Effects of Hot Isostatic Pressing on Copper Parts Additively Manufactured via Binder

Jetting

Ashwath Yegyan Kumar

Thesis submitted to the faculty of the Virginia Polytechnic Institute and State University in

partial fulfillment of the requirements for the degree of

Master of Science

In

Mechanical Engineering

Christopher Bryant Williams

Scott T. Huxtable

Hang Yu

14 February, 2018

Blacksburg, Virginia

Keywords: Additive Manufacturing, Binder Jetting, Hot Isostatic Pressing, Copper

Effects of Hot Isostatic Pressing on Copper Parts Additively Manufactured via Binder Jetting

Ashwath Yegyan Kumar

ACADEMIC ABSTRACT

Copper is a material of interest to Additive Manufacturing (AM) owing to its outstanding material

properties, which finds use in enhanced heat transfer and electronics applications. Its high thermal

conductivity and reflectivity cause challenges in the use of Powder Bed Fusion AM systems that

involve supplying high-energy lasers or electron beams. This makes Binder Jetting a better

alternative as it separates part creation (binding together of powders) from energy supply (post-

process sintering). However, it is challenging to fabricate parts of high density using this method

due to low packing density of powder while printing. This work aims to investigate the effects of

Hot Isostatic Pressing (HIP) as a secondary post-processing step on the densification of Binder Jet

copper parts. By understanding the effects of HIP, the author attempts to create parts of near-full

density, and subsequently to quantify the effects of the developed process chain on the material

properties of resultant copper parts. The goal is to be able to print parts of desired properties suited

to particular applications through control of the processing conditions, and hence the porosity.

First, 99.47% dense copper was fabricated using optimized powder configurations and process

parameters. Further, the HIP of parts sintered to three densities using different powder

configurations was shown to result in an improvement in strength and ductility with porosity in

spite of grain coarsening. The strength, ductility, thermal and electrical conductivity were then

compared to various physical and empirical models in the literature to develop an understanding

of the process-property-performance relationship.

Effects of Hot Isostatic Pressing on Copper Parts Additively Manufactured via Binder Jetting

Ashwath Yegyan Kumar

GENERAL AUDIENCE ABSTRACT

Additive Manufacturing (AM) is a technique of fabricating an object in a layer-wise fashion. The

layer-based approach provides opportunity for the manufacture of highly complex shapes. Binder

Jetting is an AM technology that creates parts by the selective jetting of a polymeric binder onto

successive layers of powdered material. In the case of metals, the printing process is followed by

sintering in an oven, which burns out the binder and densifies the part. However, this is typically

not enough to remove all the porosity in a specimen. While this enables the fabrication of a variety

of materials, the porosity in sintered parts can be a detriment to their properties. This work aims to

investigate the use of post-process Hot Isostatic Pressing (HIP) to eliminate the remaining porosity.

HIP is a technique of applying high pressures at high temperatures in an inert gas medium. The

goal of this research is to scientifically understand and quantify the effect of HIP on sintered parts

made via Binder Jetting. The research is carried out in the context of copper, which has unique

mechanical, thermal and electrical conductance properties that could be influenced by the presence

of pores. In this work, the effects of the Binder Jetting-Sintering-HIP process chain on the porosity,

and consequently the material properties, of copper parts are quantified. Resolving the issue of

porosity can enable the printing of copper parts for specialized applications from electronic

components to rocket engines. Developing a quantitative understanding can pave the way to design

specific processing conditions to fabricate not only fully dense copper parts with superior

properties, but also parts of a designed level of porosity that have specific target material

properties.

iv

ACKNOWLEDGEMENTS

I would like to express my most sincere gratitude to Dr. Williams for firstly having given me the

opportunity to work on this thesis, and for having been a constant source of guidance and advice,

for which I am indebted. Working under his supervision has been an extremely fruitful learning

experience that has not only taught me about various aspects of research, but also helped mold my

overall character. I would also like to thank the members of my committee, Dr. Huxtable and Dr.

Yu, for having provided invaluable advice and guidance towards my research.

I would next like to thank the National Science Foundation for providing the funding towards this

research (CMMI #1254287). I thank my colleague in the DREAMS lab, Yun Bai, who has been

instrumental in helping me out with running experiments and having provided countless advice

and suggestions for my research through the course of my stay here. I would also like to thank Dr.

Anders Eklund of Quintus Technologies for his knowledgeable inputs on Hot Isostatic Pressing

parameter development, which was instrumental in the success of our experiments. I thank Jue

Wang and Dr. Huxtable for having helped with running the conductivity measurement experiments

and helping me learn heat transfer concepts that were used in my research. I wish to thank Robert

Mills of the Extreme Environments, Robotics, and Materials Laboratory for having shared their

equipment for tensile testing, metallographic sample preparation and microscopy. I thank Matthew

Meeder and Dr. Al Wicks for the ideation and preliminary experiments for conductivity

measurements, which provided a platform for me to start my research.

I also thank all the members of the DREAMS lab for having given me a fantastic atmosphere to

work in, for all their inputs for my research, for having motivated me to grow as a researcher, and

for being great friends.

Last but far from the least, I express my sincerest thanks to my family back in India for having

sent me out here to learn and grow, and to several friends who have been encouraging and

supportive through my stay here.

v

Table of Contents Chapter 1: Introduction ........................................................................................................................... 1

Chapter 2: Influence of Hot Isostatic Pressing on the Density, Microstructure and Mechanical

Properties of Copper Parts Fabricated by Binder Jetting Additive Manufacturing ...................................... 3

Abstract ..................................................................................................................................................... 3

2.1 Introduction ................................................................................................................................... 3

2.1.1 Prior Work on Density Improvement in Binder Jetting ........................................................ 5

2.1.2 Hot Isostatic Pressing ............................................................................................................ 6

2.1.3 Research Objective ............................................................................................................... 7

2.2 Experimental Methods .................................................................................................................. 8

2.2.1 Materials Used ...................................................................................................................... 9

2.2.2 Processing Parameters ........................................................................................................... 9

2.2.3 Density Measurements ........................................................................................................ 10

2.2.4 Metallographic Analysis ..................................................................................................... 11

2.2.5 Tensile Testing .................................................................................................................... 12

2.3 Results ......................................................................................................................................... 12

2.3.1 Density ................................................................................................................................ 12

2.3.2 Porosity ............................................................................................................................... 13

2.3.3 Microstructure ..................................................................................................................... 17

2.3.4 Tensile Testing .................................................................................................................... 20

2.4 Closure and Future Work ............................................................................................................ 22

2.5 References ................................................................................................................................... 22

Chapter 3: Impacts of Process-Induced Porosity on Material Properties of Copper Made by Binder

Jetting Additive Manufacturing .................................................................................................................. 29

Abstract ................................................................................................................................................... 29

3.1 Introduction ................................................................................................................................. 29

3.1.1 Prior Work on Powder Bed Fusion of Copper .................................................................... 29

3.1.2 Binder Jetting of Copper ..................................................................................................... 31

3.1.3 Research Objective ............................................................................................................. 31

3.2 Modeling Porosity-Property Relationships ................................................................................. 32

3.2.1 Strength and Ductility ......................................................................................................... 32

3.2.2 Thermal and Electrical Conductivity .................................................................................. 34

3.2.3 Wiedemann-Franz Law ....................................................................................................... 35

3.3 Experimental Methods ................................................................................................................ 36

vi

3.3.1 Specimen fabrication ........................................................................................................... 36

3.3.2 Tensile Testing .................................................................................................................... 39

3.3.3 Thermal Conductivity Measurement: ................................................................................. 39

3.3.4 Electrical Resistivity Measurement..................................................................................... 40

3.4 Results ......................................................................................................................................... 41

3.4.1 Tensile Strength .................................................................................................................. 41

3.4.2 Ductility .............................................................................................................................. 42

3.4.3 Thermal Conductivity ......................................................................................................... 44

3.4.4 Electrical Conductivity ....................................................................................................... 47

3.4.5 Wiedemann-Franz Law ....................................................................................................... 48

3.5 Conclusions ................................................................................................................................. 49

3.6 References ................................................................................................................................... 50

Chapter 4: Closure ................................................................................................................................ 53

4.1 Conclusions ................................................................................................................................. 53

4.2 Summary of findings ................................................................................................................... 53

4.3 Contributions ............................................................................................................................... 54

4.4 Limitations and Future Work ...................................................................................................... 55

References (for Chapters 1 and 4)............................................................................................................... 56

Appendix A ................................................................................................................................................. 57

Appendix B ................................................................................................................................................. 67

vii

List of Figures

Figure 1.1: Schematic of the Binder Jetting process ..................................................................................... 1

Figure 2.1: A schematic of the Binder Jetting process.................................................................................. 4

Figure 2.2: Schematic of process chain ........................................................................................................ 9

Figure 2.3: Composite image showing porosity in part HIPed at 975°C for 4 hours ................................. 14

Figure 2.4: Porosity in 17µm parts ............................................................................................................. 15

Figure 2.5: Porosity in 25µm parts ............................................................................................................. 16

Figure 2.6: Porosity in bimodal parts .......................................................................................................... 17

Figure 2.7: Grain structure change upon HIP – 17µm parts ....................................................................... 18

Figure 2.8: Grain structure change upon HIP – 25µm parts ....................................................................... 18

Figure 2.9: Grain structure change upon HIP – bimodal parts ................................................................... 19

Figure 2.10: Samples of untested (top) and tested (bottom) tensile specimens .......................................... 21

Figure 3.1: A schematic of the binder jetting process ................................................................................. 31

Figure 3.2: Photograph depicting layout of parts in print bed .................................................................... 36

Figure 3.3: Schematic of process chain ...................................................................................................... 37

Figure 3.4: Porosity in 17µm specimen (YZ plane) : (a) Sintered – 16.37%; (b) HIPed – 14.17% ........... 38

Figure 3.5: Porosity in bimodal specimen (YZ Plane): (a) Sintered – 9.48%; (b) HIPed – 2.68% ............ 38

Figure 3.6: Electrical Resistivity measurement setup ................................................................................. 41

Figure 3.7: Tensile strength observed in comparison to Model Predictions ............................................... 42

Figure 3.8: Ductility – Observed vs. Model ................................................................................................ 43

Figure 3.9: Thermal Conductivity vs. Porosity: Observed data compared to models ................................ 44

Figure 3.10: Conductivity at full density (kfd) calculated for each processing condition .......................... 45

Figure 3.11: Grain Boundary Thermal Resistance calculated from Modified EMT Model ....................... 46

Figure 3.12: Electrical Conductivity vs Porosity: Observed data compared to models .............................. 47

Figure 3.13: Observed Thermal Conductivity vs. that calculated using the Wiedemann-Franz Law......... 48

viii

List of Tables

Table 2.1: Particle Size data for powders used (µm): ................................................................................... 9

Table 2.2: A comparison of density with and without oil impregnation .................................................... 13

Table 2.3: Grain size change upon HIP of various parts............................................................................. 20

Table 2.4: Tensile Strength and Ductility improvement upon HIP ............................................................ 21

Table 3.1: Powders and processing conditions used to achieve different porosities .................................. 38

1

Chapter 1: Introduction

Binder Jetting is a powder-bed based Additive Manufacturing (AM) technology that allows for a

cost-effective, scalable way of fabricating a wide variety of materials including metals, ceramics

and polymers. Part creation involves the selective jetting of a binder onto successive layers of

powdered material (Figure 1.1). This is followed by curing of the binder by supplying heat, and

depowdering any loose powder adhering to the surface by using compressed air. The ‘green’ part

thus formed is then subjected to sintering in a furnace for debinding and sintering for densification

[1]. However, it is challenging to create fully dense, homogeneous parts using this method as

sintering typically involves the infiltration of a secondary, lower melting alloy that fills the pores

by capillary action, minimizing shrinkage and helping in densification. This research attempts to

address these issues using Hot Isostatic Pressing (HIP), a technique of applying high pressures at

high temperatures to pre-sintered specimens, to remove residual porosity.

Figure 1.1: Schematic of the Binder Jetting process

The author conducts this research in the context of copper, which is a material of particular interest

for AM, owing to its excellent thermal and electrical conductivity properties. The realization of

complex geometries through AM can pave the way for novel designs for applications involving

advanced heat transfer, thermal management, electronics etc. The use of Binder Jetting for copper

helps circumvent challenges in its manufacturing using laser or electron beam based Powder Bed

Fusion (PBF) AM systems owing to its high thermal conductivity and optical reflectivity. Hence,

the overarching objective of this research is to be able to understand and quantify the effect of HIP

on Binder Jet, sintered copper parts and work scientifically towards realizing these applications

with desired thermal, electrical as well structural properties.

Towards achieving these goals, first, a proof of concept was established for the ability to fabricate

near-full density copper parts by the use of previously optimized powder configurations, printing

and sintering parameters [2] (Appendix A). Subsequently, in Chapter 2, a more detailed

2

investigation is carried out on the HIP of parts printed using three different powder configurations

in order to determine the influence of the process on the density (2.3.1), porosity (2.3.2),

microstructure (2.3.3) and mechanical properties (2.3.4) for different sintered densities. An

understanding of the effects of HIP on parts of differing levels of porosity can help in evaluating

the minimum density required for HIP to be effective in eliminating porosity, and aid in the

fabrication of parts with designed levels of target porosity.

Finally, in Chapter 3, an attempt is made to quantify the effect of such process-induced porosity

on the mechanical (3.4.1, 3.4.2), thermal and electrical properties (3.4.3 - 3.4.5) of copper. These

measurements are then positioned against models developed in the literature for Powder

Metallurgy (PM) and two-component structures, which can help develop intrinsic models

specifically suited to Binder Jetting porosity-property relationships. Such an understanding of the

process-porosity, and porosity-property correlations, can help in achieving the eventual goals of:

i. Realizing applications involving full density and the values of strength, ductility, thermal

and electrical conductivity being closest to that of pure copper; and

ii. Being able to achieve varying levels of ‘designed’ porosity and hence properties through

appropriate selection of processing conditions

3

Chapter 2: Influence of Hot Isostatic Pressing on the Density, Microstructure

and Mechanical Properties of Copper Parts Fabricated by Binder Jetting

Additive Manufacturing

Abstract

Hot Isostatic Pressing (HIP) is a technique of applying high pressures through a fluid medium at

high temperatures to enclosed powders, castings and pre-sintered metal parts to eliminate porosity.

Due to uniform photographic shrinkage expected from this process, it can be a useful post-

processing technique for complex-geometry parts fabricated using Additive Manufacturing

techniques. In order for the technique to work effectively, the parts are typically required to have

a minimum density of 92% where surface porosity is closed. While HIP has been used in

conjunction with other Powder Bed Fusion AM processes, its use for parts made using Binder

Jetting has not been investigated in detail due to the limitations of Binder Jetting in fabricating

high density parts. After previously reported success in its use in the context of Binder Jetting of

copper [1], an effort is made here to perform detailed investigations on the effect of HIP on three

different powder configurations that lead to varying levels of porosity in sintered copper

specimens. The effects of HIP on density, microstructure, tensile strength and ductility have been

investigated. The highest density achieved was 97.32% after HIP by using bimodal powders that

were printed and sintered to 90.52%. Both the tensile strength and ductility were found to improve

following HIP, which suggests that the reduction in porosity is predominant compared to the

detrimental effects of grain coarsening.

2.1 Introduction

Binder Jetting is a powder bed-based Additive Manufacturing (AM) technology that involves the

use of an inkjet printhead to pattern a binder onto a bed of powder in order to selectively bind

particles together (see Figure 2.1). This is followed by curing the binder to add sufficient strength

for part handling and depowdering. The green part can then be subjected to suitable post-

processing such as sintering or infiltration to densify and strengthen it further. Since the part

creation only involves binding powder particles, the technology can easily be adapted to a wide

variety of materials including metals, ceramics and polymers.

4

Figure 2.1: A schematic of the Binder Jetting process

The process offers a number of advantages in addition to its applicability for a wide range of

materials. Owing to its lack of reliance on any thermal processing during the primitive creation

process, Binder Jetting does not face issues with residual stresses, warping and shrinkage that are

faced in typical Powder Bed Fusion AM processes due to rapid thermal treatment. This eliminates

the need for designed support structures or anchors to be fabricated, allowing the loose powder

surrounding the printed part to act as an inherent support. Additionally, printing using Binder

Jetting is cheaper and more easily scalable than other powder bed AM processes, as scaling up

would involve merely increasing the size or number of printheads as compared to upgrading

expensive laser or electron-beam power sources. Binder Jetting thus also enables the processing

of thermally conductive and optically reflective materials such as copper [2], which can be difficult

to process using powder bed fusion technologies that employ high intensity energy sources.

One major disadvantage of the Binder Jetting AM process, however, is that the fabricated parts

often feature low density. This is due to the lack of any compaction of powder during printing,

resulting in the green parts essentially being loosely packed powder particles, held together by the

binder. This results in significant residual porosity after sintering (for e.g., 85.5% in the case of

copper [2]). One technique used to improve the density of metallic and ceramic parts is through

the infiltration of a secondary, lower melting material during sintering. This material can infiltrate

through the pores of the primary material’s matrix by capillary action after the binder pyrolyzes,

minimizing shrinkage and increasing density. This technique has been investigated in detail by

researchers for various material combinations [3 – 6]. The selection of an infiltrant material is

challenging as it involves finding a significantly lower melting material that is compatible in its

wetting properties with the parent material. The requirement that the infiltrant must have a

significantly lower melting point than the parent material also hinders the processing of

homogenous, non-alloyed structures for use in applications where the presence of a secondary

material can negatively influence the property of the overall structure. These issues with

5

infiltration necessitate research on other ways to minimize the porosity and shrinkage to achieve

enhanced material properties without infiltration.

2.1.1 Prior Work on Density Improvement in Binder Jetting

Binder Jetting involves several process and post-process sintering parameters that may influence

the density and other properties of fabricated parts. Substantial research has been conducted to

understand these relationships to achieve the best possible density and material properties. Turker

and co-authors were able to achieve 98.5% theoretical density by optimizing layer thickness and

sintering temperatures in the Binder Jetting of Inconel 718 superalloys [7]. Their process required

special process modifications such as drying after every layer is printed, and sintering in vacuum.

Gaytan et al. could achieve only as high as ~65% density in Barium Titanate using optimal binder

saturation and sintering temperatures [8], while Gonzalez and co-authors from the same research

group achieved up to ~96% density in binder jet alumina by optimizing the layer thickness, powder

particle size as well as sintering temperature [9]. Vaezi and Chua found that an increase in binder

saturation or decrease in layer thickness improved tensile strength but at the cost of surface quality

[10]. Miyanaji et al. performed an experimental design to optimize the binder saturation, drying

power and duration and the powder spread speed levels on the strength, shrinkage and dimensional

accuracy of binder jet dental porcelain [11]. They found that Z-accuracy is most significantly

improved by increasing the drying power, X/Y-accuracy by increasing spreading speed, strength

by increasing saturation and shrinkage by increasing the spreading speed. Shrestha and

Manogharan carried out a systematic experimental design to optimize binder saturation, layer

thickness, roll speed and feed-to-powder ratio in order to maximize the transverse rupture strength

of binder jet 316L Stainless Steel parts [12]. Bai and co-authors achieved 85.5% of theoretical

density in copper by optimizing powder type, binder saturation and sintering profile [2].

The use of bimodal powders in AM processes other than Binder Jetting has been explored by

various researchers [13 – 15]. In the case of Binder Jetting, Lanzetta and Sachs have studied the

influence of bimodal powders on improving the surface finish in alumina powder mixtures [16].

Verlee and co-authors [17] experimentally validated and modified the model for packing density

for Binder Jetting as presented in the context of powder metallurgy by German [18] to predict

sintered density of bimodal powder mixtures. In the case of Binder Jetting of copper, Bai and co-

authors used a bimodal powder mixture to achieve ~92% theoretical density [19]. The use of

6

nanosuspension binders has been investigated as a means to improve green density and sintering

performance of binder jet parts [20 – 22]. J. Bai and co-authors achieved up to ~89% theoretical

density in printing nanosilver suspensions in silver powder [23].

In addition to these methods, some other unique modifications to the process have been studied.

Grau et al. achieved up to 99% density in alumina using a technique called slurry-based 3DP. This

was done by dispersing sub-micron ceramic powders in a slurry, which is deposited on the build

piston instead of spreading a powder [24]. Li et al. used chemical vapor infiltration of Si3N4 to

improve the mechanical strength of porous Silicon Nitride [25]. Recently, Rabinskiy and co-

authors used a modified process involving a new binder composition, drying at high temperatures

and an additional step of mechanical compaction after the spreading of each layer, to reduce the

porosity and improve mechanical properties of Si3N4 [26].

2.1.2 Hot Isostatic Pressing

This study is concerned with investigating the use of Hot Isostatic Pressing (HIP) as a secondary

post-processing step after sintering to densify parts fabricated using Binder Jetting. HIP is a

technique of applying high, isostatic pressure using an inert gas medium (typically Argon) at high

temperatures to consolidate loose powders. Atkinson and Davies have presented the physics of the

process in detail [27]. The application of high pressures of the order of a few hundred MPa (tens

of thousands in PSI) causes the entrapped gases in the pores to overcome the surface-energy

driving force for pore closure, resulting in them dissolving in the matrix and on to the surface. In

addition to consolidation of encapsulated loose powders, the process finds other applications such

as the densification of castings and presintered powders. Photographic (dimensionally uniform)

shrinkage is expected to be observed due to the isostatic nature of fluid pressure applied. This

property makes the process potentially useful in densifying complex-geometry parts fabricated

using Additive Manufacturing. HIP is expected to work effectively when the density of the pre-

sintered parts is >92%, as this is the density at which all surface porosity is expected to be sealed

[28]. HIP has been employed in conjunction with various AM technologies to achieve high density

and improved material properties.

HIP has been successfully employed in reducing porosity of parts made using Directed Energy

Deposition methods such as Laser Engineered Net Shaping (LENS) for Ti-6Al-4V. Kobryn and

Semiatin noticed a reduction of anisotropy in Yield Strength from the stress-relieved to HIPed

7

Ti64 samples, along with a reduction in ultimate tensile strength but gain in ductility after HIP.

HIP was also found to have improved fatigue strength of these samples [29]. More recently, an

investigation by Qiu and co-authors also resulted in a reduction in porosity as well as strength after

HIP of Ti-64 made using LENS [30].

Among the Powder Bed Fusion (PBF) technologies, Selective Laser Sintering (SLS) is typically

unable to achieve the densities required for HIP to be effective. Agarwala and co-authors in the

University of Texas at Austin used HIP to densify Bronze-Nickel parts made using SLS, by

encapsulating them in glass to seal the surface connected porosity [31]. Researchers from the same

group later devised a method called SLS/HIP to achieve near-full density using containerless HIP

of SLS parts. This was done by tuning printing process parameters in such a way as to fabricate a

dense outer shell (to act as a container) of >92% density in order to avoid surface connected

porosity, while the internal bulk was processed to a density of ~80% [31 – 33]. Liu and co-authors

developed a process chain involving SLS followed by Cold Isostatic Pressing (CIP), degreasing,

sintering and finally HIP in order to achieve ~96% density in alumina parts [35].

Compared to SLS, Selective Laser Melt (SLM) is able to fabricate parts at higher densities (close

to 99%), which are better suited for HIP. Mower and co-author studied fatigue properties of HIPed

Ti-64 and 316L SS, fabricated by Direct Metal Laser Sintering (DMLS) [36]. Multiple authors

have studied the improvement in density and fatigue properties of Ti-6Al-4V fabricated using SLM

followed by HIP [36 – 38]. Similar studies were conducted for 316 Stainless Steel and Ti64 by

Leuders and co-authors [40] and for ASTM F75, a Co-Cr-Mo alloy, by Haan et al [41]. HIP has

also been studied in conjunction with Electron Beam Melting (EBM) extensively for Ti64 [41 –

46]. Other materials explored include Co-Cr-Mo alloy [48], Inconel 625 [49], Inconel 718 [45],

[50] and Ti-48Al-2Cr-2Nb, a Titanium Aluminide alloy [51]. In a vast majority of the studies

mentioned above, HIP is found to improve the density and fatigue strength and homogenize the

microstructure, but decrease the tensile strength due to coarsening/enlargement of grains caused

by high temperature processing.

2.1.3 Research Objective

There is limited reported literature on the use of HIP for Binder Jet parts, since the densities of

parts fabricated by Binder Jetting are typically less compared to other AM processes, and often

fail to reach the 92% density requirement to avoid surface-connected porosity. Kernan and co-

8

author achieved close to 100% density in WC-Co parts fabricated using slurry-based Binder Jetting

followed by sinter-HIP. Gonzalez noted an improvement the density of Binder Jet Inconel 625

from 96.51% as fabricated to 98.33% after HIP [52].

For Binder Jetting of copper, the authors of this article have previously achieved 99.47% density

using suitable HIP conditions [1]. This research work aims to further this prior study by evaluating

the HIP of parts made using various powder configurations and the effect of the process on the

density, porosity, microstructure and tensile strength of the parts fabricated by Binder Jetting.

Section 2.2 presents an overview of the experimental methods used in the investigations including

the materials (2.2.1), process parameters (2.2.2), density measurements (2.2.3), metallographic

analysis (2.2.4) and tensile testing (2.2.5). Section 2.3 presents the results of density measurements

(2.3.1), porosity analysis (2.3.2), microstructure analysis (2.3.3) and tensile testing (2.3.4), along

with brief discussions. Closure and future work are offered in Section 2.4.

2.2 Experimental Methods

The objective of this study is to evaluate the influence of HIP on the density, porosity,

microstructure and mechanical properties of Binder Jetting parts made using different powder

configurations and thus different sintered densities. While it is generally accepted that a minimum

of 92% density is required to be able to seal surface connected porosity and effectively remove

pores using HIP, the authors seek to understand the effect of HIP on parts that fall below this

density limit as well. A schematic of the process chain with the various control and experimental

variables is provided in Figure 2.2.

9

Figure 2.2: Schematic of process chain

2.2.1 Materials Used

The experiments were conducted on parts of various starting densities, by using three different

powder configurations for printing. The powders were all acquired from ACuPowder. The particle

size distributions of each of the powders used are provided in Table 2.1:

Table 2.1: Particle Size data for powders used (µm):

D10 D50 D90

17µm Powder 8 17 28

25µm Powder 16 25 37

5µm Powder 3 5.5 9

30µm Powder 15 30 37.5

The 17µm and 25µm powders were used as such, while the 5µm and 30µm powders were mixed

in the ratio of 27:73 by weight in a rotating drum for ~2 hours to create a bimodal powder mixture

which has demonstrated success in fabrication of high-density parts [1, 19].

2.2.2 Processing Parameters

The various printing, sintering and HIP parameters were set based on previous success in

fabricating and HIPing high density copper samples [1]. All tests specimens were printed using an

10

ExOne R2 3D printer. The binder used was the standard binder (PM-B-SR2-05) provided by

ExOne, as this has been known to bind well with the material of concern (copper) and leaves

minimal carbon residue after debinding [2]. The layer height for all powder types was kept constant

at 70µm. The binder saturation is defined as the ratio of binder per unit void space between powder

particles. It is an important machine parameter, which decides the X- and Y- spacing of adjacent

droplets being jetted from the nozzles. It is calculated based on user input values of powder packing

density and desired saturation percentage. The binder saturation value was set to 100%. All the

samples required for the different characterizations were fabricated in a single print in order to

maintain consistency among the samples used to study various properties.

Sintering was carried out in a box furnace, in a reducing Hydrogen atmosphere. The samples were

sintered at a temperature of 1075°C for 3h. Prior to this, debinding was carried out at 450°C for

30 minutes. All heating and cooling ramps were at 5°C/min. The sintering profile is depicted in

Figure 2.2.

All samples were post-processed using containerless HIP in a Quintus Technologies graphite

furnace equipped with the proprietary Uniform Rapid Cooling (URC®) technology to enable rapid

cooling which can aid in reducing grain growth at higher temperatures. The samples were HIPed

at a temperature of 1075°C under a pressure of 206.84 MPa for 2 hours in an Argon atmosphere.

Results have also been presented from an initial trial of HIP at 975°C under the same pressure for

4 hours.

2.2.3 Density Measurements

Density Measurements for the parts were carried out using an Archimedes Principle based

apparatus as recommended in the ASTM Standard B962-15 for measuring the density of sintered

Powder Metallurgy (PM) specimens [53]. For this purpose, multiple rectangular coupons were

printed for each powder configuration. Some of these were sent intact for HIP and some retained

for sintered part measurements with the assumption that parts within the same build are similarly

dense. Additionally, in each case, a few of the specimens were cut into two halves, with one half

retained and the other HIPed, to compare the improvement in density within the same part. The

density measurements were conducted both with and without oil impregnation (Section 2.3.1). Oil

impregnation helps avoid the formation of microbubbles on the surface parts when they are

immersed in water.

11

2.2.4 Metallographic Analysis

Sintered and HIPed specimens of each powder configuration were sectioned in the XY, YZ and

ZX cross sectional planes of each sample and polished in successively finer polishing materials in

order to smooth the surface. These polished sections were then photographed under an optical

microscope in order to study the internal porosity. The porosity was measured using a MATLAB

code written using the Image Analysis toolbox (Appendix B). This code converts the micrographs

into black and white images based on a user-defined threshold that helps distinguish the pores from

the rest of the specimen. The images loaded into this code were manually cropped in order to

exclude any regions outside the sample or large scratches that could be recognized by the software

as black regions, or pores. The code then calculates the ratio of the area of the black region, or

pores, to the total area of the section. By replicating this over several sections, a reasonable average

of the internal porosity may be obtained.

The polished samples, whose micrographic images were recorded for this analysis, were then

subsequently etched using nitric acid solution in order to study the microstructure and grain size.

In the case of 25 µm parts, the extensive porosity caused difficulties in polishing and etching due

to liquid entrapment inside the pores. This issue was overcome using a combination of the

application of a compressed air jet to clear the surface, subjecting the parts to vacuum, and allowing

the liquid entrapped to drain by gravity onto soft cloth.

Grain size calculations were performed based on the Saltykov Rectangle method, a modification

of the Jeffries planimetric analysis as described in the ASTM standard E112-13 [53, 54]. ImageJ

software was used to draw rectangles of known areas over metallographic images taken from

multiple sections. The number of grains completely enclosed (𝑛𝑖𝑛𝑠𝑖𝑑𝑒) and those intercepted by

the perimeter of the rectangle (𝑛𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡𝑒𝑑) were counted and substituted into Equation 2.1 below

to obtain the number of grains per mm2, 𝑁𝐴 :

𝑁𝐴 =[𝑛𝑖𝑛𝑠𝑖𝑑𝑒+0.5𝑛𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡𝑒𝑑+1]

𝐴 ( 2.1 )

where 𝐴 is the area of the selected region, which can be obtained using the pixel count in the

software, by appropriate scaling. The ASTM grain number 𝐺 is then calculated using Equation 2.2

below, from the ASTM standard:

𝐺 = 3.321928 𝑙𝑜𝑔(𝑁𝐴) − 2.954 ( 2.2 )

12

The average grain diameter 𝑑𝑔 is calculated as the square root of the average grain area 𝐴𝑔 per the

following equations:

𝐴𝑔 =1

𝑁𝐴 ( 2.3 )

𝑑𝑔 = √𝐴𝑔 ( 2.4 )

A 95% confidence interval was used to represent the variance in the grain number and grain size

data per the standard to account for the number of samples chosen and possible variations over the

different regions. The results from these analyses are presented in sections 2.3.2 and 2.3.3.

2.2.5 Tensile Testing

Tensile strength measurements were carried out in accordance with the ASTM standard B925 for

standard testing practices for PM specimen [56]. Flat unmachined tension test specimen

(“dogbones”) were designed and directly printed. All the specimens were printed laid flat on the

print bed (XY plane), causing the thickness of the sample to be along the build (Z) direction. Per

the ASTM Standard for Coordinate Systems nomenclature, the parts were printed oriented

randomly either in the XYZ or the YXZ direction [57]. The measurements were carried out on an

INSTRON machine with a 50kN load cell at a constant elongation rate of 0.381mm/min (strain

rate of 0.015 mm/mm/min as recommended by the standard) and a gauge length of 25.4mm (1

inch). A strain gauge was used to continually record the elongation. The tests were conducted until

the sample broke. The ultimate tensile strength and ductility results from these experiments are

presented in Section 2.3.4. Ductility is calculated based on elongation.

2.3 Results

2.3.1 Density

The density of sintered and HIPed parts (measured with and without oil impregnation) are reported

in Table 2.2 for each of the powder types.

13

Table 2.2: A comparison of density with and without oil impregnation

Part Type Density (%)

With Oil Impregnation Without Oil Impregnation

17 µm Sintered 83.63 ± 0.40 88.03 ± 2.40

HIPed 85.83 ± 0.19 86.72 ± 0.24

25 µm Sintered 77.73 ± 1.18 80.68 ± 1.96

HIPed 82.41 ± 0.33 90.24 ± 1.45

Bimodal Sintered 90.52 ± 0.30 91.53 ± 0.17

HIPed 97.32 ± 0.06 97.47 ± 0.31

These results indicate a higher variance and inaccuracy in the measurements without oil

impregnation, especially in the parts with lower density. This is due to the formation of

microbubbles on the surface of these parts due to the surface connected porosity and roughness of

binder jet parts, especially those of lower density. Henceforth in this article, the term ‘density’

shall be used in the context of values measured using oil impregnation only.

The results of improvement in density from Table 2.2 indicate that the effect of HIP on parts with

lower density is minimal. This is to be expected, as HIP has not been found to be effective in

removing porosity from parts with large amounts of porosity, especially surface porosity (Section

2.3.2). The technique is found to be effective in improving the density of the parts with higher

sintered density, namely those of the bimodal configuration. The authors have previously obtained

densities as high as 99.47 ± 0.28 % [1]. However, the results obtained in that study could not be

replicated here due to differences in the powder quality and issues with clogged inkjet nozzles

while printing samples in this study. Better quality control to replicate the same powder and

printhead conditions could help repeat those results in the future.

2.3.2 Porosity

In addition to bulk density measurements, samples of each type were sectioned at multiple

locations and polished in order to record the internal porosity using optical microscopy. Figure 2.3

shows a composite image created from three micrographs of a sample from an initial trial in the

HIP of bimodal parts at a lower temperature (975°C) for a longer holding period (4 hours) than the

rest of the samples. This image shows the presence of extensive surface porosity caused due to

14

infiltration of Argon gas into surface-connected pores in the sample, as opposed to dissolution of

entrapped gases into the matrix. This can happen in the presence of high surface porosity prior to

HIP, and with improper HIP parameters. This result shows a case where HIP failed to reduce the

porosity in binder jet parts.

Figure 2.3: Composite image showing porosity in part HIPed at 975°C for 4 hours

The parameters were then changed to a higher temperature of 1075°C and a shorter holding time

of 2 hours, which resulted in porosity reduction. Figures 2.4 and 2.5 show the polished micrographs

of 17µm and 25µm powder parts respectively. Images for each section depicted are from two

halves of the same part, one pictured after sintering and the other after HIP. These images indicate

a number of defects in the samples. For instance, a presence of excessive, non-uniform porosity

near the surface can be observed in nearly all sections. Weakly bound powder near the surface will

have been blown away during depowdering, resulting in low green density and relatively higher

surface porosity. The YZ sections of 17µm samples (Figure 2.4) indicate improper binding

between layers, which may be attributed to inefficiencies in powder spreading and packing during

printing. In some cases (e.g. 17µm – XY section, HIPed), the porosity can be observed to follow

linear patterns in the XY plane (plane of printing), along the Y direction (direction of printhead

traverse). This is due to clogged nozzles in the jetting head. It is seen that HIP does not remove

internal porosity effectively under these conditions.

15

These micrographs prove that there is a limit to the extent with which HIP can be useful in

eliminating porosity resulting from common process defects found in the Binder Jetting AM

process. Owing to the non-uniformity in the porosity of these sectioned parts, there was no fair

manner to sample sections to quantitatively determine the porosity in these parts. High-resolution

Computed tomography (CT) scanning is an alternative approach that may be used in the future to

quantify the three-dimensional porosity distribution in these cases.

Figure 2.4: Porosity in 17µm parts

16

Figure 2.5: Porosity in 25µm parts

In case of parts made using the bimodal powder, the issues mentioned above that were caused by

insufficient green part strength or powder packing were not observed. The porosity present was to

an extent that could largely be overcome by HIP, as can be seen in Figure 2.6. Due to the more

homogeneous nature of pore distribution in bimodal parts compared to the 17µm and 25µm parts,

a quantitative porosity measurement could be performed using image analysis (Section 2.2.4). The

porosity thus obtained was found to decrease from 2.90 ± 1.66 % as sintered, to 0.37 ± 0.21 %

after HIP. This quantitatively confirms the hypothesis of HIP improving the porosity in parts of a

sufficiently low starting porosity.

17

Figure 2.6: Porosity in bimodal parts

2.3.3 Microstructure

Sample micrographs for each part type and processing condition are presented in Figures 2.7 to

2.9. The grains in the case of both the 17µm (Figure 2.7) and the bimodal parts (Figure 2.9) are

largely equiaxed with significant twinning. HIP is clearly seen to coarsen this structure, as is

evidenced by the grain size measurements presented in Table 2.3.

The 25µm parts were found to form dendritic structures rather than the equiaxed grains otherwise

typically observed upon sintering (Figure 2.8). This is a natural microstructure for copper when it

is fully melted and rapidly solidified (especially in case of casting), whose formation depends on

various factors such as available surface area, oxygen content etc. [58]. While the formation of

these structures is not expected for the case of sintering and slow cooling, determining the exact

cause of their presence in these specimens with statistical accuracy requires further

experimentation that is beyond the scope of this study. However, it is observed that these dendritic

structures completely recrystallize to form equiaxed twinned grains upon HIP. The grain sizes

observed from the microstructures for these HIPed samples shows no immediately discernible,

qualitative difference in the grain sizes between the XY and YZ/XZ sections.

18

Figure 2.7: Grain structure change upon HIP – 17µm parts

Figure 2.8: Grain structure change upon HIP – 25µm parts

19

Figure 2.9: Grain structure change upon HIP – bimodal parts

The average grain sizes, as calculated using the Saltykov method along with the 95% confidence

interval, are presented in Table 2.3. The data shows that there is a significant coarsening of grains

upon HIP of 17µm and bimodal powder parts. This is to be expected due to high temperature

processing, however it is kept it at a minimal amount by rapid cooling at the end of the cycle. It is

particularly interesting to note that in the case of bimodal parts after sintering, the grain sizes are

well below those of the larger of the individual powder particles (30µm). The particle size of parent

powders in the cases of 25µm and bimodal parts are comparable to the grain sizes after HIP,

indicating only a reasonable amount of grain coarsening resulting from HIP.

20

Table 2.3: Grain size change upon HIP of various parts

Part

Type Plane

Sintered HIPed

% Theoretical

Density

Avg. Grain Diameter

± 95% CI

% Theoretical

Density

Avg. Grain Diameter

± 95% CI

17 µm XY

83.63 20.26 ± 3.01

85.83 37.05 ± 5.00

YZ/ZX 24.41 ± 5.85 30.89 ± 3.01

25 µm XY

77.73 -

82.41 24.88 ± 6.31

YZ/ZX - 23.38 ± 3.05

Bimodal XY

90.52 14.34 ± 1.34

97.32 24.08 ± 3.58

YZ/ZX 12.43 ± 2.23 27.43 ± 6.64

Differences in grain size between sections along the build plane (XY), and perpendicular to it

(YZ/ZX), were analyzed using a t-test for each part type. A statistically significant difference

between the orientations, at a 95% confidence interval, was found to be present only in the cases

of sintered bimodal specimens and HIPed 17µm specimens. This lack of a trend between sample

types could simply be a result of having had a limited number of samples (between 3 and 5 for

each) available for analysis. Randomizing the selection of regions for taking grain counts may

have brought about such a disparity. Further analysis will be required to derive physical meaning

from the grain orientation data.

2.3.4 Tensile Testing

HIP of Powder Bed Fusion AM processes typically results in a drop in strength as reported in

Section 2.1.2, due to a significant increase in the grain size. The data in Table 2.4 shows that HIP

improves the strength of all Binder Jet parts, indicating that the improvement in porosity is the

dominant factor compared to grain coarsening effects in determining strength in the case of Binder

Jetting. The maximum achievable strength from the literature for fully dense sintered copper

powder is 220 MPa [28], placing the tensile strength of 97.32% dense parts at only 80.16% of this

value. It is understandable that while HIP does improve the strength of Binder Jet parts, the

coarsened grains are going to result in a strength less than that achievable by sintering of loose

powders through traditional PM.

21

Figure 2.10: Samples of untested (top) and tested (bottom) tensile specimens

Figure 2.10 depicts a sample tension tested specimen to illustrate the scale of elongation. The

ductility is observed to increase significantly for the 25µm and bimodal powder parts, which

corresponds to an increase in their density. In case of the 17µm parts, the observed improvement

in ductility is minimal, which could be due to the comparatively smaller increase in density. The

ductility of similarly sintered pure copper powders is reported to be 45% [28]. Similar to the

strength, the ductility of all parts also falls well below this, with the most ductile specimens

(97.32% dense) at only 69.51% of this value in spite of grain coarsening.

Table 2.4: Tensile Strength and Ductility improvement upon HIP

Part

Type

Sintered HIPed

%

Theoretical

Density

Ultimate Tensile

Strength

(MPa)

% Elongation

%

Theoretical

Density

Ultimate Tensile

Strength

(MPa)

% Elongation

17 µm 83.63 115.84 ± 9.19 27.60 ± 2.58 85.83 135.31 ± 13.74 28.75 ± 1.27

25 µm 77.73 82.05 ± 5.34 12.07 ± 1.67 82.41 129.32 ± 0.94 30.38 ± 1.18

Bimodal 90.52 144.90* 17.87* 97.32 176.35 ± 6.48 31.28 ± 2.38

*Only one sample tested

The number of tested specimens was limited by available space in the print bed in addition to

damages during handling and depowdering of some green parts. The apparent premature fracture

of one set of samples (25µm, sintered) could later be traced back to a fabrication defect in these

specimens, caused by the presence of foreign particles in the powder. These results demand further

repetitions of the experiments to increase the sample size and make statistically significant

conclusions, as will be discussed in Section 2.4.

22

2.4 Closure and Future Work

Hot Isostatic Pressing (HIP) has been investigated as a secondary post-processing step in the

fabrication of high-density copper parts using Binder Jetting. The following conclusions may be

made based on the experimental results:

HIP improves density and porosity significantly only when the density of sintered parts is

at least 90%, which is achieved using a bimodal powder mixture of 30µm and 5µm powders

that has better packing properties than unimodal powders.

HIP causes dendritic structures to recrystallize into equiaxed grains, and previously

equiaxed grains to increase in size. The cause for the formation of these dendrites with the

most porous samples remains to be analyzed.

The tensile strength of samples increases upon HIP, indicating that the effect of reduction

in porosity outweighs that of grain coarsening in the case of Binder Jetting. The ductility

is found to improve when there is an increase in density due to HIP.

With careful control of powder quality and processing conditions, near-full density parts may be

printed. In future studies, copper parts will be characterized for various properties including

thermal and electrical conductivity. More work remains to be desired in order to get a complete

understanding of the mechanical properties improvement, particularly with regard to the effects of

part printing orientation on the tensile strength. The ductility improvement can be better

understood through a three-dimensional porosity distribution analysis that can be obtained using

techniques such as Computed Tomography (CT) scanning. An evaluation of the three-dimensional

pore distribution can give better insight into the nature of porosity and potentially explain the

observations regarding ductility increase. Lastly, a mathematical analysis of the variation of

material properties with porosity is part of the next phase of this study, as a step in the direction of

being able to tailor various processing parameters in order to achieve the density corresponding to

the desired properties.

2.5 References

[1] A. Kumar, Y. Bai, A. Eklund, and C. B. Williams, “Effects of Hot Isostatic Pressing on

Copper Parts Fabricated via Binder Jetting,” in Procedia Manufacturing, 2017, vol. 10, pp.

935–944.

23

[2] Y. Bai and C. B. Williams, “An exploration of binder jetting of copper,” Rapid Prototyp.

J., vol. 21, no. 2, pp. 177–185, 2015.

[3] B. D. Kernan et al., “Homogeneous steel infiltration,” Metall. Mater. Trans. A, vol. 36, no.

10, pp. 2815–2827, 2005.

[4] M. Lanzetta and M. Santochi, “Liquid-phase infiltration of thermal sintered skeletons by

low-temperature gold eutectic alloys,” CIRP Ann. - Manuf. Technol., vol. 55, no. 1, pp. 213–

216, 2006.

[5] J. Suwanprateeb and R. Chumnanklang, “Three-Dimensional Printing of Porous

Polyethylene StructureUsing Water-Based Binders,” J. Biomed. Mater. Res. B. Appl.

Biomater., vol. 78, no. 1, pp. 138–145, 2006.

[6] Z. C. Cordero, D. H. Siddel, W. H. Peter, and A. M. Elliott, “Strengthening of ferrous binder

jet 3D printed components through bronze infiltration,” Addit. Manuf., vol. 15, pp. 87–92,

2017.

[7] M. Turker, D. Godlinski, and F. Petzoldt, “Effect of production parameters on the properties

of IN 718 superalloy by three-dimensional printing,” Mater. Charact., vol. 59, no. 12, pp.

1728–1735, 2008.

[8] S. M. Gaytan et al., “Fabrication of barium titanate by binder jetting additive manufacturing

technology,” Ceram. Int., vol. 41, no. 5, pp. 6610–6619, 2015.

[9] J. A. Gonzalez, J. Mireles, Y. Lin, and R. B. Wicker, “Characterization of ceramic

components fabricated using binder jetting additive manufacturing technology,” Ceram.

Int., vol. 42, no. 9, pp. 10559–10564, 2016.

[10] M. Vaezi and C. K. Chua, “Effects of layer thickness and binder saturation level parameters

on 3D printing process,” Int. J. Adv. Manuf. Technol., vol. 53, no. 1–4, pp. 275–284, 2011.

[11] H. Miyanaji, S. Zhang, A. Lassell, A. A. Zandinejad, and L. Yang, “Optimal Process

Parameters for 3D Printing of Porcelain Structures,” Procedia Manuf., vol. 5, pp. 870–887,

2016.

[12] S. Shrestha and G. Manogharan, “Optimization of Binder Jetting Using Taguchi Method,”

24

Jom, vol. 69, no. 3, pp. 491–497, 2017.

[13] N. Karapatis, G. Egger, P. Gygax, and R. Glardon, “Optimization of powder layer density

in selective laser sintering,” in Proc. of Solid Freeform Fabrication Symposium, 1999, pp.

255–263.

[14] J. Zhou, Y. Zhang, and J. K. Chen, “Numerical simulation of laser irradiation to a randomly

packed bimodal powder bed,” Int. J. Heat Mass Transf., vol. 52, no. 13–14, pp. 3137–3146,

2009.

[15] A. B. Spierings, N. Herres, and G. Levy, “Influence of the particle size distribution on

surface quality and mechanical properties in AM steel parts,” Rapid Prototyp. J., vol. 17,

no. 3, pp. 195–202, 2011.

[16] M. Lanzetta and E. Sachs, “Improved surface finish in 3D printing using bimodal powder

distribution,” Rapid Prototyp. J., vol. 9, no. 3, pp. 157–166, 2003.

[17] B. Verlee, T. Dormal, and J. Lecomte-Beckers, “Properties of Sintered Parts Shaped by 3D-

Printing from Bimodal 316L Stainless Steel Powder Mixtures,” in Euro PM2011, 2011, pp.

357–362.

[18] R. M. German, “Prediction of sintered density for bimodal powder mixtures,” Metall. Trans.

A, vol. 23, no. 5, pp. 1455–1465, 1992.

[19] Y. Bai, G. Wagner, and C. B. Williams, “Effect of Particle Size Distribution on Powder

Packing and Sintering in Binder Jetting Additive Manufacturing of Metals,” J. Manuf. Sci.

Eng., vol. 139, no. 8, p. 81019, 2017.

[20] A. Bailey, A. Merriman, A. Elliott, and M. Basti, “Preliminary Testing of Nanoparticle

Effectiveness in Binder Jetting Applications,” 27th Annu. Int. Solid Free. Fabr. Symp., pp.

1069–1077, 2016.

[21] A. Elliott, S. AlSalihi, A. L. Merriman, and M. M. Basti, “Infiltration of Nanoparticles into

Porous Binder Jet Printed Parts,” Am. J. Eng. Appl. Sci., vol. 9, no. 1, pp. 128–133, 2016.

[22] Y. Bai and C. B. Williams, “Binderless Jetting: Additive Manufacturing of Metal Parts via

Jetting Nanoparticles,” Solid Free. Fabr. Proc., pp. 249–260, 2017.

25

[23] J. G. Bai, K. D. Creehan, and H. A. Kuhn, “Inkjet printable nanosilver suspensions for

enhanced sintering quality in rapid manufacturing,” Nanotechnology, vol. 18, no. 18, 2007.

[24] J. Grau, J. Moon, S. Uhland, M. Cima, and E. Sachs, “High green density ceramic

components fabricated by the slurry-based 3DP process,” Solid Free. Fabr. Proc., pp. 371–

378, 1997.

[25] X. Li, L. Zhang, and X. Yin, “Effect of chemical vapor infiltration of Si 3N 4 on the

mechanical and dielectric properties of porous Si 3N 4 ceramic fabricated by a technique

combining 3-D printing and pressureless sintering,” Scr. Mater., vol. 67, no. 4, pp. 380–

383, 2012.

[26] L. N. Rabinskiy, S. A. Sitnikov, V. A. Pogodin, A. A. Ripetskiy, and Y. O. Solyaev, “Binder

Jetting of Si3N4 Ceramics with Different Porosity,” Solid State Phenom., vol. 269, pp. 37–

50, 2017.

[27] H. Atkinson and S. Davies, “Fundamental aspects of hot isostatic pressing: an overview,”

Metall. Mater. Trans. A, vol. 31, no. 12, pp. 2981–3000, 2000.

[28] R. M. German, Powder Metallurgy Science (Second Edition). 1994.

[29] P. A. Kobryn and S. L. Semiatin, “Mechanical properties of laser-deposited Ti-6Al-4V,”

Solid Free. Fabr. Proc., pp. 6–8, 2001.

[30] C. Qiu, G. A. Ravi, C. Dance, A. Ranson, S. Dilworth, and M. M. Attallah, “Fabrication of

large Ti-6Al-4V structures by direct laser deposition,” J. Alloys Compd., vol. 629, pp. 351–

361, 2015.

[31] M. Agarwala, D. Bourell, J. Beaman, H. Marcus, and J. Barlow, “Post-processing of

selective laser sintered metal parts,” Rapid Prototyp. J., vol. 1, no. 2, pp. 36–44, 1995.

[32] D. W. Freitag, J. J. Beaman, and D. L. Bourell, “United States Patent 5640667,” 1997.

[33] S. Das, J. J. Beaman, M. Wohlert, and D. L. Bourell, “Direct laser freeform fabrication of

high performance metal components,” Rapid Prototyp. J., vol. 4, no. 3, pp. 112–117, 1998.

[34] S. Das, M. Wohlert, J. J. Beaman, and D. L. Bourell, “Producing metal parts with selective

laser sintering/hot isostatic pressing,” Jom, vol. 50, no. 12, pp. 17–20, 1998.

26

[35] K. Liu, Y. Shi, W. He, C. Li, Q. Wei, and J. Liu, “Densification of alumina components via

indirect selective laser sintering combined with isostatic pressing,” Int. J. Adv. Manuf.

Technol., vol. 67, no. 9–12, pp. 2511–2519, 2013.

[36] 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, 2016.

[37] S. Leuders et al., “On the mechanical behaviour of titanium alloy TiAl6V4 manufactured

by selective laser melting: Fatigue resistance and crack growth performance,” Int. J.

Fatigue, vol. 48, pp. 300–307, 2013.

[38] G. Kasperovich and J. Hausmann, “Improvement of fatigue resistance and ductility of

TiAl6V4 processed by selective laser melting,” J. Mater. Process. Technol., vol. 220, pp.

202–214, 2015.

[39] K. D. Rekedal and D. Liu, “Fatigue Life of Selective Laser Melted and Hot Isostatically

Pressed Ti-6Al-4v Absent of Surface Machining,” in 56th AIAA/ASCE/AHS/ASC

Structures, Structural Dynamics and Materials Conference, 2015, p. 894.

[40] S. Leuders, T. Lieneke, S. Lammers, T. Tröster, and T. Niendorf, “On the fatigue properties

of metals manufactured by selective laser melting – The role of ductility,” J. Mater. Res.,

vol. 29, no. 17, pp. 1911–1919, 2014.

[41] J. Haan, M. Asseln, M. Zivcec, J. Eschweiler, R. Radermacher, and C. Broeckmann, “Effect

of subsequent Hot Isostatic Pressing on mechanical properties of ASTM F75 alloy produced

by Selective Laser Melting,” Powder Metall., vol. 58, no. 3, pp. 161–165, 2015.

[42] L. Facchini, E. Magalini, P. Robotti, and A. Molinari, “Microstructure and mechanical

properties of Ti‐6Al‐4V produced by electron beam melting of pre‐alloyed powders,” Rapid

Prototyp. J., vol. 15, no. 3, pp. 171–178, 2009.

[43] A. Mohammadhosseini, D. Fraser, S. H. Masood, and M. Jahedi, “Microstructure and

mechanical properties of Ti–6Al–4V manufactured by electron beam melting process,”

Mater. Res. Innov., vol. 17, no. sup2, pp. s106–s112, 2013.

[44] S. L. Lu, H. P. Tang, Y. P. Ning, N. Liu, D. H. StJohn, and M. Qian, “Microstructure and

Mechanical Properties of Long Ti-6Al-4V Rods Additively Manufactured by Selective

27

Electron Beam Melting Out of a Deep Powder Bed and the Effect of Subsequent Hot

Isostatic Pressing,” Metall. Mater. Trans. A, vol. 46, no. 9, pp. 3824–3834, 2015.

[45] W. H. Peter et al., “Understanding the Role of Hot Isostatic Pressing Parameters on the

Microstructural Evolution of Ti-6Al-4V and Inconel 718 Fabricated by Electron Beam

Melting.” 2015.

[46] J. Hjärne and M. Ahlfors, “Hot Isostatic Pressing for AM parts,” Quintus Technologies.

2016.

[47] S. Tammas-Williams, P. J. Withers, I. Todd, and P. B. Prangnell, “The Effectiveness of Hot

Isostatic Pressing for Closing Porosity in Titanium Parts Manufactured by Selective

Electron Beam Melting,” Metall. Mater. Trans. A Phys. Metall. Mater. Sci., vol. 47, no. 5,

pp. 1939–1946, 2016.

[48] R. Kircher, A. Christensen, and K. Wurth, “Electron Beam Melted (EBM) Co-Cr-Mo Alloy

for Orthopaedic Implant Applications,” in Solid Freeform Fabrication Proceedings, 2009,

pp. 428–436.

[49] L. E. Murr et al., “Microstructural architecture, microstructures, and mechanical properties

for a nickel-base superalloy fabricated by electron beam melting,” Metall. Mater. Trans. A

Phys. Metall. Mater. Sci., vol. 42, no. 11, pp. 3491–3508, 2011.

[50] W. J. Sames, K. A. Unocic, R. R. Dehoff, T. Lolla, and S. S. Babu, “Thermal effects on

microstructural heterogeneity of Inconel 718 materials fabricated by electron beam

melting,” J. Mater. Res., vol. 29, no. 17, pp. 1920–1930, 2014.

[51] M. Seifi, I. Ghamarian, P. Samimi, P. C. Collins, and J. J. Lewandowski, “Microstructure

and Mechanical Properties of Ti-48Al-2Cr-2Nb Manufactured Via Electron Beam

Melting,” in Proceedings of the 13th World Conference on Titanium, 2016, pp. 1317–1322.

[52] J. A. Gonzalez, “Characterization and Comparison of Metallic and Ceramic Parts Fabricated

Using Powder Bed-Based Additive Manufacturing Technologies,” The University of Texas

at El Paso, 2017.

[53] ASTM B962-13, “Standard Test Methods for Density of Compacted or Sintered Powder

Metallurgy (PM) Products Using Archimedes’ Principle,” 2013.

28

[54] ASTM E112-13, “Standard test methods for determining average grain size.” 2013.

[55] S. A. Saltykov, Stereometric Metallurgy, Part 2. 1961.

[56] ASTM B925-15, “Standard Practices for Production and Preparation of Powder Metallurgy

( P / M ) Test.” 2015.

[57] ISO/ASTM 52921, “Standard terminology for Additive Manufacturing - Coordinate

systems and test methodologies.” 2013.

[58] G. F. Vander Voort, R. N. Caron, R. G. Barth, and D. E. Tyler, “Microstructures of Copper

and Copper Alloys,” in ASM Handbook, Volume 9: Metallography and Microstructures,

Vol. 9, 2004, pp. 775–788.

29

Chapter 3: Impacts of Process-Induced Porosity on Material Properties of

Copper Made by Binder Jetting Additive Manufacturing

Abstract

Additive Manufacturing (AM) can be used to manufacture copper parts featuring complex

geometries for use in advanced thermal and electrical components. Binder Jetting is a relatively

cheap and scalable AM technology that is suited for manufacturing highly conductive or reflective

materials like copper. This study aims to quantify the effects of varying amounts of porosity on

the material properties of Binder Jet copper parts, and compare them with existing models in the

literature. Towards this, copper parts of porosities ranging from 2.68% to 16.37% were fabricated

via Binder Jetting, by varying powder morphology and post-process sintering and Hot Isostatic

Pressing (HIP) conditions. Their properties have then been compared against various models in

the literature for property-porosity relations for Powder Metallurgy (PM) copper, or for general

two-component structures. This can pave the way towards developing a scientific understanding

of the process-property-performance relationship in Binder Jetting of copper, which can help in

achieving desired properties through the choice of appropriate materials and processing conditions.

3.1 Introduction

Copper is an important material with high thermal and electrical conductivities that can be used in

applications involving enhanced heat transfer and electrical and electronic components. Additive

Manufacturing (AM) of copper parts can enable the fabrication of the complex geometries that

may be required for such applications such as heat exchangers and rocket engine components with

internal cooling channels [1]. However, its high thermal conductivity makes it difficult to control

the melt pool if fabricating by laser- or electron beam-based Powder Bed Fusion (PBF) AM

technologies, and its high optical resistivity limits the choice of wavelengths of laser that may be

used [2]. Hence, processing pure, unalloyed copper powder using such systems can be challenging.

3.1.1 Prior Work on Powder Bed Fusion of Copper

Singer and co-authors provide a review on the some of the significant work on AM of copper and

copper alloys [3]. Pogson and co-authors were able to produce low-resolution, thin-walled

structures of copper on a stainless steel substrate using relatively low-power laser (by Direct Metal

Laser Re-melting, or DMLR), but their methodology was unlikely to be successful in producing

high strength parts that are commercially realizable due to insufficient melting of powders [4].

30

Lykov and co-authors were able to achieve 88.1% density and a strength of only 149 MPa in

copper parts fabricated using Selective Laser Melting (SLM) [5]. Kaden et al. studied the use of

ultrashort laser pulses in the fabrication of both thin-walled and solid cuboid structures, but these

were porous owing to large grains formed due to the laser spot diameter being roughly the same

as the powder particle size [6]. In the industry, 3T RPD, a company based in the UK, claimed to

have produced pure copper parts using Direct Metal Laser Sintering (DMLS) by suitably

modifying the machine, calibrating process parameters, and modifying support structure designs

[7]. There is yet to be published work on the exact modifications applied, or the release of a

production-ready part. The Fraunhofer Institute for Laser Technology has announced work on the

development of a green laser light of 515nm wavelength that can be better absorbed by pure copper

and copper alloys for SLM [8].

Electron Beam Melting (EBM) is an alternative that can circumvent some issues faced with laser-

copper interactions. Ramirez et al. used EBM to fabricate open- cellular structures using precursor

powder (of 99.8% purity) containing Copper Oxide (Cu2O) precipitates, which reflected as

microstructural copper oxide arrays in the fabricated part [9]. Precipitation-dislocation

architectures were observed in parts fabricated via EBM using powders of relatively low purity of

98.5% [10]. Yang et al. fabricated auxetic structures made of pure copper using EBM, but were

unable to achieve consistent dimensional accuracy and strength owing to fabrication defects

arising due to the high thermal conductivity of copper [11]. Frigola et al. circumvented issues with

the high thermal conductivity of copper, resulting in high thermal gradients and hence

delamination, to form complex parts with internal cooling channels [12]. They mention the

importance of using high vacuum, and concerns with reusing powders that can result in detrimental

oxide formation. Terrazas et al. explored the AM of multi-material structures made of Ti-6Al-4V

and Copper in tandem using EBM [13]. The copper portion was observed to be misaligned,

warranting better part registration. The microstructure was observed to transition from columnar

grains away from the interface between the materials, to an equiaxed structure near the interface,

which could have acted as a substrate to facilitate epitaxial growth. Twinning was observed after

recrystallization caused by HIP. Lodes and co-authors fabricated up to 99.5% dense copper by

optimizing process parameters of EBM, but were yet to characterize the material properties [14].

In summary, a majority of these methods developed have had challenges in fabricating high-

density copper parts with minimal specialized modifications to the processes or the machines used.

31

3.1.2 Binder Jetting of Copper

Binder jetting is an alternative to these Powder Bed Fusion AM technologies that involves

selectively jetting a binder onto layers of powder to form a ‘green’ part by curing the binder (Figure

3.1). This is followed by debinding and sintering in a furnace for densification and strength. It

effectively circumvents issues caused by the high thermal conductivity and reflectivity of copper

by separating the part creation from energy supply. The primary disadvantage of this process,

however, is its inability to sufficiently densify the green parts. This is because the spreading of

layers does not involve any compaction or agitation, resulting in powder that is loosely packed.

The residual porosity in fabricated parts can be detrimental to their material properties. In the case

of copper, Binder Jetting has been used to achieve up to 99.47% density through the use of

optimized powder morphologies [15], printing parameters and sintering profiles [16], and post-

process Hot Isostatic Pressing (HIP) [17]. Various intermediate densities achieved using different

powders and post-processing conditions have been discussed in Chapter 2.

Figure 3.1: A schematic of the binder jetting process

3.1.3 Research Objective

In this article, the authors characterize the influence of this process-induced porosity on various

material properties, namely strength, ductility, and thermal and electrical conductivity. This is

done by analyzing the properties of specimens processed to achieve varying degrees of porosity

by varying powder morphology and post-process conditions. These properties are then compared

to equations and data in the literature in the context of porous copper prepared using powder

metallurgy, or to predictions based on models for two-component structures.

A comparison with PM copper is appropriate since Binder Jetting is closest in nature to PM

processes, albeit with loosely packed powders and no compaction. However, because of this

difference between the processes, it has been seen that powder metallurgy knowledge does not

directly translate to Binder Jetting. For instance, Bai and co-authors presented that the use of

32

bimodal powders in Binder Jetting yields higher density than when using small powders which is

contradictory to PM processes involving powder compaction, where finer powders may be

favorable [15]. Similarly, two-component system based models need not directly reflect Binder

Jet part properties either. These models are instead used in this study as a reference for comparison

to Binder Jetting, and to see if any of them can model this process well. The long-term goal is to

use these models as a foundation to develop process-structure-property relationships for Binder

Jetting.

Section 3.2 presents the various models that are used to study the measured properties, including

strength and ductility (3.2.1), thermal and electrical conductivity (3.2.2) and the Wiedemann-Franz

Law that provides a relationship between the two conductivities (3.2.3). Section 3.3 presents the

experimental methods used in fabricating the specimens of varying porosity (3.3.1), testing their

tensile (3.3.2), thermal (3.3.3) and electrical conductivity (3.3.4) properties. Section 3.4 presents

the results from the analyses of strength (3.4.1), ductility (3.4.2), thermal and electrical

conductivities (3.4.3 and 3.4.4) and the Wiedemann-Franz Law (3.4.5). Conclusions of the study

and suggestions for future work are presented in Section 3.5.

3.2 Modeling Porosity-Property Relationships

The goal of this work is to map the process-induced porosity to material properties, and hence lay

a framework for developing process-structure-property relationships for Binder Jetting of copper.

This is to be done by understanding the correlation between processing conditions and porosity,

which has been explored in Chapter 2, and that between process-induced porosity and material

properties as explored in this work. The models used in developing these porosity-property

relationships are presented here.

3.2.1 Strength and Ductility

The tensile strength of porous PM specimens can be estimated to vary with density by the Equation

3.1 [18]:

𝜎 = 𝜎𝑜𝐾 (𝜌

𝜌𝑡)

𝑚

( 3.1 )

where 𝜎 is the tensile strength of the porous material, 𝜎0 is the wrought strength of the same

material (220 MPa for copper), 𝜌 and 𝜌𝑡 are the density of the porous specimen and the theoretical

density of the material respectively. 𝐾 is equivalent to a stress concentration factor due to the

33

pores, and 𝑚 is the exponential dependence of strength on porosity. Both parameters depend on

the processing conditions, and the equation is valid in the absence of any microstructural problems.

German also presents data on the strength of PM copper with various degrees of porosity in his

book [18]. This data presented by German was fitted to the model he proposed in Equation 3.1 in

order to obtain the values of parameters 𝐾 and 𝑚 empirically to be 0.9926 and 2.5150 respectively

for PM copper (with an R2 value of 0.9998). Hence, the equation for sintered PM copper is found

to be as below:

𝜎 = 218.367 (𝜌

𝜌𝑡)

2.5150

( 3.2 )

Equation 3.3 gives the dependence of ductility on porosity as proposed by Haynes [19]:

𝑍 = {

(1−𝜀)1.5

(1+𝑐𝜀2)0.5 , 𝜀 < 0.15

(1−𝜀)1.5

(1+0.152𝑐)0.5 , 𝜀 > 0.15 ( 3.3 )

where 𝑍 is the relative ductility (ratio of observed elongation of porous material to wrought

material), 𝜀 is the porosity, and 𝑐 is a coefficient representing the sensitivity of the ductility to

porosity for the material under consideration. Fitting the ductility data provided by German [18]

for various porosities under 15% into the model in Equation 3.3, the value of 𝑐 is obtained as

210.237, at an R2 value of 0.9931. The elongation of wrought copper is taken as 0.45 based on the

same data. Hence, the empirical equation for the ductility dependence on porosity for PM copper

is found to be:

𝑍 = {

(1−𝜀)1.5

(1+210.237𝜀2)0.5 , 𝜀 < 0.15

0.4177(1 − 𝜀)1.5 , 𝜀 > 0.15 ( 3.4 )

These equations and calculated empirical constants are applicable for porosities achieved through

varying degrees of sintering of powders, and do not represent the same kind of processing as

Binder Jet parts. These values are hence used here as a reference to present where the ductility of

binder jet, sintered / HIPed copper parts stand in comparison to existing Powder Metallurgy data.

This may be used as a framework for future studies that may explore deriving new equations for

the processing conditions associated with binder jetting and sintering.

34

3.2.2 Thermal and Electrical Conductivity

There are a number of models in the literature that describe the variation of thermal conductivity

with porosity. A model proposed by Aivazov and Domashnev proposes the following relation for

the variation of thermal and electrical conductivity with porosity [20]:

𝐾 = 𝐾01−𝜀

1+𝑛𝜀2 ( 3.5 )

where 𝐾 is the thermal conductivity of the porous specimen, 𝐾0 that of pure wrought copper (388

W/m-K), 𝜀 is the porosity and 𝑛 is an experimentally determined constant. In regimes of porosity

less than 0.3, the relation may be approximated as follows [18]:

𝐾 = 𝐾0(1 − 𝜔𝜀) ( 3.6 )

where 𝜔 is a constant with a value between 1 and 2, also determined experimentally. Data

presented by German for his experiments on sintering of copper powders was fitted to this model

to give the value of for 𝜔 to be 1.1360 (at an R2 value of 0.9999) for thermal conductivity, and

1.1228 for electrical conductivity (at an R2 value of 0.9999). These values, although determined

for copper at a different kind of processing, are taken as a reference to compare against the

conductivity data obtained from Binder Jet copper.

Other models discussed in the literature [21 – 23] explore the behavior from a physical perspective

by modeling the porous material as a two-phase structure, and assuming various kinds of

dispersions and shapes of one constituent (pores) with regard to the other (parent material). In all

the equations described below, 𝐾 refers to the effective conductivity, 𝑘1 and 𝑣1 refer to the

conductivity and volume fraction of the parent material (copper) and 𝑘2 and 𝑣2 to those of the

dispersed medium (pores), respectively.

The Parallel model assumes that the phases are aligned parallel to the heat flow, providing

alternating parallel conduction pathways. This gives the upper bound to the effective thermal

conductivity, as given by Equation 3.7 [24]:

𝐾 = 𝑘1𝑣1 + 𝑘2𝑣2 ( 3.7 )

35

The Maxwell-Eucken model assumes a distribution of non-interacting spherical pores within the

copper medium, for which the effective conductivity is given by the equation below [21]:

𝐾 =𝑘1𝑣1+𝑘2𝑣2

3𝑘12𝑘1+𝑘2

𝑣1+𝑣23𝑘1

2𝑘1+𝑘2

( 3.8 )

The Effective Medium Theory (EMT) model assumes a random distribution of two phases, for

which the effective conductivity is given by Equation 3.9 [23]:

𝐾 = 0.25[(3𝑣2 − 1)𝑘2 + (2 − 3𝑣2)𝑘1 + {[(3𝑣2 − 1)𝑘2 + (2 − 3𝑣2)𝑘1]2 + 8𝑘1𝑘2}]0.5 ( 3.9 )

A modified form of the EMT has also been proposed that takes into consideration grain boundary

thermal resistance effects. The equation for this is as follows [23]:

𝐾𝑝𝑜𝑙𝑦 = [1

𝑘𝑠𝑖𝑛𝑔𝑙𝑒+ 𝑛𝑅𝑡ℎ]

−1

( 3.10 )

where 𝐾𝑝𝑜𝑙𝑦 is the thermal conductivity of the polycrystalline matrix, 𝑘𝑠𝑖𝑛𝑔𝑙𝑒 is that of pure, single

crystal material in the absence of grain boundaries, 𝑛 is the number of grain boundaries per unit

length, and 𝑅𝑡ℎ is the thermal resistance of the grain boundaries.

The observed thermal and electrical conductivity are compared against that predicted from each

of the equations 3.6 – 3.9 and presented in Section 3.4.3 and 3.4.4. Additional analyses are done

by using the observed thermal conductivity for each processing condition in the EMT and modified

EMT models to calculate the conductivity of the copper matrix in each case, and further estimate

the grain boundary thermal resistance, 𝑅𝑡ℎ to compare to what is estimated in the literature [23,

24]. This is presented in Section 3.4.3.

3.2.3 Wiedemann-Franz Law

The electrical and thermal conductivities are related by a semi-empirical relation called the

Wiedemann-Franz Law, which is given by Equation 3.11:

𝐾 = 𝐿𝑇

𝜌+ 𝑏 ( 3.11 )

where 𝐿 is the Lorenz function, 𝑇 is the temperature, and 𝜌 is the electrical resistivity. 𝐿𝑇

𝜌 and 𝑏

represent the material-dependent electron and lattice (phonon) contributions to thermal

conductivity respectively. In the case of copper, previous literature has proposed the values of 𝐿

36

and 𝑏 to be 2.39 x 10-8 and 0.075 [25 – 27]. Koh and Fortini in their study of porous PM copper

determined them to be 2.307 x 10-8 and 18.6, respectively [29].

One approach to a modification of the Wiedemann-Franz Law for the specific case of Binder Jet

copper could be to fit the conductivity data to determine a modified equation [30]. However, in

this study, the observed thermal conductivity data is directly compared to predictions using Koh

and Fortini’s model for OFHC copper powders sintered to varying porosities [29]. They state that

the constants are expected to independent of porosity and dependent only on the material. It is

assumed here that the difference in powder size and type, and the processing conditions in this

study as compared to theirs, has a negligible effect on these constants. The results from this

analysis are discussed in Section 3.4.5.

3.3 Experimental Methods

3.3.1 Specimen fabrication

Figure 3.2 shows a picture of the print bed during printing, indicating the layout of various

specimens printed for testing.

Figure 3.2: Photograph depicting layout of parts in print bed:

A: Rods for electrical conductivity measurements, B. Dogbones for Tensile Testing,

C. Rectangular coupons for density measurements, D. Discs for thermal conductivity measurements

37

Previous work by the authors on using different powder configurations to explore the HIP of parts

of varying sintered densities (Chapter 2) was used in achieving differing levels of porosity. A

schematic of the process chain is illustrated in Figure 3.3, showing the process variables and

control variables for each stage of the process chain (detailed explanations in Section 2.2). To

summarize, the illustrated parameters were used to print and sinter parts using powders with

median sizes of 17µm and 25µm and a bimodal combination of powders with median sizes 30µm

and 5µm, mixed in the ratio of 73:27 by weight. Some of these were HIPed at 1075°C, 206.84

MPa (30,000 psi) for 2 hours for analysis of HIPed specimens, and others retained for analyses of

sintered specimens. Different amounts of porosity were thus achieved.

Figure 3.3: Schematic of process chain

Density measurements were done using an Archimedes principle based apparatus to measure the

weight of parts in air and in water and calculating the relative density. These specimens were oil

impregnated for accuracy in measurements (Sections 2.2.3 and 2.3.1). The processing conditions

used to obtain specimens of each density and the corresponding porosity are presented in Table

3.1.

38

Table 3.1: Powders and processing conditions used to achieve different porosities

Powder Density (% Theoretical) Porosity (%)

Bimodal HIPed 97.32 2.68

Sintered 90.52 9.48

17 µm HIPed 85.83 14.17

Sintered 83.63 16.37

25 µm HIPed 82.41 17.59

Sintered 77.73 22.27

(a) (b)

Figure 3.4: Porosity in 17µm specimen (YZ plane) : (a) Sintered – 16.37%; (b) HIPed – 14.17%

(a) (b)

Figure 3.5: Porosity in bimodal specimen (YZ Plane): (a) Sintered – 9.48%; (b) HIPed – 2.68%

39

Optical porosity characterizations (Section 2.2.4) were done by polishing samples in successively

finer grit papers and polishing cloths to remove scratches on the surface, and observed under an

optical microscope. Three samples of each type were sectioned along all three planes for analysis.

Some representative images are provided in Figure 3.4 (17µm specimens) and Figure 3.5 (bimodal

specimens). These indicated that the porosity distribution is heterogeneous in the case of 17µm

and 25µm specimens, with more pores concentrated near the surface and between layers. Such a

porosity distribution is different from those assumed in the models discussed in Section 3.2. In the

case of bimodal parts, while the porosity distribution is more uniform, the processing conditions

and powders result in microstructural issues that are not accounted for in these models (Section

2.3.3). This invalidates the direct applicability of these models to Binder Jetting, and hence only a

comparison may be made between the data available and that expected based on those models.

3.3.2 Tensile Testing

Tensile testing data has previously been reported in Section 2.3.4. The same data has been used to

model the strength and ductility as a function of porosity. As reported in that work, the testing was

carried out per the ASTM Standard E8/E8-M -16a [31] using flat unmachined test specimens

(Section 2.2.5). The specimens were printed randomly in the XYZ or the YXZ orientations (Figure

3.2) [32]. The tests were run on an INSTRON machine using a 50kN load cell at a constant strain

rate of 0.015 mm/mm/min per the recommendation in the standard (elongation rate of 0.381

mm/min) for a gauge length of 1 inch (25.4 mm). The ultimate tensile strength at break is

measured, and the ductility is reported in terms of percentage elongation of gauge length.

3.3.3 Thermal Conductivity Measurement:

The thermal conductivity of these specimens of various densities was calculated by measuring the

thermal diffusivity using a laser flash diffusivity apparatus. The procedures followed were based

on the ASTM Standard E1461-13 [33]. The samples fabricated for these measurements were thin

cylinders with a sintered diameter of 10mm and a thickness of ~2 mm (Figure 3.2). These were

then sprayed with a thin layer of graphite in order to avoid the laser flashes from reflecting off the

surface. The laser pulse on one flat surface of the cylindrical specimen causes a rise in the

temperature of the entire part, with the temperature of the opposite face being monitored by the

apparatus. The time taken for the temperature of this face to reach half the maximum value is

40

calculated. This, along with the thickness of the specimen, 𝐿 gives the thermal diffusivity,

according to Equation 3.12 below:

𝛼 =0.13879𝐿2

𝑡12

( 3.12 )

where 𝛼 is the diffusivity, 𝐿 is the thickness of the sample and 𝑡1

2

is the time taken for the

temperature of the far side to reach half of the equilibrium value. The thermal conductivity can

then be calculated from the Equation 3.13 below:

𝜆 = 𝛼𝐶𝑝𝜌 ( 3.13 )

where 𝜆 is the thermal conductivity, 𝐶𝑝 is the specific heat capacity and 𝜌 is the density of the

specimen. Here, the specific heat capacity was measured using the Laser Flash apparatus and the

density from Archimedes Principle based measurements (3.3.1).

3.3.4 Electrical Resistivity Measurement

The resistivity of the samples was measured using a four-wire measurement apparatus as is

illustrated in Figure 3.6. For this purpose, the samples printed were thin cylindrical rods (Figure

3.2). A known current is passed through a measured length of the rod, and the corresponding drop

in voltage is measured. The resistivity can be calculated using Equation 3.14:

𝜌 =𝑅∗𝐴

𝐿=

𝑉∗𝐴

𝐼∗𝐿 ( 3.14 )

where 𝜌 is the resistivity, 𝑅 is the resistance, 𝑉 is the measured voltage drop, 𝐼 is the known current

passed, 𝐴 is the cross sectional area and 𝐿 is the length of the rod.

41

Figure 3.6: Electrical Resistivity measurement setup:

A: Power source, B: Multimeter, C. Tested sample, D: Connecting wires with alligator clips

Owing to the relatively low resistivity of copper, in order for the voltage to be measurable using

the available multimeter, the length of the rod and the current passed through it had to be

maximized, while the cross sectional area had to be minimal. Hence, a current of 10A was passed

using a high-current power source. Given the constraints of the printing process, and taking into

account that the as-printed green parts must be able to withstand handling and depowdering prior

to sintering, the parts were designed to have a diameter of 2mm and a length of ~150 mm. The

results from the resistivity measurements are presented in Section 3.4.4. Conductivity values are

presented either in S/m, or as a percentage of the International Annealed Copper Standard, or IACS

(100% IACS = 5.8001e7 S/m).

3.4 Results

3.4.1 Tensile Strength

Figure 3.7 compares the strength of the material obtained for each of the processing conditions, to

the values expected based on Equation 3.2 (Section 3.2.1).

42

Figure 3.7: Tensile strength observed in comparison to Model Predictions

It is evident that the tensile strength of both sintered and HIPed copper is below the value expected

based on the models. The disparity may be expected due to microstructural differences caused by

the difference in powder types and processing conditions as compared to the model. This results

in different grain structures for each of the specimens as compared to those considered in the

model. The powder size can dictate the grain sizes in the sintered parts, and the HIP can further

recrystallize or coarsen each of these (Section 2.3.3). A detailed evaluation of the grain structures

and nature of porosity in the sintered as well as HIPed samples using 3-dimensional visualization

techniques may provide further light into the significance of these findings. Additionally, studying

the influence of printing orientation rather than randomizing them as in this study may provide

further information regarding the isotropy of strength of parts fabricated using this process chain.

3.4.2 Ductility

Figure 3.8 presents the ductility of the fabricated specimen along with the predicted values based

on the model given in Equation 3.4.

43

Figure 3.8: Ductility – Observed vs. Model

The as-sintered 25µm specimens were observed to have failed prematurely, which was later

reasoned to have been due to a fabrication defect. In addition to this, the fact that there was only

one sample in the bimodal, sintered category available for testing, warrants further investigation

for statistically significant conclusions. Hence, without quantifying the degree by which the values

observed differ from those predicted, it may only be safe to make qualitative conclusions. The

printed parts have higher ductility than the values expected from the model in the case of the 17µm

and 25µm parts (~15% and higher porosity), while the model seems to over-predict the ductility

in the case of bimodal parts (<10% porosity). In the case of parts with open porosity (>~10%), this

could be due to the distribution of porosity and layer interface bonding defects in binder jet parts

being localized to near the surface as compared to the core as seen in the sample micrographs of

17µm specimens in the YZ plane (Figure 3.4). This is different from the uniform spread of pores

that may be expected from sintered powders, as assumed in the model.

In the case of lower porosity (i.e., bimodal) parts, the surface porosity is completely sealed, with

the majority of remaining porosity homogeneously distributed in the inner regions (Figure 3.5).

This porosity distribution may be closer to what is obtained by sintering of loose powders to similar

densities, and the difference in ductility may be due to factors such as difference in grain structure,

purity of the copper powders used, oxide formation etc. Detailed investigations in the future may

help formulate more accurate mathematical relations to express the variation of ductility with

porosity for these processing conditions.

44

3.4.3 Thermal Conductivity

The results from thermal conductivity obtained using the Laser Flash method are presented in

Figure 3.9 along with a comparison to the various models presented in Section 3.2.2.

Figure 3.9: Thermal Conductivity vs. Porosity: Observed data compared to models

It is seen that all the models overestimate the conductivity in comparison to the observed data.

This is due to a multitude of reasons. Firstly, there is a disparity in the shapes and distribution of

the pores assumed in comparison to each of the models, as explained in Section 3.2.2. In addition,

none of these models account for the presence of grain boundaries of varying types and sizes that

add to the thermal resistance, or for any processing defects in the processed samples.

The observed conductivity data fitted to a linear model shows a reasonably linear trend, with the

forecasted value of conductivity for fully dense parts predicted to be 335.18 W/m-K. This value

being lower than ideally expected (388 W/m-K) can be attributed to impurity of the copper

powders used, residual carbon after the binder burn-off, and the presence of thermal resistance at

the grain boundaries. However, since each of these data points represents a different powder /

processing condition, the trend and the forecast are to be taken merely as a guideline.

Hence, instead of using such a generalized linear trend to describe the data, the EMT model

(Equation 3.9) may be used to estimate the value of 𝑘1 by substituting the value of observed

conductivity for 𝐾. The equation for 𝑘1 expressed in terms of 𝐾 is as follows:

45

𝑘1 =2𝐾2+(𝑣1−2𝑣2)𝑘2𝐾

2𝑣1𝐾+𝑘2−𝑣2𝐾 ( 3.15 )

This expression returns the conductivity of the pore-free copper matrix for each processing

condition, accounting for effects of grain boundaries, impurities and any such non-porosity-related

effects. This calculated value of 𝑘1 will henceforth be referred to as 𝑘𝑓𝑑, for conductivity at full

density for a given powder and processing condition. The Figure 3.10 below plots the value of 𝑘𝑓𝑑

calculated for each of the processing conditions.

Figure 3.10: Conductivity at full density (𝑘𝑓𝑑) calculated for each processing condition

The plot shows that for most cases, the conductivity of the pore-free matrix lies between 335 and

350 W/m-K. This can be taken to provide a physical validation for the linear trendline-based

estimate. The comparatively high conductivity for the case of as-sintered 25µm parts can be

attributed to the difference in their microstructure (dendritic) as compared to that of the other

specimens (equiaxed, twinned) which may allow for differing mechanisms of conduction and

inaccuracies in the model prediction (Section 2.3.3).

As the next step in attempting to understand the physics of heat conduction in copper processed in

this manner, an attempt is made to calculate the grain boundary thermal resistance in each of these

kinds of parts, excluding the data from sintered 25µm specimen due their having dendritic

microstructure. This is done by taking the value of 𝑘𝑓𝑑 as that of 𝑘𝑝𝑜𝑙𝑦 in Equation 3.10. Re-writing

the equation here,

46

𝐾𝑝𝑜𝑙𝑦 = 𝑘𝑓𝑑 = [1

𝑘𝑠𝑖𝑛𝑔𝑙𝑒+ 𝑛𝑅𝑡ℎ]

−1

( 3.16 )

The value of 𝑛 is estimated based on the grain size calculations done in Section 2.3.3. The total

number of grains per unit length, is approximated as being equal to the total length (1 meter)

divided by the average grain diameter, 𝑑𝑔(in meters). The number of grain boundaries encountered

per unit length is then taken to be the next closest whole number to this value. For instance, if it is

estimated that there are an average of 10,000.23 grain boundaries per meter, then the value of 𝑛 is

taken to be 10,001 m-1 per the equation:

𝑛 = 𝑐𝑒𝑖𝑙𝑖𝑛𝑔 (1

𝑑𝑔) ( 3.17 )

Re-arranging Equation 3.16 for 𝑅𝑡ℎ yields the following equation:

𝑅𝑡ℎ =1

𝑛(

1

𝑘𝑓𝑑−

1

𝑘𝑠𝑖𝑛𝑔𝑙𝑒) ( 3.18 )

where 𝑘𝑠𝑖𝑛𝑔𝑙𝑒 is taken as 388 W/m-K. The results of calculating the grain boundary thermal

resistance for each processing condition are presented Figure 3.11 below.

Figure 3.11: Grain Boundary Thermal Resistance calculated from Modified EMT Model

The data shows that the thermal resistance of grain boundaries in the cases of Binder Jet copper

subjected to sintering / HIP is up to an order of magnitude greater than that estimated from the

literature [23, 24]. This is to be expected due to the difference in processing and measurement

47

conditions between the parts in this study and that in the literature. A physical interpretation of the

quantitative differences in the estimates is beyond the scope of this work.

3.4.4 Electrical Conductivity

A similar analysis is done for the resistivity measurements to compare with the same set of models

as in Section 3.4.3, and the results are presented in Figure 3.12.

Figure 3.12: Electrical Conductivity vs Porosity: Observed data compared to models

The electrical conductivity is observed to follow a trend similar to that reported for thermal

conductivity in that the models consistently overestimate their values. However, the electrical

conductivity seems to approach a value of 5.60e7 S/m (or 96.63% IACS), which is closer to the

theoretical value of 5.8001e7 S/m (100% IACS) at full density than the thermal conductivity

projection (86.39%). However, they are expected to be analogous in their variation. This may be

seen as due to differences in the printing and measurement orientations. The electrical resistivity

specimen were all rods printed parallel to the build plane on the printer (XYZ orientation, per

[32]), with the measurements taken along the length (Y-direction). Whereas, the thermal

conductivity specimens were disks printed with the thickness along the build direction (XYZ

orientation), which is the direction of measurement (Z-direction). Hence, layer interface bonding

issues in the higher-porosity specimens (Figure 3.4) may cause disparity in these values and their

relationships. This difference is further investigated in Section 3.4.5 using the context of the

Wiedemann-Franz Law.

48

3.4.5 Wiedemann-Franz Law

The thermal conductivity measured for various porosities is plotted against that calculated by the

Wiedemann-Franz Law as described by Koh and Fortini for PM copper (Section 3.2.3) to compare

and verify the applicability of the Law to the processes under consideration (Figure 3.13).

Figure 3.13: Observed Thermal Conductivity vs. that calculated using the Wiedemann-Franz Law

The figure shows that the model underestimates the thermal conductivity for higher porosity (lower

electrical conductivity) specimens and overestimates it for lower porosity specimens. This could

potentially be because of the differences in the powder material used as compared to their model.

However, since the parts do exhibit directional porosity (layer bonding defects in higher porosity

specimens) and the measurements of thermal and electrical conductivity were taken in different

directions, it is possible that this anisotropy may have caused such a trend. Further research into

the anisotropy in porosity may be done using three-dimensional porosity visualization techniques

such as Computed Tomography (CT) scanning. Investigating the potential directional variations

of thermal and electrical conductivity will require more detailed experiments that are beyond the

scope of this work.

49

3.5 Conclusions

The effects of porosity on various material properties of copper fabricated by Binder Jetting of

three different powder types, followed by sintering and Hot Isostatic Pressing, have been

investigated in comparison to models in the literature for copper parts with porosity achieved

through other processing methods. The major findings have been summarized below:

The strength of copper parts is found to be less than that predicted by the model in Equation

3.2 (Figure 3.7), which may be attributed to the presence of microstructural issues and

differences in the grain structure due to differences in powder material and/or

manufacturing techniques.

The ductility predictions based on Equation 3.4 overestimate the elongation in the low-

porosity regimes due to microstructural and compositional differences, and underestimate

the same in the high-porosity regimes where the porosity spread is localized to the surface

as compared to uniformly distributed pores in sintered Cu powders (Figure 3.8)

The thermal conductivity has been compared to various models in the literature, all of

which overestimate the influence of porosity. A generalized linear trend, while not

conclusive, predicts a maximum thermal conductivity of 335.18 W/m-K (Figure 3.9).

This value being less than that of PM copper (388W/m-K) is due to carbon residue from

binder pyrolysis, presence of oxides and differences in grain structure for Binder Jet

copper. A validation for this value has been performed by determining an approximate

range for the maximum possible conductivity for each processing condition using the EMT

model (Figure 3.10). Further, the grain boundary thermal resistance has been calculated for

each case and found to be greater than what may be assumed based on the literature (Figure

3.11).

The electrical conductivity has been found to compare similarly with models existing in

the literature. The prediction for electrical conductivity at full density, based on a

generalized linear trend similar to that employed for thermal conductivity, is 96.63% IACS

(Figure 3.12). Using these values to estimate the thermal conductivity using the

Wiedemann-Franz Law shows a variation in the disparity between the predicted and

observed thermal conductivity (Figure 3.13). This could potentially be due to the difference

50

in measurement direction of each conductivity relative to the layer interfaces, which is

influenced by the observed directionality in porosity (Figure 3.4).

To summarize, the properties analyzed are observed to consistently fall below the predictions

based on both Powder Metallurgy and two-component structural models. In the case of a

comparison with PM copper, this is attributed to the less uniform porosity distribution, differences

in purity of powders and in grain structures arising from different powders and processing

conditions for each level of porosity achieved using Binder Jetting. A more elaborate analysis of

achieving varying degrees of porosities through simply changing the sintering or HIP conditions

for each of the powders is required to derive relations similar to PM equations (Section 3.2) for

Binder Jetting. The predictions for thermal and electrical conductivity based on models for two-

component systems are also higher than the observed values, due to each of these models assuming

a different kind of porosity distribution than that observed for each of the Binder Jet specimens.

Better approximations for both the porosity distribution and the influence of grain boundary effects

and impurities can help develop accurate, physics-based two-component models for thermal and

electrical conductivity predictions in the future.

This study has placed results for AM of copper within the context of theoretical models for other

processing methods to achieve similar levels of porosity. This presents opportunities for future

work in developing specialized models for the particular cases of Binder Jetting, and more

elaborate studies of the effects of powder types, printing orientation, measurement direction,

processing parameters etc. Developing such models will enable one to print copper parts of desired

material properties using Binder Jetting by adjusting the material and processing parameters to

achieve the corresponding degree and type of porosity.

3.6 References

[1] T. McMahan, “NASA 3-D Prints First Full-Scale Copper Rocket Engine Part,” NASA Press

Office, 2015. [Online]. Available: https://www.nasa.gov/marshall/news/nasa-3-D-prints-

first-full-scale-copper-rocket-engine-part.html.

[2] F. Sciammarella, M. J. Gonser, and M. Styrcula, “Laser Additive Manufacturing of Pure

Copper,” in Rapid, 2013.

[3] F. Singer, D. C. Deisenroth, D. M. Hymas, and M. M. Ohadi, “Additively manufactured

copper components and composite structures for thermal management applications,” 16th

IEEE ITHERM Conf., pp. 174–183, 2017.

51

[4] S. R. Pogson, P. Fox, C. J. Sutcliffe, and W. O’Neill, “The production of copper parts using

DMLR,” Rapid Prototyp. J., vol. 9, no. 5, pp. 334–343, 2003.

[5] P. A. Lykov, E. V. Safonov, and A. M. Akhmedianov, “Selective Laser Melting of Copper,”

Mater. Sci. Forum, vol. 843, pp. 284–288, 2016.

[6] L. Kaden et al., “Selective laser melting of copper using ultrashort laser pulses,” Appl. Phys.

A, vol. 123, no. 9, p. 596, 2017.

[7] “3T produces Pure Copper Heat Exchanger using metal 3D Printing.” [Online]. Available:

http://www.3trpd.co.uk/3t-success-with-pure-copper-am-production/.

[8] “Green Light for New 3D Printing Process - Fraunhofer ILT.” [Online]. Available:

https://www.ilt.fraunhofer.de/en/press/press-releases/press-release-2017/press-release-

2017-08-30.html.

[9] D. A. Ramirez et al., “Open-cellular copper structures fabricated by additive manufacturing

using electron beam melting,” Mater. Sci. Eng. A, vol. 528, no. 16–17, pp. 5379–5386,

2011.

[10] D. A. Ramirez et al., “Novel precipitate-microstructural architecture developed in the

fabrication of solid copper components by additive manufacturing using electron beam

melting,” Acta Mater., vol. 59, no. 10, pp. 4088–4099, 2011.

[11] L. Yang, O. Harrysson, H. West II, and D. Cormier, “Design and characterization of

orthotropic re-entrant auxetic structures made via EBM using Ti6Al4V and pure copper,”

in Solid Freeform Fabrication Symposium, 2011, pp. 464–474.

[12] P. Frigola et al., “Fabricating copper components with electron beam melting,” Adv. Mater.

Process., vol. 172, no. 7, pp. 20–24, 2014.

[13] C. A. Terrazas et al., “Multi-material metallic structure fabrication using electron beam

melting,” Int. J. Adv. Manuf. Technol., vol. 71, no. 1–4, pp. 33–45, 2014.

[14] M. A. Lodes, R. Guschlbauer, and C. Körner, “Process development for the manufacturing

of 99.94% pure copper via selective electron beam melting,” Mater. Lett., vol. 143, pp. 298–

301, 2015.

[15] Y. Bai, G. Wagner, and C. B. Williams, “Effect of Particle Size Distribution on Powder

Packing and Sintering in Binder Jetting Additive Manufacturing of Metals,” J. Manuf. Sci.

Eng., vol. 139, no. 8, p. 81019, 2017.

[16] Y. Bai and C. B. Williams, “An exploration of binder jetting of copper,” Rapid Prototyp.

J., vol. 21, no. 2, pp. 177–185, 2015.

[17] A. Kumar, Y. Bai, A. Eklund, and C. B. Williams, “Effects of Hot Isostatic Pressing on

Copper Parts Fabricated via Binder Jetting,” in Procedia Manufacturing, 2017, vol. 10, pp.

935–944.

[18] R. M. German, Powder Metallurgy Science (Second Edition). 1994.

[19] R. Haynes, “Effect of Porosity Content on Ductility of Sintered Metals,” Powder Metall.,

vol. 20, no. 1, pp. 17–20, 1977.

52

[20] M. I. Aivazov and I. A. Domashnev, “Influence of Porosity on the Conductivity of Hot-

Pressed Titanium-Nitride Specimens,” Powder Metall. Met. Ceram., vol. 7, no. 9, pp. 708–

710, 1968.

[21] J. K. Carson, S. J. Lovatt, D. J. Tanner, and A. C. Cleland, “Thermal conductivity bounds

for isotropic, porous materials,” Int. J. Heat Mass Transf., vol. 48, no. 11, pp. 2150–2158,

2005.

[22] J. Wang, J. K. Carson, M. F. North, and D. J. Cleland, “A new approach to modelling the

effective thermal conductivity of heterogeneous materials,” Int. J. Heat Mass Transf., vol.

49, no. 17–18, pp. 3075–3083, 2006.

[23] C. Vincent, J. F. Silvain, J. M. Heintz, and N. Chandra, “Effect of porosity on the thermal

conductivity of copper processed by powder metallurgy,” J. Phys. Chem. Solids, vol. 73,

no. 3, pp. 499–504, 2012.

[24] G. C. J. Bart, “Thermal conduction in non homogeneous and phase change media,” 1994.

[25] B. C. Gundrum, D. G. Cahill, and R. S. Averback, “Thermal conductance of metal-metal

interfaces,” Phys. Rev. B - Condens. Matter Mater. Phys., vol. 72, no. 24, p. 245426, 2005.

[26] R. W. Powell, “Correlation of Metallic Thermal and Electrical Conductivities for Both Solid

and Liquid Phases,” Int. J. Heat Mass Transf., vol. 8, no. 7, pp. 1033–1045, 1965.

[27] C. S. Smith, “The Relation Between the Thermal and Electrical Conductivities of Copper

Alloys,” Phys. Rev., vol. 48, no. 2, pp. 166–167, 1935.

[28] F. H. Schofield, “The Thermal and Electrical Conductivities of Some Pure Metals,” in

Proceedings of the Royal Society of London, Series A, 1925, vol. 107, no. 742, pp. 206–227.

[29] J. C. Y. Koh and A. Fortini, “Prediction of thermal conductivity and electrical resistivity of

porous metallic materials,” Int. J. Heat Mass Transf., vol. 16, no. 11, pp. 2013–2022, 1973.

[30] S. J. Raab, R. Guschlbauer, M. A. Lodes, and C. Körner, “Thermal and Electrical

Conductivity of 99.9% Pure Copper Processed via Selective Electron Beam Melting,” Adv.

Eng. Mater., vol. 18, no. 9, pp. 1661–1666, 2016.

[31] ASTM Int., “Standard Test Methods for Tension Testing of Metallic Materials 1,” Astm,

vol. i, no. C, pp. 1–27, 2009.

[32] ISO/ASTM 52921, “Standard terminology for Additive Manufacturing - Coordinate

systems and test methodologies,” vol. 2013, pp. 1–13, 2013.

[33] E1461, “E1461. Standard test method for thermal diffusivity by the flash method,” ASTM,

West Conshohocken, PA, vol. i, pp. 1–11, 2013.

53

Chapter 4: Closure

4.1 Conclusions

The goal of this thesis research was to develop an understanding of the process-structure-property

relationship in the Binder Jetting of copper, by investigating the effects of post-process Hot

Isostatic Pressing on the porosity of Binder Jet copper (Chapter 2), and relating the porosity to

various material properties (Chapter 3). This investigation has been conducted by an evaluation of

the effects of HIP on the density, porosity, microstructure, tensile strength and ductility of Binder

Jet copper parts of three different sintered densities, achieved using three powder configurations

(Chapter 2). This resulted six different kinds of parts, each having been subjected to a different

processing condition, that were available for study. The strength, ductility, thermal and electrical

conductivities of these were then compared to models in the literature for PM copper and/or by

treating the pore distribution in copper as a two-component system using various assumptions

(Chapter 3). These analyses provided a means of understanding the correlation between process-

induced porosity and the material properties in a scientific manner. These studies have helped

understand the improvement the density and properties of Binder Jet copper using HIP, and

provided a good framework for future studies in developing quantitative models for the process-

structure-performance relationship in Binder Jetting.

4.2 Summary of findings

Detailed investigations have been carried out towards understanding the effects of Hot Isostatic

Pressing on the properties of copper parts fabricated using Binder Jetting. The major findings are

listed below:

Copper parts of 99.47% theoretical density have been fabricated using optimized powder

configurations, printing, sintering and HIP conditions [1] (Appendix A).

The effects of HIP on the porosity, microstructure and mechanical properties of parts of

three different powder configurations has been investigated. The following conclusions

may be drawn:

o HIP has been found to be effective in removing porosity in Binder Jet copper parts

to a reasonable extent, only when the sintered density is at least 90% (Section

2.3.1). More than 93% density seems to be required for approaching full-density

[1].

54

o The reduction in porosity has been determined to be the more significant factor in

affecting the tensile strength of Binder Jet specimens, rather than the associated

grain coarsening (Sections 2.3.3, 2.3.4).

o The ductility of HIPed parts has been found to increase with densification by HIP

(Section 2.3.4).

The dependence of material properties on process-induced porosity has been investigated

by comparison with existing models in the literature. The following are the major

conclusions:

o The tensile strength of Binder Jet parts is found to be less than the existing model

for sintered copper powder due to microstructural differences (Section 3.4.1).

o The ductility, being porosity-dependent, has different behaviors compared to

existing PM model for different levels of porosity (Section 3.4.2). The

heterogeneity of porosity distribution may play a significant role here.

o The thermal conductivity has been found to be less than that predicted by various

models in the literature that describe the dependence of properties on porosity. This

is due to the presence of impurities and to the grain boundary thermal resistance,

an estimate of which has been made to be greater than that predicted by the

literature for each of the processing conditions (Section 3.4.3).

o The electrical conductivity predicted for fully dense copper parts has been found to

approach closer to the theoretical value than the thermal conductivity, which may

indicate a potential anisotropy due to differing measurement directions (Sections

3.4.4, 3.4.5).

4.3 Contributions

This work has made the following novel contributions to the scientific literature in understanding

the Binder Jetting AM process:

Near-full density, high-purity copper parts have been fabricated for the first time using a

Binder Jetting-based process chain by using post-process HIP.

While most reported literature on HIP deals with parts that are at least 92% dense and have

no surface connected porosity, this work has contributed to understanding the impact of

55

HIP on Binder Jet parts, which typically feature surface porosity and fall below this density

requirement.

This work has explored the influence of porosity induced by different post-processing

conditions, on the mechanical, thermal and electrical properties of Binder Jet copper

The strength, ductility, and thermal and electrical conductivities of Binder Jet copper have

been situated in the context of conventional Powder Metallurgy and of two-component

structures, to help understand the extents of the properties that can be achieved using

Binder Jetting AM.

4.4 Limitations and Future Work

A comprehensive attempt has been made towards understanding the effects of HIP on Binder Jet

copper specimens. The findings of this work have presented scope for further work in this area in

order to fully understand the science of the process and to be able to design the process for a desired

porosity and material properties:

Careful control of powder quality and processing conditions can help print near full density

copper parts [1], which can be characterized to quantify the highest achievable values of

various material properties.

A more comprehensive study of the effects of part orientation on the mechanical properties

can help study the impact of HIP on anisotropic parts.

A three-dimensional porosity distribution analysis of Binder Jet copper can provide further

insight into understanding the influence of heterogeneity in pore distribution on the

properties analyzed.

The role of the measurement direction of various properties needs to be evaluated in

relation to the printing orientation and porosity distribution to develop accurate quantitative

process models.

The study of additional powder configurations and processing conditions can help in

obtaining more data for porosity in Binder Jet copper. This can help in developing more

refined models specific to Binder Jetting.

Lastly, although the scope of this work is restricted to printing pure copper, similar studies can be

carried out for other metals of interest to understand material-specific behaviors.

56

References (for Chapters 1 and 4)

[1] E. Sachs, M. Cima, and J. Cornie, “Three-dimensional printing: rapid tooling and prototypes

directly form a CAD model,” CIRP Ann. - Manuf. Technol., vol. 39, no. 1, pp. 201–204,

1990.

[2] A. Kumar, Y. Bai, A. Eklund, and C. B. Williams, “Effects of Hot Isostatic Pressing on

Copper Parts Fabricated via Binder Jetting,” in Procedia Manufacturing, 2017, vol. 10, pp.

935–944.

57

Appendix A

Publication: Effects of Hot Isostatic Pressing on Copper Parts Fabricated via Binder

Jetting

Available online at www.sciencedirect.com

ScienceDirect

Procedia Manufacturing 00 (2017) 000–000

www.elsevier.com/locate/procedia

45th SME North American Manufacturing Research Conference, NAMRC 45, LA, USA

Effects of Hot Isostatic Pressing on Copper Parts Fabricated via

Binder Jetting

Ashwath Kumara, Yun Baia, Anders Eklundb, Christopher B. Williamsa1

aDepartment of Mechanical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24060, United States of

America bQuintus Tecnologies, LLC, Lewis Center, Ohio, 43035, United States of America

Abstract

Binder Jetting, an Additive Manufacturing process that fabricates parts via selective inkjet deposition of binder into a powder bed,

is capable of cost-effectively producing complex metal and ceramic components without the need for support structures or anchors.

Printed green parts are then sintered for added densification and strength. However, printed parts typically contain porosity due to

the use of coarse powders and a loosely packed powder bed. While researchers have investigated many techniques (e.g., process

parameter and powder morphology optimization) for achieving full theoretical density in binder jetted parts, 100% density is

difficult to achieve without infiltration of a secondary lower melting point material. In this work, the authors investigate the use of

Hot Isostatic Pressing (HIP) as a post-process heat treatment of sintered parts to evaluate its effect on density, porosity, and

shrinkage of parts printed in Binder Jetting. The authors conduct this investigation in the context of copper. It is demonstrated that

the use of HIP can improve the final part density from 92% (following sintering) to 99.7% of theoretical density.

© 2017 The Authors. Published by Elsevier B.V.

Peer-review under responsibility of the Scientific Committee of NAMRI/SME.

Keywords: Binder Jetting; Additive Manufacturing, Hot Isostatic Pressing, Copper

* Corresponding author. Tel.: +1-540-231-3422; fax: +1-540-231-9100.

E-mail address: cbwill@vt.edu

58

1. Introduction

Binder Jetting is an Additive Manufacturing (AM) technology that involves binding together powder particles by

selectively jetting a binder layer-by-layer to form a printed green part. Binder Jetting is able to process a wide variety

of powdered materials, including polymers [1,2], ceramics [3,4] metals [5,6], foundry sand [7], and the active

ingredients for pharmaceutical applications [8]. The printing process begins with a counter-rotating roller that spreads

a layer of powder of a set thickness from a ‘feed box’ onto a ‘build box’. After each layer is deposited, an inkjet

printhead deposits patterns of binder according to the shape of the part’s cross-section based on the model data taken

from the ‘slice file’ fed to the printer. The binder droplets penetrate into the powder through the voids between the

particles, and stitch together the particles within a layer, as well as to the previously formed layer. This process is

repeated layer after layer to form the green part. The setup is illustrated in Figure 1. The printed part is then placed

into a low-temperature oven (~200C) to cure the binder. Subsequently, the loose powder particles adhering weakly to

the edges of the green part are removed using compressed air. Since the green part is composed of bound loosely

packed powder, it is not dense enough to be mechanically strong; hence the part is then sintered in a furnace where

the binder pyrolyzes, and the high temperature causes the powder particles to sinter through atomic diffusion. This

process results in the part shrinking into the voids left by the binder, causing the part to densify and hence strengthen.

Fig. 1. Schematic Illustrating Binder Jetting [7].

A unique feature of Binder Jetting metals is the separation of powder sintering from part creation, which

circumvents issues of residual stresses being induced in the part due to rapid melting/solidification of metal powder

as found in metal Powder Bed Fusion (PBF) processes. In PBF processes, these stresses and associated warpage are

dealt with using anchors or supports. Since Binder Jetting does not involve such stresses, the powder bed is sufficient

to act as a support, thereby providing the design freedom expected from AM technologies. The independence of the

part creation and energy application steps also makes the technology inherently scalable, as printing larger parts would

not have significant additional costs as compared to metal PBF processes where upgrading and adding energy sources

is expensive. Also, a wide array of materials can be printed with this technology, as the ability to print a given material

into green parts would simply involve choosing a material that can be made in the form of a powder that flows and a

compatible binder. Post process heat treatment can be tailored by choosing suitable sintering profiles. For example,

Binder Jetting has been found to be particularly suitable for copper [9], which is challenging to process via PBF due

to its high thermal conductivity (which makes it disperse any impinging energy onto surrounding particles, thereby

resulting in a lack of control of the melt pool) and reflectivity (which places limitations on the kinds of energy sources

that can be used) [10].

While this process presents a cost-effective and scalable means of fabricating complex metal and ceramic

components, the printed parts suffer from porosity when high strength and conductivity is required. The difficulty in

creating parts of near-full density is due to the powder particles being loosely packed with insufficient compaction by

the roller, which results in the sintered parts being porous due to reduction in the availability of sintering and neck-

forming sites. Infiltration is a technique that is often incorporated in conjunction with sintering in order to achieve full

density [11,12,13]. This involves the addition of a lower melting alloy that flows into the matrix of printed metal

through guided pathways, that fills into the pores by capillary action. This reduces the shrinkage during sintering and

increases the final density. However, due to the addition of a secondary substance, the material properties change

accordingly. Thus there is a requirement for an alternative process that does not affect the composition and properties

of the part, while still being able to achieve full density.

59

Eliminating porosity via solid state sintering without infiltration is a challenge, and hence has been a focal point of

research in Binder Jetting. Investigations have been made in this direction by optimizing process parameters or making

suitable modifications to the system. Turker et al. investigated the effect of layer thickness, powder particle size and

sintering cycles in the Binder Jetting of IN 718 superalloy [14]. Liquid phase sintering mechanism or optimized

sintering parameters have been successful in improving sintered density [4,15,16]. The effect of bimodal powder

mixtures as well as printing and sintering process parameters in improving sintered density has been studied for copper

by Bai and co-authors [17]. The use of bimodal powder mixtures of 316L Stainless Steel was studied by Verlee and

co-authors [18]. A design modification to a Binder Jetting machine was applied in order to incorporate powder

compaction during printing to achieve better green part densities by Gregorski [19]. Slurry-based Binder Jetting is

another technique that uses base material in the form of slurries instead of powders. The effects of this technique in

improving the density of printed parts has been studied by Grau and co-authors [20], and by Ables [21]. Another

promising technique for densification of printed green parts is the jetting of functional inks instead of a binder. Bai

and co-authors have investigated the use of nanosilver suspensions jetted onto a silver substrate in order to improve

sintering performance and dimensional accuracy [22]. Additionally, the effect of various printing and sintering process

parameters on mechanical properties have been investigated [23]. In the case of copper, these problems have been

overcome to the extent of achieving a sintered density of 87.1% by using a bimodal powder configuration [17]. Further

refinements of the post-process sintering cycle created parts with 92% theoretical density.

In this paper, the authors explore the use of Hot Isostatic Pressing (HIP) as means for further densifying metal parts

printed with Binder Jetting. HIP has been used in both powder metallurgy and additive manufacturing industries as a

means to create finished parts that are 100% dense (Section 2). The primary goal of this work is to explore the effects

of HIP on parts made via Binder Jetting. The authors’ research is conducted within the context of printed copper parts.

In this context, achieving full density through infiltration is not acceptable, as the presence of a secondary material

would drastically alter the desired thermal, electrical and mechanical properties. Thus a secondary goal of this work

is to explore a means for additive manufacture of fully dense copper parts. Prior art in using HIP for AM parts

(predominantly in metal PBF processes) is reviewed in Section 2. An overview of the authors’ experimental methods

to test the hypothesis that HIP of Binder Jetting parts will result in full theoretical density is presented in Section 3.

Results from these experiments are presented in Section 4, with closure offered in Section 5.

2. Hot Isostatic Pressing in Additive Manufacturing

HIP is a technique widely employed in Powder Metallurgy to consolidate loose powders into a desired shape by

applying isostatic pressure using an inert gas (usually Argon) at elevated temperatures. The physics of HIP have been

investigated in detail by Atkinson and co-author [24]. In summary, the process involves the application of sufficiently

high pressure for pores entrapped in a part to overcome the surface energy driving force for pore closure. These pores

then dissolve in the matrix and diffuse to the surface of the part. Due to the pressure being isostatic in nature, it acts

perpendicular to the surface irrespective of the geometry, and the shrinkage seen in the part after HIP is expected to

be ‘photographic’, which implies that the shape of the part is retained despite the dimensional shrinkage. Apart from

powder consolidation, HIP is also used in densifying castings and pre-sintered components. Before direct-metal PBF

systems were available, researchers looked to HIP to bring green parts, formed from two-phase powder mixtures, to

full density. Currently, HIP is often used to bring metal components made via direct-metal PBF processes to 100% of

theoretical density.

HIP was found to help approach full density in a bronze-nickel alloy printed using PBF technology [25]. In addition,

a specific process, ‘SLS/HIP’ was developed by Das and co-authors [26] for achieving high density parts made of

superalloys such as Inconel Alloy 625 and Ti-6Al-4V. The process involves sintering the boundaries of each layer to

high densities (>98% of theoretical density) while printing, thereby creating a solid shell with the enclosed portion

being of density as low as 80%. This shell can essentially act as a container so the part can be HIPed as printed. Liu

and co-authors presented another process chain featuring PBF, Cold Isostatic Pressing, degreasing, sintering, and HIP

for fabricating alumina parts [27]. The use of HIP resulted in a final density of 95.94%, as compared to 32.06% density

as printed. The effectiveness of HIP in pore closure in parts made by using Electron Beam Melting (EBM) was studied

by conducting X-Ray Computed Tomography scans before and after HIP [28].

The improvement in density has been found to translate into an improvement in mechanical properties. HIP has

been found to bring about an improvement in fatigue strength of ASTM F75, a cobalt-based alloy manufactured using

a PBF process [29]. The same was found to be true of the alloys AlSi10Mg and Ti-6Al-4V [30]. A HIP process

60

developed by Peter et al. has resulted in an improvement in grain structure of Ti-6Al-4V and Inconel 718 parts

fabricated using EBM [31].

The use of HIP for green parts made via Binder Jetting has not been widely documented in the literature. The only

prior art discovered by the authors was Kernan and co-authors’ use of sinter-HIP following slurry-based Binder Jetting

of Tungsten Carbide-Cobalt (WC-Co). In their work, the use of HIP provided a means of achieving density

approaching 100% [32]. HIP requires that the outer shell of the part is dense and devoid of pores, for the internal pores

to be effectively removed. This is the reason parts are either enclosed in a container or previously sintered before HIP.

Parts fabricated via Binder Jetting are generally expected to have surface connected porosity, which could render the

use of HIP to improve density difficult. This is a possible reason for the limited prior research in this space. The goal

of this work is to address this gap by, (i) furthering the understanding of the impact of HIP on parts made via Binder

Jetting, and (ii) to exploring techniques for enabling AM of fully-dense copper artifacts.

3. Research Methods

To explore the effect of HIP on parts made via Binder Jetting, the authors evaluated the density, porosity, and

shrinkage of printed parts in their printed green state following post-process sintering, and following a subsequent

HIP process. These experiments were conducted in the context of printing high purity copper components in Binder

Jetting. The experimental methodology employed in this work, including the parameters used to form the green part

and of the post-process sintering were based on prior work done by the authors [17], and are detailed in this section.

3.1. Materials

The copper powders used in this study were gas atomized to yield a spherical shape. This shape aids in good

packing as compared to irregularly shaped powder, and also in ease of powder recoating. Additionally, spherical

particles perform better when sintering, due to easier necking as compared to irregularly shaped particles. Based on

previous work on bimodal powder mixtures to achieve maximum sintered density [17], the two copper powders were

chosen to have median particle diameter of 30µm and 5µm, and mixed in the ratio of 73:27 by weight in a rotating

drum for ~2 hours to yield the desired powder configuration.

3.2. Binder Jetting Process Parameters

The experimental specimens were printed on an ExOne R2 Binder Jetting printer, using a standard off-the-shelf

binder provided by ExOne that has been found to be compatible with the copper powders used, and to leave a minimal

amount of residue upon pyrolysis during sintering. A layer thickness of 70µm was used to print the parts. The

saturation ratio (the ratio of the amount of space between the powder particles in the bed that is occupied by binder)

is an important parameter in achieving strong and dense green parts. To set the saturation, the binder drop volume and

packing density of the powder bed are measured and entered into the machine which then adjusts the amount of jetted

binder accordingly. The value of saturation ratio must be chosen so as to be high enough to allow for sufficient binder

penetration between layers, but not so high that binder seeps into more regions of the powder bed than intended,

thereby having detrimental effects on part accuracy as well as surface finish. Based on prior investigation considering

these factors, the saturation ratio was set to 100% [17].

The parts printed for measuring density were rectangular coupons of three different kinds of target dimensions:

A. 16x16x4 mm

B. 32x8x4 mm

C. 16x8x4 mm

Following printing, all samples were sintered as described in Section 3.3. One coupon each of Type A and Type B

were cut in half; one half of each was retained and the other half was subjected to HIP. The density and microstructure

of the halves of the specimens were then evaluated (Section 4) and compared. This was done in order to study the

effect of HIP on the density of the same coupon before and after HIP, thereby accounting for any minor variations in

density between two parts from the same print. This kind of variation was a possibility in the Type C coupons, of

which two were fabricated; with one subjected to HIP and the other retained for comparison.

61

3.3. Post-process Heat Treatment process parameters

Following printing, the samples were sintered in a box furnace in a hydrogen atmosphere. The sintering profile

used is shown in Figure 2. A constant heating rate of 5°C/min was used to first raise the temperature of the furnace to

450°C, where it was held for 30 min to burn off the binder. The temperature was increased at 5°C/min to a sintering

temperature of 1075°C, where it was held for 3 hours. The furnace was then allowed to cool at a rate of 5°C/min until

it reached room temperature. The choice of this sintering profile was based on previous research by Meeder et al. in

order to optimize sintered density [33].

Fig. 2. Sintering Profile Used.

Following sintering, a subset of the specimens was subjected to HIP in an argon atmosphere using a Graphite

furnace with Quintus Technologies’ proprietary Uniform Rapid Cooling (URC®) technology. This was used in order

to speed up the HIP cycle by reducing the cooling phase, thereby minimizing the amount of time the sample is held at

higher temperatures. Grain growth, which can lead to reduction in yield strength and toughness, has been found to be

greater at higher temperatures [24]. Hence this technology can minimize grain growth during the cycle, and potentially

result in better mechanical properties than can be obtained from a cycle with a slower cooling rate. The HIP was done

in a container-less fashion, as the sintered parts were observed to not have any significant surface connected porosity

that could affect the process. The HIP parameters used were a temperature set to 1075°C and a pressure of 206.84

MPa (30,000 psi), held for 2 hours in an Argon atmosphere. Conventionally, copper is HIPed at 800 to 950°C under

a pressure of 100 MPa [24]. The pressure value was chosen to be higher than this, based on preliminary trials that

indicated a better pore closure with higher pressure. The temperature could then be raised to a slightly higher value as

well, and an initial trial had been conducted at a HIP temperature of 975°C held for four hours. However, this caused

issues with remaining porosity. Hence, the temperature in the successful trial reported here was increased to 1075°C,

with the hold time reduced to two hours to minimize grain growth.

3.4. Shrinkage, Density Measurements

Part shrinkage was calculated by measuring the dimensions of the part after sintering and after HIP, then comparing

them to the dimensions of the green part. Linear shrinkage in each of the three directions was calculated. Density

measurements were carried out using an Archimedes Principle based apparatus that calculates the density of the part

using its weight measured in air and in water (ASTM Standard B962-15) [34]. Oil impregnation is to be carried out

per the standard in order to seal any surface connected porosity that is present in the sample to be measured. However,

in this case, it was not carried out in order to avoid contaminating the parts to be HIPed. Micrographs of sintered

specimens indicated the absence of any significant surface connected porosity, and this step could hence be avoided.

62

In addition to measuring the bulk density by this method, the porosity of the internal sections was measured, the details

of which are outlined in Section 3.5.

3.5. Porosity Measurements

For porosity measurements, the samples of each type were sectioned at multiple locations (across the X-Z and Y-

Z planes), then polished. Multiple images were taken of each section under a microscope, and a MATLAB code was

written to calculate the porosity from these images using the image processing toolbox. The images were first imported

and converted to a binary black and white image, with black regions corresponding to pixels turned ‘off’ and white

regions ‘on’. The threshold for differentiating black regions (pores), from white regions (solid copper) was set by

qualitative visual observation. The porosity, which is the ratio of the area of the black regions to that of the total area

of the image, is calculated by dividing the number of pixels in the black region (returned by the function, “bwarea”),

by the total number of pixels (returned by the function, “numel”). In order to ensure any stray scratches or the

background of the images do not get counted as black regions, these features were manually cropped out of the

microscope images before analysis.

4. Results and Discussions

4.1. Shrinkage and Density

Table 1 shows the total linear shrinkage in each direction after sintering and after HIP for each of the three part

types. The X and Y dimensional shrinkages were nearly identical, and have been grouped together for the purpose of

simplification.

Table 1. Linear Shrinkage in Parts after sintering and HIP

Part Type Shrinkage after Sintering (%) Total Shrinkage after HIP (%)

X/Y Z X/Y Z

A 14.35 16.91 15.52 18.77

B 14.14 18.57 15.56 19.30

C 14.25 16.91 16.16 19.70

These results indicate that there is a slightly more shrinkage in the Z-direction as compared to the X or Y directions.

This indicates an anisotropy in the distribution of porosity, i.e. there is more porosity between layers than within a

layer. This is expected, as the particles are not consolidated across layers as well as they are within a layer, due to

limited diffusion of binder in the bed. The shrinkage may also be attributed in part to gravity effects that come into

play during sintering, that may cause greater shrinkage in the Z-direction than X/Y directions. This variation goes to

show that it is important to design for parts to be manufactured using this process chain by keeping in mind that

different compensations in the geometric dimensions will be required for each of the directions. The shrinkage in these

is accompanied by an increase in the density. Figure 3 summarizes the improvement in density in each of the parts,

by comparing the green density, sintered density, and HIP density for each type.

63

Fig. 3. Density improvement upon HIP of parts.

The mean green density of the printed specimens was measured to be 55.00%, using dimensional measurements.

Following the sintering cycle detailed in Section 3.3, the density improved to an average of 93.90%. Following the

HIP process, the measured mean bulk density was 99.47%, using the Archimedes Principle apparatus. It is noted that

the geometry (as indicated by part “Type”) did not impact the density measurements.

4.2. Porosity

The results of the porosity analysis using optical micrographs are presented in Figure 4, comparing the porosity of

parts of each type after sintering and after HIP.

Fig. 4. Porosity improvement upon HIP of parts.

The results presented above show a general trend of an overall improvement in porosity as seen from the optical

micrographs. There is a visible consistency in the range of porosity in HIPed samples, irrespective of initial sintered

55.13 54.97 54.88

93.46 93.89 94.3499.24 99.78 99.38

0

20

40

60

80

100

120

Type A Type B Type C

Den

sity

(%

Theo

reti

cal)

Part Type

Density Improvement Upon HIP

Green Density Sintered Density HIPed Density

64

porosity. The numerical averages of the porosity as obtained from the all the micrograph images of each of the

specimen are summarized in Table 2 below.

Table 2. Comparison of average porosity between sintered and HIPed parts.

Part Type Sintered Sample Porosity (%) HIPed Sample Porosity (%)

A 0.26 0.10

B 0.26 0.09

C 1.29 0.10

A decrease in average porosity in all the samples is evident from the analysis. Figure 5 shows representative

micrographs (of Type C parts) that were used in the image analysis. The porosity of these images as calculated using

the MATLAB code are in given in the description.

Fig. 5. Sample Micrographs (Type C) indicating density improvement upon HIP (a) Sintered Part, 1.88% porosity; (b) HIPed Part, 0.13%

porosity calculated from image analysis.

From these images of sectioned parts, it is clear that HIP has assisted in densifying the part to nearly full density.

There is, however, some existing residual porosity as yet unremoved after HIP, visible in some of the micrographs. It

remains to be seen if there is a scope for further improvement by tuning the HIP process parameters. The extent of

detrimental effects of this residual porosity on material properties is also yet to be seen, but it may be safely assumed

that the properties of the HIPed parts will offer improvements from those of the as-sintered parts.

5. Conclusions and Future Work

A process chain has been developed for achieving near-full density copper parts using Binder Jetting followed by

sintering and Hot Isostatic Pressing. The HIP process was successfully able to increase the density of printed copper

parts from 92% to 99.7%. The use of HIP also significantly reduced the porosity of the printed parts. In addition, it

was observed that HIP resulted in an anisotropic shrinkage (~16-20% linear shrinkage) due to graded density in the

green part.

In addition to understanding how HIP affects the density of parts created via Binder Jetting, this research also

provides the base for establishing a set protocol for Additive Manufacturing of high-density copper parts. These results

provide promise for a means of fabricating complex copper parts for end-use applications that need to utilize the

unique thermal and electrical properties of the material. The major applications of printed copper are likely to be in

the areas of enhanced heat transfer applications, advanced electrical components etc.

Future work will focus on characterizing the mechanical, thermal and electrical properties of the parts following

HIP. The complexity in the geometry of products to be used in such applications can easily be achieved using Additive

Manufacturing, but the effect of this entire process chain on the dimensional stability of such parts is yet to be

investigated. It is expected in this case that the effects of HIP in particular, on complex geometry parts will not be

65

significant, owing to the expected photographic shrinkage associated with the isostatic pressure involved.

Additionally, the effect of sintered density in the effectiveness of HIP is a potential gap to address. One could thus

investigate the minimum sintered density for HIP to work effectively.

References

[1] C. X. F. Lam, X. M. Mo, S. H. Teoh, D. W. Hutmacher, Scaffold development using 3D printing with a starch-based polymer. Materials

Science and Engineering: C, 20(1), (2002) pp.49-56.

[2] J. Suwanprateeb, R. Chumnanklang, Three‐ dimensional printing of porous polyethylene structure using water‐ based binders. Journal of

Biomedical Materials Research Part B: Applied Biomaterials, 78(1), (2006) pp.138-145.

[3] J. Will, R. Melcher, C. Treul, N. Travitzky, U. Kneser, et al., Porous ceramic bone scaffolds for vascularized bone tissue regeneration. Journal

of Materials Science: Materials in Medicine, 19(8), (2008) pp.2781-2790.

[4] S. Zhang, H. Miyanaji, L. Yang, A. Ali, J. J. S. Dilip, An Experimental Study of Ceramic Dental Porcelain Materials Using A 3D Print (3DP)

Process. In Proceeding of Solid Freeform Fabrication (SFF) Symposium, (2014) pp. 991-1011.

[5] S. Maleksaeedi, J. K. Wang, A. El-Hajje, L. Harb, V. Guneta, Z. He, F. E. Wiria, C. Choong, A. J. Ruys, Toward 3D printed bioactive titanium

scaffolds with bimodal pore size distribution for bone ingrowth. Procedia CIRP, 5, (2013) pp.158-163.

[6] C. B. Williams, J. K. Cochran, D. W. Rosen, Additive manufacturing of metallic cellular materials via three-dimensional printing. The

International Journal of Advanced Manufacturing Technology, 53(1-4), (2011) pp.231-239.

[7] D. Snelling, Q. Li, N. Meisel, C. B. Williams, R.C. Batra, A. P. Druschitz, Lightweight metal cellular structures fabricated via 3D printing of

sand cast molds. Advanced Engineering Materials, 17(7), (2015) pp.923-932.

[8] D. G. Yu, L. M. Zhu, C. J. Branford‐ White, X. L. Yang, Three‐ dimensional printing in pharmaceutics: Promises and problems. Journal of

pharmaceutical sciences, 97(9), (2008) pp.3666-3690.

[9] Y. Bai, C. B. Williams, An exploration of binder jetting of copper. Rapid Prototyping Journal, 21(2), (2015) pp.177-185.

[10] D. Q. Zhang, Z. H. Liu, C. K. Chua, Investigation on forming process of copper alloys via Selective Laser Melting. In High Value

Manufacturing: Advanced Research in Virtual and Rapid Prototyping: Proceedings of the 6th International Conference on Advanced Research

in Virtual and Rapid Prototyping, Leiria, Portugal, (2013) p. 285.

[11] B. D. Kernan, E. M. Sachs, S. M. Allen, A. Lorenz, C. Sachs, L. Raffenbeul, A. Pettavino, Homogeneous steel infiltration. Metallurgical and

Materials Transactions A, 36(10), (2005) pp.2815-2827.

[12] M. Lanzetta, M. Santochi, Liquid-phase infiltration of thermal sintered skeletons by low-temperature gold eutectic alloys. CIRP Annals-

Manufacturing Technology, 55(1), (2006) pp.213-216.

[13] K. Jiang, F. Wang, J. Zhang, Experimental study on sintering and infiltration process of metal part fabricated by mold decomposed injection

sculpturing. Rev. Adv. Mater. Sci, 33, (2013) pp.322-329.

[14] M. Turker, D. Godlinski, F. Petzoldt, Effect of production parameters on the properties of IN 718 superalloy by three-dimensional

printing. Materials characterization, 59(12), (2008) pp.1728-1735.

[15] S. M. Gaytan, M. A. Cadena, H. Karim, D. Delfin, Y. Lin, D. Espalin, E. MacDonald, R. B. Wicker, Fabrication of barium titanate by binder

jetting additive manufacturing technology. Ceramics International, 41(5), (2015) pp.6610-6619.

[16] M. J. Orange, H. A. Kuhn, P. P. Knor, T. Lizzi, Process for making nickel-based superalloy articles by three-dimensional printing. Patent WO

2015183796 A1 (2015).

[17] Y. Bai, G. Wagner, C. B. Williams, Effect of Bimodal Powder Mixture on Powder Packing Density and Sintered Density in Binder Jetting of

Metals. In Annual International Solid Freeform Fabrication Symposium, (2015) p. 62.

[18] B. Verlee, T. Dormal, J. Lecomte-Beckers, Alternative Consolidation: Properties of Sintered Parts Shaped by 3D-Printing from Bimodal 316L

Stainless Steel Powder Mixtures. In European Congress and Exhibition on Powder Metallurgy. European PM Conference Proceedings (2011)

p. 1.

[19] S. J. Gregorski, High green density metal parts by vibrational compaction of dry powder in three-dimensional printing process. Doctoral

dissertation, Massachusetts Institute of Technology, 1996.

[20] J. Grau, J. Moon, S. Uhland, M. J. Cima, E. Sachs, August. High green density ceramic components fabricated by the slurry-based 3DP

process. In Solid Freeform Fabrication Symposium, (1997) pp. 371-378.

[21] D. C. Ables, Design of a Slurry Layer Forming Station and Improved Fluid Handling System for Raster Processes in 3DP™. Masters Thesis,

Massachusetts Institute of Technology, 1995.

[22] J. G. Bai, K. D. Creehan, H. A. Kuhn, Inkjet printable nanosilver suspensions for enhanced sintering quality in rapid

manufacturing. Nanotechnology, 18(18), (2007) p.185701.

[23] M. Vaezi, C. K. Chua, Effects of layer thickness and binder saturation level parameters on 3D printing process. The International Journal of

Advanced Manufacturing Technology, 53(1-4), (2011) pp.275-284.

[24] H. V. Atkinson, S. Davies, Fundamental aspects of hot isostatic pressing: an overview. Metallurgical and Materials Transactions A, 31(12),

(2000) pp.2981-3000.

[25] M. Agarwala, D. Bourell, J. Beaman, H. Marcus, J. Barlow, Post-processing of selective laser sintered metal parts. Rapid Prototyping

Journal, 1(2), (1995) pp.36-44.

[26] S. Das, M. Wohlert, J. J. Beaman, D. L. Bourell, Producing metal parts with selective laser sintering/hot isostatic pressing. JoM, 50(12),

(1998) pp.17-20.

[27] K. Liu, Y. Shi, W. He, C. Li, Q. Wei, J. Liu, Densification of alumina components via indirect selective laser sintering combined with isostatic

pressing. The International Journal of Advanced Manufacturing Technology, 67(9-12), (2013) pp.2511-2519.

[28] S. Tammas-Williams, P. J. Withers, I. Todd, P. B. Prangnell, The Effectiveness of Hot Isostatic Pressing for Closing Porosity in Titanium

Parts Manufactured by Selective Electron Beam Melting. Metallurgical and Materials Transactions A, 47(5), (2016) pp.1939-1946.

66

[29] J. Haan, M. Asseln, M. Zivcec, J. Eschweiler, R. Radermacher, C. Broeckmann, Effect of subsequent Hot Isostatic Pressing on mechanical

properties of ASTM F75 alloy produced by Selective Laser Melting. Powder Metallurgy, 58(3), (2015) pp.161-165.

[30] T. M. Mower, M. J. Long, Mechanical behavior of additive manufactured, powder-bed laser-fused materials. Materials Science and

Engineering: A, 651, (2016) pp.198-213.

[31] W. H. Peter, P. Nandwana, M. M. Kirka, R. R. Dehoff, W. Sames, D. L. Erdman III, A. Eklund, R. Howard, Understanding the Role of Hot

Isostatic Pressing Parameters on the Microstructural Evolution of Ti-6Al-4V and Inconel 718 Fabricated by Electron Beam Melting (No.

ORNL/TM-2015/77). Oak Ridge National Laboratory (ORNL); Manufacturing Demonstration Facility (MDF), 2015.

[32] B. D. Kernan, E. M. Sachs, M. A. Oliveira, M. J. Cima, Three-dimensional printing of tungsten carbide–10wt% cobalt using a cobalt oxide

precursor. International Journal of Refractory Metals and Hard Materials, 25(1), (2007) pp.82-94.

[33] M. P. Meeder, Modeling the thermal and electrical properties of different density sintered binder jetted copper for verification and revision of

the Wiedemann-Franz law. Masters Thesis, Virginia Polytechnic Institute and State University, 2016.

[34] ASTM Standard B962, Standard Test Methods for Density of Compacted or Sintered Powder Metallurgy (PM) Products Using Archimedes’

Principle (2015).

67

Appendix B

MATLAB CODE FOR IMAGE ANALYSIS

% All image files are first manually cropped to exclude exterior regions,

scratches etc. These are saved as JPGs in subfolders for each cross section

of each specimen, with the file name as the serial number of the image within

the subfolder (e.g., 1.jpg, 2.jpg, etc.). This code works for this kind of

file/folder organization and is not generalizable without making appropriate

modifications

close all clear all clc

here = mfilename('fullpath'); [path, ~, ~] = fileparts(here); addpath(genpath(path));

% Extract images from subfolder for XY sections: cd(strcat(path,'\XY'));

PicsXY = dir(strcat(pwd,'\*.jpg')); for i=1:length(PicsXY) filename = sprintf('%d.jpg',i); fXY{i} = imread(filename); figure() imshow(fXY{i});

%convert to black and white images with threshold of 0.5, show image to

verify threshold selection

f1bw = im2bw(fXY{i},0.5); figure() imshow(f1bw);

%porosity calculation

atot1(i) = numel(f1bw); wtot1(i) = bwarea(f1bw); btot1(i) = atot1(i) - wtot1(i); por(i,1) = btot1(i)*100/atot1(i); end

%same code repeated for YZ and ZX sections; kept separate to allow for

different thresholding choices if required cd(strcat(path,'\YZ')); PicsYZ = dir(strcat(pwd,'\*.jpg')); for i=1:length(PicsYZ) filename2 = sprintf('%d.jpg',i); fYZ{i} = imread(filename2); figure() imshow(fYZ{i});

f2bw = im2bw(fYZ{i},0.5); figure() imshow(f2bw);

68

atot2(i) = numel(f2bw); wtot2(i) = bwarea(f2bw); btot2(i) = atot2(i) - wtot2(i); por(i,2) = btot2(i)*100/atot2(i); end

cd(strcat(path,'\ZX')); PicsZX = dir(strcat(pwd,'\*.jpg')); for i=1:length(PicsZX) filename3 = sprintf('%d.jpg',i); fZX{i} = imread(filename3); figure() imshow(fZX{i});

f3bw = im2bw(fZX{i},0.5); figure() imshow(f3bw);

atot3(i) = numel(f3bw); wtot3(i) = bwarea(f3bw); btot3(i) = atot3(i) - wtot3(i); por(i,3) = btot3(i)*100/atot3(i); end