer, as much as the shape of the part allows, only rotated in
each subsequent layer to reduce periodicity. The build orientation of the part is the
orientation of the part on the build plate. These parameters determine the pattern the melt
pools form in the material. In their review of anisotropy in AM materials, Kok et al [19]
concluded there three microstructural factors that influence anisotropy in SLM materials.
These are crystallographic texture, lack-of-fusion porosity, and columnar grains. With the
possible exception of crystallographic texture, these features can be controlled with build
parameters. Lack-of-fusion can be reduced by changing hatch spacing and by increasing
melt pool depth. The grain size is controllable by varying laser power and scan speed
[19], [27].
There is a second school of thought regarding anisotropy which was explored by
Shifeng et al [10]. In their work, the melt pool boundary (MPB) was considered as the
36
37
major influence on anisotropy. In this consideration, the MPBs allow for preferential slip
during deformation due to their low dislocation density [28]. This theory is supported by
Mower and Long [29], who concluded that the low fatigue strength and planar fracture of
their 316L samples was a result of the weak ‘build plane’. This theory was used to
explain the fracture behavior seen by Shifeng et al [10] as the crack path appeared to
follow the MPBs, rather than the grain boundaries.
In this work, the plastic deformation behavior of solid 304L cylinders was
analyzed. This work was initially done to support the development of a honeycomb
lattice structure failure model, published by Anandan et al [30]. During testing, the plastic
deformation of the cylinders was seen to be highly and visibly anisotropic. Further
investigation of the phenomenon has led to the investigation of the influence of build
orientation and scan pattern rotation on the elastic and plastic mechanical properties of
304L steel. Fashanu et al [31] investigated this topic from a property - parameter point of
view. This work aims to correlate property - microstructure- and parameters. A model of
the melt pool boundary network was created to facilitate this understanding, and the
results of this model were correlated to experimental test results.
2. MODELING
The melt pool boundary network (MPBN) was modeled using Blender software.
This software was chosen due to the relatively simple to use interface, ability to create
and modify complex meshes, and movie-making abilities. In this software, the scale can
be set by the user. The camera can also be set to an orthographic or a perspective mode
depending on the needs of the user. In this case, orthographic mode was selected and a
metric scale was used. Orthographic mode ensures that the mesh’s real size is imaged by
the software, rather than the perspective size. The metric scale ensures that any
measurements are in a convertible scale.
The MPBN depends on the laser scan pattern that is being used. The Renishaw
AM 250 machine used in the experimental part of this study uses a pulsed rather than the
more common continuous laser. The laser turns on and ramps to the correct wattage,
holds in place for the specified exposure time, then turns off. It then moves to the next
position at a set speed and turns on again. This means that the melt pool that forms under
the laser is not necessarily continuous. The melt pool shape will change depending on the
scan pattern used by the laser. In this case, the scan pattern was stripes. Figure 1 shows an
example of the stripes scan pattern. This figure also shows the spacing between laser
points and the hatch spacing between rows in a single layer. The hatch spacing in this
experiment was 85 microns, the point distance was 60 microns, and the layer thickness
was 50 microns.
The orientation of the developed models were based on the Renishaw builds. The
scan pattern of the laser always started the build by moving in the x or [100] direction,
and the [001] direction is the build direction or vertical axis. Each model was built with
this orientation. When planes or directions are mentioned, these are considered global
axes tied to the build plate. So, the build direction is always the [001] axis, and the layers
are in-plane with the (001) plane. The shape of the melt pool used in the models after
solidification was based on etched images of a 304L stainless steel (ss) specimen shown
in Figure 2. The idea to model the melt pool using this approach came from Li et al [16].
38
They modeled the thermal history and size change of melt pools using different input
parameters on the same machine and material as used here. The model boundary was
given a thickness to simulate the actual melt pool boundary. The melt pool boundary is
the material that solidifies first at the edges of the melt pool. In this work, the thickness
was set a single width of 0.1 microns.
39
Figure 1: The stripes scan pattern showing the input parameters used by the RenishawAM250 as modeled in this work.
The representative melt pool used in this model does not consider variations in the
melt pool size and shape. Melt pool variations can occur if the local energy density
fluctuates, border scans are used, the laser direction changes, or some other variation in
the machine occurs. Melt pool variations can include inclusions formed by residual
oxygen or carbon and lack of fusion porosity. Figure 3 shows the microstructure of a
typical 304L part as manufactured by the Renishaw AM250. Key features of the
microstructure are noted in the figure.
40
\ /
Porosity
| PMelt pool K _ „ ,, , \ Cellular boundary ^ Structure
1 0 0 p m 2 0
Figure 2: Optical images of SLM 304L in two magnifications etched with a 60:40 water: nitric acid electrolytic etchant. Relevant features have been pointed out on the images.
The first step of making the model was creating the model of a single melt pool.
The second step is to create the first layer of the model. This is done by arranging single
melt pools in a single plane to mimic the laser scan pattern. Each subsequent melt pool
annihilates any overlapping area with previous melt pools. This simulates the re-melting
that occurs. Figure 4 shows a single layer and the scan pattern that was mimicked to
create that layer. The alternating colors indicate the alternating direction the laser was
moving when that row was created.
41
(1 0 0 ) 50 [jm (0 1 0 ) Mt-.1 - 50 jjm {
Figure 3: The model of a single melt pool compared to the actual solidified melt pool inthe 304L ss used in this work.
The third step was to create multiple layers. Typically, the scan pattern is rotated
between layers. The rotation increases the ‘randomness’ of the structure. In this work,
three models were developed with different rotation angles. These were 0° (or non
rotating), 67°, and 105°. Figure 5 shows each scan pattern and the corresponding model.
Six layers were created for each model.
In steps 2 and 3, the melt pool remelting was simulated by annihilating
overlapping material. This was done by using Boolean modifiers in the Blender software.
The Boolean modifier modifies a mesh by another mesh. Figure 6 shows a mesh before
and after a Difference Boolean modifier is applied.
Once each model of the MPBN was created, it could then be analyzed. In this
work, the concentration of the MPBN was analyzed through the structure. This was done
by slicing each model in the (001) plane in increments along the [001] direction. Each
model was sliced 60 times at regular intervals. Before slicing, the model itself was cut
down into a representative volume element (RVE). This RVE contained only overlapping
42
Figure 4: The model of a single layer with two colors to indicate the change in laser direction as the melt pools were created.
67'0' 150°
§//=■
Figure 5: The scan pattern rotation is shown beside the model for that scan patternrotation.
43
Figure 6: The single melt pool mesh for the 0° hatch angle model, before and afterBooleans are applied.
melt pools and eliminated any melt pools that were along the edges of the model. Figure
7 shows each of the RVE models.
Figure 7: The three RVE models after the boundaries were removed.
The concentration of MPBN in each slice when the model is oriented with the
build direction parallel to the [001] axis only shows the concentration of MPBN along
that plane. Other orientations have other MPBN concentrations. To track this, the model
was tilted around the [100] and [010] directions 5° increments from 0° to 90°. Figure 8
shows examples of the tilted model. Each tilted model was sliced and imaged in the same
manner as the 0° model. The model, when rotated around the [100] direction by 90°, was
sliced along the (010) plane. When it was rotated around the [010] direction by 90°, it
44
was sliced along the (100) plane. Each image was fed into FIJI software. This software is
capable of measuring the color data in each pixel of an image. This was done to each of
the image slices. The background color of the images was separated from the MPBN
color of the images, and the total number of MPBN pixels was found for each image.
This number was then divided by the total number of pixels in the model at that slice to
get the concentration as a percentage.
Figure 8: The 67° hatch angle model rotated around the [100] direction by 0°, 45°, and 90° where the lines are example slices through the models.
3. PROCEDURE
The procedure used to make the samples was detailed in Fashanu et al’s [31]
work. Briefly, 304L cylinders were printed with a Renishaw AM250. The argon-gas
atomized powder was sourced from LPW. The scan pattern used was a border scan
followed by the stripes scan pattern. Octagonal prisms with a height of 27.8 mm and a
side length of 3.84 mm were printed in 4 build orientations and 3 hatch angles. Three
replications were done for each sample. The test matrix is shown in Table 1. Figure 9
shows the four build orientations as they appear on the build plate. The hatch angles were
0°, 67°, and 105° rotation between layers. The prisms were removed from the build plate
with wire electron discharge machining (EDM). They were then machined to a diameter
of 6.35 mm and a height of 6.35 mm.
The samples were characterized before testing. Hardness measurents were carried
out on one specimen from each sample set. All Vicker’s hardness measurements were
done at 9.81 N for 10 seconds. Vicker’s hardness were done in three positions near the
center of the machined top surface of the cylinders for each sample. Archimedes’ density
was performed on the samples using water as the liquid medium, and the vacuum method
to remove gas bubbles.
45
Table 1: The test matrix for each sample, where the orientation listed shows the plane normal to the compression direction. The notation for the orientation is hatch angle -
rotation direction - rotation degree.
Sample (100) (010) (001) (011)
0° 0°-[010]-90°
0°-[100]-90°
0°-[100]-0°
0°-[010]-45°
67° 67°-[010]-90°
67°-[100]-90°
67°-[100]-0°
67°-[010]-45°
O o 105°-[010]-90°
105°-[100]-90°
105°-[100]-0°
105°-[010]-45°
46
[001]
Figure 9: The four build orientations of the cylinders are shown.
Compression testing was done as described in ASTM E3 using self-aligning
platens. The samples were loaded at a rate of 5x10"3 mm per min. The cylinders were
compressed to well past the yield point. A Visual Basic for Applications (VBA) code was
used to analyze the data after testing. The ratio of the specimen height to diameter was
not suitable for modulus determination according to the standard. Camera images were
taken of each specimen after compression testing. A macro lens was used on the camera
to take close-up images of the top and sides of each specimen. The top surface images
were converted to binary where the top surface was black and the background, white. FIJI
software was used to calculate the aspect ratio of the specimens using the Feret
diameters. The Feret diameters are the longest and smallest diameters of the specimen.
The longest/smallest diameter gives the aspect ratio from the plastic deformation.
47
After testing, the samples were prepared for metallography. They were mounted
in bakelite and polished to 0.05 microns in colloidal silica. For optical microscopy, the
samples were electrolytically etched in a 60:40 nitric acid: water solution at 6 volts for
approximately 5 seconds. Vicker’s hardness measurements were done on the top surface
of a tested specimen after the images were taken. This was done in a grid to map the
hardness across the deformed specimen. This tracked the local magnitude of the plastic
deformation.
4. RESULTS AND DISCUSSION
4.1. EXPERIMENTAL WORK
The as-built sample density was not significantly different between samples. The
samples were 98% ± 1.0% dense on average. The compression testing analysis found
0.2% offset yield strength and aspect ratio. The results for each sample are shown in
Table 2. The maximum yield strength and the lowest aspect ratio for each hatch angle are
in bold font. For each hatch angle, the cylinder built either in the [100] or [010] direction
had the highest yield strength. The aspect ratio of the 67° and 105° samples were low -
with one exception equal to or below 1.10 on average. The 0° samples, except for the
[011] sample, were noticeably anisotropic. The tensile yield strength of traditionally
manufactured, annealed 304L is 210 MPa, according to the ASM specification [32],
which significantly less than the yield strength values found in this work. Additively
manufactured materials are known to have increased yield strength, due in part to residual
stresses in the material and other microstructural factors.
48
Table 2: The yield strength, aspect ratio horizonal plane of each cylinder after testing foreach specimen are shown
Hatch Angle - Rotation Axis- Rotation Angle (Plane
Normal to Compression Direction) Aspect Ratio
0.2% Offset Yield Strength (MPa)
0°-[0101-90° (100) 1.22 ± 0.008 437 ± 4.700°-[1001-90° (010) 1.53 ± 0.052 489 ± 9.940°-[1001-0° (001) 1.65 ± 0.022 439 ± 7.94
0°-[0101-45° (011) 1.07 ± 0.026 413 ± 9.2667°-[0101-90° (100) 1.18 ± 0.025 492 ± 2.0067°-[1001-90° (010) 1.10 ± 0.042 489 ± 5.0867°-[1001-0° (001) 1.04 ± 0.015 473 ± 11.667°-[0101-45° (011) 1.05 ± 0.012 491 ± 9.89105°-[0101-90° (100) 1.04 ± 0.019 544 ± 10.7105°-[1001-90° (010) 1.03 ± 0.007 516 ± 14.24105°-[1001-0° (001) 1.06 ± 0.023 498 ± 3.72
105°-[0101-45° (011) 1.10 ± 0.022 485 ± 2.08
An aspect ratio greater than 1 indicates that anisotropic plastic deformation
occurred in the samples. This can occur if the test is performed improperly, or if non
articulating platens are used. In this case, the articulating platens were used, and the
samples were sufficiently lubricated to reduce horizontal friction. Further, the degree of
anisotropy varied, even among specimens with similar strengths. The anisotropy was
consistent across specimens as well with an average standard deviation of 0.02. The
aspect ratio of the compression surface was not the only anisotropic plastic deformation
observed. Figure 10 shows the view of a compressed sample from the top. The shearing
of the cylinders is made obvious in the figures by the black arrows. The shearing was
most prominent in the 0° hatch angle specimens, and it seemed to be perpendicular to the
longest width of the compressed specimen. All specimens experienced the shearing
behavior, though it was less apparent in the samples with low aspect ratio. The 0°-[010]-
45° sample was notably different than the other 0° hatch angle samples in that the (001)
plane did not deform into a cylinder, but instead became rounded boxes, as shown in
Figure 11. Despite the low aspect ratio, the anisotropic plastic deformation was still
prevalent in this sample.
The Vicker’s hardness results for the undeformed samples were not statistically
different from each other. The average hardness was 216 ± 6 HV. According to the ASM
specification [24] for austenitic 304L steel, the Vicker’s hardness value should be 159
HV. The elevated hardness in the structure before testing supports the evidence that the
additively manufactured properties of the 304L material are significantly different from
traditionally manufactured material. The hardness was also tracked on the compression
surface of one of the specimens. The map of the hardness is shown in Figure 12. The
hardness values indicate that the sheared area of the specimen (noted on the figure) was
significantly harder than the compression surface.
4.2. MODEL WORK
It is clear from the literature that 3D printed structures depend on the way they are
printed. The mechanical property results of this study support that. The microstructure
and defect structure depend on build orientation and scan pattern. If the theory presented
by this work is true - that the melt pool boundary network microstructural feature
impacts the mechanical properties of the material - then the model of the MPBN may
49
50
Figure 10: Anisotropy in the 0°-[100]-0° specimens after compression testing, where the black arrows indicate the direction of shear plastic deformation and the white arrows
indicate the maximum width.
help predict strong and weak combinations in the parts. There are, of course, other
microstructural features that impact the mechanical properties of the part. These features
are slowly becoming well-documented in the literature. The features include grain size,
crystallographic orientation, and porosity distribution. The MPBN is, according to the
theory presented here, another microstructural feature that is a significant feature that can
directionally alter properties. Depending on the material, the MPBN may be more or less
influential.
51
Figure 11: The 0°-[010]-45° specimens sheared as the other 0° hatch angle specimens did, but the (001) anisotropy was not elliptical.
When the percent MPBN per slice was plotted over the length of the model, there
were two types of lines produced. Figure 13 shows an example of each one. Type One
was a pseudo-sinusoidal increase and decrease. Type Two was a relatively flat line. The
flat line indicates that the MPBN concentration was relatively constant in that orientation.
So, theoretically, the impact of the MPBN in that direction on the mechanical properties
could be consistent. The sinusoidal MPBN concentration was fluctuating between layers
semi-consistantly. The impact of the MPBN network in that direction will be,
presumably, similar to a composite with a strong and weak phase. Depending on the
microstructure, the MPBN in this direction could act similarly to a composite that has
52
Figure 12 A hardness map of the 0°-[100]-90° specimen after compression testing. Theorientation of the specimen is marked.
layers of material like corrugated cardboard. Since the material only consists of 304L
steel, the ‘composite’ is made of two ductile steels of (presumably) very similar
properties, so the composite-like structure’s impact on the properties may be minimal. In
all three hatch angles, the orientation that was guaranteed to produce a Type One line was
the orientation parallel with the build direction.
One way to analyze the data is to use a rule of mixtures approach to the strength.
That is, the higher the concentration of the ‘weak phase’ - the melt pool boundaries - the
weaker the material will be. So, since the material was compressed uniaxially, the plane
normal to the compression direction is the plane of interest. In Figure 14, the average
53
50.0%
45.0%
40.0%
0 5 .0 %PQ§ 3 0 .0%
25.0%
2 0 .0%
15.0%20% 30% 40% 50% 60% 70% 80%
Percent Position in ModelFigure 13: There were two types of lines produced when the MPBN concentration was plotted over the length of the model - flat and pseudo-sinusoidal. Examples of each are
shown.
yield strength for each sample was plotted against the concentration of the melt pool
boundaries through the plane normal to the compression direction. The negative linear
correlation is weak (R2-value of only 0.39), indicating other factors at play. However, the
negative correlation does lend some veracity to the theory presented here.
After compression, the MPBN distorted at the edges of the part. This is easiest to
see in the 0° hatch angle samples as the melt pool boundary network is stacked in parallel
lines due to the non-rotating scan pattern. The cylinders barreled out non-uniformly as
previously described. The previously smooth edges rippled as well, as shown in Figure
15. The build direction is indicated in each image of Figure 15 for context. In Figure 15.B
and C, the rippling roughly corresponds to the melt pool tracks in a single layer (B) and
54
C3
560540520500480460440420400
y = -667x + 6581 R2 = 0.39
J___ I___ I___ I___ |___ I___ I___ I___ I___ |___ I___ I___ I___ I___ |___ I___ I___ I___ I___ |___ I___ I___ I___ I___ |___ I___ I___ I___ I___ |___ I___ I___ I___ I___ |___ I___ I___ I___ L+ + + + H20.0% 22.0% 24.0% 26.0% 28.0% 30.0% 32.0% 34.0% 36.0%
[MPBN] in Plane Normal to Compression Direction
Figure 14: The yield strength vs concentration of the MPNB in the compression directionare correlated.
the layers (C), as if the features were compressed independently of the other layers. The
rippling also appeared along the melt pool track (C) possibly corresponding to individual
melt pools. This phenomenon was not limited to the edges of the cylinder, though it is
easiest to see there. The semi-independent plastic deformation of the melt pools occurred
within the bulk of the cylinder as well. This can be seen in the etched microstructure - the
MPBN is distorted similarly to the edges of the cylinder in all three figures. This
phenomenon is not limited to the 0° hatch angle samples, occurring in all the samples.
This implies that the MPBN was acting as a barrier to load transfer. Figure 16 shows a
before and after compression view of the MPBN and cellular structure. Before
compression testing, the MPBN was clearly visible in the material, and the cellular
structure appeared as imperfect hexagons or similar size. After compression, the
hexagons stretched and distorted in some places. The MPNB after compression testing
was less clearly defined.
55
A: 0°-[010]-90°
200 [jm
C: 0°-[100]-0°
Figure 15: The edges of the cylinders rippled after compression testing. The orientationof the part is marked.
Anisotropic plastic deformation of the parts, the result of which was the greater
than 1 aspect ratio of the compressed parts, was most obvious in the 0° hatch angle
56
Figure 16: A close view of the melt pool boundary and cellular structure before (A) andafter (B) compression testing.
samples. Figure 17 shows the side and top view of one compressed cylinder (0°-[100]-
0°), and Figure 12 shows the top view of a 0°-[100]-90° cylinder. Figure 18 shows the
side and top view of a 0°-[010]-90° cylinder. For the 0°-[100]-0° cylinder, the direction
of greatest width was parallel to the [100] direction and the direction of the lowest width
was the [010] direction. The 0°-[100]-0° sample had the highest aspect ratio. For the 0°-
[100]-90° specimen, the largest and smallest widths were the [100] and [001] directions,
respectively. For the 0°-[010]-90° specimen, the largest and smallest widths were the
[001] and [010] directions, respectively. These results in conjunction with the aspect ratio
data can be used to determine a ranking for these directions. The directions that caused
the most to the least plastic deformation: [100], [001], [010]. This behavior correlates to
the pseudo-sinusoidal behavior of the MPBN in those directions. Table 3 shows the
maximum and minimum peaks of the MPBN in those three directions for the 0° hatch
angle samples. The difference between the maximum and minimum MPBN in the planes
equates to the rank-ordered directions. The other hatch angles did not experience the
57
Figure 17: A 0°-[100]-0° specimen after compression testing shown from the side and the top of the cylinder. The orientation of each image is marked.
anisotropic behavior to the same degree as the 0° hatch angle samples. For the 67° hatch
angle samples, the three planes of interest had a low difference between the peaks, which
may explain why they did not experience much of an aspect ratio difference. For the 105°
hatch angle samples, very few planes exhibited the pseudo-sinusoidal behavior. The
maximum amplitude of the pseudo-sinusoidal behavior seems to be correlated with the
deformation in that direction.
58
Figure 18: A 0°-[010]-90° specimen after compression testing shown from the side and the top of the cylinder. The orientation of each image is marked.
5. CONCLUSIONS
In this work, anisotropic deformation in SLM 304L was investigated using a
model of the melt pool boundary network. The melt pool boundary network is a thin
phase of material that forms from the first material to solidify after the laser moves away
59
Table 3: The maximum, minimum, and difference between the pseudo-sinusoidal peaksfor the three directions of interest.
Hatch Angle Direction
Percent Maximum [MPBN1 in Plane
Normal to Direction
Percent Minimum
[MPBN1 in Plane Normal to Direction
Difference Between
Maximum and Minimum
0° [1001 68% 9% 59%0° [0011 44% 2 1 % 23%0° [0101 31% 18% 13%67° [1001 30% 26% 2 2%67° [0011 46% 24% 8%67° [0101 31% 24% 7%105° [1001 39% 15% 23%105° [0011 N/A N/A N/A105° [0101 N/A N/A N/A
from an area. The MPBN was modeled using a representative volume element containing
six layers of material. Three models were created with different hatch angles of 0°, 67°,
and 105°. The concentration of the MPBN in each model was tracked across multiple
planes. Experimental samples were printed using the same three hatch angles. Four build
directions were printed for each hatch angle. The samples were tested in compression.
The yield strength of these materials increased with increasing hatch angle. The build
orientations with the highest yield strengths were in the (0 0 1) plane as well, either built in
the [100] or [010] direction, depending on the hatch angle. The yield strength of the
samples was correlated to the concentration of the MPBN normal to the compression
direction of the experimental samples. A linearly negative correlation was found with an
R2 value of 0.39. The edges of the compressed cylinders were indented and outdented
after testing. Etched micrographs indicated that the rippling behavior occurred
corresponds to the melt pools. The compression tested samples were observed to have
local anisotropic plastic deformation, particularly in the 0° hatch angle samples. The
aspect ratio was used to quantify the degree of anisotropy per sample. The change in
hardness across the compression surface of one specimen confirmed the presence of
anisotropic plastic deformation. The MPBN concentration exhibited a pseudo-sinusoidal
behavior in the planes related to this anisotropic deformation. The amplitude of the
pseudo-sinusoidal curve seems to be related to the extent of the deformation experienced
by the specimens in that direction.
60
REFERENCES
[1] G. N. Levy, R. Schindel, and J. P. Kruth, “Rapid Manufacturing and Rapid Tooling With Layer Manufacturing (Lm) Technologies, State of the Art and Future Perspectives,” CIRP Ann. - Manuf. Technol., vol. 52, no. 2, pp. 589-609, 2003.
[2] R. Huang et al., “Energy and emissions saving potential of additive manufacturing: the case of lightweight aircraft components,” J. Clean. Prod., vol. 135, pp. 15591570, 2016.
[3] L. E. Murr et al., “Fabrication of Metal and Alloy Components by Additive Manufacturing: Examples of 3D Materials Science,” J. Mater. Res. Technol., vol.1, no. 1, pp. 42-54, Apr. 2012.
[4] W. E. Frazier, “Metal Additive Manufacturing: A Review,” J. Mater. Eng. Perform., vol. 23, no. 6 , pp. 1917-1928, Jun. 2014.
[5] M. Mazur, M. Leary, S. Sun, M. Vcelka, D. Shidid, and M. Brandt, “Deformation and failure behaviour of Ti-6Al-4V lattice structures manufactured by selective laser melting (SLM),” Int. J. Adv. Manuf. Technol., vol. 84, no. 5-8, pp. 13911411, 2016.
[6] P. Kohnen, C. Haase, J. Bultmann, S. Ziegler, J. H. Schleifenbaum, and W. Bleck, “Mechanical properties and deformation behavior of additively manufactured lattice structures of stainless steel,” Mater. Des., vol. 145, pp. 205-217, May 2018.
[7]
61
C. Yan, L. Hao, A. Hussein, and D. Raymont, “Evaluations of cellular lattice structures manufactured using selective laser melting,” Int. J. Mach. Tools Manuf., vol. 62, pp. 32-38, 2012.
[8] B. Brown, W. Everhart, and J. Dinardo, “Characterization of bulk to thin wall mechanical response transition in powder bed AM,” Rapid Prototyp. J., vol. 22, no. 5, pp. 801-809, Aug. 2016.
[9] E. Chlebus, B. Kuznicka, T. Kurzynowski, and B. Dybala, “Microstructure and mechanical behaviour of Ti—6Al—7Nb alloy produced by selective laser melting,” Mater. Charact., vol. 62, no. 5, pp. 488-495, May 2011.
[10] W. Shifeng, L. Shuai, W. Qingsong, C. Yan, Z. Sheng, and S. Yusheng, “Effect of molten pool boundaries on the mechanical properties of selective laser melting parts,” J. Mater. Process. Technol., vol. 214, no. 11, pp. 2660-2667, Nov. 2014.
[11] D. Wang, C. Song, Y. Yang, and Y. Bai, “Investigation of crystal growth mechanism during selective laser melting and mechanical property characterization of 316L stainless steel parts,” Mater. Des., vol. 100, pp. 291-299, 2016.
[12] M. S. Pham, B. Dovgyy, and P. A. Hooper, “Twinning induced plasticity in austenitic stainless steel 316L made by additive manufacturing,” Mater. Sci. Eng. A, vol. 704, pp. 102-111, Sep. 2017.
[13] T. Vilaro, C. Colin, and J. D. Bartout, “As-Fabricated and Heat-Treated Microstructures of the Ti-6Al-4V Alloy Processed by Selective Laser Melting,” Metall. Mater. Trans. A, vol. 42, no. 10, pp. 3190-3199, May 2011.
[14] A. S. Wu, D. W. Brown, M. Kumar, G. F. Gallegos, and W. E. King, “An Experimental Investigation into Additive Manufacturing-Induced Residual Stresses in 316L Stainless Steel,” Metall. Mater. Trans. A, vol. 45, no. 13, pp. 6260-6270, Dec. 2014.
[15] S. A. Khairallah, A. T. Anderson, A. Rubenchik, and W. E. King, “Laser powder- bed fusion additive manufacturing: Physics of complex melt flow and formation mechanisms of pores, spatter, and denudation zones,” Acta Mater., vol. 108, pp. 36-45, Apr. 2016.
[16] L. Li, C. Lough, A. Replogle, D. Bristow, R. Landers, and E. Kinzel, “Thermal Modeling of 304L Stainless Steel Selective Laser Melting,” in Solid Freedorm Fabrication, 2017, pp. 1068-1081.
62
[17] P. Guo, B. Zou, C. Huang, and H. Gao, “Study on microstructure, mechanical properties and machinability of efficiently additive manufactured AISI 316L stainless steel by high-power direct laser deposition,” J. Mater. Process. Technol., 2017.
[18] L. Thijs, F. Verhaeghe, T. Craeghs, J. Van Humbeeck, and J.-P. Kruth, “A study of the microstructural evolution during selective laser melting of Ti-6Al-4V,” Acta Mater., vol. 58, no. 9, pp. 3303-3312, May 2010.
[19] Y. Kok et al., “Anisotropy and heterogeneity of microstructure and mechanical properties in metal additive manufacturing: A critical review,” Mater. Des., vol. 139, pp. 565-586, Feb. 2018.
[20] T. DebRoy et al., “Additive manufacturing of metallic components - Process, structure and properties,” Prog. Mater. Sci., vol. 92, pp. 112-224, Mar. 2018.
[21] B. E. Carroll, T. A. Palmer, and A. M. Beese, “Anisotropic tensile behavior of Ti- 6Al-4V components fabricated with directed energy deposition additive manufacturing,” Acta Mater., vol. 87, pp. 309-320, Apr. 2015.
[22] A. H. Maamoun, M. Elbestawi, G. K. Dosbaeva, and S. C. Veldhuis, “Thermal post-processing of AlSi10Mg parts produced by Selective Laser Melting using recycled powder,” Addit. Manuf., vol. 21, pp. 234-247, May 2018.
[23] E. Liverani, S. Toschi, L. Ceschini, and A. Fortunato, “Effect of selective laser melting (SLM) process parameters on microstructure and mechanical properties of 316L austenitic stainless steel,” J. Mater. Process. Technol., vol. 249, pp. 255-263, Nov. 2017.
[24] A. Yadollahi, M. Mahmoudi, A. Elwany, H. Doude, L. Bian, and J. C. Newman, “Effects of crack orientation and heat treatment on fatigue-crack-growth behavior of AM 17-4 PH stainless steel,” Eng. Fract. Mech., vol. 226, 2020.
[25] K. Guan, Z. Wang, M. Gao, X. Li, and X. Zeng, “Effects of processing parameters on tensile properties of selective laser melted 304 stainless steel,” Mater. Des., vol. 50, pp. 581-586, 2013.
[26] B. Song et al., “Differences in microstructure and properties between selective laser melting and traditional manufacturing for fabrication of metal parts: A review,” Frontiers of Mechanical Engineering, vol. 10, no. 2. 2015.
[27] Z. Wang, T. A. Palmer, and A. M. Beese, “Effect of processing parameters on microstructure and tensile properties of austenitic stainless steel 304L made by directed energy deposition additive manufacturing,” Acta Mater., vol. 110, pp. 226-235, May 2016.
63
[28] Y. M. Wang et al., “Additively manufactured hierarchical stainless steels with high strength and ductility,” Nat. Mater., vol. 17, no. 1, pp. 63-71, Oct. 2017.
[29] T. M. Mower and M. J. Long, “Mechanical behavior of additive manufactured, powder-bed laser-fused materials,” Mater. Sci. Eng. A, vol. 651, pp. 198-213, Jan. 2016.
[30] S. Anandan et al., “Failure In metal honeycombs manufactured by selective laser melting of 304 L stainless steel under compression,” Virtual Phys. Prototyp., vol. 14, no. 2, pp. 114-122, Apr. 2018.
[31] O. Fashanu et al., “Effect of SLM Build Parameters on the Compressive Properties of 304L Stainless Steel,” J. Manuf. Mater. Process., vol. 3, no. 2, p. 43, Jun. 2019.
[32] ASM Aerospace Specification Metals Inc, “AISI Type 304L Stainless Steel,”ASM Material Data Sheet. [Online]. Available:http://asm.matweb.com/search/SpecificMaterial.asp?bassnum=MQ304L. [Accessed: 03-Jan-2019].
64
III. OPTIMIZATION AND CHARACTERIZATION OF NOVEL INJECTION MOLDING PROCESS FOR METAL MATRIX SYNTACTIC FOAMS
Myranda Spratt; Joseph W. Newkirk
Materials Science and Engineering Department, Missouri University of Science andTechnology, Rolla, MO 65409
ABSTRACT
Metal matrix syntactic foams are particulate composites comprised of hollow or
porous particles embedded in a metal matrix. These composites are difficult to
manufacture due primarily to the lightweight, relatively fragile filler material. A method
of low-pressure injection molding metal matrix syntactic foams was developed in this
work. First, an aqueous binder was optimized for low-pressure injection molding. A
three-component simplex design of experiment was used to model the density response to
changing the ratio of water to agar to glycerin. The model predicted the maximum
density was at a binder composition (in volume percent) of 7% agar, 4% glycerin, and
89% water. Second, this binder was used to manufacture copper matrix syntactic foams
with 0, 5, 10, and 15 volume % porous silica as the filler material. The solids loading for
these compositions decreased with increasing filler material from 55 to 44 volume %,
likely due to binder filling the pores in the porous silica particles. Finally, the sample
quality after injection molding was characterized. Only 0.11 ± 0.06 volume % carbon
remained in the samples. Silica particles were well-dispersed in the samples after
sintering, and they did not appear to be fractured. The specific strength of the copper
matrix material increased with increasing porous silica additions.
65
Keywords: Metal-matrix Composites; Metal Foams; Injection molding; Molding
compounds
1. INTRODUCTION
Lightweight structural composites are in high demand in certain industries,
especially in aerospace. These materials are generally two-phase composites of some
kind. For example, an aluminum-silicon carbide composite increases the strength of the
aluminum and slightly increases the density for an overall increase in the specific
strength and stiffness. Other types of composites combine a matrix and air - foams
sacrifice strength for drastically decreasing the density of the material. Syntactic foams
combine the best of both types of composites. Metal matrix syntactic foams (MMSF) are
3-phase particulate composites where hollow particles are suspended in a metal matrix.
[1] The pores in MMSFs are encased in the hollow spheres, which allows for both tight
control over the density and pore size of the material and restricts the porosity from
interacting with the matrix material directly. Rohatgi et. al. [2] summarized the typical
compression properties of MMSF in their review paper. MMSF exhibit other properties
as well as a result of being both foams and composites. For example, Mondal et. al. [3]
found that the wear rate of an aluminum - cenosphere MMSF was comparable to an Al-
SiC composite at low loads. Dou et. al. [4] found that MMSFs might be useful
electromagnetic shielding materials.
There are several methods of manufacturing MMSFs and several challenges in
doing so. Stir casting and pressure infiltration are the most popular methods [2]. Powder
metallurgy routes such as injection molding and press and sinter are less common. The
main challenge in manufacturing these materials is in incorporating the filler into the
matrix without segregation, sphere fracture, or negative chemical interactions between
the matrix and filler. In their review of the types of hollow spheres used in the literature,
Szlancsik et. al. [5] found that the hollow particles, or filler, are usually a glass or ceramic
material, though metal particles and expanded clay have also been utilized. In particular,
fly ash cenospheres remain a popular filler as they would otherwise be a waste-product
and can withstand higher temperatures than glass fillers [2, 6 , 7], The filler materials, as
they have very low densities, tends to float in molten metal which can cause density
gradients. The hollow particles can fracture during processing as well. [8] Material
compatibility is another challenge. Most glasses, other than pure silica, will likely melt
below the sintering or melting temperature of the metal. Diffusion of the matrix material
into the filler and vice versa can also occur, which can cause chemical reactions to occur.
[9, 10] In glasses, this can also cause the melting temperature of the filler to decrease.
Oxide ceramics might not wet strongly with the matrix. The two powders should bond
strongly so that load partitioning can occur. [11] Neville and Rabiei [12] worked around
these issues by using metal hollow particles as the filler. Majlinger et al [13] studied the
effect of using multiple types of particles in the same material. Lehmhus et. al. [14]
compared hollow glass to cenosphere filler in a steel alloy (316L) matrix.
A key part of MMSF research is the development of novel techniques to improve
the manufacturing of MMSF materials. For example, Weise et al [15] and Yang et al [16]
used similar methods to ensure an even distribution of hollow particles during infiltration
of the metal matrix. They sintered the filler before infiltration - Weise et al [15] using
66
67
glass filler and Yang et al [16] using ceramic spheres. Augmentation of a manufacturing
technique is one method of working around the manufacturing weaknesses of MMSF
materials. Other researchers, such as Orbulov [17] with pressure infiltration, optimize a
manufacturing process for MMSF materials. Another method is to use a novel
manufacturing technique such as done by Shiskin et al [18]. They sputter-coated the
hollow spheres with copper then used spark plasma sintering to densify the parts. Further,
a manufacturing technique that mitigates many of the manufacturing issues can be
selected.
Injection molding is one of those techniques that need little process optimization
to effectively make MMSF materials. Hollow particle flotation is not an issue in injection
molding as the powder is held in place by a binder. Particle fracture can still be an issue,
but gentle comminution and injecting procedures can be used. Weise et. al. [19, 20] used
injection molding to produce an iron alloy matrix with hollow glass microsphere
syntactic foams. They were able to sinter the iron matrix to the point where little matrix
porosity remained. However, the glass did appear to soften and wick into residual
porosity during sintering. This is a material compatibility problem solved by using
ceramic powders rather than glass ones. Hollow ceramic powders that are small enough
to use in injection molding and of good quality are difficult to find without requesting
specialty powders. Injection molding requires fine powders to form stable slurries that
can be injected into molds. As described by Szlancsik et. al. [5], most hollow ceramic
powders are in the 1 to 10 mm size range, not the <45 |im range that is typically used for
injection molding processes.
68
Injection molding requires careful selection of powders and a compatible binding
agent. The powder shape and size will influence both sintering and slurry rheology. Fine
powder sinter better than coarse powders due to their increased surface area. This
increase in surface area may negatively affect slurry fluidity. Bimodally distributed
powders achieve higher green densities as small particles fit between larges ones and fill
space. It is mathematically appropriate to have the fine particles in the bimodal
distribution be 1/7th of the coarse particles. For MMSFs, this might mean that the matrix
particles should be, on average, 7 times smaller than the hollow particles, or vice versa.
The choice of binder will determine the binder removal and burnout procedure. The
binder usually leaves some carbon or oxides behind, so minimizing the amount of binder
needed to create a flowable slurry is necessary.
Aqueous binders leave very little material leftover after the water backbone has
been removed. One such binder is a water-based gel is comprised of agar, glycerin, and
water. [21] The basic formula for this binder is agar and water which, when heated, forms
a polysaccharide gel. The glycerin in the formula acts as a gel strengthener. The binder
has similar rheological properties to a thermoplastic according to Labropoulos et. al. [22].
It has an added benefit of being more environmentally friendly than waxy binders as
usually more than 90% of the binder is water. The general formula for this binder was the
subject of or used in several patents [23-25]. It was also produced commercially
(Honeywell Powderflo Technologies, for example). Chen et. al. [26] produced NiTi
foams with an agar binder, though they used sucrose as the gel strengthener rather than
glycerin.
The goals of this work were to develop an aqueous binder for low-pressure
injection molding of metal matrix syntactic foams and successfully manufacture a copper
matrix syntactic foam using that binder and process. The composition of the agar-
glycerin-water binder was optimized to obtain the highest green and sintered relative
density in a bronze syntactic foam. The relative density was selected as a simple way to
check the quality of the parts. This optimized binder was then used to injection mold
copper matrix syntactic foams with a porous silica filler. Four samples were made with
an increasing volume fraction of filler. The quality of the sintered samples was analyzed.
The MMSFs were tested in flexure to determine the mechanical properties and compared
to the literature.
69
2. MATERIALS AND METHODS
2.1. POWDER CHARACTERIZATION
Powder characterization was performed on the syntactic foam materials to verify
morphology, and particle size distribution. Table 1 lists the syntactic foam materials used
in this experiment with the manufacturer-supplied chemistry and density. The particle
size distribution was characterized via particle measuring and counting from SEM
images. Each powder was dusted over the surface of a carbon dot on a flat specimen
holder. The powder was then mounted into an ASPEX scanning electron microscope
(SEM). The ASPEX has an automated feature analysis (AFA) that will identify and
measure inclusions or particles. At least 10,000 particles were measured for each powder,
except the two glasses. The borosilicate glass and porous silica powders could not be
70
distinguished from the noise in the ASPEX and so the particle ‘features’ were instead
analyzed by manual measurement. This was done using FIJI (Fiji Is Just ImageJ)
software. The AFA included the aspect ratio of each particle, which (along with visual
inspection of the SEM images) was used to determine morphology.
Table 1: Details provided by the manufacturer on each metal and ceramic powder used inthis experiment
Material Supplier Chemistry Density(g/cc)
Borosilicate Glass MO-SCICorp.
70-85 % SiO2
10-15% B2O3
5-10% Na2O 2-5% Al2O3
0.15
Porous Silica (P-S) MO-SCICorp.
SiO2 1.475
Copper Royal Metal Powders Inc
99.8% Copper 0.06%
Hydrogen loss
8.94
Bronze Royal Metal Powders Inc
88.46% Copper 11.3% Tin
0.24%Phosphorous
8.73
2.2. BINDER DEVELOPMENT
In this study, an aqueous binder was developed using a mixture model. The model
is discussed in the calculation section of this paper. The binder components were
deionized water, agar, and glycerin. The mixture model requires 10 samples with
different binder compositions to optimize the composition. The solids loading remained
constant at 55 vol%, and the solid composition was held constant as well. The target
71
sintered sample was a 30/70 volume ratio of hollow borosilicate spheres to bronze matrix.
Specimens were made in 50 g batches by manually mixing the binder and the solid
powders. To make the binder, boiling water was mixed with the glycerin and agar in the
appropriate combinations. The mixture was stirred until all the dry ingredients dissolved
into the solution, about 5 minutes. The solid components were heated to 100°C then
mixed into the binder. Each specimen was poured into a 50*20*15 cm silicone mold.
Three repetitions were done for each sample, and all specimens were made and measured
in a completely random order. Sample post-process characterization is discussed in
Section 2.4.
A binder-only trial in the injection mold machine was done to determine the rate
of water evaporation during use. In this test, the binder components were added to the
reservoir. The reservoir was then heated, while mixing, to 80°C. Five binder-only
specimens were injection molded every 30 minutes for 120 minutes. The specimens were
dried at 120°C for 12 hours. The mass of the specimens before and after drying was
measured and used to find the liquid loss of the binder. These values were then compared
over time to find the change in mass loss over time.
2.3. INJECTION MOLDING
A Peltsman MIGL-28 was used to injection mold the samples in this experiment.
This low-pressure injection mold machine includes a feedstock reservoir that allows for
compounding of the feedstock before injection molding. The reservoir is paddle mixed at
a fixed rate of 58 rpm. The reservoir temperature was set at 80°C for this experiment. Gas
pressure is applied to the reservoir during molding which pushes material through a
72
heated tube to the mold. The tube and orifice ring were set to 85°C. The gas pressure was
set at 67 kPa. The pressure was held for 10 seconds per specimen. The specimens were
demolded after approximately 30 seconds. The mold used in this experiment was a water-
cooled iron alloy rectangular mold. The dimensions were 6.21*0.99*0.66 cm.
Four compositions were injection molded in this study. Table 2 shows the volume
fraction of each component that was added to the reservoir, out of the total volume.
Thirty specimens were injection molded for each sample. Each sample was prepared by
mixing the binder on a hot plate, then adding the prepared binder to the pre-warmed IM
reservoir. The powders were added to the reservoir pre-heated to 100°C. For some
samples, extra binder was added to decrease slurry viscosity. The slurry was mixed for 1
hour after the last addition before injection molding began.
Table 2: The composition of each slurry in the injection molding reservoir before injection molding is listed, all units in volume percent
Component 0% P-S 5% P-S 10% P-S 15% P-S
Agar 3% 3% 4% 4%Glycerin 2% 2% 2% 2%Water 40% 44% 50% 50%BinderTotal
45% 49% 56% 56%
P-S 0% 3% 4% 7%Copper 55% 49% 40% 38%
Solids Total 55% 51% 44% 44%
73
2.4. SAMPLE POST-PROCESSING AND CHARACTERIZATION
After injection molding, the ‘wet’ samples were placed in a furnace at 120°C.
Samples were dried for at least 12 hours before sintering. The ‘dry’ specimens underwent
debinding and sintering in an atmospheric furnace with an attached retort. Debinding was
done at 450°C for 1 hour in flowing air. A steel plate was placed in the furnace beside the
samples to help getter escaping carbon and oxygen. Sintering was done directly after
debinding by dwelling at 450°C for 1 hour in flowing argon, then ramping to the
sintering temperature at a rate of 120 at 120°C/hour. Samples remained under flowing
argon throughout the sintering time. Bronze matrix samples were held at 900°C for 1
hour. Copper matrix samples were held at 1000°C for 10 hours. Samples were removed
from the furnace after cooling.
Characterization included density measurements and microstructural evaluation.
The geometric density of each specimen was measured after each processing stage (wet,
dry, and sintered). The geometric density calculation is mass over volume. Volume was
estimated as the length, width, and height of each specimen, measured with calipers. The
mass was measured using a laboratory balance. Sintered density was measured by
Archimedes’ method as described in ASTM C373-18 [27]. The vacuum method was
used. Theoretical density at each stage of processing was done by rule of mixtures using
the measured amounts of each component added to the injection molding reservoir for
each sample. Specimens were prepared for microstructure observation by mounting
specimens in bakelite and polishing. The ASPEX SEM was used for standardless energy-
dispersive x-ray spectroscopy (EDS) mapping of each sample. The automated feature
analysis function was used on the polished surface of one specimen from each sample as
well. This analysis measured the chemistry (using EDS), area, and average diameter of
10,000 features. The features were sorted by chemistry. The total area of the features and
the total area observed were used to calculate the total area percent of features in the
specimen.
The samples were the correct size and shape for 3-point bend testing. ASTM
C1161-18 [28] was used for this test, with test configuration B. Fifteen specimens were
tested for each sample. A fully articulating fixture was used. A strain rate of 0.5 mm/min
was used. Pre-loading was not done for these tests. Visual Basic for Excel code was used
to remove the toe region from each data set and determine the 0 .2% offset yield stress.
The 0.2% offset yield stress was calculated by first finding the slope of the elastic region.
Then, a line was created offset by 0.2% from the stress-strain curve. The intersecting
between the new line and the stress-strain curve was taken as the 0 .2% offset yield stress.
74
3. CALCULATION
The agar-glycerin-water binder used in this experiment was developed using a
mixture model. A 3-component simplex design was used. The wet, dry, sintered relative
density responses to altering the ratios of agar to glycerin to water were modeled. Ten
sample compositions are used in this model. The overall composition of each sample was
45 vol% binder, 33 vol% bronze, and 22 vol% hollow borosilicate glass. Three
replications were made for each sample. The 30 specimens were made and measured
completely randomly. JMP software was used to design this experiment and calculate the
response surfaces.
The experiment was designed by first selecting the lower bound of each
component. The lower limit of agar was determined after tests showed that agar does not
gel with sufficient strength below 4%. The glycerin was only a gel strengthener so the
lower limit was set at 0%. Finally, after about 15% agar, the agar became increasingly
difficult to completely dissolve into the water so the lower limit on the water was set at
85%. The compositions for the test samples were determined using these lower bounds.
The standard composition array is shown in Table 3. This table shows three pseudo
components which exist at the corners of the ternary response surface. The pseudo
components are related to the pure components by Equation 1. The ten compositions with
the pure components used in this experiment are shown in Table 4.
75
Table 3: The general composition of each sample for the model calculation, all units involume percent
PseudoComponents
1 2 3 4 5 6 7 8 9 10
A 100% 0% 0% 50% 50% 0% 33% 67% 17% 17%B 0% 100% 0% 50% 0% 50% 33% 17% 67% 17%C 0% 0% 100% 0% 50% 50% 33% 17% 17% 67%
Table 4: The composition of each binder for the model calculation, all units in volumepercent
PureComponents
1 2 3 4 5 6 7 8 9 10
Agar 15% 4% 4% 10% 10% 4% 8% 1 1 % 6% 6%Glycerin 0% 1 1 % 0% 6% 0% 6% 4% 2% 7% 2%Water 85% 85% 96% 85% 91% 91% 89% 87% 87% 92%
76
The cubic response model (Equation 2) uses seven of the ten measured responses
to generate a response surface. Each b-value in the model corresponds to a mathematical
combination of measured responses, as shown by Equations 3 through 9. The other three
samples are used to test the lack-of-fit. Lack-of-fit measures the accuracy of the predicted
response values. The variance for the lack of fit (Slf2) was calculated by finding the
variance between the measured responses of samples 8 through 10 and the model
prediction of compositions 8 through 10. This is shown in Equation 10. The model must
also be tested to determine its viability. This was done by an F-ratio test. The selected
confidence interval for this test was 95%. The error variance (Serror2) was then calculated
with Equation 11. The F-ratio was calculated by Equation 12. This F-ratio was then
compared to the tabulated value. If the calculated F-ratio is less than the tabulated F-ratio,
then the model is considered a reasonable representation of the response surface of the
mixture within a set confidence interval. The tabulated F-ratio, in this case, has 3
specimen degrees of freedom, and 10 sample degrees of freedom.
i= iLB* ( 1 - iLB) * x t Equation 1
where i is the actual component amount (agar, glycerin, or water), Ilb is the lower bound
of component i, and Xi is the amount of the pseudo-component.
Cubic Response Model = b1x1 + b2x2 + b3x3 + b12x1x2 + b13x1x3 +
b23x2x3 + b123x1x2x3 Equation 2
where xi, X2, and X3 are the amount of pseudo-components A, B, and C, respectively, and
bi, b2, b3, bi2, bi3, b23, and bi23 are related to the measured response values as indicated in
Equations X4 to X10.
b i = 7 i Equation 3
77
b2 = y 2 Equation 4
^3 = y3 Equation 5
= 4y4 - 2 (y1 + y 2) Equation 6
= 4y5 - 2 (y1 + y3) Equation 7
= 4y6 - 2 (y2 + y3) Equation 8
* (y4 + y5 + y6) + 3(yi + y2 + y3) Equation 9
where yi to y 7 are the average measured response of samples one through seven.
s If = - y Vrd2) Equation 10
where yoi is the observed average response at the ith checkpoint (samples 8 through 10),
andypri is the predicted response at the ith checkpoint.
S2 =u erro rS p o o le d Equation 11
where Spooled2 is calculated with Equation 13 and is the pooled variance of all the data,
and n is the number of replications per sample.
F = 4 ^ -
c2 = 1 L c2‘-’pooled n _ ^ £ i *J £
Equation 12
Equation 13
where Si2 is the variance of the ith sample.
e r r o r
4. RESULTS AND DISCUSSION
4.1. POWDER CHARACTERIZATION
The particle size distributions by count, volume, cumulative count, and
cumulative volume percent for each of the four powders are shown in Figure 1. Table 5
shows the 10th, 50th, and 90th percentiles by count (D10, D50, D90, respectively) and
average for each powder. The size of the copper and porous silica powders were similar.
The borosilica and bronze powders appear to have similar sizes from the table, however,
the borosilica powder has far more particles between 10 and 20 microns than the bronze
powder as shown in Figure 1. The bronze powder was sieved by the manufacture to
remove fines, however, some less than 10 micron particles remained (~10%) which has
skewed the results. Due to the large density difference between the heavy metal powders
and the lightweight glass powders, these two composite powders will not stay mixed, if
they can be mixed at all. This can be solved by using a binder to facilitate mixing.
However, segregation while the binder is liquid will be a concern. In the future, it would
be appropriate to sieve out specific particle sizes such that the composite powders are
approximately 1:7 in size ratio. This will aid in mixing and may help prevent some
amount of segregation.
The particle size and morphology of the metal powders will affect how easy they
are to process and sinter. Irregular particles sinter marginally better than spherical
particles due to their higher surface area, but they are much less flowable. Figure 2 shows
example images of each powder that were used to determine the particle size distribution.
All the powders used here were spherical, as Figure 2 shows clearly. The size of the
copper powder is suitable for an injection molding and sinter operation. However, the
bronze powder is larger than would be typical for a sintering operation, especially a
pressure-less sintering operation as was used here. This is one reason that the bronze
powder used in the binder optimization procedure was not used for the injection molding
78
process.
79
25%
20%
100%
5%
0%
25%
20%
15%o10%
5%
0%
80%CD PLh CD>JB 3so
60%
40%
20%
0%o o o o o o o o o o o o o o o o
b Particle Diamete^^m) 100%
80%
60%
40%
20%
0%
cCDO—a>Ph0)>
"3£3O
O O O O O O O O O O O O O O O O
25% T C
20%
Particle Diameter (^m)
15%o10%
5%
0%
25%
20%
15%
10%
0 2 4 6Particle Diameter (^m)
5%
0%
100% Count Perce
80%_60%
: > 40%
: 3 20%
= O
ESSS3 VolumePercent
CumulativePercent
0%) — — Cumulativef
Volume100% Percent
80%
60%
40%
20%
0%
s<DoCDPLhCD.>J33£o
0 4 8 1216 2024 28 32 3640 44 48 52 56 60Particle Diameter (^m)
Figure 1: The particle size distribution by count and volume is shown for the four powders used in this study, a) Bronze, b) Borosilica, c) Copper, d) Porous Silica
80
Table 5: The 10th, 50th, 90th percentiles and average particle size in microns for eachpowder, all units in microns.
Powder D10 D50 D90 AverageBronze 5 56 93 57
Borosilicate 20 46 84 50Copper 3 6 13 7
P-S 8 16 29 17
Figure 2: SEM images of each powder are shown: a) Bronze, b) BorosiPorous Silica.
ica, c) Copper, d)
4.2. BINDER OPTIMIZATION
The binder development results include the cubic response equation and graph for
each measured response. Table 6 shows the b-values for each response which are applied
to Equation 2 (the cubic response equation) to calculate the response surfaces. Figure 3
81
shows response surfaces calculated from the cubic response equations for the wet (A),
dry (B), and sintered relative density (C), as well as the open porosity after sintering (D).
The area of highest relative density appears in the same general area of each graph. The
optimized composition was selected to be 7% agar, 4% glycerin, and 89% water. The
predicted values for each response given this composition are shown in the last column of
Table 6 . The error, pooled variance, lack-of-fit, and F-statistic for the mixture model are
shown in Table 7. With a 95% confidence interval, the tabulated F-statistic is 3.71. All
the F-statistics were below this value, so they are considered reasonable estimations of
the response. Because the solids loading and solid composition were kept constant across
all samples, this experiment should only reveal the trends of the binder. However, the
actual density depends strongly on the powder chemistry and particle size. This is
especially true of the sintered density. The b-values used here are valid only for this
experiment, but the trends should be relatively universal.
The binder development work was done using the bronze and hollow borosilicate
powders. During sintering, the borosilica glass reacted with the tin in the bronze alloy.
Tin oxide and silica inclusions were spotted in the bronze. This material compatibility
issue did not invalidate the binder development, as the trends found in this study are
considered to be material agnostic. This did prompt the matrix change to copper for the
injection molding portion of the study. Further, the melting temperature of the borosilica
powder was much lower than the sintering temperature of copper, so the filler material
was changed to the porous silica powder.
82
Volume Fraction AgarVolume Fraction Agar
o.87
._o 0.09
$■ 0.06
0.03
0.15 0.12 0.09 0.06
Volume Fraction Agar
0.40.45
0.50.55
0.6
Figure 3: The response surfaces for the a) wet density, b) dry density, c) bulk sintered density, d) open porosity after sintering where the legend indicates the relative density
fraction for a, b and c and the fraction of open porosity to total volume for d.
The binder used in this experiment was described by German and Bose [21] as an
aqueous gel useful for injection molding at low temperatures. The binder solidifies below
about 50°C and should be kept below 100°C (boiling) to prevent bubbles from forming in
the gel. This binder was chosen for this process specifically for several reasons. First, the
Peltsman MIGL-28 IM machine used in this study has a maximum operating temperature
of 120°C, so a relatively low-temperature binder had to be selected. Further, this binder is
easily cleaned with friction and warm water and is non-toxic. Finally, this binder is
approximately 90% water, which is both environmentally friendly and easy to remove
from the samples. Once the water is removed, the small amount of remaining binder is
removed by a binder burnout step. Though pre-made feedstock was not necessary for this
experiment, as the IM machine included a mixing apparatus as part of the machine,
industrial MIM machines may require pre-mixed feedstock. This binder is also capable of
being re-liquidized after comminution with the powders if the water has not been
removed. Scale-up of this binder to industrial processes is therefore possible.
83
Table 6: The cubic model b-values calculated for each response and the predicted valuebased on the optimized composition
Response b1 b2 b3 b12 b13 b23 b123 PredictedValues
WetRelativeDensity 0.674 1.106 1.127 0.295 0.447 -0.125 3.487 118%
DryRelativeDensity 0.674 1.051 1.148 0.350 0.393 0.136 4.201 123%SinteredRelativeDensity 0.587 0.777 0.838 0.368 0.367 0.127 2.197 91.6%Open
Porosity0.617 0.482 0.455 -0.276 -0.321 -0.086 -1.207 40.0%
Figure 4 shows the amount of water over time in binder-only injection molded
samples. During the first 10 to 20 minutes of the trial, the specimens varied in
composition (not shown). The agar and glycerin were added to the water cold, so the agar
agglomerated in the reservoir. It took approximately 30 minutes after reaching the
84
working temperature of 80°C for the agar to fully incorporate into the solution. In future
trials, the binder was mixed on a hotplate before adding it to the reservoir to prevent this
issue. The composition of the binder remained steady up to 2 hours after reaching the
working temperature. This indicates that the reservoir was sealed enough to prevent
significant amounts of water evaporation during usage. No water additions were therefore
necessary during use.
Table 7: The error for each response and relevant variance calculations are tabulated.R esponse n Spooled2 S 2Serror Sl f 2 F(calc)
Wet Relative
D ensity
3
0.029 0.010 0.002 0.159
D ry Relative
D ensity
3
0.034 0.011 0.000 0.035
Sintered Relative
D ensity
3
0.029 0.010 0.002 0.159
Open Porosity 3 0.0147 0.0049 0.0003 0.0653
4.3. INJECTION MOLDING
For the injection molding of syntactic foams, the binder is a critical part of
stopping segregation and promoting good mixing between the dissimilar powders that
make up the composite. The way this is determined is by tracking the composition of
85
Figure 4: The percent mass loss in each binder-only injection molded specimen over time
each specimen after molding. This can then be compared to the input composition of the
slurry. The volume and weight of each sample were measured three times: after injection
molding (wet), after removing all the water (dry), and after sintering (sintered). These
three measurements form the basis of the calculation to determine composition. The
appendix shows the calculation steps used to determine the composition for each of the
thirty specimens in the samples.
The porosity in the porous silica particles after sintering was not assumed to be
the same as before sintering. Instead, this value was found by measuring the average size
of the porous silica particles in the composite after sintering and comparing that to the
average size of the powder. The AFA data from the feature analysis performed on the
5%, 10%, and 15% P-S specimens were used to determine the average size of the P-S
particles after sintering. The average particle size of the porous silica particles after
sintering was 9.0 |im. Compared to the 17.4 |im average of the powder before sintering,
this change is significantly large. The particles had, on average, 45% porosity before
sintering. The difference in average size after sintering was 52%, so it appears that little
to no porosity remained intact in the porous silica spheres. There is significant error in
using this technique to estimate the porosity in the silica particles so this may not be
accurate. For example, the ASPEX tends to under-predict the size of glass materials. The
silica powder was measured manually, in contrast. For this calculation, the sphere
porosity was taken as 0% after sintering.
The results of finding the composition of each specimen are summarized in Table
8. Figures 5 and 6 show the data for each specimen. Figure 5 shows the percentage of
binder out of the total wet volume per specimen. Each specimen was assigned a number
from 1 to 30 after they were injection molded. The specimen number loosely correlates to
time, approximately 1 to 2 minutes per specimen. The binder concentration was relatively
stable over time. The amount of binder was generally low compared to what was added to
the reservoir (Table 2). The binder may have segregated from the powders in the
reservoir or during molding. Figure 6 shows the percentage of porous silica in each
specimen. The amount of porous silica was high on average, particularly in the case of
the 15% P-S sample. The filler concentration in the 5% and 10% P-S samples seemed to
fluctuate almost sinusoidally over time. The 15% P-S specimens did not exhibit this
behavior, remaining more consistent except at the ends of the graph. The compositional
fluctuations did not appear to be a segregation issue, as in that the composition would
have decreased over time. There may have been pockets of high concentrations of filler
that contributed to the fluctuating composition, which would indicate poor mixing of the
86
87
Table 8: The average volume percent of binder and porous silica in each sample afterinjection molding, as calculated
Sample % Binder % Porous Silica100% Cu 40% ± 0.6% N/A5% P -S 40% ± 2.3% 6% ± 1.9%
10% P -S 50% ± 2.3% 11% ± 3.2%15% P -S 53% ± 0.9% 24% ± 3.4%
slurry. Otherwise, it is possible that the slurry segregation was reduced or mitigated by
the constant agitation of the slurry in the IM reservoir.
The amount of matrix porosity in the wet specimens was assumed to be negligibly
small. However, three specimens from the 5% P-S sample were found to have significant
matrix porosity. These specimens were each found to have a large hole in their center,
likely due to an injection molding error. They were removed from the data, due to this
defect. The other specimens did not appear to have this issue.
4.4. BINDER BURNOUT AND SINTERING
The relative density of each sample after sintering is shown in Table 9. These are
average values calculated from the average filler volume percent of each sample in Table
8. The samples did not sinter to full density despite holding at over 90% of the melting
temperature of copper for 10 hours. This is due either to a binder burnout issue or a result
of water corrosion affecting the copper powder. Figure 7 shows an element map from a
0% P-S specimen. This image shows the distribution of copper and oxygen in the
specimen. The oxygen seems to surround the porosity, creating a shell of copper oxide.
The atmosphere used in this experiment was flowing air. A forming gas would be more
Cal
cula
ted
Bin
der
Am
ount
(V
olum
e Pe
rcen
t)
88
A 0% P-S
• 5% P-S
X 10% P-S
- 15% P-S
Specimen Number
Figure 5: The percentage of binder in each specimen versus specimen number, where specimen number is the order in which specimens were injection molded
• 5% P-S
X 10% P-S
- 15% P-S
Figure 6: The percentage of porous silica in each specimen versus specimen number, where specimen number is the order in which specimens were injection molded.
89
appropriate, as it appears the oxygen in the air reacted not just with the carbon in the
binder, but also with the copper powder. The formation of copper oxide allowed porosity
to remain in the sample during sintering.
Table 9: The relative density of each sample after each stage of sample post-processing shows the approximate amount of porosity in each sample
S a m p le
Wet Relative
Density
Dry Relative
Density
Sintered Relative
Density
1 0 0 % Cu 101% ± 1.1% 64% ± 0.5% 83% ± 1.3%
5 % P -S 99% ± 3.1% 70% ± 4.9% 88% ± 3.3%
1 0 % P -S 98% ± 2.3% 60% ± 3.2% 82% ± 2.6%
1 5 % P -S 111% ± 2.4% 70% ± 2.6% 91% ± 2.3%
*•
A, %' v,
^ ^ ©ACopper &
%
Figure 7: A element map of a 0% P-S specimen with the carbon, copper, and oxygen elements identified, the image was taken near the center of a specimen.
The amount of carbon present in the samples was found using the ASPEX AFA
analysis. One specimen from each sample was tested. The concentration of carbon was
90
0.11% ± 0.08% (volume percent) on average. This indicates that the carbon in the binder
was successfully removed from the samples.
The porous silica particles appeared to be well dispersed in the samples. Figure 8
shows a micrograph of a 15% porous silica specimen. This figure shows the that the
particles appear well dispersed. Some of the particles did appear to wick into the
surrounding porosity during sintering. Other particles were surrounded by porosity and
remained spherical. Bright spots in the silica may indicate areas of porosity as the edges
of the pores in the silica charge under the SEM. Alternatively, these could be copper that
migrated into the silica particles. The silica particle did not appear fragmented, most were
either spherical or wicked into nearby pores. The well-dispersed particles indicate good
local mixing during injection molding.
Figure 8: A micrograph of a 15% P-S specimen showing the dispersion of the porous silica particles and the morphology of the particles after sintering, the image was taken
near the center of a specimen.
91
4.5. MECHANICAL TESTING
The results of the 3-point bend testing are shown in Figure 9 compared to other
metal matrix syntactic foams. The matrix material and density for each MMSF were used
to normalize the data to better compare MMSF materials with different matrix materials.
The normalized density shows how much matrix porosity and filler porosity is in the
composite. The normalized strength is a measure of the matrix strengthening done by the
filler. For a lightweight structural material, the optimal MMSF, from a design standpoint,
should be in the upper left corner of the graph. Generally, above 1 in the strength axis and
below 1 in the density axis should be a design target. The MMSF materials made here
meet both of these targets, though they are on the low end of both targets.
5. CONCLUSIONS
Copper and porous silica (P-S) powders were used in a low-pressure injection
molding process to make rectangular specimens. Four compositions were made - 0, 5,
10, and 15 volume percent P-S with the remainder being copper. The powders used in
this experiment were expected to dry or wet mix poorly due to their similar size and
highly dissimilar densities. The binder optimized in this experiment worked well with the
syntactic foam materials. The agar-glycerin-water binder was optimized to achieve the
highest density. The optimized composition was 7% agar, 4% glycerin, and 89% water.
The specimens in each sample composition experienced some compositional fluctuations
over time. There was no strong time dependence, however, so this was unlikely to be a
segregation issue. The overall average filler concentration was higher than expected for
92
'J
Composite Density / Matrix Density
A This Work •[29]0[30]0[14]= [16]A [18]X [13]
Figure 9: The strength and density of several MMSF materials from this work and the literature [13], [14], [16], [18], [29], [30], where each composite’s strength was
normalized by the average strength of the matrix material and each composite density was normalized by the density of the matrix material for the sake of comparison. Note
that some literature values were estimated from figures to the best of the author’s ability.
each sample. The binder burnout procedure left 0.11% ± 0.08% carbon remaining in the
specimens, a reasonably low amount. The silica particles were well-dispersed in
specimens, indicating good mixing during injection molding. The yield strength of the
material was within the optimal range for a lightweight structural composite.
CONFLICT OF INTEREST
On behalf of all authors, the corresponding author states that there is no conflict
of interest.
93
APPENDIX
The concentration of each component in each specimen was required to be
calculated. The components were as follows: copper, silica, water, additives, matrix
porosity, and sphere porosity. The porous silica contained on average 45% porosity, and
this was kept separate from the matrix porosity. The data used to find these values was
the mass and volume measurements at each stage of post-processing. The stages were
wet, dry, and sintered. In each stage, the components were different. For example, there
was no water in the samples after the drying stage. Equations A1.1 to A1.6 show the
equations generated for each mass and volume measurement. The density of each
component was also assumed to be known. The density of the components was assumed
to be known, per Table 1, and assuming that the porous silica was 45% porosity, the
density of water was 1 g/cm3 and the density of the additives was also 1 g/cm3. This
calculation also assumes that the pores in the porous silica completely filled with binder
during the wet stage. To calculate the volume of sphere porosity after drying, Equations
A1.7 and A1.8 were used.
Mw = MCu + Msm + Mwater + Madd Equation A1.1
VW ^Cu + ^water + Vadd + ^Wmp Equation A1.2
Md = MCu + MSm + Madd Equation A1.3
Kd ^Cu + ^add + ^Dmp + ^Dsp Equation A1.4
Ms = MCu + MSm Equation A1.5
Vs = VCu + Vsm + Vsmp Equation A1.6
Pps = (PsmVsm')/ (VTsp + VSM) Equation A1.7
94
^Dsp ^Tsp Kv/ ( ^add + Kv) Equation A1.8
where M, V, and p represent mass, volume, and density, respectively. Subscripts W, D, S,
Cu, SM, water, and add represent wet, dry, sintered, copper, silica material, water, and
additives. Xmp andXsp represents matrix porosity and sphere porosity where X can be
W, D, S, or T (total).
It was assumed that the matrix porosity in subsequent stages depended on the
previous stage. Equations A1.9 and A1.10 show these relationships. Notably, all sphere
porosity was assumed to be converted to matrix porosity as the sphere porosity collapsed
during sintering.
^Dmp ^water + ^Wmp + ^extra Equation A1.9
^Smp ^ Dmp + ^Tsp ^ Kr/ ( ^add + Vw) + Vadd) Equation A1.10
where A represents the extent to which the matrix densified during the sintering
operation. In the 0% P-S sample, this value was found to be 18.0 ± 0.807 %. This value
was assumed to be constant across all specimens. Equations A1.1 through A1.10 were
used to calculate the concentration of each component in each injection molded
specimen.
REFERENCES
[1] N. Gupta and P. K. Rohatgi, “Metal Matrix Syntactic Foams: Processing, Microstructure, Properties and Applications,” in Metal Matrix Syntactic Foams: Processing, Microstructure, Properties and Applications, DEStech Publications, Inc, 2014, p. 370.
[2] P. K. Rohatgi, N. Gupta, B. F. Schultz, and D. D. Luong, “The Synthesis, Compressive Properties, and Applications of Metal Matrix Syntactic Foams,” JOM, vol. 63, no. 2, pp. 36-42, 2011.
[3] D. P. Mondal, S. Das, and N. Jha, “Dry sliding wear behaviour of aluminum syntactic foam,” Mater. Des., vol. 30, no. 7, pp. 2563-2568, Aug. 2009.
95
[4] Z. Dou, G. Wu, X. Huang, D. Sun, and L. Jiang, “Electromagnetic shielding effectiveness of aluminum alloy-fly ash composites,” Compos. Part A Appl. Sci. Manuf., vol. 38, no. 1, pp. 186-191, 2007.
[5] A. Szlancsik, B. Katona, A. Kemeny, and D. Karoly, “On the Filler Materials of Metal Matrix Syntactic Foams,” Materials (Basel)., vol. 12, no. 12, p. 2023, Jun. 2019.
[6] D. P. Mondal, S. Das, N. Ramakrishnan, and K. U. Bhasker, “Cenosphere filled aluminum syntactic foam made through stir-casting technique,” Compos. Part A Appl. Sci. Manuf., vol. 40, no. 3, pp. 279-288, 2009.
[7] U. S. Agrawal and S. P. Wanjari, “Physiochemical and engineering characteristics of cenosphere and its application as a lightweight construction material- A review,” Mater. Today Proc., vol. 4, no. 9, pp. 9797-9802, Jan. 2017.
[8] X. Xue and Y. Zhao, “Ti matrix syntactic foam fabricated by powder metallurgy: Particle breakage and elastic modulus,” JOM, vol. 63, no. 2, pp. 36-40, 2011.
[9] Y. Lin, Q. Zhang, and G. Wu, “Interfacial microstructure and compressive properties of Al-Mg syntactic foam reinforced with glass cenospheres,” J. Alloys Compd., vol. 655, pp. 301-308, 2016.
[10] I. Orbulov and K. Majlinger, “On the microstructure of ceramic hollow microspheres,” Period. Polytech. Mech. Eng., vol. 54, no. 2, p. 89, 2010.
[11] D. K. Balch and D. C. Dunand, “Load partitioning in aluminum syntactic foams containing ceramic microspheres,” Acta Mater., vol. 54, no. 6, pp. 1501-1511, 2006.
[12] B. P. Neville and A. Rabiei, “Composite metal foams processed through powder metallurgy,” Mater. Des., vol. 29, no. 2, pp. 388-396, 2008.
[13] K. Majlinger and I. N. Orbulov, “Characteristic compressive properties of hybrid metal matrix syntactic foams,” Mater. Sci. Eng. A, vol. 606, pp. 248-256, 2014.
[14] D. Lehmhus et al., “Quasi-static and Dynamic Mechanical Performance of Glass Microsphere- and Cenosphere-based 316L Syntactic Foams,” Procedia Mater. Sci., vol. 4, no. 0, pp. 383-387, 2014.
96
[15] J. Weise, V. Zanetti-Bueckmann, O. Yezerska, M. Schneider, and M. Haesche, “Processing, Properties and Coating of Micro-Porous Syntactic Foams,” Adv. Eng. Mater., vol. 9, no. 1-2, pp. 52-56, 2007.
[16] Q. Yang, J. Cheng, Y. Wei, B. Yu, Z. Miao, and P. Gao, “Innovative compound casting technology and mechanical properties of steel matrix syntactic foams,” J. Alloys Compd., p. 156572, Aug. 2020.
[17] I. N. Orbulov, “Metal matrix syntactic foams produced by pressure infiltration— The effect of infiltration parameters,” Mater. Sci. Eng. A, vol. 583, no. 0, pp. 1119, 2013.
[18] A. Shishkin, M. Drozdova, V. Kozlov, I. Hussainova, and D. Lehmhus, “Vibration-Assisted Sputter Coating of Cenospheres: A New Approach for Realizing Cu-Based Metal Matrix Syntactic Foams,” Metals (Basel)., vol. 7, no. 1, p. 16, Jan. 2017.
[19] J. Weise, J. Baumeister, O. Yezerska, N. Salk, and G. B. D. Silva, “Syntactic Iron Foams with Integrated Microglass Bubbles Produced by Means of Metal Powder Injection Moulding,” Adv. Eng. Mater., vol. 12, no. 7, pp. 604-608, 2010.
[20] J. Weise, D. Lehmhus, J. Baumeister, R. Kun, M. Bayoumi, and M. Busse, “Production and Properties of 316 L Stainless Steel Cellular Materials and Syntactic Foams.,” Steel Res. Int., vol. 85, no. 3, pp. 486-497, Mar. 2014.
[21] R. M. German and A. Bose, Injection molding of metals and ceramics. Metal Powder Industries Federation, 1997.
[22] K. C. Labropoulos, S. Rangarajan, D. E. Niesz, and S. C. Danforth, “Dynamic Rheology of Agar Gel Based Aqueous Binders,” J. Am. Ceram. Soc., vol. 84, no.6, pp. 1217-1224, Dec. 2004.
[23] M. Sekido, H. Nakayama, M. Nukaya, and Y. Noro, “Injection-molding of metal or ceramic powders,” 14-May-1992.
[24] M. S. Zedalis, B. C. Sherman, and J. C. LaSalle, “Process for debinding and sintering metal injection molded parts made with an aqueous binder,” 27-Aug- 1998.
[25] A. J. Fanelli and R. D. Silvers, “Process for injection molding ceramic composition employing an agaroid gell-forming material to add green strength to a preform,” 15-May-1986.
97
[26] G. Chen, P. Cao, G. Wen, N. Edmonds, and Y. Li, “Using an agar-based binder to produce porous NiTi alloys by metal injection moulding,” Intermetallics, vol. 37, pp. 92-99, Jun. 2013.
[27] ASTM International, “C373-18 Standard Test Methods for Determination of Water Absorption and Associated Properties by Vacuum Method for Pressed Ceramic Tiles and Glass Tiles and Boil Method for Extruded Ceramic Tiles and Non-tile Fired Ceramic Whiteware Products.” ASM International, West Conshohocken,PA, 2018.
[28] ASTM International, “C1161-18 Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature.” ASTM International, West Conshohocken, PA, 2018.
[29] N. Wang, X. Chen, Y. Li, Y. Liu, H. Zhang, and X. Wang, “Preparation and compressive performance of an A356 matrix syntactic foam,” Mater. Trans., vol. 59, no. 5, pp. 699-705, 2018.
[30] G. A. Rocha Rivero, B. F. Schultz, J. B. Ferguson, N. Gupta, and P. K. Rohatgi, “Compressive properties of Al-A206/SiC and Mg-AZ91/SiC syntactic foams,” J. Mater. Res., vol. 28, no. 17, pp. 2426-2435, 2013.
98
IV. INFLUENCE OF FILLER MATERIAL AND VOLUME FRACTION ON COPPER MATRIX SYNTACTIC FOAMS
Myranda Spratt, Joseph W. Newkirk
Materials Science and Engineering Department, Missouri University of Science andTechnology, Rolla, MO 65409
ABSTRACT
Syntactic foams are composite materials with hollow filler materials embedded in
the matrix. Metal matrix syntactic foams have lightweight structural applications due to
their high stiffness and strength to weight ratio. Two fillers were investigated in this
work, both with a copper matrix material. The two fillers were selected to be as different
as possible to contrast their influence on the properties of the syntactic foam. The foams
were made using injection molding with an aqueous binder. Porous silica syntactic foams
were made with 5, 10, and 15 volume percent filler, and Hollow alumina syntactic foams
were made with 10, 30, and 50 volume percent filler. Characterization of the samples
after processing found that both filler materials were compatible with the process and the
matrix material, though the porous silica powder out-performed the hollow alumina.
Mechanical testing included three-point bending and Charpy impact testing. The porous
silica improved the specific elastic properties of the copper more than the hollow alumina
particles. The alumina bubble at 10% filler maximized the impact energy absorption of
the samples tested.
99
1. INTRODUCTION
Syntactic foam materials combine two material design techniques - composites
and foams. Syntactic foams have no matrix material requirements, but the filler must be a
porosity-containing particle. The filler is usually a hollow particle, but it can also be
porous. The porous filler - usually a ceramic or glass - provides the joint benefits of
porosity and the filler material such as decreased density and increased stiffness. [1], [2]
If the filler is sufficiently bonded to the matrix, load partitioning between the matrix and
the filler will occur [3]. For that reason, MMSFs tend to perform better than metal foams
in that they have higher strength and stiffness to weight ratios [1], [4]-[8]. MMSFs are
also operational at higher temperatures than many composite materials, such as carbon
fiber composites [9]. The energy absorption capabilities of MMSF materials are quite
promising as well. The reinforcement of the porosity in the MMSF microstructure helps
to absorb energy during failure of these materials [10], [11].
Despite the properties that MMSFs show, they are not widely utilized. Syntactic
foams can be designed for specialty applications where complex behaviors are needed,
though for metal matrix syntactic foams (MMSFs) these applications are generally only
theoretical at this time. Proposed applications include lightweight armor [12], lightweight
electronic packaging (due to superior EM shielding) [13], biomedical implants (Ti-
matrix, primarily) [14], [15], and more. However, they are difficult to manufacture. The
hollow spheres used as the filler material have a necessarily low density in comparison to
the matrix material, often well below 1 g/cm3. They are therefore buoyant and float
during processing unless constrained or kept in motion. Methods of mitigating the
100
segregation issue include using binders as in injection molding [16], [17] or rapid casting
and solidification as in stir casting [11], [18], [19]. Mixing apparaus used to keep spheres
from floating can also fracture them, another potential problem. Material compatibility
with the matrix and the processing temperatures can also impede manufacturability
Often, the casting or sintering temperature of the matrix material will allow deleterious
reactions to occur to the filler. These can include the filler itself melting, corrosion of the
filler by the matrix material, or unwanted interfacial reactions [1], [4], [8], [11], [20]-[25]
Typical filler materials used in MMSF manufacturing are hollow glass spheres,
naturally occurring or by-product materials, ceramic hollow spheres, and expanded clay
particles [20], [26]-[32]. Some researchers also use metal hollow spheres [12], [33]. A
cost-benefit analysis of the more typical types of filler material was performed by
Szlancsik et al [34]. The glass spheres and naturally occurring or by-product materials
(which usually have a glassy component) are inexpensive. They are available in average
particle sizes of less than 100 microns, which can benefit may processes. However, they
tend to melt at temperatures low enough to preclude the use of high sintering or melting
temperature matrix materials such as steel [21], [35]-[37] Glasses also tend to rapidly
corrode in contact with molten metals such as aluminum [3], [31], [38]. Ceramic hollow
spheres are less likely to melt or otherwise react negatively with the matrix material.
However, the available ceramic spheres are usually above 100 microns in size, usually on
the order of millimeters [1], [34], [39]. So, the feasible choices in filler are limited to
those that will be compatible with the matrix material, the process, and will meet design
criteria.
101
Injection molding is one method of manufacturing metal matrix syntactic foams
[16], [30], [40]. This process, and powder metallurgy processes in general, are not often
utilized in the literature to make MMSF materials. Metal powders are expensive and lab-
scale injection molding machines are not widely available. There are a few major benefits
to this process, however, that merit its investigation. First, the composition of the
composite is entirely designable. Second, the binder will retard segregation of the filler
and matrix. Finally, sintering temperatures are lower than casting temperatures. High
melting temperature alloy matrix materials, or uncastable alloys, can be used with this
method. The downside is that sintering often takes much longer than the rapid cooling of
pressure infiltration or stir casting. In this method, the filler has to withstand high
temperature and densification pressures for a significant time. Many filler materials -
glass or glass-containing hollow spheres - tend to melt or diffuse into the matrix at the
required sintering temperatures. Refractory ceramic hollow particles that are available
will not typically melt at the required temperatures. However, these tend to be quite large,
which may make injection molding difficult.
In this work, two filler materials - one glass and one ceramic - were paired with a
copper matrix. The influence of the filler material on the manufacturability and properties
of the MMSF was analyzed. Low-pressure injection molding was used to manufacture
the samples. The two fillers were selected to be as different as possible - the size,
porosity distribution, and material of these fillers were on opposing extremes. The copper
matrix was selected as a high-density metal with a sintering temperature higher than
aluminum, but lower than the melting temperature of the glass filler. The high-density
metal contrasts with the low-density filler to allow small volume fractions of filler to
have a significant influence on the composite density.
102
2. EXPERIMENTAL PROCEDURE
In this experiment, copper matrix syntactic foams were prepared from powder
using two different filler materials. The gas atomized pure copper powder was
manufactured by Royal Metal Powders Inc. No sintering aids were added to this powder.
Porous silica (P-S) powder from Mo-Sci Corp was one filler material. This powder was
spherical with 45% porosity, per the manufacturer. The porosity in this powder was
distributed throughout each particle. Alumina bubble (AB) from Washington Mills was
the second filler. These particles were hollow, as is typical of syntactic foam filler
material. The density and average cell wall size of the alumina bubble particles was not
given by the manufacturer. The density was measured directly by mounting AB powder
in epoxy and grinding the spheres to approximately the center of the average particle size.
ImageJ software was used to measure the cell wall thickness and particle size of the AB
particles from SEM images. This data was then used to calculate the average density of
the AB powder. Powder characterization also included particle size distribution via
scanning electron microscope (SEM) for each material. The powders were imaged, and
then automated software was used to determine the diameter of each particle in each
image for up to 10,000 particles per powder. The powder was prepared for imaging by
brushing a small (<1g) sample of the powder over a carbon dot. The non-conductive
powders were then sputter-coated with palladium.
103
Seven syntactic foam compositions were injection molded in a Peltsman MIGL-
28 low-pressure injection molding machine. The volume percent of filler and copper of
each sample as input into the machine is listed in Table 1, as is the solids loading of each
sample. The MIGL-28 uses a reservoir with an attached paddle mixer to combine the
binder and powders. Gas pressure is applied to the reservoir which pushes the material
through a thin tube, through an orifice, and into the mold. The reservoir, tube, and orifice
are heated individually. In this experiment, the temperature was held at 85 °C for each
sample. The gas pressure was held for 30 seconds per specimen, and the specimens were
demolded after 1 to 2 minutes. A water-cooled 10 x 6.5 x 60 cm rectangular mold was
used.
Table 1: The solids loading, filler, and copper volume percent of each sample as inputinto the injection molding machine.
Sam ple % Solids Loading % Filler % Copper0% Filler 55% 0% 100%5% P-S 51% 5% 95%10% P-S 44% 10% 90%15% P-S 44% 15% 85%10% AB 54% 12% 88%30% AB 53% 24% 76%50% AB 56% 55% 45%
A water-based gel binder was used in this experiment. The composition was 7%
agar, 4% glycerin, and 89% water. The solids loading of each composition started at
60%, and was adjusted as necessary to achieve an injection moldable viscosity. The
slurry was mixed for 1 hour after the last adjustment. Thirty rectangular specimens were
injection molded from each sample. Each specimen was labeled in the order they were
injection molded.
Post-processing binder burnout and sintering steps occurred for each sample in
separate runs with the same equipment. Each specimen was measured for mass and
volume during each post-processing stage. The mass was measured using a laboratory
balance, and the volume was measured geometrically using calipers. The stages were
‘wet’, ‘dry’, and ‘sintered’. The wet stage occurred just after injection molding. In the
‘dry’ stage, the samples were placed in a 120 °C furnace for 12 hours to remove water
from the binder. The remaining binder was removed in a binder burnout step before the
sintering procedure. Binder burnout occurred at 450 °C for 1 hour under flowing air. The
samples were then ramped at 2°C/min to 1000 °C under flowing argon. They were held at
the sintering temperature for 10 hours.
The samples were characterized before mechanical testing. The extent of binder
segregation over time during injection molding was determined by calculating the mass
loss of each specimen from the wet to the sintered stage. Equation 1 was used to
determine this value for each specimen. A regression analysis was then used to determine
if the binder separated from the solids. Segregation between the filler and copper powder
was done by a regression analysis of the sintered relative density versus time. The
relative density was found by Equations 2 and 3. Microstructural analysis of the sintered
samples was done with extra specimens which were injection molded after the thirty test
specimens were made. The specimens were mounted in bakelite and ground down to the
midpoint of the width of the specimens. EDS mapping was done for the copper-porosity,
copper-silica, and copper-alumina interfaces to determine if interfacial reactions or
104
105
detectable diffusion occurred between the materials. The particle size of the porous silica
particles in the copper matrix was measured using the same automated feature analysis
used to characterize the powder.
Mioss (Mwet ^sintered)/'^wet Equation 1
where MioSS is the percent mass lost after sintering, Mwet is the mass at the wet stage, and
Msintered is the mass at the sintered stage of each specimen.
Prelative Pspecimen' Ptheoretical Equation 2
where preiative is the relative density, pspecimen is the specimen density, and ptheoreticai is the
theoretical density calculated from Equation 3.
Pstage 2 Pcomponents * V^components Equation 3
where the subscript stage indicates the wet, dry, or sintered stage of processing, the
subscript components indicates the components present in the stage of processing, and
V% indicates volume percent.
Two mechanical tests were performed on the samples. The specimens in each
sample were split into two sets of fifteen randomly. The dimensions of the samples were,
on average, 54 ± 2.7 x 8.6 ± 0.49 x 6.1 ± 0.31 mm. One set of fifteen for each sample was
tested by three-point bending. ASTM C1161 was used as a guide. The samples were
tested without machining. A fully-articulating fixture was used in this test, which was
required due to slight warpage in some of the specimens. Test configuration B was used
with a support span of 40 mm and a load span of 20 mm. A strain rate of 0.5 mm/min was
utilized for each test. The yield strength and modulus of each sample were determined by
visual basic for excel (VBA) code. The second set of fifteen specimens for each sample
was tested via notched Charpy impact testing. ASTM E23 was used as a guide. The
samples were not machined before use and did not comply with the standard size for a
Charpy specimen. They therefore cannot be truly compared to other Charpy impact tests
beyond this paper. The notch was the standard V-notch. Due to the nature of the
composite samples, the surface finish standard was not followed. The filler particle size
was larger than the surface finish requirements in both cases. Absorbed energy was
recorded for each specimen from the machine’s digital readout.
3. RESULTS AND DISCUSSION
3.1. POWDER CHARACTERIZATION
The particle size distribution of each powder is shown in Figure 1. Each bin of the
histogram counts the number of particles greater than the label and less than the next bin
size. Figure 1.a is the copper powder size distribution. The copper had an average particle
size of 7.0 ± 4.3 microns and a median particle size of 6.0 microns. Figure 1.b is the
alumina bubble histogram. The AB powder had an average particle size of 343.1 ± 225.0
microns and a median particle size of 278.9 microns. There were significant fragments of
spheres in this powder, which contributed to the data’s skew to the left. The powder had
the widest distribution of the three powders, and it had the largest average size. Figure 1.c
shows the porous silica powder size distribution. The average particle size of the P-S
powder was 12.0 ± 6.0 microns, and the median size was 11.1 microns. This powder was
approximately double the copper powder in size, and the distributions had a similar
shape. Figure 2 shows SEM images of each powder. All the powders were spherical. On
close inspection, the AB particles appear to be comprised of fine alumina particles, as
106
107
15% x
| 10% <u
fin
5%OO
0%0 1 2 3 4 5 6 7 8 9 1011 1213 1415
o o o o o o o o o o o o o o o o
15%
10%<ufin
5%o
0%
Particle Diameter (^m)c r
—
1 / 1 -i L M lJ l + _ L + _i_
100%80%
60%40%20%
0%
c<uo!-h<UPin<U_>
100% 80%
60%
40% _ 20% |
0% 5
C(L)O!-h(L)Pin(L)_>33ao
c(L)O!-h<u
Pin(L)_>33ao
0 2 4 6 8 1012141618202224262830
CountPercent
^ — CumulativePercent
Particle Diameter (^m)Figure 1: The particle size distribution by count percent and cumulative count percent for
each powder where a) is copper, b) is alumina bubble, and c) is porous silica.
shown in Figure 2.d. The fine alumina particles were not fully densified with significant
porosity in the cell walls. Figure 3 is one of the SEM images used to measure the average
diameter and wall thickness of the AB particles. On average, the density was 1.36 ± 0.39
g/cm3. The large standard deviation appears to arise from the fact that the wall thickness
was relatively constant (33 ± 14 |im) regardless of the particle size. Further, the density
108
100 urn100 um
1000 200um um■
Figure 2: SEM images of the three powders - a) copper, b) silica, c) and d) alumina - show the morphology and relative size of the powders.
calculation assumed that the cell walls were fully dense alumina, which was not the case.
The 1.36 g/cm3 value for density was used to determine filler volume percent for the AB
syntactic foam specimens.
3.2. SAMPLE CHARACTERIZATION
Table 2 shows the wet and sintered density and relative density of each sample.
The standard deviation for each average is included. The wet density was higher than
expected overall, which indicates that that expected amount of binder and/or filler was
109
Figure 3: The cross-sections of AB powder particles were used to determine the densityof the AB powder.
not present in the part or that the filler density was incorrect. Both options likely occurred
to some degree. The expected filler density could have changed if particles filled with
binder or copper particles during mixing or if the AB powder contained fragments. The
P-S particles, not being hollow, would retain their density upon fragmenting unless all
pores were exposed. The AB particles are hollow thin-walled porous shells made from
very fine alumina particles. The average size of the pores in the cell wall was 6 ± 7
microns, which is approximately the same size as the copper powder. So, copper particles
can also easily enter the AB particles, as could the binder. The porous silica could have
absorbed some binder as well from surface connected porosity, increasing their effective
density. The amount of binder in the samples was measurable quantity that was corrected
in the theoretical density calculation in Table 3. This was done by comparing the mass
after the wet stage and after the sintering stage. Then the actual volume percent of the
binder was found and used to calculate a new theoretical wet density for each sample.
110
This was used to find a new relative wet density. The samples are closer to 100% overall
with this correction. The remaining error is likely attributable to the filler density. In this
case, it seems the porous silica was less suceptable to filler density error than the alumina
bubble. It likely has less fragments and did not fill with copper or binder to the extent of
the AB particles.
Table 2: The density and relative density for each stage and sample are shown with theirstandard deviation.
Sam ple Wet SinteredDensity(g/cm3)
RelativeDensity
Density(g/cm3)
RelativeDensity
0%Filler
5.84 ± 0.07 109% ± 1.28%
7.44 ± 0.12 83% ± 1.29%
5% P-S 5.42 ± 0.04 111% ± 0.87%
7.49 ± 0.15 87% ± 1.77%
10% PS
4.36 ± 0.05 104% ± 1.22%
6.59 ± 0.26 80% ± 3.14%
15% PS
4.64 ± 0.04 115% ± 0.9%
6.44 ± 0.21 82% ± 2.71%
10%AB
5.7 ± 0.36 119% ± 7.49%
7.4 ± 0.27 92% ± 3.32%
30%AB
5.39 ± 0.13 127% ± 3.09%
6.79 ± 0.13 95% ± 1.85%
50%AB
4.32 ± 0.32 138% ± 10.08%
4.26 ± 0.2 89% ± 4.14%
Segregation during injection molding could occur over time and it could be a
relatively steady-state separation. Both the binder and the filler could separate from the
matrix during mixing. Table 3 shows the steady-state separation of the binder and solid
material for each sample, as calculated in the previous analysis. The amount of binder
that was in each sample, on average, after injection molding was less than the amount of
111
binder that was expected for all samples. A regression analysis on each sample found that
over time binder segregation occurred significantly in the samples with alumina bubble
particles. The effect size of this increase in binder amount over time was small overall.
The other samples did not show significant evidence of segregation of the binder over
time. Segregation of the filler was measured from the density of the sintered specimens
over injection molding timeThis analysis found no evidence of significant density
changes over time in all samples, indicating that the filler amount was constant. The
steady-state separation of the filler could not be measured, as there were too many
unknowns to calculate the actual volume percent of filler in each specimen.
Table 3: The volume percent of binder in each sample was used to calculate a more accurate relative density for each sample.
Input Volume % Actual Volume % New Wet Relative Binder Binder Density
0% Filler 45% 39% 100%5% P-S 49% 41% 100%10% P-S 56% 51% 96%15% P-S 56% 51% 106%10% AB 46% 32% 98%30% AB 47% 36% 110%50% AB 44% 32% 120%
The remaining filler porosity after sintering is an important metric to judge filler
quality. To measure this for the porous silica, the average size of the silica before
sintering and after was compared. The size of the particles decreased by 33%, on average.
The expected porosity in the PS particles was 45%, so 12% porosity remained in the
silica on average. Some silica particles partially melted and wicked into nearby matrix
porosity, which caused the noted decrease in porosity. The alumina bubble porosity was
determined from SEM images, one of which is shown in Figure 4. The AB particles were
filled with copper particles, likely during injection molding. On average, the AB particles
contained 35% porosity, 34% alumina, and 30% copper. As the image shows, however,
some of the AB particles were filled with copper while others contained no copper
particles. Fragments of AB particles were also present in the microstructure. The AB
particles, in general, did not collapse and retained more porosity than the porous silica
after sintering. However, the AB particles were observed to have many fragmented
particles, while the porous silica did not.
112
1000 [im
Figure 4: Some of the AB particles in the copper matrix filled with copper powder.
In Table 2, the relative density of the samples after sintering is low, between 82%
and 95%. Matrix porosity is apparent in the microstructure, confirming that the matrix
did not densify sufficiently. EDS mapping of a 0% filler specimen, shown in Figure 5,
113
shows the likely reason for the high amount of porosity. Oxygen was found surrounding
many of the pores in the specimen. This corrosion could have occurred during the water-
removal step or the binder-burnout step of the drying stage. In either case, the copper
oxide layer prevented densification of the matrix during sintering. The matrix of the
syntactic foam made in this experiment is functionally a low porosity copper foam rather
than a dense matrix. Further, Figure 5 shows the copper oxide coated some of the internal
pores in a pseudo-syntactic foam manner, creating almost in-situ ceramic hollow spheres.
The exact amount of copper oxide present in the samples is unknown, but it is not an
insignificant amount. This behavior could be further investigated in systems where the
oxide of the metal matrix is stiff and strong, such as alumina, as a method of creating
metal matrix syntactic foams without the need for added hollow spheres. For example,
Yu et. al. [41] investigated this method of making metal matrix syntactic foams with AlN
as the in-situ filler.
The filler-copper interfaces were EDS mapped in Figures 6 and 7. Figure 6 shows
the silica-copper microstructure. The EDS did not detect noticeable silica diffusion in the
copper matrix. However, the silica particles did appear to be surrounded by porosity,
possibly from the glass shrinkage that occurred. The particles in this microstructure
image did not fall out of the matrix during grinding and polishing, however, so they were
not loose. Figure 7 shows the alumina-copper interface. There was also no evidence of
alumina diffusion into the copper matrix.
114
© #
Figure 5: Elemental mapping of a 100% copper specimen shows the distribution ofcopper and oxygen in the sample.
3.3. MECHANICAL TESTING
One purpose of metal matrix composites to which they are particularly suited, at
least theoretically, is lightweight structural materials. To that purpose, the elastic
properties of the composite are of particular interest. The specific flexural modulus and
specific 0.2% offset yield strength are shown in Figure 8. The porous silica filler
substantially improved the specific yield strength of the copper matrix. The alumina
bubble filler did not improve the specific yield strength, but they also did not negatively
115
Figure 6: EDS map of the elements present in a P-S and copper specimen; elements areindicated on the figure.
affect the specific yield strength, except for the 50% AB sample. The specific modulus
was neutrally or negatively affected by the filler. The AB samples decreased in specific
modulus as the filler amount increased. The P-S samples did not significantly change in
specific modulus as the volume percent filler increased.
There are models in the literature derived for foams that can be used to investigate
the modulus property trends further. Zhu et al [42] developed a model for open-celled
foams (Equation 1).
116
Figure 7: EDS map of the elements present in an AB and copper specimen; elements areindicated on the figure.
EC = ES * ( 1009*p ) Equation 1^1 + 1.514 * p '
where Ec, Es, and p represent composite foam modulus, solid modulus, and relative
density, respectively. Because the copper matrix includes significant amounts of copper
oxide, the modulus of the copper matrix could not be taken from literature. The
experimentally determined modulus for the 100% copper sample was assumed to be
equal to Ec. Es was then calculated using the relative density of the sample. The Es of the
117
54.5
m
4o ^3.5
f t3
2.52 ± 2s
1.5s 1 1f t
0.5 4 0
i
f
T
KTO Specific Modulus
• Specific Yield
1 1
8.0
7.0
6.0bD
5.0S
4.0
3.0 o2.0
f tGO
1.0
0.0100% Cu 5% P-S 10% P-S 15% P-S 10% AB 30% AB 50% AB
Figure 8: The specific modulus and 0.2% offset yield strength for each sample, note that modulus is in GPa and yield strength is in MPa.
was found to be 53.7 GPa, which is approximately half of the expected tensile modulus
for pure copper.
The solid modulus for the composite samples was calculated using two methods.
For the first method, rule of mixtures was used to calculate the modulus from the solid
components of the composite. The alumina was assumed to have a modulus of 375 GPa,
which is the expected modulus of pure alumina. The silica was assumed to have a
modulus of 73 GPa, which is the modulus of dense SiO2 glass. In this method, the
theoretical density of the composite was used to calculate the relative density of the
material. This method of calculating the modulus assumes that the filler material assumed
some of the load, though it does not necessarily account for the internal nature of the
filler porosity. The second method used to calculate Ec was to assume that Es is equal to
the copper matrix alone. In this method, the theoretical density used to calculate the
118
% Porous Silica
Figure 9: The specific modulus of the porous silica and alumina bubble samples contrasted with a model for metal foams by Zhu et al [42].
relative density for each sample assumed the filler material had a density of zero. Figure
9 shows the results of the models when compared to the porous silica and alumina bubble
samples. The porous silica samples remained between the models. This indicates that
load partitioning to the silica did occur, though not as much as could have been. This is
likely due to the porosity surrounding the silica particles in the matrix. The alumina
119
bubble samples match reasonably with the model where Es was assumed to be equal to
the copper matrix. This indicates that little to no load transfer occurred in these samples
during the elastic part of the test.
The results of the notched Charpy impact testing can be seen in Figure 10. The
porous silica particles negatively impacted absorbed energy. The 10% AB sample
improved the absorbed energy and the specific absorbed energy significantly. The 30%
AB sample improved specific absorbed energy, and the absorbed energy was similar to
the 0% filler sample. The 50% AB sample had the lowest absorbed energy of any sample.
These results indicate that the alumina bubble particles imparted some impact resistance
to the copper matrix at low volume fraction AB. When the AB particles began to form a
continuous network (beyond 30% AB), the impact resistance decreased.
4. CONCLUSIONS
Two filler materials were used to create two sets of copper matrix syntactic
foams. The volume percent of each filler increased over three increments - 5, 10, and
15% for the porous silica filler and 10, 30, and 50% for the alumina bubble filler. A
control sample with no filler was also made. The two fillers were different in size,
material, and porosity distribution. The copper powder had an average particle size of 7.0
± 4.3 microns. The AB powder had an average particle size of 343.1 ± 225.0 microns.
Finally, the porous silica powder had an average particle size of 12.0 ± 6.0
microns. The alumina bubble cell walls were not dense, and the cell wall porosity
allowed copper particles to enter the internal porosity. This resulted in an after-sintering
120
% Porous Silica
Figure 10: The absorbed energy of each sample
inter-particle porosity of approximately 34%, down from 66%. The porous silica partially
melted and wicked into nearby porosity. This resulted in an inter-particle porosity of
approximately 12%, down from 45%. All the samples contained significant matrix
porosity with a copper oxide coating the pore surfaces. The flexural yield strength and
modulus and the Charpy impact energy were tested for all seven samples. The porous
silica increased the specific yield strength of the syntactic foam as more filler was added
to the matrix. The sample yield strength increased from 4.9 ± 0.44 MPa/g*cm-3 at 0%
filler to 7.0 ± 0.43 MPa/g*cm-3 at 15% P-S. The alumina bubble samples’ specific yield
121
did not depend on the filler amount. The specific moduli of the porous silica samples also
did not appear to significantly depend on the filler amount. The specific modulus of the
alumina bubble samples decreased with filler amount. The moduli were compared to an
open-cell foam model, using two methods of calculating the moduli. The porous silica
particles did appear to take on some of the load, according to the model. The impact
testing results showed that of the samples tested, the absorbed energy maximized at 10%
AB. The porous silica particles negatively affected absorbed energy.
ACKNOWLEDGEMENTS
This research did not receive any specific grant from funding agencies in the
public, commercial, or not-for-profit sectors.
REFERENCES
[1] P. K. Rohatgi, N. Gupta, B. F. Schultz, and D. D. Luong, “The Synthesis, Compressive Properties, and Applications of Metal Matrix Syntactic Foams,”JOM, vol. 63, no. 2, pp. 36-42, 2011.
[2] M. Labella, V. C. Shunmugasamy, O. M. Strbik, and N. Gupta, “Compressive and thermal characterization of syntactic foams containing hollow silicon carbide particles with porous shell,” J. Appl. Polym. Sci., vol. 131, no. 17, p. n/a-n/a, 2014.
[3] D. K. Balch and D. C. Dunand, “Load partitioning in aluminum syntactic foams containing ceramic microspheres,” Acta Mater., vol. 54, no. 6, pp. 1501-1511,2006.
[4] G. Castro and S. R. Nutt, “Synthesis of syntactic steel foam using gravity-fed infiltration,” Mater. Sci. Eng. A, vol. 553, pp. 89-95, 2012.
122
[5] L. P. Zhang and Y. Y. Zhao, “Mechanical Response of Al Matrix Syntactic Foams Produced by Pressure Infiltration Casting,” J. Compos. Mater., vol. 41, no. 41, p. 17, 2007.
[6] P. K. Rohatgi, J. K. Kim, N. Gupta, S. Alaraj, and A. Daoud, “Compressive characteristics of A356/fly ash cenosphere composites synthesized by pressure infiltration technique,” Compos. Part A Appl. Sci. Manuf., vol. 37, no. 3, pp. 430437, 2006.
[7] D. P. Mondal, M. D. Goel, and S. Das, “Compressive deformation and energy absorption characteristics of closed cell aluminum-fly ash particle composite foam,” Mater. Sci. Eng. A, vol. 507, no. 1-2, pp. 102-109, 2009.
[8] G. A. Rocha Rivero, B. F. Schultz, J. B. Ferguson, N. Gupta, and P. K. Rohatgi, “Compressive properties of Al-A206/SiC and Mg-AZ91/SiC syntactic foams,” J. Mater. Res., vol. 28, no. 17, pp. 2426-2435, 2013.
[9] D. P. Mondal, N. Jha, A. Badkul, S. Das, and R. Khedle, “High temperature compressive deformation behaviour of aluminum syntactic foam,” Mater. Sci. Eng. A, vol. 534, pp. 521-529, Feb. 2012.
[10] L. C. Zou, Q. Zhang, B. J. Pang, G. H. Wu, L. T. Jiang, and H. Su, “Dynamic compressive behavior of aluminum matrix syntactic foam and its multilayer structure,” Mater. Des., vol. 45, 2013.
[11] A. Daoud, M. T. A. El-Khair, M. Abdel-Aziz, and P. Rohatgi, “Fabrication, microstructure and compressive behavior of ZC63 Mg-microballoon foam composites,” Compos. Part A Appl. Sci. Manuf., vol. 67, no. 9, pp. 1842-1853,2007.
[12] M. Garcia-Avila, M. Portanova, and A. Rabiei, “Ballistic performance of composite metal foams,” Compos. Struct., vol. 125, pp. 202-211, Jul. 2015.
[13] X.-F. X. Meng, D.-H. D. Li, X. X.-Q. Shen, and W. Liu, “Preparation and magnetic properties of nano-Ni coated cenosphere composites,” Appl. Surf. Sci., vol. 256, no. 12, pp. 3753-3756, Apr. 2010.
[14] X. Xue and Y. Zhao, “Ti matrix syntactic foam fabricated by powder metallurgy: Particle breakage and elastic modulus,” JOM, vol. 63, no. 2, pp. 36-40, 2011.
[15] X. B. Xue, Y. Y. Zhao, V. Kearns, and R. L. Williams, “Mechanical and Biological Properties of Titanium Syntactic Foams,” in TMS, 2010, vol. 2, pp. 129-136.
123
[16] J. Yu, M. Rombouts, and G. Maes, “Cracking behavior and mechanical properties of austenitic stainless steel parts produced by laser metal deposition,” Mater. Des., vol. 45, pp. 228-235, Mar. 2013.
[17] L. Peroni et al., “Investigation of the mechanical behaviour of AISI 316L stainless steel syntactic foams at different strain-rates,” Compos. Part B Eng., vol. 66, no. 0, pp. 430-442, 2014.
[18] D. P. Mondal, S. Das, N. Ramakrishnan, and K. U. Bhasker, “Cenosphere filled aluminum syntactic foam made through stir-casting technique,” Compos. Part A Appl. Sci. Manuf., vol. 40, no. 3, pp. 279-288, 2009.
[19] P. K. Rohatgi, D. Weiss, and N. Gupta, “Applications of Fly Ash in Synthesizing Low-Cost MMCs for Automotive and Other Applications,” JOM, vol. 58, no. 11, pp. 71-76, 2006.
[20] Y. Lin, Q. Zhang, and G. Wu, “Interfacial microstructure and compressive properties of Al-Mg syntactic foam reinforced with glass cenospheres,” J. Alloys Compd., vol. 655, pp. 301-308, 2016.
[21] J. Weise, J. Baumeister, O. Yezerska, N. Salk, and G. B. D. Silva, “Syntactic iron foams with integrated microglass bubbles produced by means of metal powder injection moulding,” Adv. Eng. Mater., vol. 12, no. 7, pp. 604-608, Apr. 2010.
[22] D. Lehmhus et al., “Quasi-static and Dynamic Mechanical Performance of Glass Microsphere- and Cenosphere-based 316L Syntactic Foams,” Procedia Mater. Sci., vol. 4, no. 0, pp. 383-387, 2014.
[23] R. A. Palmer, K. Gao, T. M. Doan, L. Green, and G. Cavallaro, “Pressure infiltrated syntactic foams—Process development and mechanical properties,” Mater. Sci. Eng. A, vol. 464, no. 1-2, pp. 85-92, 2007.
[24] J. A. Santa Maria, B. F. Schultz, J. B. Ferguson, and P. K. Rohatgi, “Al-Al2O3 syntactic foams - Part I: Effect of matrix strength and hollow sphere size on the quasi-static properties of Al-A206/Al2O3 syntactic foams,” Mater. Sci. Eng. A, vol. 582, no. 0, pp. 415-422, 2013.
[25] I. N. Orbulov and J. Dobranszky, “Producing metal matrix syntactic foams by pressure infiltration,” Period. Polytech. Mech. Eng., vol. 52, no. 1, pp. 35-42,2008.
[26] B. G. Kutchko and A. G. Kim, “Fly ash characterization by SEM-EDS,” Fuel, vol. 85, no. 17-18, pp. 2537-2544, 2006.
124
[27] Q. Zhang and G. Wu, “Three dimensional microstructure and image-based simulation of a fly ash/al syntactic foam using X-ray micro-computed tomography,” J. Wuhan Univ. Technol. Mater. Sci. Ed., vol. 28, no. 1, pp. 99-103, 2013.
[28] A. Daoud, M. T. A. El-Khair, A. Y. Shenouda, E. Mohammed, and P. K. Rohatgi, “Microstructure, tensile properties and electrochemical behavior of Pb alloy-45 vol.% fly ash microballoon composites,” Mater. Sci. Eng. A, vol. 526, no. 1-2, pp. 225-234, 2009.
[29] D. P. Mondal et al., “Titanium-cenosphere syntactic foam made through powder metallurgy route,” Mater. Des., vol. 34, pp. 82-89, Feb. 2012.
[30] J. Weise, J. Baumeister, O. Yezerska, N. Salk, and G. B. D. Silva, “Syntactic Iron Foams with Integrated Microglass Bubbles Produced by Means of Metal Powder Injection Moulding,” Adv. Eng. Mater., vol. 12, no. 7, pp. 604-608, 2010.
[31] D. D. Luong, V. C. Shunmugasamy, N. Gupta, D. Lehmhus, J. Weise, and J. Baumeister, “Quasi-static and high strain rates compressive response of iron and Invar matrix syntactic foams,” Mater. Des., vol. 66, pp. 516-531, Feb. 2015.
[32] D. Lehmhus et al., “Quasi-static and Dynamic Mechanical Performance of Glass Microsphere- and Cenosphere-based 316L Syntactic Foams,” Procedia Mater. Sci., vol. 4, pp. 383-387, Jan. 2014.
[33] A. Rabiei and L. J. Vendra, “A comparison of composite metal foam’s properties and other comparable metal foams,” Mater. Lett., vol. 63, no. 5, pp. 533-536, Feb.2009.
[34] A. Szlancsik, B. Katona, A. Kemeny, and D. Karoly, “On the Filler Materials of Metal Matrix Syntactic Foams,” Materials (Basel)., vol. 12, no. 12, p. 2023, Jun. 2019.
[35] L. Peroni, M. Scapin, M. Avalle, J. Weise, and D. Lehmhus, “Dynamic mechanical behavior of syntactic iron foams with glass microspheres,” Mater. Sci. Eng. A, vol. 552, pp. 364-375, Aug. 2012.
[36] L. Peroni et al., “High Strain Rate Tensile and Compressive Testing and Performance of Mesoporous Invar (FeNi36) Matrix Syntactic Foams Produced by Feedstock Extrusion,” Adv. Eng. Mater., Nov. 2016.
[37] D. Lehmhus et al., “Quasi-static and Dynamic Mechanical Performance of Glass Microsphere- and Cenosphere-based 316L Syntactic Foams,” Procedia Mater. Sci., vol. 4, pp. 383-387, Jan. 2014.
125
[38] M. Spratt, J. W. Newkirk, and K. Chandrashekhara, “Aluminum Matrix Syntactic Foam Fabricated with Additive Manufacturing,” in Solid State Symposium, 2017.
[39] L. Licitra, D. D. Luong, O. M. Strbik Iii, and N. Gupta, “Dynamic properties of alumina hollow particle filled aluminum alloy A356 matrix syntactic foams,” Mater. Des., vol. 66, Part B, no. 0, pp. 504-515, 2015.
[40] J. Weise, D. Lehmhus, J. Baumeister, R. Kun, M. Bayoumi, and M. Busse, “Production and Properties of 316 L Stainless Steel Cellular Materials and Syntactic Foams.,” Steel Res. Int., vol. 85, no. 3, pp. 486-497, Mar. 2014.
[41] P. Yu and M. Qian, “Metal injection moulding of in-situ formed AlN hollow sphere reinforced Al matrix syntactic foam parts,” Mater. Chem. Phys., vol. 137, no. 2, pp. 435-438, Dec. 2012.
[42] H. X. Zhu, J. F. Knott, and N. J. Mills, “Analysis of the elastic properties of opencell foams with tetrakaidecahedral cells,” J. Mech. Phys. Solids, vol. 45, no. 3, pp. 319-343, Mar. 1997.
126
SECTION
2. CONCLUSIONS AND FUTURE WORK
In this dissertation, lightweight structures were designed, fabricated,
characterized, and their mechanical properties evaluated. A honeycomb lattice structure
and copper matrix syntactic foams were focused on in this work. The honeycomb lattice
structure is a 2D periodic cellular structure. This structure was manufactured using
powder bed selective laser melting from 304L, an austenitic stainless steel. The copper
matrix syntactic foams were manufacturing using an injection molding process. A water-
based binder was optimized for this process. Two fillers were compared to each other, a
porous silica filler and an alumina bubble filler.
The honeycomb lattice structure was evaluated in the as-built condition. Four
honeycomb samples were printed, with the wall thickness increasing from 0.2 to 0.5 mm
in 0.1 mm increments. The processing parameters used in this experiment were optimized
already to reduce porosity in the material but not for thin features. The quality of the
samples after building was analyzed by determining the extent of dimensional mismatch.
The dimensional mismatch was caused primarily by particles welded to the surfaces of
the honeycomb. The effective wall thickness of the honeycombs was increased by these
particles. The wall thickness of the honeycomb was shown to significantly affect the
strength of the material. This work was done to support model validation work by
Anandan et al [30] and Hussein et al [85].
Part of the model validation work involved finding the solid compression
properties of SLM 304L steel. During compression testing of the material, anisotropic
plastic deformation behavior was observed. Research into this phenomenon led to a
microstructure feature that has previously received little attention from the literature. This
feature, the melt pool boundary, is unique to additively manufactured material. It forms a
continuous network throughout the as-built material. This network was modeled using 3D
software to better visualize the network and calculate the concentration of the melt pool
boundary network (MPBN) in various orientations through the material. The
concentration of the MPBN was correlated to yield strength data in various build
orientations. Overall, the yield strength decreased linearly with increasing MPBN
concentration. The anisotropic plastic deformation was also linked to the concentration of
the MPBN. Directions that deformed more strongly than other directions under the same
load had higher MPBN concentration.
The lattice structure work could be extended in several ways. Property predication
via finite element analysis of the honeycomb structure would be improved by using the
effective wall thickness or by adding a variable wall thickness to the model. The
dimensional analysis work could be the basis for this improvement. This work also
highlights an important way of improving dimensional accuracy for thin-walled
structures. For thin features, the melt pool width and overlap set the strut size or wall
thickness that can be printed. This limit can be applied to future lattice structure unit
cells. The melt pool boundary network analysis can be extended to other materials and
scan pattern to see if the correlations found here hold true.
127
128
Copper matrix syntactic foams were binder injected molded with a custom binder.
The binder was a water-based gel. The binder composition was optimized using a 3-
component mixture model for maximum relative density. The optimized composition was
7% agar, 4% glycerin, and 89% water. Three samples were made with 5, 10 and 15
volume percent porous silica filler. A 0% filler sample was also made. The copper matrix
syntactic foams did not show significant evidence of segregation during injection
molding, and the binder was removed with little remaining carbon (0.11% ± 0.08%). The
matrix porosity in the samples was high (>10%) due to oxidation of the copper at the pore
surfaces. The mechanical properties of the syntactic foams increased with decreasing
density, which is ideal for these materials.
The final study compared the porous silica filler to an alumina bubble filler. The
two fillers were selected to be extremely different so as to contrast the influence they had
on the foam properties. The alumina bubble powder had an average particle size over 35
times larger than the copper powder, and the silica powder was double the size of the
copper. The porous silica spheres had distributed porosity, while the alumina particles
were thin-walled shells with porous walls. The porous shell walls allowed copper powder
to enter some alumina particles, reducing internal porosity. The copper had a sintering
temperature lower than the melting temperature of both materials. The silica did soften
during sintering, which partially collapsed some of its porosity. The alumina did not
noticeably react with the copper during sintering. The copper-alumina bubble syntactic
foams improved the impact resistance over the copper-only sample at 10% alumina
bubble. The porous silica samples did not improve impact resistance. The specific yield
strength of the porous silica increased with decreasing density. The alumina bubble’s
specific yield strength stayed relatively constant with decreasing density.
The metal matrix syntactic foam development started in this dissertation could be
extended in several ways. The binder that was developed could be used in other metal-
filler systems, in particular. This work could also be scaled-up to industrial-sized
injection molding machines. During this experiment, pores in the matrix material were
found to have an oxide coating which impeded densification. This could be a method of
making in-situ metal matrix composites. Future work could include using this binder
system to make in-situ MMSF materials with a metal matrix that has a high strength
oxide material, such as aluminum. The impact of the filler materials on different matrix
materials could also be explored.
129
130
BIBLIOGRAPHY
[1] M. F. Ashby, “The properties of foams and lattices.,” Philos. Trans. A. Math. Phys. Eng. Sci., vol. 364, no. 1838, pp. 15-30, Jan. 2006.
[2] L. Yang, D. Cormier, H. West, O. Harrysson, and K. Knowlson, “Non-stochastic Ti-6Al-4V foam structures with negative Poisson’s ratio,” Mater. Sci. Eng. A, vol. 558, no. 0, pp. 579-585, 2012.
[3] J. Banhart, “Light-Metal Foams-History of Innovation and Technological Challenges,” Adv. Eng. Mater., vol. 15, no. 3, pp. 82-111, Mar. 2013.
[4] J. Banhart, “Manufacture, characterisation and application of cellular metals and metal foams,” Prog. Mater. Sci., vol. 46, no. 6, pp. 559-632, 2001.
[5] N. Gupta and P. K. Rohatgi, Eds., Metal Matrix Syntactic Foams: Processing, Microstructure, Properties and Applications. DEStech Publications, Inc, 2014.
[6] M. F. Ashby, Materials Selection in Mechanical Design Third Edition. 2005.
[7] M. F. Ashby, “Material Property Charts,” in Materials Selection in Mechanical Design, Elsevier, 2011, pp. 57-96.
[8] W. S. Sanders and L. J. Gibson, “Mechanics of hollow sphere foams,” Mater. Sci. Eng. A, vol. 347, no. 1-2, pp. 70-85, 2003.
[9] P. K. Rohatgi, N. Gupta, B. F. Schultz, and D. D. Luong, “The Synthesis, Compressive Properties, and Applications of Metal Matrix Syntactic Foams,”JOM, vol. 63, no. 2, pp. 36-42, 2011.
[10] M. G. Rashed, M. Ashraf, R. A. W. Mines, and P. J. Hazell, “Metallic microlattice materials: A current state of the art on manufacturing, mechanical properties and applications,” Mater. Des., 2016.
[11] P. Kohnen, C. Haase, J. Bultmann, S. Ziegler, J. H. Schleifenbaum, and W. Bleck, “Mechanical properties and deformation behavior of additively manufactured lattice structures of stainless steel,” Mater. Des., vol. 145, pp. 205-217, May 2018.
[12] C. Yan, L. Hao, A. Hussein, and D. Raymont, “Evaluations of cellular lattice structures manufactured using selective laser melting,” Int. J. Mach. Tools Manuf., vol. 62, pp. 32-38, 2012.
131
[13] J. . Tu et al., “Preparation and properties of TiB2 nanoparticle reinforced copper matrix composites by in situ processing,” Mater. Lett., vol. 52, no. 6, pp. 448-452, Feb. 2002.
[14] J. P. Kruth, X. Wang, T. Laoui, and L. Froyen, “Lasers and materials in selective laser sintering,” Assem. Autom., vol. 23, no. 4, pp. 357-371, Dec. 2003.
[15] L. Thijs, F. Verhaeghe, T. Craeghs, J. Van Humbeeck, and J.-P. Kruth, “A study of the microstructural evolution during selective laser melting of Ti-6Al-4V,” Acta Mater., vol. 58, no. 9, pp. 3303-3312, May 2010.
[16] E. Chlebus, B. Kuznicka, T. Kurzynowski, and B. Dybala, “Microstructure and mechanical behaviour of Ti—6Al—7Nb alloy produced by selective laser melting,” Mater. Charact., vol. 62, no. 5, pp. 488-495, May 2011.
[17] D. Wang, C. Song, Y. Yang, and Y. Bai, “Investigation of crystal growth mechanism during selective laser melting and mechanical property characterization of 316L stainless steel parts,” Mater. Des., vol. 100, pp. 291-299, Jun. 2016.
[18] Y. M. Wang et al., “Additively manufactured hierarchical stainless steels with high strength and ductility,” Nat. Mater., vol. 17, no. 1, pp. 63-71, Oct. 2017.
[19] W. Shifeng, L. Shuai, W. Qingsong, C. Yan, Z. Sheng, and S. Yusheng, “Effect of molten pool boundaries on the mechanical properties of selective laser melting parts,” J. Mater. Process. Technol., vol. 214, no. 11, pp. 2660-2667, Nov. 2014.
[20] M. Mazur, M. Leary, S. Sun, M. Vcelka, D. Shidid, and M. Brandt, “Deformation and failure behaviour of Ti-6Al-4V lattice structures manufactured by selective laser melting (SLM),” Int. J. Adv. Manuf. Technol., vol. 84, no. 5-8, pp. 13911411, 2016.
[21] G. Castro and S. R. Nutt, “Synthesis of syntactic steel foam using gravity-fed infiltration,” Mater. Sci. Eng. A, vol. 553, pp. 89-95, 2012.
[22] L. P. Zhang and Y. Y. Zhao, “Mechanical Response of Al Matrix Syntactic Foams Produced by Pressure Infiltration Casting,” J. Compos. Mater., vol. 41, no. 41, p.17, 2007.
[23] P. K. Rohatgi, J. K. Kim, N. Gupta, S. Alaraj, and A. Daoud, “Compressive characteristics of A356/fly ash cenosphere composites synthesized by pressure infiltration technique,” Compos. Part A Appl. Sci. Manuf., vol. 37, no. 3, pp. 430437, 2006.
[24] D. P. Mondal, M. D. Goel, and S. Das, “Compressive deformation and energy absorption characteristics of closed cell aluminum-fly ash particle composite foam,” Mater. Sci. Eng. A, vol. 507, no. 1-2, pp. 102-109, 2009.
132
[25] G. A. Rocha Rivero, B. F. Schultz, J. B. Ferguson, N. Gupta, and P. K. Rohatgi, “Compressive properties of Al-A206/SiC and Mg-AZ91/SiC syntactic foams,” J. Mater. Res., vol. 28, no. 17, pp. 2426-2435, 2013.
[26] A. Daoud, M. T. A. El-Khair, A. Y. Shenouda, E. Mohammed, and P. K. Rohatgi, “Microstructure, tensile properties and electrochemical behavior of Pb alloy-45 vol.% fly ash microballoon composites,” Mater. Sci. Eng. A, vol. 526, no. 1-2, pp. 225-234, 2009.
[27] M. Garcia-Avila, M. Portanova, and A. Rabiei, “Ballistic performance of composite metal foams,” Compos. Struct., vol. 125, pp. 202-211, Jul. 2015.
[28] R. A. Palmer, K. Gao, T. M. Doan, L. Green, and G. Cavallaro, “Pressure infiltrated syntactic foams—Process development and mechanical properties,” Mater. Sci. Eng. A, vol. 464, no. 1-2, pp. 85-92, 2007.
[29] A. Szlancsik, B. Katona, A. Kemeny, and D. Karoly, “On the Filler Materials of Metal Matrix Syntactic Foams,” Materials (Basel)., vol. 12, no. 12, p. 2023, Jun. 2019.
[30] S. Anandan et al., “Failure In metal honeycombs manufactured by selective laser melting of 304 L stainless steel under compression,” Virtual Phys. Prototyp., vol. 14, no. 2, pp. 114-122, Apr. 2018.
[31] M. R. Karamooz Ravari, M. Kadkhodaei, M. Badrossamay, and R. Rezaei, “Numerical investigation on mechanical properties of cellular lattice structures fabricated by fused deposition modeling,” Int. J. Mech. Sci., vol. 88, pp. 154-161, 2014.
[32] V. Br0tan, O. Fergani, K. S0rby, and T. Welo, “Mechanical Properties of Biocompatible 316L Steel Rhombic Dodecahedron Lattice Structures,” in Solid Freedom Fabrication, 2016, pp. 2087-2094.
[33] S. Josupeit, P. Delfs, D. Menge, and H. Schmid, “Manufacturability and Mechanical Characterization of Laser Sintered Lattice Structures,” pp. 2077-2086, 2016.
[34] L. Yang, “A study about size effects of 3D periodic cellular structures,” pp. 21812193, 2016.
133
[35] L. Yang, O. Harrysson, D. Cormier, H. West, H. Gong, and B. Stucker, “Additive Manufacturing of Metal Cellular Structures: Design and Fabrication,” JOM, vol. 67, no. 3, pp. 608-615, Mar. 2015.
[36] I. Maskery et al., “A mechanical property evaluation of graded density Al-Si10- Mg lattice structures manufactured by selective laser melting,” Mater. Sci. Eng. A, vol. 670, pp. 264-274, Jul. 2016.
[37] L. J. Gibson, “Modelling the Mechanical Behavior of Cellular Materials,” Mater. Sci. Eng. A, pp. 1-36, 1989.
[38] M. Yamashita and M. Gotoh, “Impact behavior of honeycomb structures with various cell specifications - Numerical simulation and experiment,” Int. J. Impact Eng., vol. 32, no. 1-4, pp. 618-630, Dec. 2006.
[39] P. Colombo and H. P. Degischer, “Highly porous metals and ceramics,” Mater.Sci. Technol., vol. 26, no. 10, pp. 1145-1158, 2010.
[40] Y. Zhou, H. Peng, M. Xuan, and X. Zuo, “The Microstructure and Properties of 304L Stainless Honeycombs Fabricated by Extruding and Sintering,” 2012.
[41] E. Fitzgerald and W. Everhart, “The Effect of Location on the Structure and Mechanical Properties of Selective Laser Melted 316L Stainless Steel,” Proc. 27th Annu. Int. Solid Free. Fabr. Symp., pp. 574-583, 2016.
[42] W. Everhart, P. Korinko, M. Reigel, M. Morgan, and J. Bobbitt, “CHARACTERIZATION OF SELECTIVE LASER MELTED 304L,” in MS&T, 2016.
[43] H. Bikas, P. Stavropoulos, and G. Chryssolouris, “Additive manufacturing methods and modelling approaches: a critical review,” Int. J. Adv. Manuf. Technol., vol. 83, no. 1-4, pp. 389-405, Mar. 2016.
[44] P. Hanzl, M. Zetek, T. Baksa, and T. Kroupa, “The Influence of Processing Parameters on the Mechanical Properties of SLM Parts,” Procedia Eng., vol. 100, pp. 1405-1413,Jan. 2015.
[45] L. E. Murr et al., “Fabrication of Metal and Alloy Components by Additive Manufacturing: Examples of 3D Materials Science,” J. Mater. Res. Technol., vol. 1, no. 1, pp. 42-54, Apr. 2012.
[46] L. Thijs, M. L. Montero Sistiaga, R. Wauthle, Q. Xie, J.-P. Kruth, and J. Van Humbeeck, “Strong morphological and crystallographic texture and resulting yield strength anisotropy in selective laser melted tantalum,” Acta Mater., vol. 61, no.12, pp. 4657-4668, 2013.
134
[47] J. A. Cherry, H. M. Davies, S. Mehmood, N. P. Lavery, S. G. R. Brown, and J. Sienz, “Investigation into the effect of process parameters on microstructural and physical properties of 316L stainless steel parts by selective laser melting,” Int. J. Adv. Manuf. Technol., vol. 76, no. 5-8, pp. 869-879, Feb. 2015.
[48] K. V. Yang, Y. Shi, F. Palm, X. Wu, and P. Rometsch, “Columnar to equiaxed transition in Al-Mg(-Sc)-Zr alloys produced by selective laser melting,” Scr. Mater., vol. 145, pp. 113-117, Mar. 2018.
[49] D. Dai et al., “Influence of scan strategy and molten pool configuration on microstructures and tensile properties of selective laser melting additive manufactured aluminum based parts,” Opt. Laser Technol., vol. 99, pp. 91-100, Feb. 2018.
[50] A. T. Sutton, C. S. Kriewall, M. C. Leu, and J. W. Newkirk, “Powders for Additive Manufacturing Processes: Characterization Techniques and Effects on Part Properties,” Solid Free. Fabr. Proc., pp. 1004-1030, 2016.
[51] L. Cordova, M. Campos, and T. Tinga, “POWDER CHARACTERIZATION AND OPTIMIZATION FOR ADDITIVE MANUFACTURING”
[52] L. Cordova, M. Campos, and T. Tinga, “ASSESSMENT OF MOISTURE CONTENT AND ITS INFLUENCE ON LASER BEAM MELTING FEEDSTOCK”
[53] V. Gunenthiram et al., “Experimental analysis of spatter generation and melt-pool behavior during the powder bed laser beam melting process,” J. Mater. Process. Technol., vol. 251, pp. 376-386, Jan. 2018.
[54] C. Qiu et al., “Influence of processing conditions on strut structure and compressive properties of cellular lattice structures fabricated by selective laser melting,” Mater. Sci. Eng. A, vol. 628, pp. 188-197, Mar. 2015.
[55] L. Yang et al., “Design for Additively Manufactuerd Lighweight Structure: A Perspective,” Solid Free. Fabr. Proc., vol. 6, no. 1, pp. 1-17, 2016.
[56] B. Brown, W. Everhart, and J. Dinardo, “Characterization of bulk to thin wall mechanical response transition in powder bed AM,” Rapid Prototyp. J., vol. 22, no. 5, pp. 801-809, Aug. 2016.
[57] A. Daoud, M. T. A. El-Khair, M. Abdel-Aziz, and P. Rohatgi, “Fabrication, microstructure and compressive behavior of ZC63 Mg-microballoon foam composites,” Compos. Part A Appl. Sci. Manuf., vol. 67, no. 9, pp. 1842-1853, 2007.
135
[58] A. Rabiei and L. J. Vendra, “A comparison of composite metal foam’s properties and other comparable metal foams,” Mater. Lett., vol. 63, no. 5, pp. 533-536, Feb. 2009.
[59] M. Taherishargh, I. V. Belova, G. E. Murch, and T. Fiedler, “Low-density expanded perlite-aluminium syntactic foam,” Mater. Sci. Eng. A, vol. 604, pp. 127-134, 2014.
[60] Sudarshan and M. K. Surappa, “Dry sliding wear of fly ash particle reinforced A356 Al composites,” Wear, vol. 365, no. 3-4, pp. 349-360, 2008.
[61] Sudarshan and M. K. Surappa, “Synthesis of fly ash particle reinforced A356 Al composites and their characterization,” Mater. Sci. Eng. A, vol. 480, no. 1-2, pp. 117-124, 2008.
[62] O. Couteau and D. C. Dunand, “Creep of aluminum syntactic foams,” Mater. Sci. Eng. A, vol. 488, no. 1-2, pp. 573-579, 2008.
[63] J. B. Ferguson, J. A. Santa Maria, B. F. Schultz, and P. K. Rohatgi, “Al-Al2O3 syntactic foams—Part II: Predicting mechanical properties of metal matrix syntactic foams reinforced with ceramic spheres,” Mater. Sci. Eng. A, vol. 582, no. 0, pp. 423-432, 2013.
[64] M. Y. He, M. Kiser, B. Wu, and F. W. Zok, “Influence of thermal expansion mismatch on residual stressesand flow response of microballoon composites,” Mech. Mater., vol. 23, no. 2, pp. 133-146, 1996.
[65] D. P. Mondal et al., “Titanium-cenosphere syntactic foam made through powder metallurgy route,” Mater. Des., vol. 34, pp. 82-89, Feb. 2012.
[66] J. Weise, N. Salk, U. Jehring, J. Baumeister, D. Lehmhus, and M. A. Bayoumi, “Influence of Powder Size on Production Parameters and Properties of Syntactic Invar Foams Produced by Means of Metal Powder Injection Moulding,” Adv. Eng. Mater., vol. 15, no. 3, pp. 118-122, Mar. 2013.
[67] J. A. Santa Maria, B. F. Schultz, J. B. Ferguson, and P. K. Rohatgi, “Al-Al2O3 syntactic foams - Part I: Effect of matrix strength and hollow sphere size on the quasi-static properties of Al-A206/Al2O3 syntactic foams,” Mater. Sci. Eng. A, vol. 582, no. 0, pp. 415-422, 2013.
[68] I. N. Orbulov and J. Dobranszky, “Producing metal matrix syntactic foams by pressure infiltration,” Period. Polytech. Mech. Eng., vol. 52, no. 1, pp. 35-42,2008.
136
[69] Y. Lin, Q. Zhang, and G. Wu, “Interfacial microstructure and compressiveproperties of Al-Mg syntactic foam reinforced with glass cenospheres,” J. Alloys Compd., vol. 655, pp. 301-308, 2016.
[70] J. Weise, J. Baumeister, O. Yezerska, N. Salk, and G. B. D. Silva, “Syntactic iron foams with integrated microglass bubbles produced by means of metal powder injection moulding,” Adv. Eng. Mater., vol. 12, no. 7, pp. 604-608, Apr. 2010.
[71] D. Lehmhus et al., “Quasi-static and Dynamic Mechanical Performance of Glass Microsphere- and Cenosphere-based 316L Syntactic Foams,” Procedia Mater. Sci., vol. 4, pp. 383-387, Jan. 2014.
[72] D. D. Luong, V. C. Shunmugasamy, N. Gupta, D. Lehmhus, J. Weise, and J. Baumeister, “Quasi-static and high strain rates compressive response of iron and Invar matrix syntactic foams,” Mater. Des., vol. 66, pp. 516-531, Feb. 2015.
[73] B. G. Kutchko and A. G. Kim, “Fly ash characterization by SEM-EDS,” Fuel, vol. 85, no. 17-18, pp. 2537-2544, 2006.
[74] Q. Zhang and G. Wu, “Three dimensional microstructure and image-based simulation of a fly ash/al syntactic foam using X-ray micro-computed tomography,” J. Wuhan Univ. Technol. Mater. Sci. Ed., vol. 28, no. 1, pp. 99-103, 2013.
[75] S. Mohanty and Y. P. Chugh, “Development of fly ash-based automotive brake lining,” Tribol. Int., vol. 40, no. 7, pp. 1217-1224, 2007.
[76] X. Xue and Y. Zhao, “Ti matrix syntactic foam fabricated by powder metallurgy: Particle breakage and elastic modulus,” JOM, vol. 63, no. 2, pp. 36-40, 2011.
[77] P. K. Rohatgi, R. Q. Guo, H. Iksan, E. J. Borchelt, and R. Asthana, “Pressure infiltration technique for synthesis of aluminum-fly ash particulate composite,” Mater. Sci. Eng. A, vol. 244, no. 1-2, pp. 22-30, 1998.
[78] J. Weise, J. Baumeister, O. Yezerska, N. Salk, and G. B. D. Silva, “Syntactic Iron Foams with Integrated Microglass Bubbles Produced by Means of Metal Powder Injection Moulding,” Adv. Eng. Mater., vol. 12, no. 7, pp. 604-608, 2010.
[79] J. K. Cochran, “Ceramic hollow spheres and their applications,” Curr. Opin. Solid State Mater. Sci., vol. 3, no. 5, pp. 474-479, 1998.
[80] K. Majlinger and I. N. Orbulov, “Characteristic compressive properties of hybrid metal matrix syntactic foams,” Mater. Sci. Eng. A, vol. 606, pp. 248-256, 2014.
137
[81] “Production and Properties of 316L Stainless Steel Cellular Materials and Syntactic Foams,” 2003.
[82] P. Yu and M. Qian, “Metal injection moulding of in-situ formed AlN hollow sphere reinforced Al matrix syntactic foam parts,” Mater. Chem. Phys., vol. 137, no. 2, pp. 435-438, Dec. 2012.
[83] K. Myers, P. Cortes, B. Conner, T. Wagner, B. Hetzel, and K. M. Peters,“Structure property relationship of metal matrix syntactic foams manufactured by a binder jet printing process,” Addit. Manuf., vol. 5, pp. 54-59, 2015.
[84] M. A. Ben Trad, V. Demers, R. Cote, M. Sardarian, and L. Dufresne, “Numerical simulation and experimental investigation of mold filling and segregation in low- pressure powder injection molding of metallic feedstock,” Adv. Powder Technol., vol. 31, no. 3, pp. 1349-1358, Mar. 2020.
[85] R. Hussein et al., “Effective elastic moduli of metal honeycombs manufactured using selective laser melting,” Rapid Prototyp. J., vol. 26, no. 5, pp. 971-980, Feb. 2020.
138
VITA
Myranda Shea Spratt, nee Ferris, was born in Columbia, Missoui. She earned an
Associate of Science in Pre-Engineering in May 2012 from East Central Community
College in Union, Missouri. She then went to Missouri University of Science and
Technology to earn her Bachelor of Science in Ceramic Engineering in May 2015.
Myranda accepted the Chancellor’s Distiguished Fellowship and began the Ph.D.
degree program at Missouri S&T after receiving her bachelor’s degree. Her research was
focused on additive manufacturing and lightweight composites. In December 2020,
Myranda received her Doctor of Philosophy in Materials Science and Engineering from
Missouri University of Science and Technology.