Brigham Young University Brigham Young University

BYU ScholarsArchive BYU ScholarsArchive

Theses and Dissertations

2021-08-02

Feasibility and Impact of Liquid/Liquid-encased Dopants as Feasibility and Impact of Liquid/Liquid-encased Dopants as

Method of Composition Control in Laser Powder Bed Fusion Method of Composition Control in Laser Powder Bed Fusion

Taylor Matthew Davis Brigham Young University

Follow this and additional works at: https://scholarsarchive.byu.edu/etd

Part of the Engineering Commons

BYU ScholarsArchive Citation BYU ScholarsArchive Citation Davis, Taylor Matthew, "Feasibility and Impact of Liquid/Liquid-encased Dopants as Method of Composition Control in Laser Powder Bed Fusion" (2021). Theses and Dissertations. 9256. https://scholarsarchive.byu.edu/etd/9256

This Thesis is brought to you for free and open access by BYU ScholarsArchive. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of BYU ScholarsArchive. For more information, please contact ellen_amatangelo@byu.edu.

Feasibility and Impact of Liquid/Liquid-Encased Dopants as Method of

Composition Control in Laser Powder Bed Fusion

Taylor Matthew Davis

A thesis submitted to the faculty ofBrigham Young University

in partial fulfillment of the requirements for the degree of

Master of Science

Nathan B. Crane, ChairTracy W. Nelson

David T. Fullwood

Department of Mechanical Engineering

Brigham Young University

Copyright © 2021 Taylor Matthew Davis

All Rights Reserved

ABSTRACT

Feasibility and Impact of Liquid/Liquid-Encased Dopants as Method ofComposition Control in Laser Powder Bed Fusion

Taylor Matthew DavisDepartment of Mechanical Engineering, BYU

Master of Science

Additive manufacturing (AM) – and laser powder bed fusion (LPBF) specifically – con-structs geometry that would not be possible using standard manufacturing techniques. This geo-metric versatility allows integration of multiple components into a single part. While this practicecan reduce weight and part count, there are also serious drawbacks. One is that the LPBF processcan only build parts with a single material. This limitation generally results in over-designing someareas of the part to compensate for the compromise in material choice. Over-designing can lead todecreased functional efficiency, increased weight, etc. in LPBF parts. Methods to control the mate-rial composition spatially throughout a build would allow designers to experience the full benefitsof functionality integration. Spatial composition control has been performed successfully in otherAM processes – like directed energy deposition and material jetting – however, these processesare limited compared to LPBF in terms of material properties and can have inferior spatial reso-lution. This capability applied to the LPBF process would extend manufacturing abilities beyondwhat any of these AM processes can currently produce. A novel concept for spatial compositioncontrol – currently under development at Brigham Young University – utilizes liquid or liquid-encased dopants to selectively alter the composition of the powder bed, which is then fused withthe substrate to form a solid part.

This work is focused on evaluating the feasibility and usefulness of this novel compositioncontrol process. To do this, the present work evaluates two deposition methods that could be used;explores and maps the laser parameter process space for zirconia-doped SS 316L; and investigatesthe incorporation of zirconia dopant into SS 316L melt pools. In evaluating deposition methods,inkjet printing is recommended to be implemented as it performs better than direct write materialextrusion in every assessed category. For the process space, the range of input parameters overwhich balling occurred expanded dramatically with the addition of zirconia dopant and shifted withchanges in dopant input quantities. This suggests the need for composition-dependent adjustmentsto processing parameters in order to obtain desired properties in fused parts. Substantial amountsof dopant material were confirmed to be incorporated into the laser-fused melt tracks. Individualinclusions of 100 nm particles distributed throughout the melt pool in SEM images. Howewver,EDX data shows that the majority of the incorporated dopant material is located around the edgesof the melt pools. Variations of dopant deposition, drying, and laser scanning parameters shouldbe studied to improve the resulting dopant incorporation and dispersion in single-track line scans.Area scans and multi-layer builds should also be performed to evaluate their effect on dopantcontent and dispersion in the fused region.

Keywords: additive manufacturing, AM, laser powder bed fusion, LPBF, direct metal laser melt-ing, DMLM, spatial composition control, liquid doping, melt pool characteristics, process maps

ACKNOWLEDGMENTS

I would like to express my gratitude to my advisor, Dr. Nathan Crane, for his insights,

guidance, and encouragement which have been invaluable to me and my work. The support and

resources given to me by my committee members, Dr. Tracy Nelson and Dr. David Fullwood,

were greatly impactful towards the completion of my thesis.

Corey Smithson’s contributions made this work possible. I express my appreciation for his

hard work and attention to detail with the polishing and etching of samples.

I am grateful for Paul Minson’s expertise and the time he took to teach me about scanning

electron microscopes and energy-dispersive x-ray spectroscopy. Thank you to Clint Bybee, Therin

Garrett, and Jason Redding for their help with maintaining and operating the Concept Laser LPBF

system. I also express my appreciation to Serah Hatch, Clayton Young, and Taylor Greenwood

for allowing me to use their direct write setup and for their assistance throughout that endeavor. I

thank Colton Inkley and Trent Colton for their efforts and assistance in using their inkjet deposition

setup. In addition to these, I would also like to thank Derek Black, John Hunt, McKay Sperry, Nick

Wallace, and any other friends or classmates that provided a listening ear, advice, or encouragement

through the many difficulties I faced in conducting this research.

I would like to acknowledge the divine guidance and strength that I have received during

this pursuit.

The encouragement and prayers for my success given by my parents, in-laws, and other

family and friends helped me greatly throughout my time here at BYU.

Finally, and most of all, I would like to thank my wife, Lauren, for her unfailing love and

support through the challenges associated with completing this work.

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Laser Powder Bed Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Challenge/Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Chapter 2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.1 Metal AM Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.1.1 Directed Energy Deposition . . . . . . . . . . . . . . . . . . . . . . . . . 62.1.2 Powder Bed Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1.3 Binder Jetting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2 Property Control Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.2.1 Parameter-based Property Control . . . . . . . . . . . . . . . . . . . . . . 102.2.2 Composition-based Property Control . . . . . . . . . . . . . . . . . . . . 13

2.3 Spatial Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3.1 Uniform Multi-material . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.3.2 One-dimensional Multi-material . . . . . . . . . . . . . . . . . . . . . . . 152.3.3 Three-dimensional Multi-material . . . . . . . . . . . . . . . . . . . . . . 15

2.4 Proposed Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Chapter 3 Evaluation of Liquid Doping Methods for Use in Laser Powder Bed Fusion 183.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1.1 New Approach to Multi-material Laser Powder Bed Fusion . . . . . . . . 193.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.2.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.2.2 Direct Write . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.2.3 Inkjet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.3.1 Deposition Quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.3.2 Integration Feasibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.3.3 System Productivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Chapter 4 Influence of Liquid-encased Zirconia Dopant on Melt Pool Characteris-tics in Laser Powder Bed Fusion . . . . . . . . . . . . . . . . . . . . . . . 37

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

4.2.1 Materials and Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

iv

4.2.2 Inkjet Deposition and Single-track Laser Scans . . . . . . . . . . . . . . . 404.2.3 Sample Preparation and Measurement . . . . . . . . . . . . . . . . . . . . 43

4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.3.1 Visual Classifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.3.2 Profile Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.3.3 Melt Pool Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Chapter 5 Integration of Liquid-encased Zirconia Dopant into 316L Stainless SteelMelt Pools in Laser Powder Bed Fusion . . . . . . . . . . . . . . . . . . . 56

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

5.2.1 Materials and Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.2.2 Inkjet Deposition and Single-track Laser Scans . . . . . . . . . . . . . . . 585.2.3 Sample Preparation and Measurement . . . . . . . . . . . . . . . . . . . . 615.2.4 Analysis of EDX Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685.3.1 Dopant Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695.3.2 Dopant Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Chapter 6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 836.1 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6.1.1 Deposition Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 836.1.2 Processing Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 846.1.3 Dopant Incorporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856.2.1 Deposition Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856.2.2 Processing Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856.2.3 Dopant Incorporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 866.2.4 Material Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 866.2.5 Property Improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

Appendix A Lack of Mixing Between Built Material and Build Plate . . . . . . . . . . 99

Appendix B Parameter Sets Used in Chapter 4 . . . . . . . . . . . . . . . . . . . . . . 101

Appendix C Derivation of Zirconia Fraction Value from Zirconium Fraction Valuefor a Given Pixel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

Appendix D Derivation of Predicted Melt Pool Concentrations . . . . . . . . . . . . . . 106

Appendix E Implementation of Box Separation for dIndex Calculations . . . . . . . . 108

v

Appendix F Cumulative Distribution Graphs . . . . . . . . . . . . . . . . . . . . . . . 109

Appendix G MATLAB Script for Zirconia Content and Dispersion Calculations Basedon EDX Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

vi

LIST OF TABLES

3.1 Direct write values using alumina slurry . . . . . . . . . . . . . . . . . . . . . . . . . 233.2 Direct write values using zirconia slurry . . . . . . . . . . . . . . . . . . . . . . . . . 243.3 Inkjet deposition parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.4 Inkjet deposition calculated results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

4.1 Direct write values using alumina slurry . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.1 Inkjet deposition lime spacing and resulting dopant quantity metrics . . . . . . . . . . 595.2 Parameter sets used for single-track laser scans . . . . . . . . . . . . . . . . . . . . . 61

B.1 Parameter sets used for single-track laser scans . . . . . . . . . . . . . . . . . . . . . 101

vii

LIST OF FIGURES

1.1 Typical LPBF process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Novel multi-material LPBF process . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1 Multi-material DED process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Typical LPBF process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.3 Laser scan parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.4 Multi-material LPBF system designed by Wei et al. . . . . . . . . . . . . . . . . . . . 16

3.1 Direct write liquid dopant deposition method . . . . . . . . . . . . . . . . . . . . . . 203.2 Inkjet liquid dopant deposition method . . . . . . . . . . . . . . . . . . . . . . . . . . 213.3 Direct write deposition system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.4 Pocket walls form small, independent powder beds . . . . . . . . . . . . . . . . . . . 253.5 Single-nozzle inkjet deposition system . . . . . . . . . . . . . . . . . . . . . . . . . . 263.6 Resultant direct write slurry deposits . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.7 Resultant inkjet slurry deposits in powder . . . . . . . . . . . . . . . . . . . . . . . . 283.8 Skipping during direct write deposition . . . . . . . . . . . . . . . . . . . . . . . . . 303.9 Skipping during single-line direct write deposition . . . . . . . . . . . . . . . . . . . . 303.10 Recently-deposited alumina slurry of varying concentrations . . . . . . . . . . . . . . 313.11 Errors present in resulting inkjet-deposited powder beds . . . . . . . . . . . . . . . . . 32

4.1 Power-velocity process map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384.2 Images of powder bed pocket platforms and walls . . . . . . . . . . . . . . . . . . . . 414.3 Images of zirconia-doped powder beds and single-track laser scans . . . . . . . . . . . 424.4 Examples of five melt track classification levels . . . . . . . . . . . . . . . . . . . . . 444.5 Image of etched melt pool cross-section with characteristic melt pool measurements . . 444.6 Dopant input vs. parameter process maps for scan speed, laser power, and laser spot size 464.7 Melt pool and corresponding zirconium content map from Chapter 5 . . . . . . . . . . 474.8 Dopant input vs. energy density process map . . . . . . . . . . . . . . . . . . . . . . 484.9 Profile roughness vs. energy density for varying dopant levels . . . . . . . . . . . . . 504.10 Melt pool depths and widths for various energy densities . . . . . . . . . . . . . . . . 524.11 Melt pool heights for various energy densities . . . . . . . . . . . . . . . . . . . . . . 534.12 Illustration of possible coater blade damage when melt pool height is greater than twice

the layer height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5.1 Images of powder bed pocket platforms and walls . . . . . . . . . . . . . . . . . . . . 595.2 Images of powder bed pocket platforms and walls . . . . . . . . . . . . . . . . . . . . 605.3 EDX data for wt% Zr and corresponding etched melt pool . . . . . . . . . . . . . . . 625.4 Image analysis process for determining dopant content and distribution in melt pool . . 645.5 Examples of three different types of melt tracks . . . . . . . . . . . . . . . . . . . . . 685.6 Etched images of unreliable melt pools . . . . . . . . . . . . . . . . . . . . . . . . . . 695.7 Etched images of reliable melt pools . . . . . . . . . . . . . . . . . . . . . . . . . . . 695.8 Graph of measured zirconia content vs. input dopant amount . . . . . . . . . . . . . . 705.9 Graph of measured zirconia content vs. predicted dopant content . . . . . . . . . . . . 71

viii

5.10 Noise-reduced colormaps of melt pools at varying dopant input levels using variousparameter sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

5.11 Averaged colormaps of melt pools at varying dopant input levels using various param-eter sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

5.12 Cumulative distributions of zirconia content in melt pools by pixel value . . . . . . . . 775.13 Zoomed in version of cumulative distributions of zirconia content in melt pools by

pixel value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785.14 SEM backscatter images and EDX analysis of individual zirconia particles . . . . . . . 795.15 dIndex values for zirconia dispersion in the melt pool using 1x1 pixel and 2x2 pixel

box sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 805.16 Other dispersion metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

A.1 EDX maps of main elements in SS 316L over fused material area . . . . . . . . . . . . 99A.2 Cr and Ni content with depth in LPBF-built platform . . . . . . . . . . . . . . . . . . 100

F.1 Cumulative distributions for parameter set 7 of zirconia content in melt pools by pixelvalue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

F.2 Cumulative distributions for parameter set 8 of zirconia content in melt pools by pixelvalue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

F.3 Cumulative distributions for parameter set 9 of zirconia content in melt pools by pixelvalue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

F.4 Cumulative distributions for parameter set 10 of zirconia content in melt pools by pixelvalue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

F.5 Cumulative distributions for parameter set 11 of zirconia content in melt pools by pixelvalue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

F.6 Cumulative distributions of zirconia content in melt pools by pixel value for sampleswith 4.8 wt% zirconia input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

ix

CHAPTER 1. INTRODUCTION

1.1 Laser Powder Bed Fusion

Laser powder bed fusion (LPBF) is an additive manufacturing (AM) technique that uses a

laser to selectively melt and fuse powders in a layer-by-layer fashion to form completed parts (see

Figure 1.1). The layer-by-layer nature of LPBF enables the creation of complex geometries and

internal features that are costly or impossible to make using conventional manufacturing methods

(e.g. machining, casting, etc.) [1]. The ability to create these types of geometries opens up many

design opportunities such as weight reduction and optimization [2], biomimetic design [3, 4], and

functionality integration – combining the functions of multiple components into a single part [1].

Build Chamber Powder Chamber

Finished Part

Unmelted Powder

Scanner System Laser

Coater Blade

Powder Reservoir

Figure 1.1: Typical LPBF process. At the beginning of each layer, the build chamber moves down,the powder chamber moves up, and the coater blade moves fresh powder from the powder chamberto the build chamber. A laser then selectively melts powder to create a solid layer. This is repeateduntil the complete part is formed.

1

While LPBF is utilized with metals, polymers, and ceramics, this work will focus on its application

to metals.

1.2 Challenge/Motivation

While the benefits of LPBF are impressive, the results have been limited because of the

single-material nature of powder bed systems. In many applications, a distribution of mechani-

cal, thermal, and chemical properties throughout a single part is necessary to fulfill the required

functions [5, 6]. One example of this is a turbine blade in a jet engine [7]. This type of turbine

blade requires high temperature resistance and high stiffness on the exterior, while the interior re-

quires higher thermal conductivity to cool the blade during use. Additionally, the base of the blade

requires high ductility and high fatigue life. Current LPBF methods (and most conventional man-

ufacturing methods) require that a single material be chosen that comes closest to satisfying this

collection of competing requirements and then be subjected to extensive post-processing to meet

final specifications. Alternatively, a varied material composition across a part would allow for the

creation of location-specific properties for required functions.

In recent years, efforts have been made to create composite structures in LPBF by mixing

multiple material powders or by using composite powders such as oxide dispersion strengthened

steels [8–12]. A part made from these composites can exhibit property improvement compared to

the same part made using the standard material. However, the resultant properties are still uniform

across the part geometry. Many researchers have also begun to explore the possibility of creating

functionally graded materials (FGMs) in LPBF by using different powder compositions during

different parts of the build. However, these FGMs are still typically limited to variation in one di-

mension (the build direction) created by changing out the powder hopper partway through the build

(e.g. steel for the first 200 layers, then alumina for the next 200 layers). Due to the complexity of

changing the powder source, these FGMs are also limited to large composition changes that can

cause defects in the parts due to poor adhesion between dissimilar materials [13–15]. Other ap-

proaches that attempt to vary the composition in three dimensions present additional complications

with powder recycling and extremely long build times [16–18]. While other metal AM processes

like binder jetting (BJ) and direct energy deposition (DED) both have the capability to create spa-

2

tial composition variations, LPBF is still largely preferred for uniform material applications in

industry because it yields superior properties.

In addition to the benefits that typically accompany AM parts, spatial composition control

in LPBF would lead to products with superior performance capabilities (strength, hardness, wear

resistance, etc.) and increased weight reduction. The proposed research focuses on necessary steps

toward three-dimensional spatial composition control in LPBF that does not add significantly to

the build time or require extensive post-processing.

1.3 Objective

A new concept that is being developed at Brigham Young University involves the use of

liquid dopant or dopant in a liquid carrier added to a standard powder bed to create small composi-

tion changes (<10%) in select regions (see Figure 1.2). It is anticipated that the small composition

change will affect the properties of a built part in select regions – such as hardness, wear resistance,

A B C D

Coater BladeInkJet Nozzle Array

Dopant in Liquid Precursor Dopant without

Liquid Precursor

Material with Dopant Without

Dopant

TopView

SideView

Spread Powder,Deposit Dopant

Evaporate Liquid Precursor

Fuse Material and Dopant

Finished Part

Figure 1.2: Novel multi-material LPBF process. Images on top row are top views, images onbottom row are side cut views. (A) Coater blade spreads powder layer while inkjet deposits dopantin liquid precursor in select areas on top of fresh powder layer. (B) Low power laser scan evaporatesliquid precursor leaving behind dopant in powder layer. (C) High power laser scan fuses togetherpowder layer with dopant in select areas. (D) Finished layer, ready for additional powder layers.

3

corrosion resistance, etc. – without the negative impacts seen with large composition changes. The

dopants used in steels could include oxides, other dispersion strengthening agents, or solid solution

strengthening agents to increase hardness and strength [8, 19–25]; chromium or nickel to increase

corrosion resistance [26]; tungsten, titanium, or chromium to increase hardness by forming car-

bides [27, 28]; and titanium oxide or sulfur to change the microstructure by altering the melt pool

geometry [29–31]. The liquid deposition in the powder bed will be accomplished via an inkjet

printing nozzle or similar system attached to the coater blade. The inkjet printing system will not

add significantly to the build time because it can run simultaneous to the recoating process that

happens after each layer. With the spatial control available in inkjet printing, the dopant will only

be deposited in areas that will be solidified as part of the build. This means that the surrounding

powder will retain its original composition allowing for full recycling of the unused powder. By

controlling the amount of dopant added to specific locations, the designer will be able to control

and adjust the local properties of selected regions within the part.

This concept is based on several assumptions which must be validated to demonstrate the

feasibility of the process. These important assumptions include:

1. Liquid dopant can be introduced into the powder bed in a controlled, accurate manner.

2. Laser processing will combine the dopant and powder into a single, metal-matrix composite

with the base powder being the matrix material.

3. Dopant inclusions will be of a size, distribution, and uniformity to have a beneficial impact

on the resulting microstructure.

4. Dopant inclusions will have a significant, beneficial impact on the properties of the resulting

part.

5. The deposition system can be integrated with the existing LPBF coater blade system.

6. Control of the deposition system can be synchronized with the controls of the commercial

LPBF system.

The intent of the current research is to address assumptions 1, 2, and 3 stated above. The

main objective is to assess the feasibility of composition control in LPBF-manufactured parts using

4

liquid or liquid-encased dopants and to assess the potential magnitude of property variation in a

representative material system. Specifically, it is of interest to understand the limits of the dopant

deposition methods and the properties of the resulting material system that develops through laser

processing. Dopant quantity and dispersion are of particular importance.

1.4 Thesis Outline

This thesis is organized as follows: Chapter 1 outlines the background and motivation and

then provides a brief overview of a novel method of spatial property variation in metal LPBF. It

concludes with the objectives for the current research.

Chapter 2 provides background information on metal AM as well as previous attempts at

multi-material and spatial property control in metal AM.

Chapter 3 presents an experimental evaluation of two different methods of depositing liquid

dopants for use in LPBF. This includes direct write and inkjet printing techniques adapted for use

with dopants contained in colloidal slurries. It details the impacts of certain key parameters for

both methods as well as limitations and suggestions for implementation.

Chapter 4 provides data and analysis on melt pool characteristics in zirconia-doped SS

316L powder beds in LPBF for a variety of parameter sets and zirconia concentrations. Process

maps are presented for varying zirconia concentrations, laser powers, scan speeds, spot sizes, etc.

Recommendations are made as to the best parameter sets for future work.

Chapter 5 investigates the effects of zirconia-doped SS 316L processed using LPBF. Focus

is placed on the resulting concentrations of dopant in the melt pool compared to the input dopant

amounts. Resulting distribution of the dopant and its effect on microstructure is examined.

Chapter 6 reviews the findings of the current research and presents a summary of the con-

clusions made. Directions for future research are provided.

5

CHAPTER 2. BACKGROUND

2.1 Metal AM Processes

In AM, there are three main processes that are widely preferred in industry for processing

metals: directed energy deposition (DED), powder bed fusion (PBF), and binder jetting (BJ) [32].

2.1.1 Directed Energy Deposition

In DED, an energy source (laser or electron beam) and a stream of powder or wire feedstock

are directed at the same point on the surface of a solid base or pre-existing part. There, the new

material and substrate surface are melted by the energy source and solidify as the source moves

away to form a solid part (see Figure 2.1).

This method is often used for repairing broken tooling and adding coatings of different

material to a part because it does not have to build up in the same, one-dimensional, layer-wise

manner as most AM processes [34]. Since the material is introduced through the nozzle with the

energy source, depositions can be made in any area on an existing part that is accessible to the noz-

zle. The nozzle can even be mounted on a 5-axis CNC machine to allow greater access to complex

parts and combined with machining tools to create a hybrid additive/subtractive manufacturing

system [35]. It is important to note that this method of adding material to a component has been

shown to exhibit superior bonding between the new and old material than conventional methods

of material coating (chemical vapor deposition, physical vapor deposition, thermal spray coating,

etc.) [34]. Another benefit of DED is that the resulting properties of a part can be controlled by

manipulating the microstructure and composition. The microstructure can be accurately predicted

as a function of the processing parameters (laser power, speed, etc.) used during deposition and

the composition can be regulated by adjusting the type and proportions of materials that are used

at a given time [35].

6

Figure 2.1: Multi-material DED process. Powders of the materials of interest are mixed beforeentering the nozzle where they are guided into the path of the laser beam. The material is meltedby the laser and solidifies on the substrate in the desired region. Used with permission. Image fromBrueckner et al. [33].

While DED does have impressive advantages over other types of AM, there are some draw-

backs that limit its use in other industrial applications. Since the process requires a substrate on

which to deposit the new material, it is often necessary to add dense supports for parts with any

sort of overhang or internal structure to avoid part deformation. This, along with limited access to

internal geometries, limits the complexity of structures that can be fabricated using DED [35]. The

DED process can also be slow relative to other metal AM processes because each section of de-

posited material must have sufficient time under the energy source to melt completely along with a

section of the substrate surface. The larger weld pool (compared to LPBF) also limits the minimum

feature size and creates coarser microstructures [36]. Additional time is required for post-process

7

finishing because of the poor resolution and surface finish as well as post-process heat treatments

to mitigate the residual stresses that are often present in DED parts [35].

2.1.2 Powder Bed Fusion

In metal PBF, a layer of powder is spread over a build chamber and selectively scanned by

an energy source (laser or electron beam) to create a solid layer. This is repeated multiple times

until a complete part is formed (see Figure 2.2). The present work focuses on laser powder bed

fusion (LPBF).

PBF is the most widely used of the metal AM methods because of its ability to process a

variety of materials in very complex geometries with excellent resulting material properties [37].

Its current uses range from forming jewelry out of precious metals to creating high-consequence

aerospace components out of superalloys [34]. The high resolution demonstrated in PBF fabrica-

tion of jewelry also makes possible the production of custom, anatomically-matched dental and

Build Chamber Powder Chamber

Finished Part

Unmelted Powder

Scanner System Laser

Coater Blade

Powder Reservoir

Figure 2.2: Typical LPBF process. At the beginning of each layer, the build chamber moves down,the powder chamber moves up, and the coater blade moves fresh powder from the powder chamberto the build chamber. A laser then selectively melts powder to create a solid layer. This is repeateduntil the complete part is formed.

8

medical implants and tool fixtures. Meanwhile, the complex geometric capabilities of PBF pro-

cesses allow for the creation of intricate internal features such as cooling channels and optimized,

weight-reduced structures.

Like DED, the process characteristics of PBF that afford some of its greatest advantages

over other manufacturing techniques also limit its capabilities. The full melting of the material

creates instabilities in the powder bed that require support structures to form correctly [38]. Support

structures are also necessary for heat transfer from the melted material away from the part and for

the resulting residual stresses that are common in larger parts [37, 39]. For these residual stresses

to be mitigated, post-process heat treatments may also be required. Another downside to powder

bed processes, in general, is that the spreading of powder for each layer – in increments as small as

30 um – can cause extremely long build times [37]. Also, changing materials in the middle of the

build is costly/time intensive and is one of the only ways to introduce non-uniform composition

into the PBF process.

2.1.3 Binder Jetting

BJ is a layer-by-layer powder bed system (like PBF), that uses an inkjet printhead to selec-

tively deposit a liquid binder in the metal powder instead of melting powder with an energy source.

A heat source then cures the binder to hold the metal powder that it surrounds to create a “green”

part which is later sintered to improve density and mechanical properties [40, 41].

One of the most common uses for BJ in the metal industry is the creation of sand molds

for casting non-standard parts such as custom or discontinued automotive parts as well as casting

molds for small lot sizes [34]. These sand molds can reach much larger build envelopes than can be

achieved with PBF but with similar accuracies [32]. BJ can also be more cost-effective than other

AM methods because of its higher speed and lower power consumption [40]. Since inkjet printing

is already a well-established technology, speed can be increased in BJ by simply adding more

nozzles making it more scalable than other metal AM processes. Less power is consumed because

the process does not require an inert atmosphere or high-power energy source. Another benefit of

BJ is that support structures aren’t needed because the powder beds are self-supporting [40]. This

allows for more parts to be processed in a single build by nesting the parts throughout the height of

the build. Inkjet deposition is also advantageous for manipulating the concentration by including

9

additives in the binder. In the same way inkjet printers can produce full-color designs across a part,

BJ has the potential to create a variety of compositions at any given point within a part [40].

Despite all of the advantages available in BJ, it is sometimes more difficult to implement

in industry because it requires extensive post-processing [40]. Green parts must be sintered to

form solid parts and may need to be infiltrated with another metal to create fully dense parts.

The sintering step allows for a wider range of materials to be used that are not suitable for melt

processing, however, shrinkage during sintering limits the part accuracy of BJ parts and must be

accounted for during the design stage [32].

2.2 Property Control Methods

Of these three methods, DED and BJ are both capable of spatially controlling the com-

position throughout a part. However, LPBF produces parts with the mechanical properties most

comparable to those achieved using conventional engineering-grade materials [37]. LPBF is gener-

ally preferred in industry because of its superior properties when using standard AM materials [32];

however, it cannot compete with the compositional flexibility of DED and BJ. If the spatial com-

position control available in BJ and DED could be applied to the LPBF process, it would provide

the capability to create parts with vastly superior properties to those that can currently be made

using any one of these processes. This section describes current capabilities in altering mechanical

properties in LPBF-produced parts by controlling the processing parameters and the composition.

2.2.1 Parameter-based Property Control

The main processing parameters that can be changed in LPBF include laser power, scan

speed, scan spacing, and laser spot size (see Figure 2.3); all of which have a substantial effect on

the resulting properties of a part [42–46]. These parameters all have a direct effect on the energy

density – the amount of energy that is delivered to a given area within the powder bed. This energy

melts the powder to form a melt pool which then solidifies.

Leaving the other parameters constant, increasing the laser power will increase the effective

energy density because more power is being delivered to the same powder area in the same amount

of time [47]. Conversely, increasing the scan speed will decrease the amount of time for which the

10

Laser exposure vector

Scan speed direction

Laser spot size

Melt pool width

Scan spacing

Figure 2.3: Illustration of consecutive laser scans with associated processing parameters.

laser is delivering energy to a certain area of powder, effectively decreasing the energy density [47].

Increasing the spot size will also decrease the energy density by distributing the laser energy over

a greater area [48]. The scan spacing is slightly different from the rest of the parameters because

it controls how much of the melt pool is remelted rather than directly controlling the size/shape of

the melt pool. An increase in scan spacing will effectively spread the laser energy over a greater

area creating an effective decrease in energy density [49].

By controlling the amount of energy delivered to the powder bed, the user controls the size

and shape of the individual melt pools which has a direct effect on the resulting microstructures

and, therefore, properties of the part [50–53]. Clymer et al. [42] showed that the yield strength

of SS 316L could vary between 325 and 700 MPa just by adjusting the laser power and scan

speed. However, the effects of these processing parameters are limited by the property range of

the material composition (i.e. there is no set of processing parameters that will give AlSi10Mg

the strength of Ti6Al4V). In order to obtain properties outside of the range that is attainable by

manipulating processing parameters, the composition must be altered.

Energy Density

Linear, areal, and volumetric ED have all been used to describe the energy input in LPBF;

however, each definition has its particular flaws and virtues. Linear ED (EDL) is usually defined

as

11

EDL

[J

mm

]=

Pv

[W

mm/s

](2.1)

where P is the laser power and v is the scan speed resulting in units of energy per distance. The

benefit of using the EDL is that P and v are the process parameters that are most commonly changed

to influence the resulting part properties and is often shown as a complete design space [42]. While

EDL does not consider other parameters like laser spot diameter (dS), it has been shown to correlate

with single-track melt pool characteristic changes when dS is constant [54].

Areal ED (EDA) is based on the same relationship as EDL, but adds dS to the denominator

to define the energy intensity [55]. This intensity is measured as the laser power density (PA) as

PA

[W

mm2

]=

P

π (dS/2)2

[W

mm2

](2.2)

used by Yadroitsev et al. and Leung et al. [38, 56]. PA can be combined with the laser dwell time

(td) defined as

td [s] =ds

v

[mm

mm/s

](2.3)

and simplified to obtain

EDA = PAtd =

(PAS

)(dS

v

)=

PvdS

(4π

)(2.4)

with units of energy per area. The constant, 4/π , can be omitted for comparative purposes resulting

in

EDA

[J

mm2

]=

PvdS

[W

(mm/s)(mm)

](2.5)

which is used often in the literature [55, 57, 58]. Using the power density and dwell time is a

sensible way to derive the energy input; however, this equation does not provide any insight to

how the energy reacts with the powder and is not relatable between different materials [58].

Volumetric ED (EDV ) can be construed in multiple ways. Most often, it is defined as a

combination of laser power (P), scan speed (v), hatch spacing (hd), and layer thickness (t) as

12

EDV

[J

mm3

]=

Pvhdt

[W

(mm/s)(mm)(mm)

](2.6)

to give a measurement of J/mm3 [55, 59, 60]. Although this gives the units of energy per volume,

hd only influences the energy input for sequential line scans when the hatch distance is smaller

than the powder consolidation zone (i.e. area scans) [49] and t is often deemed irrelevant because

the melt pool of any given laser scan is not confined to the height of the powder layer [61].

Of these three methods for measuring energy density, all have been shown on some oc-

casions to fit experimental data trends in some way [54, 62, 63]. However, these metrics are still

subject to change based on laser type and the material being used [64]. More complex associa-

tions of laser processing parameters to resulting melt pool structures have also been made in the

literature [58, 64, 65].

2.2.2 Composition-based Property Control

Significant improvements to mechanical properties can be made through small changes

in composition. Oxide dispersion strengthened (ODS) steels, for example, have significantly im-

proved high-temperature strength compared to ordinary steels even though the oxide nanoparticles

make up less than 1 wt% of the overall composition [19, 21, 22, 66]. Adding a small amount of

materials like titanium oxide, sulfur, or selenium to a laser weld can also change the geometry

of the melt pool that is formed by the laser [29–31]. This change in the melt pool geometry, in

turn, influences the resulting microstructure and mechanical properties of the weld. In casting,

inoculants are also used to stimulate nucleation during solidification in an effort to produce more

equiaxed, isotropic grains [67]. Small composition changes are also commonly seen in similar

alloying elements that are used for different applications. 316L stainless steel differs from the

standard SS 316 by a mere 0.05 wt% of carbon; but this difference has proven to be imperative in

reducing sensitization when used in welding or similar applications [26]. In each of these cases,

small changes in composition create sizable effects in the resulting properties of the part.

While the standard materials for LPBF – Ti-6Al-4V, AlSi10Mg, SS 316L, SS 17-4PH –

are all useful in different applications, many researchers have begun to explore other material

systems that could be used in LPBF [8, 10, 24, 27, 68, 69]. Many of these material systems would

13

be classified as metal matrix composites that have a main “matrix” material and some amount of

an additive with no variation throughout the part. Usually, these composites are created by mixing

powders together by ball milling or roller mixing [9, 11, 70], or by using powder derived from a

pre-alloyed version of the composite by gas or water atomization [8, 11]. 316L stainless steel, for

example, has been shown to have increased hardness as well as yield and tensile strengths with the

addition of 1 to 3 wt% alumina powder to the raw LPBF materials [9].

2.3 Spatial Control

The various attempts at multi-material fabrication in LPBF can be separated into 3 cat-

egories: uniform, one-dimensional, and three-dimensional. Uniform multi-material (UMM) in-

cludes most attempts that have been made to create composites or new alloys by virtue of the

various AM processes. These UMM strategies were highlighted in Section 1.2. One-dimensional

multi-material (1MM) defines attempts to create functionally graded materials (FGMs) that vary

the material composition along a single axis (usually the build direction). FGMs were also touched

on in Section 1.2, but will be explored further in the current section. Three-dimensional multi-

material (3MM) encompasses attempts to vary the composition throughout the build independent

of location or build direction.

Binder jetting and directed energy deposition are both examples of AM processes that ex-

hibit 3MM capabilities [28]. In BJ, some nozzles can be supplied by a binder filled with nanoparti-

cles [41] or metal salts [71] while others deliver a standard binder. This allows for complete control

over which areas within the final part include the added material while the entire green part is still

held together with the necessary binder. With DED, multiple hoppers can be used in conjunction

to create a system that varies the composition based on how much each hopper is opened at any

given point in the build (see Figure 2.1) [33].

While some of these forms of spatial control can be implemented with changes in process-

ing parameters to effect property variation (see Section 2.2.1), the current work focuses primarily

on spatial composition control to create property variation throughout a part.

14

2.3.1 Uniform Multi-material

In recent years, efforts have been made to create composite structures in LPBF by mixing

multiple material powders or by using powders made from composites such as oxide dispersion

strengthened steels [8–12]. A part made from these composites can exhibit property improvement

compared to the same part made using the standard material. However, the resultant properties are

still uniform across the part geometry. These types of parts would be classified as UMM because

there is no controlled variation in the resulting properties throughout the part.

2.3.2 One-dimensional Multi-material

As mentioned previously, some researchers have also begun to implement the principles of

FGMs using different AM processes as an alternative to traditional FGM manufacturing [6]. DED

has typically been one of the main AM methods used to create FGMs because of its versatility in

powder selection (i.e. different powder hoppers can be opened and closed as desired as seen in

Figure 2.1) [28]. However, one of the major limiting factors of this method is that deposited ma-

terial often has trouble bonding to surfaces of differing compositions or it may form cracks during

solidification leading to major defects and quality issues [28]. In LPBF, FGMs have typically been

fabricated by changing the powder source after a certain number of layers [13–15]. These LPBF-

manufactured FGMs also struggle with material bonding, but in some cases have proved to be

capable of creating parts with good adhesion between different materials [14]. While these meth-

ods have some potential benefits, they are still limited in terms of spatial composition variation and

build rates. The composition changes are limited to a single dimension and only implement large

changes in material composition. These methods also usually take more time because of their more

intricate powder handling systems – more than one powder bed, powder-mixing hoppers, etc. –

which increases the overall cost significantly.

2.3.3 Three-dimensional Multi-material

Chivel [16] introduced an interesting concept for multi-material laser powder bed fusion

(MM-LPBF) that varied the selected material across three dimensions by using materials with

15

different sized powders. However, the method of powder recycling that he presents would be diffi-

cult at best to actually separate the powders completely. Wei et al. [17] surpassed Chivel’s original

ideas to develop a MM-LPBF system that uses a vacuum to remove base material powder in certain

regions and a selective dispenser to deposit powder where base powder was removed (see Figure

2.4). This system has since been adapted for use with metal-metal [72, 73], metal-glass [18], and

metal-polymer [74] material systems. Although this system does provide 3-dimenional property

variation, it also increases build time significantly. In addition to the normal laser scan time and

spreading of a new powder layer, this process also requires vacuuming time to clear powder in

certain areas as well as the time to deposit powder of a different material in those same areas with

a scanning deposition nozzle [17]. Since the initial powder layer spreading takes a significant por-

tion of the build time in standard LPBF, this process could more than double the build time in many

cases. Furthermore, this method has still only been used with vastly different materials; in many

cases, large composition changes have been found to produce challenges such as delamination

(poor adhesion) and thermally induced stresses (due to differing expansion coefficients) [14].

Figure 2.4: Multi-material LPBF system illustration. In this system, powder is moved from thesupply chamber to the build chamber like a standard LPBF machine. The laser then scans theportion of the layer that should use the base material. The vacuum sucker is then used to removematerial in select regions which are then refilled by the powder dispenser with the secondary pow-der. The roller is used to level the new powder and the laser scans the remaining portion of thelayer. Used with permission. Original image by Wei et al. [17].

16

2.4 Proposed Work

The current studies have as their purpose to validate the assumptions stated and advance

the novel composition control concept described in Section 1.3. This is divided into three exper-

imental inquiries. The proposed work involves, first, evaluating and recommending a deposition

method to be used as the mode of dopant delivery for the composition control concept; second,

investigating the effect that the addition of dopant will have on the LPBF processing window; and

third, determining how well the dopant is incorporated into the fused material.

17

CHAPTER 3. EVALUATION OF LIQUID DOPING METHODS FOR USE IN LASERPOWDER BED FUSION

3.1 Introduction

Additive manufacturing (AM) – and laser powder bed fusion (LPBF) in particular – is well

known for its ability to construct geometric forms that would not be possible using standard man-

ufacturing techniques [1, 4, 32]. This geometric versatility has inspired a design practice known as

functionality integration – where multiple components are condensed to create a single part that

performs all of the functions that were previously attributed to the individual components. While

this practice can be quite useful in terms of weight reduction, part count reduction, and meeting

spatial confinements; there are also serious drawbacks [2]. One of these drawbacks occurs in LPBF

when different functions integrated into a single part have extremely different requirements (e.g.

high fracture toughness on the interior and high hardness on the exterior), but the process can only

sustain a single material [6]. This dilemma generally results in over-designing some areas of the

part to compensate for the compromise in material choice. Over-designing can lead to decreased

functional efficiency, increased weight, decreased fatigue life, etc. in LPBF parts. Creating meth-

ods to control the material composition spatially throughout a build would allow for designers to

mitigate the negative effects and experience the full benefits of functionality integration.

The literature reports multiple attempts to expand LPBF processes into the multi-material

regime. One approach is to simply switch out the powder feedstock for a different material at a

certain point in the build [14, 15]. While this has been done successfully in some cases, it greatly

increases the overall build time. Because of the time sacrifice, it is generally limited to a single,

large composition change during the build. This discrete change in composition can be problem-

atic. Materials with vastly different properties create extra residual stresses from thermal expansion

and contraction during fabrication and poor adhesion at the inter-material surface due to poor wet-

ting and insufficient mixing [13, 14]. Some studies have shown that adhesion can be improved by

18

remelting the inter-material zone multiple times to improve mixing between the dissimilar materi-

als, but this adds even more time to the process [14]. In addition to quality concerns, this method

is still limited to material change in one direction and few changes throughout the part.

Other research has been done on methods of three-dimensional variation of composition.

Wei et al. [17] have implemented a process that uses a vacuum to remove base powder in select ar-

eas and a separate hopper and nozzle to deposit powders of different compositions in the excavated

areas. This novel method has been shown to work with glass-metal, metal-metal, ceramic-metal,

and metal-polymer systems and has shown significant improvement in terms of spatial control

throughout all three dimensions [17, 18, 72–75]. The method of Wei et al. has the downside of

allowing powder of different compositions that is not fused to remain in the powder bed and con-

taminate the unfused powder that would be recycled. Since powder recycling is one of the main

factors that contributes to the economic feasibility of LPBF, this method significantly increases

the cost of the overall process [37]. The other significant cost issue that is affected by this multi-

material method is the increase in time necessary to complete each layer. Vacuuming out and

replacing powder can more than double the overall process time that originally consisted of only

powder spreading and laser scanning. Large composition changes in three dimensions present the

same issues observed for large composition changes in one dimension.

3.1.1 New Approach to Multi-material Laser Powder Bed Fusion

Many of the issues that are present in these current multi-material LPBF (MM-LPBF) ef-

forts could be mitigated or eliminated by using liquid or liquid-encased dopants as the means of

altering composition in a way that does not significantly add to the build time. The nature of the

LPBF process provides two distinct opportunities for the introduction of liquid dopant to the pow-

der bed. The first of these opportunities occurs after a layer is fused by the laser but before the

next layer of powder is spread. The second is just after the powder is spread but before the layer is

fused.

In the first case, liquid dopant could be deposited on the solid substrate using either a

direct write (controlled extrusion through an orifice) or droplet-based deposition method like inkjet

printing (see Figure 3.1). The liquid would be dried either by the heat in the build chamber or using

an exterior heat source (such as a lamp or the laser at low power) after which the next layer of

19

A B C D E

Fused Layer Deposit Dopant Spread Powder Fuse Material Cross Section

NeedleCoater blade

Fused materialDopant

Single-trackmelt poolsPowder

Laser beam

Figure 3.1: Direct write liquid dopant deposition method. The liquid dopant is deposited usingan open-tip needle directly onto the previous (fused) layer. Once the dopant is dry, the powderis spread over the top and the fusing process is carried out and the base and dopant materials aremixed.

powder would be spread. This method provides a distinct advantage of the dopant being covered

by powder of the base material, which would reduce the amount of dopant that is evaporated or

ejected as spatter during the fusing process. Since the dopant is initially concentrated in a specific

area (between the powder and substrate) this may also affect the final location and dispersion of

the dopant throughout the melt pool. Direct write can also be used with a wide range of suspension

properties and has little risk of clogging in the nozzle. However, direct write can be slow since it

is generally only used with a single nozzle.

In the second case, the powder layer would be spread and then the liquid dopant would

be deposited into the spread powder before the layer is fused (see Figure 3.2). Deposition in this

method could only be done using a droplet-based system like the inkjet printing technology used in

binder jetting – another AM technique. An inkjet printhead could be mounted on the coater blade

system to print while spreading the next layer of powder – minimizing the impact on the total build

time.

This work focuses on evaluating the feasibility of depositing liquid dopants as part of the

LPBF process by exploring the implications of exploiting the two deposition options described.

These situations are replicated in a way that does not permanently alter the existing LPBF system.

Direct write of a liquid dopant onto a solid substrate mimics the deposition onto a fused layer

before the next layer of powder is spread, while inkjet deposition into a powder bed represents

deposition after a powder layer is spread but before the layer is fused together. Simplified versions

20

BA

Spread PowderFused Layer

C D E

Inkjet Print Head Dopant in

Liquid Precursor

Dopant Fused Powder

Deposit Dopant Evaporate Liquid Fuse Material

Coater Blade

New Powder Layer

Figure 3.2: After the powder is spread, liquid dopant is deposited into the powder in the form ofmicro-scale droplets from an inkjet printhead. Once the dopant and powder bed are dry, the fusingprocess is carried out.

of these two methods are performed with careful observation of any requirements or outcomes that

could significantly affect integration of the method to the LPBF process. Conclusions are made

as to the deposition quality, integration feasibility, and system productivity of the two methods in

relation to their application in LPBF.

3.2 Methods

3.2.1 Materials

Two dopant material systems were selected based on the availability of stable commercial

suspensions and the potential for property enhancement in SS 316L. One material system that has

previously been studied with LPBF is alumina (Al2O3) reinforcement in a stainless steel 316L

matrix [9]. In this study, Li et al. showed that the addition of 1 to 3 wt% alumina to stainless steel

processed using LPBF improved hardness in all cases and improved yield and tensile strengths

in the case of 1-wt% alumina addition. While this study by Li et al., and most multi-material

studies in LPBF, have been done using physically mixed matrix and additive powders, the property

improvements shown demonstrate that the alumina/SS 316L material system is a useful material

system for testing composition adjustment using alternative doping techniques.

Another material system of interest is zirconia (ZrO2) reinforcement in a steel matrix.

Koopmann et al. [14] showed that zirconia powder can be processed reasonably well using LPBF

and that it can have good adhesion with steel when the interface between the two materials un-

21

dergoes sufficient mixing during the laser processing. While their research was focused more on

large composition changes, the good mixing between the two materials and processability of zir-

conia indicate that it could be a suitable test subject for feasibility of composition adjustment using

alternative doping techniques.

In both the direct write and inkjet deposition methods, SS 316L powder (CL 20ES, Concept

Laser, 29.9 µm, spherical) is used as the base material, to which either alumina or zirconia is added

in the form of a water-based slurry using one of the two deposition methods described. The alumina

slurry is Gamma B 0.05 µm Alumina from LECO with a 10 wt% concentration alumina with an

added 10 wt% propylene glycol. The zirconia slurry is ZR100/20 from NYACOL with 20 wt%

colloidal zirconia with a mean particle size of 100 nm. The alumina and zirconia concentrations

were measured to be 10.6 and 25.4 wt% respectively. The LPBF machine used is a Concept Laser

M2 Cusing Multilaser.

3.2.2 Direct Write

Direct write deposition describes a system in which a material is extruded directly onto a

substrate through an orifice (see Figure 3.1). The material creates a liquid bridge with the substrate

that moves with the needle and leaves a liquid track behind. This technique would be difficult to

apply to a powder bed as the fluid meniscus would likely move the powder during liquid deposition;

however, it is suitable for deposition on a solid substrate. In the current study, the direct write

deposition is applied to deposit alumina or zirconia slurry onto a solid plate made of SS 316L to

simulate the case in which dopant is deposited on a previously fused layer.

A custom direct write system (see Figure 3.3) was used for printing [76–78]. The system

uses a 3-axis CNC stage and a stepper motor to control the plunger of a syringe with an open-tip

needle to deposit the slurry. The needle used was 25 gauge with a 0.305 mm inner diameter. The

plate was leveled with respect to the x- and y-axis movement of the system and the needle was

zeroed to the surface of the flat plate. Deposition was performed with the needle tip at a distance

of 0.10 to 0.15 mm above the plate surface. Areas and lines were deposited at a flow rate of

0.0454 mm3/mm while the needle traveled at speeds of 80 mm/s relative to the plate surface. Area

depositions were performed in a concentric, rectangular pattern. The spacing between tracks was

varied to control the total amount of slurry deposited in a given area. The dopant concentration

22

Figure 3.3: Direct write deposition system (image from Romero et al. [78] used with permission).The three-axis stage controls the deposition location while the stepper motor above the syringecontrols the rate of material ejection from the needle onto the substrate.

in the slurries was also varied by adding distilled water to dilute the slurry to a predetermined

concentration. The combination of these two adjustments was used to vary the overall dopant

concentration amounts as described in Table 3.1 for the alumina depositions and Table 3.2 for the

zirconia depositions.

After the slurry was deposited, the plate was baked at 200 ◦C for 30 minutes to evaporate

the remaining moisture and propylene glycol (boiling point ∼190 ◦C) additive before installation

to the LPBF machine. Using the LPBF machine’s coater system and build chamber controls, a

layer of powder was spread over the dried alumina by adding 500 µm of powder and removing

powder in 100 and 50 µm increments to get to a 50 µm powder layer. This was done to determine

Table 3.1: Deposition parameters and calculated deposition results for direct write deposition ofalumina slurries over an area of 37 x 25 mm.

Area # Slurry Concentration Volume deposited Area density of alumina(wt%) (mm3) deposited (µg/mm2)

1 3.0 100.28 3.1662 2.0 100.28 2.1203 1.0 100.28 1.0644 0.5 100.28 0.533

23

Table 3.2: Deposition parameters and calculated deposition results for direct write deposition ofzirconia slurries over an area of 37 x 25 mm.

Area # Slurry Concentration Volume deposited Area density of alumina(wt%) (mm3) deposited (µg/mm2)

1 20 97.5 26.4332 10 97.5 11.9073 10 97.5 11.9074 10 97.5 11.907

if the alumina deposits were stable during the powder coating step or if they would break down

and contaminate the powder.

Direct write single-line depositions were also performed to investigate the difference be-

tween small- and large-area depositions. In these cases, the substrate was heated during deposition

to a temperature of 50 ◦C rather than baking the substrate after deposition to simulate a heated

build plate during the LPBF process.

3.2.3 Inkjet

The inkjet method is suitable for printing dopants into spread powder layers. While printing

into powder can be challenging [79,80], it is successfully done in both the binder jetting and multi

jet fusion/high speed sintering commercial processes. In order to simulate dopant deposition onto

loose powder, a solid SS 316L plate was used as a base onto which walls were built to create

isolated pockets of powder. These pockets reduce the error in quantifying the amount of dopant

deposited by limiting the area into which the dopant can spread. The pocket walls were built in

the LPBF machine with a laser power of 370 W , a scan speed of 1350 mm/s, and a spot size

of 130 µm. The walls were built by depositing a 25 µm layer of powder, fusing the pocket walls,

depositing another 25 µm powder layer (50 µm total) and fusing the pocket walls again (see Figure

3.4). These 25 µm increments are smaller than typical to ensure total fusing of the pocket walls

with the base plate so that each pocket was totally separate from the others. Once these pockets

were built, the powder inside the pockets was left undisturbed while the plate was taken out of the

LPBF machine and transported to the inkjet printing station.

24

1 2 3

4 5 6

separated powder beds

pocket walls

Figure 3.4: Pockets walls printed 50 µm tall on substrate to separate powder beds of the sameheight. Powder beds before inkjet deposition are fairly uniform. Pocket walls and individualpowder beds are labeled. Individual pocket labels are also included for future reference.

A simple inkjet setup (see Figure 3.5) was used to deposit colloidal zirconia slurry into

the walled-off powder beds [79, 80]. This setup uses a pressurized chamber connected to a single,

80 µm-nozzle, piezo-electric, drop-on-demand print head (MicroFab Technologies, Inc.; Part #

MJ-AB-01-80-8MX) controlled in coordination with the movements of the stages to create lines

of consistently-spaced droplets. The nozzle released droplets at a rate of 1000 Hz while moving in

the x-direction at a speed of 60 mm/s creating a droplet spacing of 60 µm. The plate was baked

after deposition at 180 ◦C for 30 minutes to evaporate the liquid from the slurry. When multiple

passes were necessary to achieve the desired concentration, the plate was baked between passes as

well.

Inkjet Calculations

The concentration of dopant in the powder bed was controlled by varying the line spacing

and number of lines for each pocket (see Table 3.3). These parameters were chosen based on an

estimate of how much dopant would be integrated into the melt pool during laser processing. Some

of the values are purposefully higher than might be typically needed in order to test the upper limits

of the deposition method.

25

Pressurized Chamber

Piezo-electric Nozzle

Linear

Motion Stages

Figure 3.5: Single-nozzle inkjet deposition system. The three axis stage controls the depositionlocation while the piezo-electric nozzle controls the droplet size and ejection frequency into thepowder bed or onto the substrate.

Table 3.3: Deposition parameters for inkjet deposition of zirconia slurry.

Pocket # # of passes # of lines/pass Line spacing (mm)1 3 1060 0.02262 3 774 0.03103 1 1506 0.01594 1 1506 0.01595 1 886 0.02716 1 290 0.0830

Droplet volume was measured by jetting the zirconia slurry in a stationary position for 10

minutes into a 5 mL beaker of known mass. The beaker was then baked at 180 ◦C for 30 minutes

– leaving only the zirconia – and weighed again. Subtracting the original mass of the beaker from

the mass of the beaker and zirconia gives the mass of zirconia deposited over the time interval

(mz). This can then be used to estimate the amount of slurry deposited (msl) by msl = mz/c where

c is the zirconia concentration in the slurry. The number of droplets can be calculated as nd = t ∗ f

where t is the deposition time and f is the droplet frequency. The slurry mass and number of

droplets can then be used to calculate the mass and volume of an average droplet (md and Vd

26

respectively) by md = msl/nd and Vd = md/ρsl where ρsl is the density of the slurry provided by

the manufacturer (1.22 g/cm3). The average amount of zirconia per droplet (mzd ) can also be

calculated as mzd = mz/nd using the same values for mz and nd as above. Using these calculations,

the droplets were measured to have a volume of about 0.145 nL/droplet and to contain about 35.4

ng of zirconia in each droplet.

The droplet volume described above was used to calculate estimates of the amount of slur-

ry/dopant deposited into the powder beds (see Table 3.3). Total volume was obtained by calculating

the number of droplets in each line and multiplying by the number of lines for the deposition. The

total saturation is the measure of the amount of space between powder particles that is filled by the

slurry. The powder packing fraction was assumed to be 50%, which is typical for non-compacted

powder beds [81]. An area density is also calculated by dividing the total mass of zirconia de-

posited over the area that it was deposited. This total mass of zirconia is obtained in similar

fashion to the total volume by using the total number of droplets in a given deposition.

Table 3.4: Predicted volume, saturation, and density calculations for the inkjet parametersmentioned in Table 3.3. Pocket numbers are defined in Figure 3.4

Pocket # Total volume Powder bed Area density of zirconiadeposited (mL) saturation (%) deposited (µg/mm2)

1 0.1794 1281.8 78.192 0.1309 935.4 57.063 0.0850 607.2 37.044 0.0850 607.2 37.045 0.0499 356.8 21.766 0.0163 116.5 7.11

3.3 Results and Discussion

The resulting deposits from the two deposition methods can be seen in Figure 3.6 and

Figure 3.7. Since there are no quantitative metrics readily available to evaluate these deposition

methods, qualitative visual observations of these resulting deposits are the basis for many of the

claims as to the value of the two methods. Other observations were also made by the authors during

the deposition processes that are included in the ensuing discussion.

27

1 2

43

12

34

Figure 3.6: Dried alumina (left) and zirconia (right) deposits performed using direct write deposi-tion.

1 2 3

4 5 6

Figure 3.7: Resulting deposits from inkjet deposition method into small, enclosed powder bedusing various dopant densities.

Three criteria were selected to evaluate and compare the two deposition methods. These

criteria are deposition quality, integration feasibility, and system productivity. Deposition quality

refers to the uniformity of the deposited material as well as the accuracy of quantity and location of

the deposited material. Irregularities, non-uniformities, and inaccuracies would diminish the value

28

that a deposition method could provide. Integration feasibility refers to how well the deposition

method could be integrated into the LPBF process. This includes reliability and repeatability of

the deposition method. It also involves any ways in which the deposition method would disrupt,

interfere with, or otherwise negatively impact the LPBF process as a whole. System productivity

refers to the overall value that a combined deposition/LPBF system would provide. This includes

build time increases, feasible concentration ranges, and other things that would need to be consid-

ered when implementing the full deposition/LPBF system. The direct write and inkjet deposition

methods are both evaluated according to these three criteria by observing various characteristics of

the deposition processes in action as well as qualities of the resulting dopant deposits.

3.3.1 Deposition Quality

Direct Write

The quality of the direct write process was mostly impacted by inconsistencies observed

during deposition and in the resulting dopant structures after drying.

The first non-uniformity was observed in moments when the needle tip was too far from the

deposition surface. In these instances, the slurry would “skip” small sections of the deposition area

because the slurry would build up around the edge of the needle instead of wetting and depositing

on the substrate. The skipping would eventually end when the size of the slurry bead on the needle

was large enough to make contact with slurry that had already been deposited or the substrate

itself. Areas where skipping occurred can easily be seen in recently-deposited areas as small holes

or lines in the deposition areas (see Figure 3.8).

Skipping was also observed in single-line depositions (see Figure 3.9). The effect was

slightly different, however, as there were generally no nearby depositions that could be used to

reestablish contact between the slurry in the needle and the substrate surface. When this happened,

much more slurry built up on the needle tip than during the area depositions. This resulted in

larger sections of the deposition area without dopant and large bead-like areas shortly after the

blank sections.

29

1 1 2

3 4

Figure 3.8: Photos of liquid zirconia slurry in different concentrations shortly after deposition bydirect write. Some instances of “skipping” are indicated by red arrows. A large area in deposition3 did not deposit due to silicone contamination on the surface that altered the plate wetting.

beads

skip

skip

Figure 3.9: Direct write deposition of zirconia in single-track lines. Some instances of “skipping”are indicated by brackets. Beads resulting from the skipping are indicated with arrows.

Another type of non-uniformity was observed when the dopant material moved toward

the center of the deposition areas during the drying/baking stage. As the edges of the deposition

areas began to dry, the solid particles were transported inward ahead of the solid/liquid interface

[82]. The literature attributes this type of capillary push to the decreasing height of the liquid

wedge near the contact line [83, 84]. The result was a higher-concentration area in the center of

the deposition area with low dopant concentrations near the edges (see Figure 3.10). While this

trend was consistent for all of the direct write depositions, the magnitude of the resulting dopant

concentration disparity varied with the input slurry concentration. When the slurries were diluted

to lower concentrations, less of a gradient was observed throughout the deposition area.

30

1 12 12

34

2 1

3

Figure 3.10: Recently-deposited liquid alumina slurry diluted to different concentrations. Eachimage shows the deposition of a unique concentration to a different quadrant on the substrate. Leftto right: 3 wt%, 2 wt%, 1 wt%, and 0.5 wt% slurry. Dopant agglomeration near the centers of thedeposition areas is observed as more white (alumina) than the exterior regions of the depositions.

One last concern about the quality of the direct write method is the structure and stability

of the resulting dopant deposit – especially in the highly-concentrated regions. In the alumina

deposits, the highly-concentrated centers were quite stable despite some small cracks. The zirconia

deposits, however, formed extremely fragile, highly discontinuous structures in their centers that

delaminated from the substrate in most cases (see Figure 3.6). These structures would be likely to

break and move under almost any application of force. Since this structure fragility was only seen

in the zirconia direct write depositions, this can be categorized as a reflection on the properties of

the slurry/dopant and the drying conditions.

Overall, the deposition quality of the direct write method is not promising as it is non-

uniform and can produce structurally unsound deposits in the case of zirconia dopant.

Inkjet

While the inkjet deposition method has the potential to have great uniformity, there are also

many opportunities for error. In general, the samples with dopant deposited using this method are

fairly uniform (see pockets 3, 4, and 5 in Figure 3.11) as long as simple errors can be avoided. One

of these errors is clearly seen near the bottom of pocket 4 and just above the center of pocket 3.

In these cases, the nozzle was clogged temporarily during the deposition resulting in a few missed

lines where dopant should have been deposited. Another error seen in pocket 5 is contamination

from an outside source as flakes of residue material fell from the nozzle structure into the pocket

during deposition. These errors all affect the uniformity of the resulting deposition; however, they

could be resolved with a more controlled inkjet system commonly used in industrial applications.

31

1 2 3

4 5 6

skipped lines

increased roughness

contaminates

high non-uniformity

Figure 3.11: Zirconia-doped powder beds at different concentrations with pocket 1 being the high-est concentration and pocket 6 being the lowest (see Table 3.4 for detailed concentration esti-mates). Contaminates and skipped lines are the result of user-error during deposition while in-creased roughness and non-uniformity of dopant concentration are important results that will im-pact the LPBF process.

While the mid-range depositions appear uniform, the higher-density depositions (see pock-

ets 1 and 2 in Figure 3.11) show more variance in the uniformity of the dopant concentration

throughout the pocket. This is due to the increased movement of the liquid and powder during the

deposition stage. The greater amount of liquid in the pocket created a pool of liquid, in which the

316L powder particles could flow.

The uniformity of the inkjet method can be somewhat limited in the higher range of sat-

urations; however, increasing the dopant concentration in the slurry or depositing the material in

multiple passes with drying between passes could expand the feasible range of doping deposits.

3.3.2 Integration Feasibility

Direct Write

In practice, there are a few things that would need to be considered before implementing

the direct write method in an LPBF system.

32

The first of these is the needle tip distance from the surface. During the direct write op-

eration, it was observed that the skipping mentioned above could be avoided by keeping the tip

of the needle within about 0.15 mm of the substrate surface during deposition. This requires an

extremely flat substrate which is not always the case with built layers in the LPBF process. During

the depositions, it was noted that little to no skipping occurred when the substrates were level to

within about 0.05 mm.

The incorporation of a direct write system into an LPBF machine would also require sig-

nificant modifications. This would entail an additional motion axis for the nozzle to move in the

direction perpendicular to the coater blade or two motion axes if the system moves independent

from the coater system. A vertical motion axis may also be warranted to ensure that the nozzle

tip does not come in contact with the metal powder at any point in the process and that it can be

lowered to the correct height as mentioned previously.

Another important consideration is whether the dopant will contaminate the powder that

will not be fused as part of the final build. The delaminated flakes of dried zirconia slurry would be

broken off during powder spreading and could travel to almost any part of the build chamber. This

would be problematic as the composition could be changed in the wrong locations within the part,

unfused powder could be contaminated, and powder flowability could decrease due to the different

shape and size of the zirconia inclusions. However, this problem is specific to the properties of the

dopant. The plate with alumina depositions was inserted into the LPBF machine and powder was

spread over the top. After removing the plate and clearing the powder, it was determined that the

powder spreading process did not cause any visible physical damage to the deposited alumina.

Powder contamination could also be a concern when the dopant is still in liquid form. If

the liquid in the slurry does not evaporate before the next layer of powder is spread, the powder

could be contaminated as mentioned above. The moisture could also severely limit the flowability

of the powder or even cause build-ups on the coater blade. Both of these cases could initiate the

formation of defects that would propagate through multiple layers if not the entire build [55,85,86].

Inkjet

There are a few things that were noted in the inkjet process that may affect the way it is

implemented in LPBF applications.

33

One of these observations was the dramatic change in color in the higher-density deposi-

tions. While all of the depositions did show some change in color, the change in pockets 1 and

2 were the most pronounced. This may be due to the dopant being present on the top surface of

the powder bed at higher concentrations compared to the other depositions. This is of particular

importance for LPBF because the added material being on top of the material could change the ef-

fective absorptivity of the powder bed. This change would require a shift in processing parameters

to achieve the same part quality.

Another feature that could impact the LPBF process is the powder bed roughness observed

in pocket 6 (see Figure 3.11). In binder jetting, this increase in roughness has been described as a

first layer phenomenon that occurs when the droplets are too widely spaced [87]. This can cause

poor lamination throughout the first few layers in binder jetting. Similarly, in LPBF, increased

roughness in one layer can cause balling and other defects in subsequent single-track scans or fused

layers and can propagate throughout the part [55, 85, 86]. This could prove prohibitive to low-

saturation depositions using the inkjet method; however, slurries or solutions with lower dopant

particle densities could be used to achieve smaller composition changes while maintaining the

same saturation levels.

While these effects on the LPBF process may require some adjustment, they are not pro-

hibitive as other processes (such as binder jetting) have successfully addressed them through print-

ing parameter adjustments [79, 80, 87].

Inkjet printheads can also be fitted across a wide span and could easily cover the entire

width of the coater blade. This would make it so that no additional motion axis is required for

incorporation into the LPBF system.

3.3.3 System Productivity

Direct Write

While using liquid dopants to alter composition may be quite attractive in an LPBF setting,

the benefits of the method being used must outweigh the associated costs to make a reasonable

claim at increasing value.

34

The main cost that this method incurs is the extra time that it will take. Since the deposition

step cannot occur simultaneous to any of the other LPBF process steps, every layer where dopant

is needed would take considerably longer to complete. This time addition could be in the range

of a few seconds to a few minutes for each layer depending on the traverse speeds, the amount

of dopant to be deposited, and the area over which it was to be deposited. The dopant then also

needs to dry before continuing – to avoid introducing moisture to the delicate powder process as

mentioned above. This wait could add more seconds or minutes onto each layer time depending

on the size and shape of the deposits.

The benefits are also limited by the non-uniformity of the resulting dopant deposits. In

order to maintain a somewhat uniform deposition, the direct write method as performed in this

study would be limited to less than 0.6 µg/mm2 of alumina (or 0.3 wt% of the deposited material

when measured with a 50 µm layer of powder) for any area depositions. While this may be

useful to alter some properties, it is on the low side of small composition change. Studies have

shown that useful small composition changes are generally between 1 and 5 wt% for dispersion

strengthening [9,21]. However, the amount of dopant required to affect other changes in a material

may be significantly higher.

In this case, the added time costs and small concentration range are extremely limiting

when considering implementation of a direct write deposition system into a LPBF process.

Inkjet

The inkjet system has some clear advantages over the direct write system. The added

time cost for this method will be close to nothing. Since inkjet printhead arrays can be scaled to

perform at very fast speeds and the printhead can operate simultaneous to the powder spreading,

the deposition step will add very little if any time to the overall process. Considerations may need

to be made for drying the liquid dopant; however, a heated build chamber could reduce that time

cost significantly as well.

Though there are many benefits to using an inkjet deposition system with LPBF, there are

some limitations as well. One specific limitation found during this study is that the alumina slurry

did not work in the inkjet setup. With the small-aperture piezo-electric nozzles that were used with

the inkjet system, a 5.00 µm filter was used to keep larger particles from damaging the nozzles.

35

This filter made it impossible for the alumina slurry to pass through the system, meanwhile the

zirconia slurry had no problems. This illustrates the need for careful preparation of the dopant to

be used with an inkjet deposition system.

Generally, the inkjet deposition system has the potential to add significant value to the

LPBF process. The problems and limitations can be addressed if the property enhancements are

sufficient.

3.4 Conclusions

Overall, inkjet printing is a better option for liquid-dopant deposition in LPBF because of

its superior uniformity, high resolution, minimal time addition to the process, and its history of

use with powder beds (e.g. binder jetting). The direct write method could be beneficial in spe-

cific circumstances (e.g. single tracks of dopant), but would be much more difficult to implement

across the rough surfaces often encountered in LPBF. The only clear advantage that the direct

write method could claim over inkjet deposition is that the dopant is underneath the powder and

would have a minimal effect on laser absorption. The inkjet method can deposit liquid dopants in

concentrations between about 10 and 40 µg/mm2 while the direct write method is limited to less

than 0.6 µg/mm2 for large areas. Choice of material that is suitable for the different methods is

of paramount importance as zirconia proved to not be useful in the direct write method while the

alumina slurry could not be used in inkjet printing. This limitation may be overcome by using inks

that are formatted specifically for inkjet printing.

36

CHAPTER 4. INFLUENCE OF LIQUID-ENCASED ZIRCONIA DOPANT ON MELTPOOL CHARACTERISTICS IN LASER POWDER BED FUSION

4.1 Introduction

Laser powder bed fusion (LPBF) is one of the premier additive manufacturing (AM) pro-

cesses used in industry today [32]. Known for its impressive properties and complex geometric

capabilities [3, 4, 88], LPBF has been the subject of much interest in the biomedical, automotive,

and aerospace fields [1–3, 89]. While LPBF often produces accurate parts with good properties,

there is still an incomplete understanding of common defects, such as balling, and their influence

on final part properties [44, 46, 90]. Balling is of particular importance because it is a source of

other common defects and process problems including increased surface roughness, pore forma-

tion, delamination due to poor layer adhesion, non-uniform powder layer deposition, and coater

blade damage [48, 55, 85, 86]. As industries look to use LPBF for higher-consequence parts, it

will be increasingly important to fully understand how and when balling occurs as well as how to

mitigate it and its effects.

Most research on balling in LPBF has been focused on finding acceptable parameter regimes

for specific materials in which balling does not occur. This is generally done with single track

laser scans, whose melt pool geometries are indicative of resulting part microstructure and proper-

ties [91, 92]. The melt pool geometries are changed by altering parameters that affect the spatial

energy input to the powder bed. While most studies have typically only varied laser power and scan

speed, other parameters – such as laser spot size, scan strategy, hatch spacing, and layer thickness

– also have a significant impact on spatial energy input [39, 51, 59, 93, 94].

The characteristics of the resulting melt pools are used to categorize each parameter set.

Categories that are often used in the literature include lack-of-fusion, balling, conduction, and

keyholing [53, 54]. Lack-of-fusion – also referred to as “under-melt” or “insufficient melting” –

occurs when the melt pool does not wet/adhere well to the previous layer or substrate (usually due

37

to insufficient energy input) leaving very little of a melt pool attached to the surface [53,95]. LPBF

operation in the lack-of-fusion regime can cause interlayer porosity and delamination which can

both have an immense negative effect on final properties. Balling is observed when the melt pool

divides into periodic bumps or spheres due to instabilities present while the melt pool is in liquid

form [55–57, 85]. The conduction regime represents an ideal melt pool that is stable (straight,

consistent size) and nearly cylindrical with good adhesion to the substrate [56, 85, 95]. Keyholing

occurs when the pressure on the melt pool (caused by evaporation) forces the liquid to burrow

into the substrate or previous layer creating a large and oddly-shaped melt pool [96]. When the

evaporation pressures against the liquid surface are high enough, instabilities allow the depression

in the melt pool to collapse. Keyholes often contain porosity created when gasses are trapped

during these collapses and pushed to the bottom of the melt pool [97]. These regimes are often

illustrated on process maps where they can be related to parameter changes. The most common

process map uses laser power and scan speed as the variables to show the effect they can have

on the resulting melt pool characteristics (see Figure 4.1). Energy density (ED) is a metric that

combines basic parameters in a way that describes the overall energy input by the laser [65]. Using

ED then allows for comparison against a greater number of variables (e.g. laser spot size, powder

layer height, powder particle size, etc.) rather than being limited to power and scan speed.

Figure 4.1: Process map by Francis [63] using laser power and scan speed to illustrate the effectsthat parameter changes have on melt pool characteristics.

38

One of the main opportunities for LPBF to develop is through expansion into the multi-

material realm [98]. As new material combinations are used to create LPBF parts, melt pool

characterization will be extremely important as a tool to predict appropriate processing param-

eters. It has been shown that even a 0.5 wt% addition of ceramic nanoparticles to AlSi10Mg

powder requires a change in process parameters to obtain samples of the same quality as the plain

AlSi10Mg [99]. Parameter prediction will only be possible, however, if there is a coherent under-

standing of how different composition changes alter the characteristics of the melt pools. With this

understanding, it will be possible to interpolate between known data to estimate the appropriate

parameters for any material composition that will be used. This can be done because different

materials exhibit similar trends in melt pool characteristics [86]. For example: low power and high

scan speed with a large spot size will generally not produce a coherent melt pool, while high power

and low scan speed with a small spot size will generally produce a keyholing melt pool. To be able

to interpolate across different materials in this way, much more will need to be known about the

effects of specific additives across multiple materials.

Some preliminary work has been done with a few material/additive combinations. Increas-

ing oxygen content in the build chamber has been found to increase the likelihood of balling for a

given set of process parameters and adding deoxidants to the material has been found to decrease

the likelihood of balling [85, 86]. Other additives have also been shown to have an effect on the

resulting fused material. TiB2 in AlSi10Mg, for example, expanded the range of parameters that

produce optimal melt pools [95]. However, the effects of and on a given material depend heavily

on the material’s properties such as thermal conductivity [56]. For example, Nb and Cr were shown

to have vastly different effects when added to a base titanium powder because of their positive and

negative enthalpies of mixing respectively [98].

Using LPBF for multi-material applications is currently of interest in the academic com-

munity because it promises more efficient components and better use of resources [5, 6, 98]. Parts

will become more efficient as the properties are spatially-tailored to match the specific functions

being performed. Meanwhile, fewer unique powders will need to be stored if the compositions

can be modified during the LPBF process. As many researchers are looking to expand LPBF into

the multi-material regime, it will be crucial to understand how defects like balling will translate to

the novel material systems that are being used. A first step to approaching this understanding is to

39

map the process space with a base material as well as the base material with different amounts of

dopant material. These process maps will provide insight as to the changes that the dopant mate-

rial creates and can be used as a baseline for models that can be adapted to other material systems

in the future. The present work compares process maps for SS 316L and SS 316L with zirconia

added via inkjet printhead deposition in various densities. Parameters varied to create the process

maps include laser power, scan speed, and laser spot size.

4.2 Methods

4.2.1 Materials and Equipment

The current study was performed using a Concept Laser M2 Cusing Multilaser LPBF sys-

tem with dual 400 W lasers. The base material is a SS 316L powder (CL20 ES, Concept Laser,

Spherical, 29.9 µm) and the added dopant material is 100 nm colloidal Zirconia slurry (NYACOL

ZR100/20) measured to be 25 wt% zirconia particles by measuring the mass of slurry in a beaker

before and after evaporating the liquid at 180 ◦C for 12 hours. This zirconia is shown in 1 to work

well with the inkjet deposition process described below. Zirconia ceramic powder has also been

shown to perform well in LPBF processing on its own and to adhere well to a steel substrate during

LPBF processing [14].

4.2.2 Inkjet Deposition and Single-track Laser Scans

An adapter plate and small build plates were built to adapt the large build chamber of the

M2 to enable fabrication on smaller removable plates. The small plates were machined out of

1018 mild steel. On this small plate, platforms were built from SS 316L to a height of 0.5 mm to

create foundations on top of which the deposition of dopant into powder would occur (see Figure

4.2-left). The 0.5 mm height was shown to be sufficient to mitigate the effects of the different

material (1018 Base, 316 SS powder) for the majority of melt pools (see Appendix A). After the

tenth layer of the foundations, another 50 µm layer of powder was spread and only the edges of

the foundation areas were scanned to form walls that would separate the powder on top of each

40

Figure 4.2: Left: Platforms built to separate small powder beds from small build plate. Right: Onelayer of powder spread on top of platforms with walls built around the edges to form pockets tocontain inkjet depositions.

foundation as its own small, isolated powder bed (see Figure 4.2-right). Once the small powder

beds were separated, the small plate was removed (leaving the adapter at the finished position).

The zirconia slurry was deposited into the small powder beds using a 3 axis stage to scan

an 80 µm-nozzle, piezo-electric, drop-on-demand print head (MicroFab Technologies, Inc.; Part #

MJ-AB-01-80-8MX) over the surface. The printhead is connected to a pressure controller. Two

different concentrations (see Table 4.1) were deposited into two of the larger isolated powder beds

while the third served as a baseline with no zirconia added. The lower dopant input (4.8 wt%) was

chosen for its similarity to composition changes (1 – 5 wt%) that have shown useful results with

similar material systems [9,24,100]. The larger dopant input (17.2 wt%) was selected to accentuate

Table 4.1: Inkjet deposition line spacing and resulting dopant quantity metrics for two of the threelarger powder bed pockets (see Figure 4.3). Zirconia in powder bed is a metric to describe the

input dopant amount in a simplified way. This is comparable to the way powders aredescribed when multiple compositions are mixed together [9].

Position Y-spacing Volume of slurry Mass of zirconia/ Zirconia in powder(µm) deposited/layer (µL) area (µg/mm2) bed (wt%)

Left 74.4 15.48 10.096 4.8Bottom 18.0 63.84 41.645 17.2

41

processing differences compared to standard SS 316L. The line spacing necessary to achieve these

concentrations are shown in Table 4.1 along with the amount of zirconia dopant present calculated

as a mass fraction of total deposited material (dopant and 316L powder). Because this is a single

layer test, the actual melt pool zirconia content will be lowered because the melt pool extends into

significantly below the unfused powder on the surface. After deposition of the zirconia slurry into

the powder beds, the plate was baked at 50 ◦C for four hours, followed by five hours at 150 ◦C

before being left to cool in the oven to room temperature. The purpose of the baking step was to

evaporate the liquid introduced in the slurry, leaving just the zirconia dopant in the steel powder

(see Figure 4.3-left); the baking temperature was increased incrementally to limit the size of the

temperature gradients in the plate which can cause transport of the liquid through the powder.

After baking, the small plate with zirconia-doped powder beds was returned to the adapter

plate in the M2 in its original orientation and at the same height at which it was removed. The

Figure 4.3: Left: Powder beds with deposited dopant after plate has been baked (liquid fromdopant slurry all evaporated). Note that the bottom large powder bed was filled with slurry, butthe slurry stopped jetting between 1/2 and 2/3 of the way through the powder bed. Dopant levelsfor each pocket are indicated. Right: Single-track laser scans in zirconia-doped and non-dopedpowder beds. The large pockets on the exterior each contain single-track scans for 54 differentparameter sets which are the basis for the data in this work. Individual single-track scans can beseen in Figure 4. The 15 smaller pockets were used for a similar study investigating the amount ofzirconia dopant that is incorporated into the melt pools.

42

machine was then used to create 108 single-track laser scans for each dopant level (see Figure

4.3-right) using 54 different parameter sets (see Appendix B) with 2 independent scans for each

parameter set. These parameter sets were chosen to obtain a distribution of areal energy densities

by varying the laser power (190 – 395 W ), scan speed (150 – 1650 mm/s), and laser spot size (50

– 190 µm). The areal energy density is defined as EDA = P/(v∗ dS) where P is power in W , v is

scan speed in mm/s, dS is laser spot size in mm, and EDA is measured in J/mm2. The variations in

power, speed, and spot size mentioned resulted in an ED range from 1.08 to 26.33 J/mm2.

4.2.3 Sample Preparation and Measurement

After laser fusion, the small plate was removed from the LPBF machine and 3D images

were collected of each set of line scans at 100x magnification using the 3D stitching feature on the

Keyence VHX-7000. From these 3D images, profile morphology metrics were obtained for each

line scan in the VHX software. These metrics have been used in the literature to analyze the effects

of different processing parameters on the resulting part geometry in LPBF [46].

Each single-track scan was also classified based on observed physical features by three

independent observers. These classification groups were: 1 – Lack-of-fusion, 3 – Balling, and 5 –

Continuous; with 2 and 4 being transition classifications between the main categories (see Figure

4.4). Similar classifications are often used in the literature as a method of creating process maps for

different materials and is a well-known method of evaluating the parameter space in LPBF [101].

The median value of the three independent classifications was taken for each melt track, and the

values for the two melt tracks made with identical parameters were averaged to obtain an overall

classification value for each independent set of parameters at each deposition level.

After all non-destructive measurements were taken, the samples were sectioned using a

waterjet cutter, polished using a standard polishing recipe for SS 316L, and electrolytically etched

in 10% oxalic acid for 45 seconds with a 6 V voltage differential to reveal the grain structure and

melt pool boundaries. The melt pool geometries for the parameter sets with higher energy density

(sets 37-54) were imaged using an Olympus GX51 optical microscope and measured using the

accompanying Pax-It image analysis software (see Figure 4.5). Melt pool width (wmp) is defined

as the distance between the two furthest points in the melt pool along the profile of the substrate

surface. Melt pool depth (dmp) is defined as the orthogonal distance from the interpolated substrate

43

Figure 4.4: Examples of each possible melt track classification level. Class 1 is for lack-of-fusionwhere there is little to no adherence of the melt track to the substrate. Class 3 is for melt tracksthat exhibit the balling phenomenon. Class 5 is for continuous melt tracks. Classes 2 and 4 are formelt tracks that fall in between the main classification groups (e.g. class 4 has some balling, butwith significant elongated segments indicative of a transition towards continuous melt tracks).

Figure 4.5: Image of etched melt pool cross-section with characteristic melt pool measurements.L1 is the width of the melt pool at the substrate plane, L2 is the height of the melt pool from thesubstrate plane, and L3 is the depth of the melt pool from the substrate plane.

surface to the lowest point of the melt pool. Melt pool height (hmp) is defined as the orthogonal

distance from the interpolated substrate surface to the highest point of the melt pool.

44

4.3 Results and Discussion

4.3.1 Visual Classifications

The classification data for each parameter set was used to create process maps for the

parameter-dopant design space. These maps show how adding dopant and/or changing the param-

eters can alter the overall results of the LPBF process. This understanding can then inform the

selection of parameter sets for work within the range of the data.

The scan speed process map (Figure 4.6-top) illustrates that the addition of zirconia in-

creases the range of scan speeds for which balling is likely to occur. This range also increases

when varying only laser power or laser spot size as well (see Figure 4.6-middle and -bottom re-

spectively). While varying the laser power or spot size alone does not cover the entire range of

the melt pool types, it is apparent that the balling regime is significantly larger when zirconia is

added to the 316L powder. The expansion of the balling regime could allow for increased speed

and improved properties during the LPBF process.

Since the balling regime expanded in both directions, it is not certain what the ultimate

cause of the change is or if there are multiple factors. One possible reason for expansion of the

balling regime toward higher energy densities could be that zirconia particles decrease the effective

surface tension of the melt pool [102]. This would be supported by the finding in Chapter 5 where

the majority of the zirconia incorporated into the melt pool was found near the exterior edges –

edges of the freestanding portion of the melt pool that are not in contact with the substrate (see

Figure 4.7). Since a system naturally moves toward lower free energy (including surface energy)

the location of the zirconia near the surfaces indicates that the surface tension is lower when the

zirconia is near the surface compared to other locations within the melt pool.

Conversely, the balling regime could have expanded into the lower range of energy den-

sities because of increased absorptivity by the zirconia-doped steel powder. This would instigate

more powder melting and adhesion to the substrate at lower energy inputs compared to the non-

doped steel powder. Since the parameter sets included in this study were selected based on known

reactions of SS 316L, the transition to the lack-of-fusion regime in the zirconia-doped melt pools

was not observed. Parameter sets based on higher scan speeds, lower power, and increased spot

size should be implemented to observe the full range of melt pool types. If melt pools can be

45

Figure 4.6: Process maps showing the dopant input/individual parameter design space and re-sulting melt track characteristics. Top: scan speed/dopant quantity design space. Middle: laserpower/dopant quantity design space. Bottom: laser spot size/dopant quantity design space. Thedotted line on each map indicates a parameter set that is generally used for most of the laser scansin the bulk of a part. The ranges over which the given parameter set/dopant quantity combinationproduce the balling phenomenon are emphasized. Approximate divisions between the 5 main clas-sification levels are marked by vertical lines of the corresponding color.

46

Figure 4.7: Melt pool and corresponding zirconium content map from Chapter 5. As mentioned,the majority of the zirconium is shown to inhabit the exterior edges of the melt pool. These are theedges of the melt pool that are not in contact with the surrounding substrate.

produced with the same qualities but at higher speeds, overall productivity of the entire LPBF pro-

cess could be improved. It will also be important to see if this improved quality obtained at lower

speeds can be consistent when applied in area scans and multilayer builds instead of single-track

laser scans.

When evaluating all of the measured data as a function of the energy density (see Figure

4.8) the same expanded balling range is observed. However, it shows a more complete picture

of the design space. One important contribution to be gained from this perspective is that the

energy density boundary at which the melt tracks begin to become continuous remains relatively

unchanged between the no-dopant data and the high-dopant quantity data ( 2% difference in energy

density at the regime boundary) while the minimum energy density required to create continuous

melt pools for the low-dopant quantity data is increased by about 30%.

This trend is also seen for the corresponding boundary of the balling regime – that the low-

dopant condition shows balling at higher energy densities than either the no-dopant or high-dopant

quantity data. While these boundary shifts occur at higher energy densities than are used for laser

scans in the bulk of the part, they are very close to the “Lattice/Single-track Parameter Set” noted

47

Figure 4.8: Process map showing the dopant input/energy density design space with resulting melttrack characteristics. The dotted lines indicate commonly-used parameter sets for areas in thebulk of a part and for lattice or single-track scans. The ranges over which the given parameterset/dopant quantity combination produce the balling phenomenon are emphasized. Approximatedivisions between the 5 main classification levels are marked by vertical lines of the correspondingcolor.

in Figure 4.8. This parameter set is used for vector-based parts which are generally composed of

non-contiguous single-track laser scans arranged in a pattern (as opposed to volume-based parts

that create solid areas using adjacent laser scans). Since this parameter set is generally used in

single tracks, it is of paramount importance that it creates a solid melt track and completely wets

the underlying fused material without the additional energy provided by a neighboring melt track.

Since it is necessary for this setting to produce fully continuous melt pools, it may be necessary to

adjust the parameters for some levels of doping to make sure that a reliably continuous melt pool

is established. It will be important to understand how the magnitude and direction of this shift

changes in relation to the amount of dopant added to the powder bed.

The seemingly contradictory nature of the mentioned regime shift could be explained as a

competition between the mechanisms that were explained previously. Decrease in surface tension

could have more of an effect when the dopant level is low, while the increased absorptivity might

dominate with higher dopant levels. These phenomena require further study including more dopant

levels and a broader range of parameter sets. To be sure that all melt pool regimes are represented,

48

a similar experiment to the current study is recommended using areal energy densities as low as 0.5

J/mm2, but it may not be necessary to use as many parameter sets that have high energy densities (¿

12 J/mm2). Very small dopant levels should be used to understand how much zirconia is required

before it begins to have an effect on the melt pools. It would also be useful to test dopant levels

at smaller intervals to better understand the trends that occur. Intervals as small as 0.5 or 1 wt%

zirconia in the powder beds may be appropriate.

A parameter set often used for the bulk interiors of LPBF parts is marked on each of the

plots in Figure 4.6 and Figure 4.8. This bulk parameter set is of interest because it is one of the

main factors that controls the quality (density, strength, etc.) of the interior regions in an LPBF

part. While this parameter set is known by the authors to create high-quality parts when used

in multi-layer area scans, single-track scans made using the bulk parameter set exhibit standard

balling behavior. This difference may be because, in area scans, adjacent melt tracks are able to

wet previous melt tracks in addition to the substrate allowing for less instabilities in the individual

melt tracks. This would result in more uniform and continuous area scans compared to single-track

scans performed with the same parameter set. It would be valuable to directly test this hypothesis

to quantify the shift in melt pool characteristics that takes place when using area scans instead of

single-track scans.

For this study, the implications of the relationship between area scans and single-track

line scans are significant. A study of area scans could determine if the expanded balling-regime

observed in single-track studies indicates an increase or decrease in the processing window for

area-scanning. An expansion of the processing window would allow for higher scan speeds or

lower laser powers to be used with the zirconia dopant to obtain fused areas of similar quality

while decreasing the build time and/or energy expended.

4.3.2 Profile Morphology

The arithmetic mean roughness (Ra for a profile, Sa for an surface) is the most commonly

used of the morphology metrics for characterizing height fluctuations [55]. Since the morphology

data was collected before the powder was removed, they are more accurate for the continuous melt

tracks than those that may include sections of powder bed in the measurement. The Ra values of

the continuous melt pools showed a general trend toward decreasing roughness as energy density

49

increased (see Figure 4.9). The samples with no dopant exhibited lower overall roughness than

those with added dopant but the roughest surfaces were those with the lower zirconia content (4.8

wt%) rather than the high zirconia content.

This increased roughness could be due in part to a change in the powder bed morphology

when the dopant is added. In this case, the surface roughness, (Sa), after printing the zirconia but

before laser scanning increased by an average of 9.8 µm in the 4.8 wt% powder bed and decreased

by an average of 1.3 µm in the 17.2 wt% powder bed. This is commonly seen in binder jetting

where inkjet deposition significantly disturbs the powder bed on the first layer, but can be avoided

by adjusting inkjet printing parameters such as droplet size, spacing, and velocity [87]. Since

the melt tracks in the 17.8 wt% powder bed still exhibited an increased level of profile roughness

compared to those in the non-doped powder bed, increased powder bed roughness is likely not

the only factor behind this change. Another possible cause of this increased roughness could be

irregularities caused by the non-uniformity of the zirconia deposit. If significant material transport

occurs before the liquid carrier for the dopant is evaporated, there may be areas with more or less

Figure 4.9: Arithmetic mean profile roughness (Ra) measurements plotted as a function of arealenergy density (log scale) for the three deposition quantities (no dopant added, 4.8 wt% zirconiain the powder bed, and 17.2 wt% zirconia in the powder bed). Dotted lines represent best-fitexponential curves for the data set of the corresponding color. While the three data sets all followa similar general trend at higher energy densities, there is much more spread and inconsistency atlower energy densities.

50

zirconia content 3. If the zirconia content affects the amount of energy absorbed or the way the

melt pool forms in any significant way and that zirconia content varies throughout the powder

bed, then it follows that there would be more variation in a laser track performed in this highly

non-uniform powder bed than there would be in a laser track performed in a non-doped powder

bed.

Greater understanding of the difference between these mechanisms and which one has a

greater effect on the final melt track roughness may be achieved by using liquids with varying

dopant concentrations. By this method, the amount of liquid deposited – and therefore change

in powder bed roughness – can be kept constant while varying the dopant amount or the dopant

amount can be maintained while varying the amount of liquid that is used to perform the deposition.

Further investigation into the effects of specific properties on the resulting melt track roughness is

also warranted. This could be achieved by choosing a base parameter set that consistently forms

continuous melt pools and varying each of the parameters from those base values independently.

4.3.3 Melt Pool Dimensions

The basic melt pool dimensions measured from the cross sections were width at the sub-

strate surface plane, height from the substrate surface plane, and depth from the substrate surface

plane. These three dimensions were measured for melt tracks made with energy density greater

than 5 J/mm2, almost all of which were classified as continuous or nearly continuous. The melt

pool depths and widths were comparable across the range of energy densities explored (see Figure

4.10), but an interesting difference was observed in the melt pool heights between zirconia-doped

and non-doped powder (see Figure 11). Adding zirconia consistently increased the height of the

melt pool above the surface.

51

Figure 4.10: Melt pool depths (left) and widths (right) for various energy densities compared acrossdifferent dopant levels. Depths are measured from the plane of the substrate surface to the deepestpoint in the melt pool. Widths are measured using the two points at which the melt pool boundaryintersects the substrate surface plane. Trendlines are linear across a log-scale domain.

Since the depths and widths of the melt pools do not show separation between the data

sets (see Figure 4.10), it could be concluded that the amount of added dopant material did not

significantly affect the way that the melt pool interacted in the substrate. Rather, the added dopant

had more of an effect on the region of melt pool not in direct contact with the substrate. This effect

is demonstrated by the significant change in average melt pool height when dopant is added (see

Figure 4.11). This could be a result of the effect that zirconia has on the surface tension and laser

energy absorption, mentioned previously. This type of surface tension alteration has been shown

to change the shape of the melt pool [31] and the increased absorption would increase the amount

of material that is fused together. Since the zirconia content and dispersion data from 5 show that

the amount of zirconia near the top edges of the melt pool increases with increasing concentration,

the effects of surface tension on the melt pool shape are expected to amplify with increased dopant

52

Figure 4.11: Melt pool heights for various energy densities compared across different dopant lev-els. Heights are measured from the plane of the substrate surface to the highest point in the meltpool. Trendlines are linear across a log-scale domain.

input (see Figure 4.7). Laser energy absorption is also expected to increase with added zirconia.

Deeper understanding of the correlation between melt pool height and size with quantity of dopant

added could be found by investigating a greater range of dopant quantities as mentioned previously.

The linear trendlines show a significant gap of close to 50 µm between the no-dopant and

dopant-added data sets while the high- and low-dopant data sets are much nearer to each other.

The average height of the no-dopant melt pools is 72.4 µm compared to 116.0 and 132.5 µm for

the 4.8 and 17.2 wt% melt pools respectively. At a 50 µm layer height, the zirconia-added melt

pools, on average, are more than double the height of the powder layer with which they were

made. This increased melt pool height will be problematic when it exceeds twice the powder

layer height (100 µm for a 50 µm layer height). If fused material is more than twice as tall as

the set layer height, it will likely come into contact with the coater blade while the next layer of

powder is being spread (see Figure 4.12). If the coater blade is rubber, this can create tears in

the blade leading to inconsistencies in the spread powder and a compounding effect of deformities

propagating through the part. If the coater blade is made of steel, some damage could be done

to the blade or the part. As the blade crosses a protruding melt pool, the melt pool will either be

deformed, detached, or transfer stress to other areas of the fused part. Any of these scenarios could

have significant negative consequences for the overall part.

53

Figure 4.12: Illustration of possible coater blade damage when melt pool height is greater thantwice the layer height. Left: Powder is spread over a substrate with layer height t. Middle: Twomelt pools are formed. One has height > 2t, the other has height < 2t. Right: Coater blade canspread powder over shorter melt pool without a problem, but is damaged when it makes contactwith taller melt pool.

4.4 Conclusions

The goal of this work was to illustrate the changes that occur to the acceptable processing

windows in LPBF when a dopant is added to the base material. Zirconia in SS 316L will not be

representative of all material systems, but it suggests a possible magnitude of change in any given

material system and demonstrates a need to understand the effects that composition changes have

on the LPBF build process. Some of the changes for this specific example are:

1. Adding zirconia to SS 316L tended to increase the roughness of single-track scans across

a wide range of processing parameters. This may be indicative of non-uniformity of the

deposited dopant and powder bed roughness increased by the inkjet deposition process.

a. Experimentally varying the dopant concentration in the liquid carrier and the amount

of liquid deposited will provide insight as to the strength of these causes.

2. The range of processing parameters with which balling occurs in LPBF is significantly en-

larged when zirconia dopant is added. This expansion may be the result of competing mech-

anisms like decreased surface tension in the melt pools and increased energy absorption.

a. Further investigation using a greater range of dopant content (4-5 values) and a lower

range of energy densities (0.5-12 J/mm2) is necessary to understand how parameters

should be changed to obtain similar results.

54

3. The mechanism creating more balling in the lower energy density parameter sets is more

dominant at higher dopant inputs, while the mechanism creating more balling at higher en-

ergy density parameter sets is more dominant at lower dopant inputs. The parameter sets

used to do lattice supports and single-track line scans may need to be changed to obtain the

same results when dopant is added.

a. Further investigation using a greater range of dopant content (4-5 values) and a lower

range of energy densities (0.5-12 J/mm2) is necessary to understand how parameters

should be changed to obtain similar results.

4. Zirconia-doped melt pools exhibit taller melt pools while their width and depth are relatively

unchanged compared to their non-doped counterparts.

a. This may be problematic for multi-layer builds as the heights of the melt pools with

dopant are more than twice the layer height used.

Overall, the effects of laser fusion on zirconia-doped SS 316L powder are substantial and

must be taken into consideration if this process is to be implemented to build full parts. Under-

standing the interactions for this specific material system will lead to a greater understanding of

other material systems (dopant/base material) and will allow for continued expansion into multi-

material LPBF and specifically into 3D spatial composition control in LPBF.

55

CHAPTER 5. INTEGRATION OF LIQUID-ENCASED ZIRCONIA DOPANT INTO316L STAINLESS STEEL MELT POOLS IN LASER POWDER BED FUSION

5.1 Introduction

Laser powder bed fusion (LPBF) is the most popular and the most developed of the several

metal additive manufacturing (AM) technologies [32]. This process uses a laser to selectively melt

and fuse metal powders, layer-by-layer, to form completed parts. Like other AM processes, the

layer-wise nature of the process allows for increased geometric complexity compared to traditional

manufacturing methods [3, 4, 88]. This geometric complexity has been highly sought after in

recent years in the biomedical, automotive, and aerospace fields because of the design opportunities

that it provides [1–3, 89]. These opportunities include weight reduction, biomimetic design, and

functional integration – where multiple components of an assembly are combined into a single,

integrated part [1–4].

While LPBF offers remarkable opportunities for advancement, the process also comes with

new challenges. One such challenge lies in the single-material nature of the LPBF process. This

challenge is exemplified when considering a functionally integrated part. This integrated part

performs multiple functions, comparable to an assembly of components where each component

performs a single function. In an assembly, these components would each be optimized individu-

ally and assigned a material that best suits their function. However, an integrated part created using

LPBF can only be built using a single material. This would limit the extent to which the integrated

part could be optimized since material choice will greatly influence the geometric constraints. In

this scenario, qualities such as weight reduction, wear resistance, and corrosion resistance may be

sacrificed in some areas to ensure full functionality of the component. To realize its full potential,

an integrated part would require a distribution of properties that can vary spatially to address local

requirements [5, 6].

56

Many recent studies with multi-material LPBF (MM-LPBF) have been focused on metal-

matrix composites which implement small composition changes as a means of creating improve-

ments in mechanical properties [9, 75, 100, 103–107]. One of the most significant examples of a

metal matrix composite is oxide dispersion strengthened (ODS) steels, where the high-temperature

strength is significantly higher than ordinary steels even though the oxide particles make up less

than 1 wt% of the overall material [19, 21, 22, 66]. Other examples of small composition change

inducing property improvement have been seen in laser welding where small amounts of titanium

oxide, selenium, or sulfur have a substantial effect on the melt pool geometry [29–31] and in cast-

ing where inoculants are added to stimulate nucleation during cooling for more equiaxed grain

formation [67]. While there are many other methods of improving properties by changing the

composition, this work will focus on adding small amounts of oxide materials to steels which has

exhibited positive results in LPBF when adding 1-3 wt% alumina to SS 316L [9]. With this type of

composition change, a homogenous dispersion of the reinforcement material is essential to create

the desired property improvements [105].

These composites can greatly improve the resulting properties, but the properties are still

relatively homogenous across the resulting part. This is because the powders used are either derived

from a composite material directly by gas or water atomization [8, 11] or are the result of mixing

a metal powder with the powder of a reinforcement material by ball milling or roller mixing [9,

11, 70]. Some researchers have experimented with switching powders during a print to create a

property gradient in the part, but these attempts have been limited to large composition changes

and are only capable of altering the composition in one dimension [13–15]. Others have designed

and built new systems that incorporate methods of locally removing powder of one material and

replacing it with that of another material [16,17,74]. While these novel systems can create material

gradients in three dimensions throughout a part, they add substantially to the overall processing

time and are still limited to large composition changes.

A new method of spatially-varied, multi-material LPBF that is being explored at Brigham

Young University is based on doping the standard LPBF powder bed with a liquid or liquid-encased

dopant to create small composition changes in the resulting parts. This would allow for small, local

composition variations without adding substantially to the build time of the part. Previous work by

the authors 3 has shown that inkjet printing is an effective method of liquid deposition that can be

57

integrated into the LPBF process without substantial increase in process time. The current study

investigates the effects of single-track laser scans in powder beds with varying amounts of zirconia

added via inkjet deposition of zirconia slurry to determine how well the dopant is incorporated

into the melt pool. The compositions of the resulting melt pools are analyzed and compared with

predictive calculations for zirconia content. Dispersion of the zirconia within the melt pools is also

examined. Recommendations are made as to the feasible range of implementation for this method

of composition tuning.

5.2 Methods

5.2.1 Materials and Equipment

Previous work by the authors 3 indicated that a colloidal zirconia (ZR100/20, NYACOL)

was suitable for use with inkjet deposition into a steel powder bed. This zirconia slurry had an

average particle size of 100 nm and was measured to be 25 wt% zirconia particles by weighing in

a beaker before and after evaporation of the liquid carrier at 180 ◦C for 12 hours. Zirconia has also

been shown by Koopman et al. [14] to be processed well using LPBF and to have good mixing

with steel when introduced as a large composition change. This study uses the same colloidal

zirconia slurry (ZR100/20, NYACOL) as the source of zirconia dopant and uses SS 316L powder

(CL 20ES, Concept Laser, 29.9 µm, spherical) as the base material. The LPBF system used is a

Concept Laser M2 Cusing Multilaser with dual 400 W lasers.

5.2.2 Inkjet Deposition and Single-track Laser Scans

A small base plate was machined from 1018 mild steel and secured in an adapter plate to fit

the larger build chamber. Sections on this small plate were built up to a height of 0.5 mm (0.5 mm is

shown to be a sufficient height to avoid the effects of using a base plate of a different material (see

Appendix A) to form foundations for the deposition areas (see Figure 5.1-left). On top of these

foundations, powder was spread in a 50 µm layer and walls were built from that powder layer to

define pockets of loose powder that acted as individual powder beds (see Figure 5.1-right). Each

pocket could be doped with Zirconia without the possibility of dopant transport from one pocket

58

Figure 5.1: Left: Platforms built to separate small powder beds from small build plate. Right: Onelayer of powder spread on top of platforms with walls built around the edges to form pockets tocontain inkjet depositions.

to another. Once the pocket walls were built, the small plate was removed from the adapter for

deposition of the dopant by inkjet printing.

Inkjet deposition of the zirconia slurry into the small powder beds was performed with a

simple 3 axis stage and a single, 80 µm-nozzle, piezo-electric, drop-on-demand print head (Micro-

Fab Technologies, Inc.; Part # MJ-AB-01-80-8MX) connected to a pressurized chamber [79, 80].

Three different dopant levels (see Table 5.1) were each deposited into identical powder beds (3

repetitions of each doping level) while 6 powder beds were left untouched to serve as a base-

line (no zirconia deposition). The inkjet deposition was performed using a droplet frequency of

1000 Hz; the speed in the x-direction (during deposition) was set to 60 mm/s and the speed in

Table 5.1: Inkjet deposition line spacing and resulting dopant quantity metrics for the threecolumns of powder bed pockets that have zirconia added in Figure 5.2.

Position Y-spacing Volume of slurry Mass of zirconia/ Zirconia in powder(µm) deposited/layer (µL) area (µg/mm2) bed (wt%)

Left 74.4 15.48 10.096 4.8Middle 37.2 9.62 20.124 9.1Right 18.0 63.84 41.645 17.2

59

Figure 5.2: Left: Powder beds with deposited dopant after plate has been baked (liquid fromdopant slurry all evaporated). Each column of 3 small powder beds are replicates of each other.The two outside columns are both control groups with no dopant added to the powder beds. Right:Single-track laser scans are performed in the doped and non-doped powder beds using 11 differentparameter sets. In both images, the larger powder beds at the top, bottom, and left of the plate wereused for a study investigating the effect of dopant on melt pool characteristics (Chapter 4).

the y-direction (between lines) was set to 5 mm/s. The line spacing (y-direction) was adjusted to

achieve the desired dopant contents in the powder beds as described in Table 5.1. Table 5.1 also

enumerates several metrics for describing the amount of dopant that was deposited into the powder

beds. Of these metrics, the amount of zirconia in the powder bed calculated by weight percent is

of particular importance for comparison to other multi-material LPBF applications that report the

composition change as weight percent of the material in the powder bed [9,100,106]. Note that in

a single layer test, the doped powder will mix with undoped 316 SS from the previous, fused layer

below the powder which will reduce the resulting dopant concentration in the melt pool.

Following the inkjet deposition, the small plate was baked to evaporate the liquid precursor

in the slurry, leaving just the zirconia particles in the steel powder (see Figure 5.2-left). While in

an actual system, the liquid might be evaporated using laser energy, these samples were dried at a

slow rate in an oven. The oven was set to 50 ◦C for the first four hours and 150 ◦C for the next

five hours, then allowed to cool in the oven to room temperature. The temperature was increased

in stages to limit transport of the liquid dopant due to temperature gradients.

60

Table 5.2: Parameter sets used for single-track laser scans. Variable parameters are laser power,laser spot size, and scan speed. Calculated areal energy density is based on the basic

processing parameters.

Position Change from Laser Laser Spot Scan Speed Areal Energy DensityDefault Power (W ) Size (µm) (mm/s) P/(vdS) (J/mm2)

1 S+1 370 160 1350 1.7132 V+1 370 130 1425 1.9973 Original 370 130 1350 2.1084 P+1 395 130 1350 2.2515 S-1 370 100 1350 2.7416 V-7 370 130 825 3.4507 V-9 370 130 675 4.2178 V-11 370 130 525 5.4219 S-3 370 50 1350 5.48110 V-14 370 130 300 9.48711 V-15 370 130 225 12.650

After inkjet printing, the small base plate was returned to the LPBF machine and secured in

the same position and at the same height that it was during the fusing of the pocket walls. Single-

track laser scans were then performed on the zirconia-doped and undoped powder beds using a

range of 11 different parameter sets (see Table B.1) to ensure that some of the tracks resulted in

conduction-style melt pools. The parameter sets varied in laser power (P), laser scan speed (v),

and laser spot size (dS).

5.2.3 Sample Preparation and Measurement

Cross sections of the single-track scans were obtained from the small build plate using a

waterjet cutter to section them. Samples were then prepared by suspending the specimens in epoxy

with the cross-section of the melt pools facing the polishing surface. All samples were ground and

polished to a mirror finish using a standard metallurgical polishing recipe for SS 316L. Residues

from polishing materials (alumina and silica polishing slurries) were removed by sonicating the

samples in micro-organic soap and distilled water followed by sonicating in ethanol. The polished

surfaces of the samples were then coated with a thin (∼15 A) layer of carbon to mitigate charging

of the epoxy in the scanning electron microscope (SEM).

61

The cross-sections of each single-track melt pool were examined using a ThermoScientific

Verios G4 UC SEM. A combination of standard SEM imaging and energy dispersive x-ray spec-

troscopy (EDX) was used to image the melt pools and map the zirconium content inside the melt

pools. Standard SEM images of the melt pools were taken with a resolution of 1024 x 680 pixels at

350x magnification. EDX maps of the zirconium content were taken with a resolution of 256 x 170

pixels at 350x magnification with over 200 frames of data and between 4-5M counts. The EDX

maps were output as grayscale images with each pixel’s intensity related to the wt% of zirconium

measured for that pixel (see Figure 5.3-left).

After SEM data collection, the samples were etched to reveal the grain structure and melt

pool geometry (see Figure 5.3-right). Electrolytic etching was performed using 10% oxalic acid

to reveal the microstructure. The voltage differential was 6 V between the sample surface and the

acid and was applied for approximately 45 seconds. The etched melt pools were then observed and

Figure 5.3: Left: EDX data for Zr wt% in each pixel converted from grayscale for easier visualdistinction. Some noise can be seen in the region immediately surrounding the melt pool. This isan effect of the epoxy that the samples are in and is excluded from the final data. Right: Etchedcross section of melt pool. The melt pool boundaries inside the part are used to designate the areain which the Zr content data is collected.

62

measured using an optical microscope (Olympus GX51 metallurgical microscope). Images of the

melt pools were recorded with a resolution of 1600x1200 pixels at 200x magnification.

5.2.4 Analysis of EDX Data

All image analyses, calculations, and data preparation were done in MATLAB (see Ap-

pendix G for full MATLAB script).

Dopant Content

Zirconia content in the melt pools was calculated using the three types of images that

were mentioned above (SEM, EDX, and optical melt pool). First, the SEM image was resized to

match the EDX image. Then, the optical image of the etched melt pool was scaled and rotated to

match the SEM image by manual input of corresponding points between the two images. Once the

optical image was the same orientation and size as the SEM and EDX images, the etched image

was overlayed onto the SEM image to see the original melt pool surface geometry simultaneously

with the underlying melt pool boundary (see Figure 5.4-left). The melt pool was then manually

traced to create a mask which was used to separate the data inside the melt pool from the rest (see

Figure 5.4-middle). The number of pixels inside the melt pool were summed and multiplied by

the pixel area (calculated using the SEM image size and resolution) to obtain the melt pool area

for each sample. The width was calculated by multiplying the pixel side length with the maximum

number of pixels in the melt pool found in a single row.

After separating the data, the zirconium content data from inside the melt pool (see Figure

5.4-right) was converted to wt% zirconia ( fZrO2) data for each pixel using

fZrO2 =fZrMZrO2

fZrMZrO2 +(1− fZr)MZr(5.1)

where fZr is the zirconium fraction measured for each pixel, MZr is the molar mass of zirconium

(91.224 u), and MZrO2 is the molar mass of zirconia (123.218 u). This calculation operates under

the assumptions that all of the zirconium in the melt pool was in the form of zirconia and that the

data in each pixel represents a volume. Further explanation of the derivation of Equation 5.1 can

63

Figure 5.4: Image analysis process for determining dopant content and distribution in melt pool.Left: Etched melt pool image overlayed with original SEM image. Middle: Mask-subtractedoriginal SEM image. Right: EDX ZrO2 data inside the melt pool with wt% zirconia markedaccording to accompanying color bar scale (0-7 wt% values are white to minimize noise).

be found in Appendix B. The total mass of each pixel volume (mpx) was calculated by

mpx = ( fZrO2)ρZrO2 +(1− fZrO2)ρSS (5.2)

where ρZrO2 is the density of zirconia and ρSS is the density of SS 316L. The values used for ρZrO2

and ρSS are 5.68 and 7.99 g/cm3, respectively. This calculation assumes that the only material in

the melt pool besides zirconia is SS 316L. The total mass of zirconia per pixel volume (mZrO2) was

then calculated as

mZrO2 = ( fZrO2)mpx (5.3)

which can then be used to calculate the overall weight fraction of zirconia in the melt pool (FZrO2)

by

FZrO2 =∑mZrO2

∑mpx(5.4)

where ∑mZrO2 is the sum of mZrO2 for all of the pixels in the melt pool – which equals the total

mass of zirconia in the melt pool – and ∑mpx is the sum of mpx for all of the pixels in the melt pool

– which equals the total mass of the melt pool. The dividend of these two sums is equivalent to the

mass fraction since the pixel volumes of each term cancel out.

64

Noise in the EDX data of an entire melt pool was compensated for using the resulting FZrO2

values obtained for samples from the control group (to which no dopant was added) to estimate the

cumulative noise. Since the FZrO2 for these melt pools should be 0, they were used as a baseline

to adjust the data by subtracting the average no-dopant FZrO2 from each value obtained. It is also

noteworthy that the designation of the melt pool area was done by manually tracing the image.

Applying an automatic image processing tool to recognize and define the melt pool range could be

quite helpful in eliminating human error from this process.

The pixel by pixel EDX data required a different approach to compensating for the noise.

Since the pixel by pixel noise quantity is not known, the data cannot be corrected on a pixel basis.

Instead, the control samples were utilized to estimate the probability distribution of measuring

different zirconia levels due to noise. The cumulative distributions of the control sample melt

pools were used to calculate the pixel value for the 95th percentile pixel noise value. The average

of the 95th percentile noise values across all the zero dopant melt pools (about 7 wt%) was used

as a cutoff value for images representing the individual pixel values in the melt pool. All pixels

with values below the cutoff are white (see Figure 5.4-right). Thus, pixels that are colored have a

greater than 95% chance of representing added zirconia.

Predicted values for the zirconia fraction in each melt pool were calculated by assuming

that the entire amount of zirconia deposited above a given melt pool was incorporated into the

melt pool during fusion. By calculating the amount of zirconia and steel powder that should be

incorporated into the melt pool and comparing that to the actual size of the melt pool, predicted

zirconia fraction (PZrO2) can be calculated using

PZrO2 =ϕZrO2wMPρZrO2

ϕZrO2wMPρZrO2 +ρSS(AMPρZrO2−ϕZrO2wMP)(5.5)

where ϕZrO2 is the amount of zirconia deposited per unit area, wMP is the measured width of the

melt pool, AMP is the cross-sectional area of the melt pool, and ρZrO2 and ρSS are the densities of

zirconia and SS 316L, respectively. Full derivation of predicted melt pool concentrations (Equation

5.6) can be found in Appendix D. In this case, ϕZrO2 is calculated using the inkjet deposition

parameters, z is 50 µm, fp is assumed to be 0.5, wMP and AMP are measured for each melt pool as

noted previously, and ρZrO2 and ρSS are 5.68 and 7.99 g/cm3, respectively. It is important to note

65

that Equation 5.6 only takes into account the amount of dopant that is directly above the widest

point of the melt pool; it does not consider the dopant material that could be introduced into the

melt pool from the denudation zones.

Dopant Dispersion

Cumulative distributions of the zirconia contribution by pixel value were calculated using

the same approach as the overall weight fraction of zirconia (see Equation 5.4), with the excep-

tion of the number of pixels included in the sum. For each pixel value, the amount of zirconia

contributed to the melt pool by pixels of equal or lesser value is described as

FZrO2(x) =∑

x0 mZrO2

∑mpx(5.6)

where the numerator is the sum of zirconia mass in pixels of equal or lesser value than x. This

results in a cumulative distribution of the overall wt% zirconia with the ultimate value for each

function representing the total, non-noise-adjusted weight fraction of zirconia in the melt pool.

The contribution of dopants can be observed as a deviation in the cumulative distribution function

from the non-doped case. This approach proved to be most effective at detecting low levels of

dopant dispersed in the melt pools.

The spatial distribution of the zirconia across the melt pool was also calculated using the

EDX data extracted from the melt pools. One of the metrics used to quantify this is an adaptation

of the dispersion distribution index (dIndex) described by Haslam and Raeymaekers [108]. They

showed that the dIndex is especially valuable compared to established ASTM standards when the

range of dispersion results are large and when the whole sample must be considered. The dIndex is

calculated by dividing the region of interest into equal sections and comparing the average content

values for each section. This takes the form of

dIndex =12

[1− s(b)

max(s(b))+

bmax(b)

](5.7)

where b represents the set of content values for each box, b is the average content value of all of the

boxes, max(b) is the highest box content value, s(b) is the standard deviation of the box content

66

values, and max(s(b)) is the maximum possible standard deviation of the set (i.e. when half of

the values are 1 and the other half are 0). Possible dIndex values can range from 0 to 1 where

1 represents perfect dispersion. The dIndex was adapted to describing melt pool distributions by

including the entire melt pool in the calculation, rather than just a small sample. This was done

by dividing the entire EDX image into a grid with boxes either the size of a single pixel or 2x2

pixels (see Appendix E for elaborated discussion). The zirconia content was calculated for each

box as described above using only the values from pixels inside the melt pool boundaries. The

fraction of pixels that were included in this calculation was recorded for each box. The set of

zirconia content values for each box was used as the b vector in Equation 5.7. From this set of

values, the standard deviation (s(b)), average (b), and maximum value (max(b)) were calculated.

The maximum possible standard deviation (max(s(b))) was calculated using the number of boxes

used (nb) as

max(S(b)) =

√nb2 |1−0.5|2 + nb

2 |0−0.5|2

nb−1(5.8)

representing the standard deviation if half of the values were at the maximum possible value (1)

and the other half were at the minimum possible value (0). Since the distance between either the

maximum or minimum values to the average value will be the same (0.5), this can be simplified to

max(s(b)) =

√nb ∗0.52

nb−1(5.9)

and used in Equation 5.7 along with the other values derived from the set of weight fraction values

for the boxes (b).

The dispersion of the dopant throughout the melt pool was also analyzed by calculating the

variance of the pixel dopant content values in the melt pool. The resulting variance values for each

melt pool were then compared. Another comparison between different melt pools was made based

on the fraction of pixels that contained dopant content values over 50 wt%.

67

Figure 5.5: Three examples of typical melt track formation of different types. Top: Lack-of-fusion occurs when the input energy is insufficient to create melt pools large enough to adhere tothe substrate below. Middle: Balling is when the melt track divides up into periodic, sphericalbumps due to instabilities in the liquid melt pool. Bottom: Continuous melt tracks are consistentin size and can exhibit melt pool cross sections that are somewhat circular (conduction mode) orvertically-elongated (keyhole mode).

5.3 Results and Discussion

The resulting melt tracks exhibited a wide range of characteristics that were highly depen-

dent on the laser processing parameters as described in Chapter 4. While some of the melt tracks

were continuous, others exhibited severe balling and lack-of-fusion (see Figure 5.5).

The analysis methods involved in EDX are based on an assumption that the specimen is

basically homogenous throughout the electron interaction volume and that there are no geometric

effects on the measured values [109]. Since some of the melt tracks are not continuous, measuring

EDX data would be meaningless as the electrons could pass in and out of the melt track resulting in

a combination of data for the melt track present and the epoxy directly beneath it. This description

is characteristic of the majority of the melt pools produced using parameter sets 1-6 (see Table

B.1). When these samples were polished for SEM imaging, the melt pools were quite abnormal or

non-existent (see Figure 5.6). Parameter sets 7-11 all produced continuous melt pools that could

be measured using EDX elemental analysis (see Figure 5.7). The data for both the dopant content

and dopant distribution discussed below are extracted from parameter sets 7-11 that generated

continuous melt pools.

68

1a 2a 3a 4a 5a 6a

1b 2b 3b 4b 5b 6b

Figure 5.6: Poor/non-existent melt pools from which meaningful data could not be extracted basedon the specimen requirements for EDX elemental analysis mentioned. Note that melt pools madeusing the original parameter set (# 3) are among those that were not evaluated further.

7a 8a 9a 10a 11aFigure 5.7: Representative set of melt pools from parameter sets 7-11. These melt pools were moreconsistent compared to those produced using the other parameter sets (see Figure 5.6).

5.3.1 Dopant Content

The resulting zirconia content values for parameter sets 7-11 given the various input quan-

tities of dopant are shown in Figure 5.8. The data follow the expected trend of increased dopant

content with increased dopant input. The trendline for the measured melt pool concentrations has

a slightly lower slope than the trendline produced using the predicted concentrations for each melt

pool. This indicates that, while the majority of the dopant was incorporated into the melt pool, a

small portion of the deposited dopant did not make it into the melt pool. If some of the dopant

is not incorporated into the melt pool, the resulting composition will be less than desired and the

change in properties may not be as substantial as intended. Understanding the amount of dopant

necessary to attain a specific composition (and therefore properties) in the fused material will be

vital to the realization of effective composition control using this deposition method.

69

y = 0.114x

Measured

y = 0.1574x

Predicted

-1

0

1

2

3

4

5

6

0 5 10 15 20

Do

pan

t C

on

ten

t (w

t% Z

rO2

in M

elt

Poo

l)

Input Dopant (wt% ZrO2 in Deposited Material)

Parameter Set 7

Parameter Set 8

Parameter Set 9

Parameter Set 10

Parameter Set 11

Figure 5.8: Resulting zirconia content in the melt pool by wt% compared to the amount of zirconiadeposited as a wt% of the total deposited material (steel powder included). The trendline for thepredicted data overestimates the amount of dopant that is measured by about 40%. Individual datapoints for the predicted values are not shown.

Some of the dopant loss could be attributed to spatter (material ejection from the melt pool)

which has been known to cause material loss directly by expelling particles at the laser focal point

and, at times, indirectly by expelling those particles at a low enough angle to impinge directly

on the powder bed causing a denudation zone [110, 111]. While the direct ejection of material at

the laser/melt pool interface is consistent across the parameter space, indirect powder loss could

be avoided by adjusting the parameters to control the shape of the melt pool. A wider range of

parameter sets with understood spatter tendencies should be investigated with deposited dopant to

determine how much the dopant is affected as opposed to the base powder.

Another possible mechanism for the loss of dopant material is evaporation of the zirco-

nia. Evaporation of the material in the melt pool occurs often in LPBF and is the driving force

behind the elongated melt pools that define keyhole-mode laser melting [96, 97, 112]. While it

is unlikely that this would affect zirconia in the same way as SS 316L (evaporation temperatures

differ by ∼1200 ◦C), evaporation of the material may be a contributing factor. Analysis of zirco-

nia content in melt pools formed using parameter sets known to produce less material evaporation

70

y = 0.6023x

Measured = Predicted

-0.5

0.5

1.5

2.5

3.5

4.5

5.5

0 1 2 3 4 5

Do

pan

t C

on

ten

t (w

t% Z

rO2

in M

elt

Poo

l)

Predicted Dopant Content (wt% ZrO2 in Melt Pool)

Parameter Set 7

Parameter Set 8

Parameter Set 9

Parameter Set 10

Parameter Set 11

Figure 5.9: Measured dopant content compared directly with corresponding predicted content val-ues rather than with the amount of zirconia deposited (see Figure 5.8). These predicted values takeinto account the width and cross-sectional area of the resulting melt pools as described in Equation5.6. A trendline for the set of data points is shown as well as a line that indicates where the mea-sured dopant content would be equal to its corresponding predicted value. A small fraction of thedata points lies above this line, indicating high-efficiency in terms of dopant material use.

(non-keyhole mode) would be appropriate to investigate the extent to which evaporation is a factor

in the amount of zirconia incorporated into the melt pools.

In comparing the zirconia content data directly with their corresponding predicted content

values (see Figure 9), it is interesting to note that there are a significant number of samples for

which the measured dopant content exceeds the corresponding predicted value. The mechanisms

involved in these variations need to be understood to make this type of dopant delivery consistent

enough for industrial applications.

One possible reason for the measured values being higher than the predicted values is the

incorporation of material from the denudation zone around the melt pools. The predicted dopant

content was simplified to only include the material directly above the melt pool. However, it is

known that the material from the region immediately surrounding a single-track laser scan can be

incorporated into the melt pool [56,113]. In instances where the inclusion of surrounding material

dominates the expulsion of material by spatter or evaporation, this could result in higher dopant

71

concentrations in the melt pool than predicted. This phenomenon could be investigated further

by generating single-track scans in doped powder beds at varied distances from each other. At

close proximities, it would be expected that the resulting melt pools have less incorporated dopant

because of the denudation zones created by neighboring melt tracks; conversely, solitary melt

tracks unaffected by their neighbors should have slightly more incorporated dopant.

While these results are encouraging and supported by the literature, there is also a spread

between data points that can cause skepticism. This spread in the data is understandable for a few

reasons. First, the measurements are very sensitive to the melt pool boundary definition because

much of the zirconia is along the boundaries. A small deviation in defining the boundaries could

cause significant errors in the dopant calculations. More uniform distributions or larger dopant

areas will facilitate more accurate measurements. Additionally, the inkjet deposition process used

to deliver the dopant to the powder bed is relatively unrefined for the material that was used.

Another source of error is the transport in the powder bed during drying 3. This material transport

may be mitigated by better controlling the temperature of the powder bed or by depositing smaller

amounts of dopant at a time and evacuating all moisture from the powder bed between depositions.

Despite these sources of uncertainty, the data are quite promising because they show dopant

content values in ranges that have been specified to be useful in controlling properties through com-

position change [9]. This is great step toward spatial composition control for small composition

adjustments rather than being limited to abrupt material boundaries that can cause serious defects

in a part. It will be important to continue refining this process in an effort to create accurate and

precise composition changes.

5.3.2 Dopant Dispersion

Qualitatively, it is observed that the majority of the zirconia is agglomerated near the edges

of the melt pool that are protruding from the substrate (see Figure 5.10). This is especially true

of the melt pools with higher concentrations of zirconia. The measured zirconia in the non-doped

samples is believed to be noise since no zirconia was added to those samples. To help visually

reduce this noise seen in Figure 5.10 and to facilitate comparisons between treated and untreated

samples, the measured zirconia concentration was averaged in a 15 x 15 pixel region. Repre-

sentative comparisons using this method are shown in Figure 5.11. The results do not show a

72

v=675 mm/s

v=525 mm/s

ds=50 𝜇m

v=300 mm/s

v=225 mm/s

0 wt% 4.8 wt% 9.1 wt% 17.2 wt%

73

Figure 5.10: Resulting noise-reduced colormaps (all Zirconia levels below the 95% noise thresholdare white) of melt pools at varying dopant input levels using various parameter sets. Dopant inputlevels are labeled above the column of corresponding images by the zirconia wt% in the powderbed. The several parameter sets used to create these specific melt pools only varied from theoriginal parameter set (P = 370 W , v = 1350 mm/s, ds = 130 µm) by the parameter labeled tothe left of the images. The original parameter set was not analyzed due to its inconsistencies asmentioned previously. Each pixel is colored according to its zirconia content as described by thecolor bar on the right. It is noteworthy that the zirconia tends to agglomerate near the edges of themelt pools and that it mostly remains in the tops of the melt pools.

clear contrast between the zirconia concentration in the bulk of the melt pool and the untreated

regions outside of the melt pool. The majority of the zirconia is located on the protruding melt

pool boundaries. This agglomeration could be detrimental to mechanical properties rather than

improving them as ODS steels and other dispersion strengthened alloys rely on small particles to

create grain boundary pinning [114]. Agglomeration, or poor dispersion of the dopant, signifi-

cantly limits the applicability of this doping technique for zirconia in stainless steel at these dopant

levels.

The poor distribution at higher dopant contents could be a result of poor miscibility between

the ceramic dopant (ZrO2) and metal base material (SS 316L). This would be supported by the

work of Koopmann et al. [14] who found that ceramic/metal interfaces in LPBF can have extremely

poor bond strength. Koopmann’s study also found that this bond strength can be improved by

re-melting the interface area to increase the amount of mixing that takes place between the two

materials.

This concept of re-mixing could be implemented in the current application by scanning

each single-track multiple times as a way to encourage dispersion of the dopant throughout the

melt pool. Another method that may promote increased mixing is to create area scans instead of

single-track scans. The overlap of the tracks in this case may provide enough re-melting to mix

and disperse the dopant throughout the melt pools more uniformly. With these area scans, it is

also possible that the mixing pulls dopant from formed melt pools into new adjacent melt pools

in significant amounts. This would warrant specific attention to possible disparities of dopant

concentrations across the area scanned with respect to the direction in which successive tracks are

scanned.

74

v=675 mm/s

v=525 mm/s

ds=50 𝜇m

v=300 mm/s

v=225 mm/s

0 wt% 4.8 wt% 9.1 wt% 17.2 wt%

0

0.5

1

1.5

2

2.5

3

3.5

75

Figure 5.11: Resulting colormaps created using local averaging to reduce influence of measure-ment noise. Images represent melt pools at varying dopant input levels using various parametersets. Dopant input levels are labeled above the column of corresponding images by the zirconiawt% in the powder bed. The several parameter sets used to create these specific melt pools onlyvaried from the original parameter set (P = 370 W , v = 1350 mm/s, ds = 130 µm) by the parameterlabeled to the left of the images. The original parameter set was not analyzed due to its inconsis-tencies as mentioned previously. Each pixel is colored according to the average zirconia content ofthe 15 x 15 pixel square with the pixel of interest at the center. The pixel colors correspond to thecolor bar on the right. Pixels with values less than 0.5 wt% show as white and pixels with valuesabove 3.5 wt% are marked as grey. The epoxy areas around the protruding sections of the meltpools generally show high zirconia content because of increased noise due to lower levels of otherelements measured (e.g. iron, nickel, etc.) It is noteworthy that the bulk interiors of the melt poolsdo not differ significantly from the bulk area outside of the melt pools. This indicates that little tono zirconia is present in these bulk regions within the melt pools.

Understanding the final location of these zirconia agglomerates within the fused material

will also be impactful to how this doping approach is implemented. If it is known that most of

the dopant stays near the top of the first layer that receives dopant, it will be easier to control the

exact location along the build direction at which the composition begins to change. Since most

melt pools have depths greater than the height of a deposited powder layer, fusing a second layer

on top of the initial single-track or area scans would be another method that may improve mixing

between the dopant and base material. As the area above the doped melt pool is scanned, the

majority of the melt pool will be remelted and become part of the new melt pool. For the new melt

pool, this would change the effective introduction location of the dopant from the added material

to inside the “substrate” material. This change could allow for higher dopant concentrations in

the bottom part of the second-layer melt pools and improve the overall dispersion of the dopant

throughout the part. Additionally, when the dopant is added to each successive layer, the overall

dopant concentration will be higher than seen in this study for the same input amounts. This means

that much lower dopant input levels would be required to reach dopant levels similar to those found

in ODS steels [19, 21, 22, 66]. These lower dopant levels could also facilitate better distribution

throughout the fused material.

The cumulative distributions of zirconia content were plotted for multiple samples that

were created using the same parameter set and various input dopant levels (see Figure 5.12). These

distributions showed that all of the samples have similar amounts of low-zirconia content pixels.

76

AverageNoise

0

1

2

3

4

5

6

0 3 6 9

12

15

18

21

24

27

30

33

36

39

42

45

48

51

54

57

60

63

66

69

72

75

78

81

84

87

90

93

96

99

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 11

0B-11a

1B-11a

2B-11a

4B-11a

Figure 5.12: Cumulative distribution of zirconia content in the melt pool by pixel value in wt%.All represented samples were created using parameter set 11 (P = 370 W , v = 225 mm/s, anddS = 130 µm). Four unique samples are measured for each of the four input dopant levels. Thisillustrates the relative contribution to the overall zirconia content amount by pixels of a givenvalue. For most of the samples, the majority of the zirconia is contained in pixels with large valuesindicating grouping or agglomeration of the zirconia and poor overall dispersion. Similar graphsfor parameter sets 7-10 can be found in Appendix F.

The amount of zirconia in the melt pool that is attributed to these low-content pixels is close to the

average noise threshold for the entire melt pool evaluated previously. The distributions also exhibit

increasing amounts of zirconia in high-content pixels with increasing input dopant amount. Since

high-content pixels correspond to areas with high zirconia concentrations, samples with more high-

content pixels can be said to have more dopant agglomeration and poorer overall dispersion. This

shift in the distribution is seen in samples with greater input dopant quantities and suggests that

the more of the dopant is segregated in these cases.

While the majority of the zirconia in the melt pools is found in higher-concentration regions

(see Figure 5.12), a closer look at the cumulative distributions of the zirconia content shows that the

zirconia-doped distributions depart from their non-doped counterparts close to the noise threshold

(see Figure 5.13). The zirconia contribution from these modest-concentration areas indicates that

77

AverageNoise

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 11

0B-11a

1B-11a

2B-11a

4B-11a

Figure 5.13: Zoomed in frame of Figure 5.12. Spreading the data reveals that the zirconia-dopedsamples behave differently than non-doped samples around and immediately after reaching theaverage noise threshold. The non-doped samples level off rather quickly with little to no additionalzirconia contributed by pixels with values greater than 30 wt% while the zirconia-doped samplesdecrease their slopes slightly, but do not level off. This is an indication of substantial amountsof lower-value pixels in the zirconia-doped samples compared to the non-doped samples. Similargraphs for parameter sets 7-10 can be found in Appendix F.

there is local dispersion of zirconia throughout the melt pools. This type of dispersion will be vital

to the potential use of this doping method to create dispersion-strengthened alloys with improved

properties. Future work using lower dopant levels and possible re-melting may be most favorable

for achieving dispersed zirconia.

Further investigation of the melt pools using backscatter SEM imaging reveals areas in

which individual zirconia particles can be observed (see Figure 5.14). The composition of the

particles was confirmed to have high zirconium and oxygen contents and can be assumed to be

zirconia. These zirconia particles are circular in shape and measure around 100 nm in diameter.

The zirconia particles also appear to be spaced out rather than agglomerated in a large mass. This

local dispersion confirms that some local dispersion does occur throughout the melt pool. Since

the zirconia particles in the slurry were originally around 100 nm, this finding also indicates that

78

500 nm

100 nm

Figure 5.14: Left: Backscatter SEM image showing individual zirconia particles. Particles areround with sizes about the same as the manufacturer-provided slurry particles sizes. Right: EDXanalysis of area immediately surrounding a single zirconia particle confirms the presence of zirco-nium in the region sampled.

there was little to no change in the size of the input zirconia particles. Understanding this relation-

ship between input dopant properties and resulting dopant morphology will be helpful in selecting

dopants to be used with this deposition method.

The lack of change in zirconia particle size indicates that some of the zirconia (if not all)

remains solid throughout the melting and solidification of the melt pool. This stands to reason

since the temperature regime in which SS 316L is a liquid (∼1400 ◦C –∼2815 ◦C) barely includes

the temperature regime in which zirconia begins to melt (∼2715 ◦C). The spacing between the

zirconia particles may be a result of the initial charge on the particles that is used to keep the

particles from agglomerating in the slurry. If the individual particles maintain their similar charges

after introduction into the melt pool, they will repel each other similar to their behavior in the

slurry. Further investigation of dopants with varied initial charge, particle size, concentrations, and

melting temperatures is warranted.

The dIndex values obtained with both the 1x1 and 2x2 pixel box sizes show a similar

trend of decreasing dIndex (worsening dispersion with more agglomeration/less uniformity) as

the observed concentration of dopant in the melt pool increases (see Figure 5.15). This poor

distribution is generally caused by agglomeration of the dopant near the edges of the melt pool

79

Figure 5.15: Resulting dIndex values for zirconia dispersion in the melt pool compared to themeasured zirconia content of the melt pool. A dIndex value of 1 represents perfect dispersion.The general trend points toward a less-disbursed distribution as the amount of zirconia in the meltpools increases. Left: Box size of 1x1 pixel is approximately 681x larger than the average expectedzirconia particle size. Right: Box size of 2x2 pixels is approximately 2,725x larger than the averageexpected zirconia particle size.

as noted in Figure 5.10. This poor dispersion at higher zirconia concentrations may be a limiting

factor to the amount of dopant that can be added to a material using this doping method and the

resulting material properties that can be achieved using zirconia in SS 316L.

The dIndex trends were confirmed by evaluating two more metrics that characterize the

dispersion of the dopant in the melt pools: elevated zirconia content in individual pixels and the

variance of the data across the melt pool (see Figure 5.16). Elevated zirconia content was measured

by determining the fraction of the pixels in a given melt pool that contain more than 50% zirconia

by weight. Plotting this data against the measured zirconia content in the melt pools showed

that the fraction of pixels with highly-concentrated zirconia increased at a rate of about 1.3% for

every 1 wt% increase of zirconia content in the melt pool (see Figure 5.16-right). The variance of

the melt pool data was calculated using statistical variance of the reported wt% zirconia for each

pixel inside the melt pools (see Figure 5.16-left). In comparing these metrics against the observed

zirconia contents of the entire melt pools, a clear trend of increasing agglomeration and decreasing

80

Figure 5.16: Other dispersion metrics confirm the trends observed using the dIndex data (disper-sion/uniformity gets worse as dopant concentration increases). Left: The fraction of pixels withmore than 50 wt% zirconia are likely to be agglomerates given the low overall concentrations ofdopant in the melt pools. Right: Statistical variance of the data set including all of the pixels in themelt pool is larger for less-uniform melt pools.

uniformity can be seen. This clearly confirms the trends reported by the dIndex values for both

box sizes.

A possible reason for this decreasing dispersion with increased dopant content may be a

natural limit of how much zirconia can be well-dispersed in a stainless steel matrix – similar to a

solubility limit. In that case, it will be important to characterize that limit by repeating similar test-

ing to the current study at smaller concentration intervals. Greater understanding of the solubility

could also be gained by focusing on smaller dopant concentrations where there seems to be less

dopant agglomeration and greater uniformity.

Poor miscibility and solubility of zirconia in stainless steel also warrants investigation into

other material systems where the dopant and base would either mix together more readily or where

the dopant is soluble in the base material. If the segregation observed in this study is material

dependent, other material systems may prove to be much more feasible using this doping method.

81

5.4 Conclusions

Single-track laser scans were performed on a zirconia-doped SS 316L powder bed where

the zirconia was deposited at various quantities using an inkjet deposition system. The quantity and

distribution of the zirconia that was incorporated into the melt pools were measured. Significant

amounts of the deposited zirconia were incorporated into the melt pools. The resulting melt pools

had resulting zirconia concentrations similar to other work that has shown promising property

improvements when adding small amounts of oxides to steel in LPBF. Some dispersion of the

zirconia is noted, but the majority of the zirconia is found in agglomerates near the top surfaces of

the melt pool. The amount of zirconia included in these agglomerate areas increases with increased

zirconia content. This increased agglomeration corresponds to a decrease in overall dispersion of

zirconia in the melt pool.

Future work with single-track scans in zirconia-doped SS 316L should focus on smaller in-

tervals between dopant concentrations, parameter sets that reduce spatter and material evaporation,

and material systems with greater miscibility and solubility. Performing area scans and multi-layer

builds or implementing re-melting techniques to improve dispersion and understand steady-state

dopant concentration levels is also recommended.

82

CHAPTER 6. CONCLUSION

6.1 Summary and Conclusions

This thesis addresses several main issues surrounding the novel concept for composition

control in laser powder bed fusion using liquid or liquid-encased dopants. Particularly, deposition

methods are evaluated, necessary changes to processing conditions are identified, and proof of

concept is performed by creating substantial composition changes in LPBF-processed melt-tracks.

6.1.1 Deposition Methods

In Chapter 3, two different liquid deposition methods – direct write extrusion and inkjet

printing – were evaluated in consideration for use with this novel composition control process.

The deposition methods were compared under three criteria by observing various characteristics

of the processes in action and qualities of the resulting dopant deposits. The three criteria used

were deposition quality, integration feasibility, and system productivity. From this evaluation,

inkjet deposition was recommended to be used to introduce the dopant because it outperformed

the direct write method in every considered category.

Inkjet printing showed good deposition quality because the observed defects (e.g. increased

roughness, skipped lines) were correctable in ways that were known and feasible. It is also feasible

to integrate inkjet deposition into a LPBF process because it has good spatial resolution and can

operate without significant alterations to any other components in the LPBF system. The overall

productivity of an LPBF system with an integrated inkjet printhead would be improved compared

to a standard LPBF system because the added inkjet capabilities would allow for higher-value

products without creating extra time delays or contaminating any powder that is not included in

the final part. The inkjet deposition method was also observed to deposit dopant in quantities con-

83

sistent with the literature’s description of effective concentration levels for property improvement

in many metal matrix composites.

6.1.2 Processing Windows

In Chapter 4, the process space was explored and mapped for various laser parameters and

dopant levels. This was done by creating single-track laser scans using 54 different parameter sets

in powder beds with three different dopant input quantities. Each resulting melt track was evaluated

with respect to its melt pool characteristics and categorized accordingly. The relationship between

input energy density and resulting melt pool category was compared for the different dopant input

quantities.

For samples in which zirconia dopant was added to the SS 316L powder, the range of

energy densities over which balling occurred expanded dramatically while the range over which

continuous melt pools were formed changed very little. The expansion of the balling regime may

be caused by competing physical mechanisms – such as decreased surface tension and increased

laser energy absorption – that alter the forming melt pool in different ways. The competition

between these two mechanisms is dependent on the amount of dopant that is added. At lower

dopant quantities, the mechanism that shifts the balling regime toward higher energy densities

dominates. Conversely, the mechanism that shifts the balling regime toward lower energy densities

dominates at higher input dopant levels. These dissimilar regime shifts at different dopant levels

indicate a need for composition-dependent adjustments to the processing parameters in order to

obtain desired results. Different adjustments may be necessary for different material systems.

6.1.3 Dopant Incorporation

In Chapter 5, the incorporation of dopant into the melt pool was investigated using single-

track laser scans in zirconia-doped powder beds. The incorporation of dopant into the melt pool

was measured in terms of the amount of dopant measured in the melt pool as well as in terms of

the dispersion of the included zirconia throughout the melt pool.

For zirconia-doped SS 316L, it was shown that an average of 72% of the predicted amount

of zirconia was measured in the melt pools. This indicates that dopant deposited using inkjet print-

84

ing can be successfully integrated into a fused part despite some material losses. Some dispersion

of the dopant throughout the melt pool was observed. However, the majority of the zirconia ag-

glomerates near the top surfaces of the melt pools. The amount of zirconia in these agglomerate

regions increased with increased dopant input, leading to an overall decrease in the fraction of

incorporated zirconia that was well-dispersed. These poor distributions could be improved with

other material systems that exhibit greater miscibility, lower overall dopant concentrations, and/or

re-melting strategies.

6.2 Future Work

As this novel process continues to develop, there are a lot of factors that still need to be

understood. These factors range from inkjet deposition development to dopant content and distri-

bution in fused parts to the effects that these composition changes have on final properties.

6.2.1 Deposition Methods

Although inkjet printing is a well-developed process, it is still somewhat difficult to use

liquids with solid particles in an inkjet system. Further development of inkjet printing using slurries

or other possible dopant carriers will be central to the full development of the type of spatial

composition control that is the subject of this work. A vital component of this development will be

the understanding of how the solid particles interact with the powder bed and the extent to which

particles are transported after deposition. Many different materials will need to be tested to ensure

that a broad range of composition options are available for this type of composition control. It will

also be important to emphasize methods to expand the feasible range of concentrations and input

quantities that can be used while maintaining good powder bed morphology and uniformity.

6.2.2 Processing Windows

Understanding the mechanisms that affect the LPBF processing windows will also be valu-

able for scaling this process to build full-scale parts. For the zirconia/SS 316L material system, a

lower range of energy densities with a greater range of input dopant quantities would give greater

85

insight into the way that these mechanisms impact melt pool formation and the balance of the

mechanisms’ comparative influence.

6.2.3 Dopant Incorporation

With the dopant content and distribution, the results that were obtained in this study are

promising. However, improving the dispersion of the dopant throughout the melt pool will be

necessary for achieving the desired dispersion-strengthening effects. Similar experiments using

area and multi-layer scans will be informative of how the dispersion will change when building

a full part. Additional techniques for improving dispersion should be investigated. An example

of such a technique would be direct re-mixing of the melt pools using subsequent laser scans.

Further, the way that dopant content changes along the melt track should also be investigated.

This could be done by depositing dopant in one section of the powder bed and scanning the laser

from the doped area to the non-doped area. Values for dopant content could then be collected and

compared based on the distance of the sample area from the doped region. A similar technique

should be implemented with area scans and multi-layer builds to understand the way the dopant is

transported throughout the fused regions in two dimensions and three dimensions, respectively.

6.2.4 Material Systems

Other types of dopants and dopant/base material combinations should be investigated.

Dopants of interest for use with steels could include other oxides and dispersion strengthening

agents, solid solution strengthening agents, titanium or other carbide formers, and chromium or

nickel for increased corrosion resistance. Other dopants that have substantial effects on the melt

pool formation or resulting microstructure may also be of interest.

6.2.5 Property Improvements

Finally, the entire purpose of composition control is to improve the properties when the

composition is changed. This improvement can be understood by studying the microstructure and

the impact that the composition changes create. It will also be necessary to test the mechanical

properties of the parts with changed composition. This will begin with tests on properties like

86

hardness, wear resistance, and corrosion resistance. However, when full parts can be made, the

full array of properties (i.e. strength, modulus, etc.) will need to be tested to fine tune the process

and make it capable of creating properties specific to any end use.

87

REFERENCES

[1] Kover, A., 2018. Transformation in 3d: How a walnut-sized part changed the way ge avia-tion builds jet engines. 1, 18, 37, 56

[2] Bockin, D., and Tillman, A. M., 2019. “Environmental assessment of additive manufactur-ing in the automotive industry.” Journal of Cleaner Production, 226, pp. 977–987 Ia9uaTimes Cited:0 Cited References Count:50. 1, 18, 37, 56

[3] Zhang, B. Q., Pei, X., Zhou, C. C., Fan, Y. J., Jiang, Q., Ronca, A., D’Amora, U., Chen,Y., Li, H. Y., Sun, Y., and Zhang, X. D., 2018. “The biomimetic design and 3d printingof customized mechanical properties porous ti6al4v scaffold for load-bearing bone recon-struction.” Materials & Design, 152, pp. 30–39 Gh2ea Times Cited:23 Cited ReferencesCount:57. 1, 37, 56

[4] du Plessis, A., Broeckhoven, C., Yadroitsava, I., Yadroitsev, I., Hands, C. H., Kunju, R.,and Bhate, D., 2019. “Beautiful and functional: A review of biomimetic design in additivemanufacturing.” Additive Manufacturing, 27, pp. 408–427 Hw9ff Times Cited:1 CitedReferences Count:143. 1, 18, 37, 56

[5] Gupta, A., and Talha, M., 2015. “Recent development in modeling and analysis of function-ally graded materials and structures.” Progress in Aerospace Sciences, 79, pp. 1–14 Cy0bsTimes Cited:162 Cited References Count:192. 2, 39, 56

[6] El-Galy, I. M., Saleh, B. I., and Ahmed, M. H., 2019. “Functionally graded materialsclassifications and development trends from industrial point of view.” SN Applied Sciences,1(11) Jk6kx Times Cited:4 Cited References Count:199. 2, 15, 18, 39, 56

[7] Gibson, I., Rosen, D., and Stucker, B., 2015. Design for Additive Manufacturing. SpringerNew York, New York, NY, book section Chapter 17, pp. 399–435. 2

[8] Hunt, R. M., Kramer, K. J., and El-Dasher, B., 2015. “Selective laser sintering of ma956oxide dispersion strengthened steel.” Journal of Nuclear Materials, 464, pp. 80–85 Com-pilation and indexing terms, Copyright 2019 Elsevier Inc. 20151900824292 Damage resis-tance High temperature Joining process Net-shape parts Nuclear structural materials Oxidedispersion strengthened steel Selective laser sintering Ultimate tensile strength. 2, 4, 13, 14,15, 57

[9] Li, X., Willy, H. J., Chang, S., Lu, W., Herng, T. S., and Ding, J., 2018. “Selective lasermelting of stainless steel and alumina composite: Experimental and simulation studies onprocessing parameters, microstructure and mechanical properties.” Materials & Design,145, pp. 1–10 Ga2by Times Cited:17 Cited References Count:49. 2, 14, 15, 21, 35, 41, 57,60, 72

88

[10] Shishkovsky, I., Sherbakov, V., Ibatullin, I., Volchkov, V., and Volova, L., 2018. “Nano-sizeceramic reinforced 3d biopolymer scaffolds: Tribomechanical testing and stem cell activity.”Composite Structures, 202, pp. 651–659 Sp. Iss. SI Gs6qb Times Cited:0 Cited ReferencesCount:59. 2, 13, 15

[11] Simonelli, M., Aboulkhair, N. T., Cohen, P., Murray, J. W., Clare, A. T., Tuck, C., andHague, R. J. M., 2018. “A comparison of ti-6al-4v in-situ alloying in selective laser meltingusing simply-mixed and satellited powder blend feedstocks.” Materials Characterization,143, pp. 118–126 Sp. Iss. SI Gx9dp Times Cited:6 Cited References Count:31. 2, 14, 15,57

[12] Hsu, T.-H., Chang, Y.-J., Huang, C.-Y., Yen, H.-W., Chen, C.-P., Jen, K.-K., and Yeh, A.-C., 2019. “Microstructure and property of a selective laser melting process induced oxidedispersion strengthened 17-4ph stainless steel.” Journal of Alloys and Compounds, 803,pp. 30–41 Compilation and indexing terms, Copyright 2019 Elsevier Inc. 20192607113056Geometrically necessary dislocations Microstructure and properties Microstructure engi-neering Microstructure properties Oxide dispersion strengthened Oxide dispersion strength-ening Selective laser melting (SLM) Transformation induced plasticity. 2, 15

[13] Demir, A. G., and Previtali, B., 2017. “Multi-material selective laser melting of fe/al-12sicomponents.” Manufacturing Letters, 11, pp. 8–11. 2, 15, 18, 57

[14] Koopmann, J., Voigt, J., and Niendorf, T., 2019. “Additive manufacturing of a steel–ceramicmulti-material by selective laser melting.” Metallurgical and Materials Transactions B,50(2), pp. 1042–1051 Compilation and indexing terms, Copyright 2019 Elsevier Inc.20190706488443 Adhesive strength Complex structure Direct manufacturing Multi materi-als Multi-material process Porous structures Selective laser melting (SLM) Steel substrate.2, 15, 16, 18, 19, 21, 40, 57, 58, 74

[15] Nadimpalli, V. K., Dahmen, T., Valente, E. H., Mohanty, S., and Pedersen, D. B. “Multi-material additive manufacturing of steels using laser powder bed fusion.” In euspen’s 19thInternational Conference & Exhibition, The European Society for Precision Engineeringand Nanotechnology, pp. 240–243. 2, 15, 18, 57

[16] Chivel, Y., 2016. “New approach to multi-material processing in selective laser melting.”Physics Procedia, 83, pp. 891–898. 2, 15, 57

[17] Wei, C., Li, L., Zhang, X. J., and Chueh, Y. H., 2018. “3d printing of multiple metallicmaterials via modified selective laser melting.” Cirp Annals-Manufacturing Technology,67(1), pp. 245–248 Gm8iy Times Cited:8 Cited References Count:19. 2, 16, 19, 57

[18] Zhang, X. J., Wei, C., Chueh, Y. H., and Li, L., 2019. “An integrated dual ultrasonic selectivepowder dispensing platform for three-dimensional printing of multiple material metal/glassobjects in selective laser melting.” Journal of Manufacturing Science and Engineering-Transactions of the Asme, 141(1) Hc7cr Times Cited:0 Cited References Count:23. 2, 16,19

[19] Ukai, S., Harada, M., Okada, H., Inoue, M., Nomura, S., Shikakura, S., Asabe, K., Nishida,T., and Fujiwara, M., 1993. “Alloying design of oxide dispersion strengthened ferritic steel

89

for long life fbrs core materials.” Journal of Nuclear Materials, 204, pp. 65–73 Ma980Times Cited:234 Cited References Count:7. 4, 13, 57, 76

[20] El-Genk, M. S., and Tournier, J.-M., 2005. “A review of refractory metal alloys and me-chanically alloyed-oxide dispersion strengthened steels for space nuclear power systems.”Journal of Nuclear Materials, 340(1), pp. 93–112 909cc Times Cited:195 Cited ReferencesCount:62. 4

[21] Klueh, R. L., Shingledecker, J. P., Swindeman, R. W., and Hoelzer, D. T., 2005. “Oxidedispersion-strengthened steels: A comparison of some commercial and experimental al-loys.” Journal of Nuclear Materials, 341(2-3), pp. 103–114 934dh Times Cited:341 CitedReferences Count:25. 4, 13, 35, 57, 76

[22] Walker, J. C., Berggreen, K. M., Jones, A. R., and Sutcliffe, C. J., 2009. “Fabrication offe-cr-al oxide dispersion strengthened pm2000 alloy using selective laser melting.” Ad-vanced Engineering Materials, 11(7), pp. 541–546 484zs Times Cited:29 Cited ReferencesCount:27. 4, 13, 57, 76

[23] Kimura, A., Kasada, R., Iwata, N., Kishimoto, H., Zhang, C. H., Isselin, J., Dou, P., Lee,J. H., Muthukumar, N., Okuda, T., Inoue, M., Ukai, S., Ohnuki, S., Fujisawa, T., and Abe,T. F., 2011. “Development of al added high-cr ods steels for fuel cladding of next generationnuclear systems.” Journal of Nuclear Materials, 417(1-3), pp. 176–179 852um TimesCited:162 Cited References Count:24. 4

[24] Boegelein, T., Dryepondt, S. N., Pandey, A., Dawson, K., and Tatlock, G. J., 2015.“Mechanical response and deformation mechanisms of ferritic oxide dispersion strength-ened steel structures produced by selective laser melting.” Acta Materialia, 87, pp. 201–215 Compilation and indexing terms, Copyright 2019 Elsevier Inc. 20150500479234Anisotropic behaviours Electron back scatter diffraction Mechanical and physical proper-ties Oxide dispersion strengthened alloys Oxide dispersion strengthened ferritic steels Ox-ide dispersion strengthened steel Rapid prototyping techniques Selective laser melting. 4,13, 41

[25] Shakerin, S., Hadadzadeh, A., Amirkhiz, B. S., Shamsdini, S., Li, J., and Mohammadi,M., 2019. “Additive manufacturing of maraging steel-h13 bimetals using laser powder bedfusion technique.” Additive Manufacturing, 29, p. 100797. 4

[26] Davis, J. R., ed., 1998. Metals Handbook Desk Edition. ASM International. 4, 13

[27] Yap, C. Y., Chua, C. K., Dong, Z. L., Liu, Z. H., Zhang, D. Q., Loh, L. E., and Sing, S. L.,2015. “Review of selective laser melting: Materials and applications.” Applied PhysicsReviews, 2(4), p. 041101 Dc0kr Times Cited:420 Cited References Count:222. 4, 13

[28] Bandyopadhyay, A., and Heer, B., 2018. “Additive manufacturing of multi-material struc-tures.” Materials Science and Engineering R: Reports, 129, pp. 1–16 Compilation andindexing terms, Copyright 2019 Elsevier Inc. 20181705046039 3-D printing Economic ben-efits Environmental adaptation Multi-material structure Multiple materials Performance im-provements Research and development Single components. 4, 14, 15

90

[29] Heiple, C. R., and Roper, J. R., 1990. CHAPTER 1 - The Geometry of Gas Tungsten Arc,Gas Metal Arc, and Submerged Arc Weld Beads., Vol. 8 of Materials Processing: Theoryand Practices Elsevier, pp. 1–34. 4, 13, 57

[30] Lu, S., Fujii, H., Sugiyama, H., Tanaka, M., and Nogi, K., 2002. “Weld penetration andmarangoni convection with oxide fluxes in gta welding.” MATERIALS TRANSACTIONS,43(11), pp. 2926–2931 628bb Times Cited:67 Cited References Count:22. 4, 13, 57

[31] Gan, Z., Yu, G., He, X., and Li, S., 2017. “Surface-active element transport and its effecton liquid metal flow in laser-assisted additive manufacturing.” International Communica-tions in Heat and Mass Transfer, 86, pp. 206–214 Fe4kw Times Cited:31 Cited ReferencesCount:29. 4, 13, 52, 57

[32] Milewski, J. O., 2017. On the Road to AM. Springer Series in Materials Science. SpringerInternational Publishing, Cham, book section Chapter 3, pp. 35–48. 6, 9, 10, 18, 37, 56

[33] Brueckner, F., Riede, M., Muller, M., Marquardt, F., Willner, R., Seidel, A., Lopez, E.,Leyens, C., and Beyer, E., 2018. “Enhanced manufacturing possibilities using multi-materials in laser metal deposition.” J. Laser Appl, 30(3) Gs7om Times Cited:10 CitedReferences Count:17. 7, 14

[34] Milewski, J. O., 2017. Additive Manufacturing Metal, the Art of the Possible. Springer Se-ries in Materials Science. Springer International Publishing, book section Chapter 2, pp. 7–33. 6, 8, 9

[35] Gibson, I., Rosen, D., and Stucker, B., 2015. Directed Energy Deposition Processes.Springer New York, book section Chapter 10, pp. 245–268. 6, 7, 8

[36] Herzog, D., Seyda, V., Wycisk, E., and Emmelmann, C., 2016. “Additive manufacturingof metals.” Acta Materialia, 117, pp. 371–392 Dv5xz Times Cited:743 Cited ReferencesCount:150. 7

[37] Gibson, I., Rosen, D., and Stucker, B., 2015. Powder Bed Fusion Processes. Springer NewYork, New York, NY, book section Chapter 5, pp. 107–145. 8, 9, 10, 19

[38] Leung, C. L. A., Marussi, S., Atwood, R. C., Towrie, M., Withers, P. J., and Lee, P. D., 2018.“In situ x-ray imaging of defect and molten pool dynamics in laser additive manufacturing.”Nature Communications, 9(1), p. 1355 Leung, Chu Lun Alex Marussi, Sebastian Atwood,Robert C Towrie, Michael Withers, Philip J Lee, Peter D eng Research Support, Non-U.S.Gov’t England Nat Commun. 2018 Apr 10;9(1):1355. doi: 10.1038/s41467-018-03734-7.9, 12

[39] Koutny, D., Palousek, D., Pantelejev, L., Hoeller, C., Pichler, R., Tesicky, L., and Kaiser, J.,2018. “Influence of scanning strategies on processing of aluminum alloy en aw 2618 usingselective laser melting.” Materials, 11(2) Fz4az Times Cited:10 Cited References Count:29.9, 37

[40] Gibson, I., Rosen, D., and Stucker, B., 2015. Binder Jetting. Springer New York, booksection Chapter 8, pp. 205–218. 9, 10

91

[41] Ziaee, M., and Crane, N. B., 2019. “Binder jetting: A review of process, materials, andmethods.” Additive Manufacturing, 28, pp. 781–801 Jh3ms Times Cited:16 Cited Refer-ences Count:405. 9, 14

[42] Clymer, D. R., Cagan, J., and Beuth, J., 2017. “Power–velocity process design charts forpowder bed additive manufacturing.” Journal of Mechanical Design, 139(10) Fp0lw TimesCited:11 Cited References Count:34. 10, 11, 12

[43] Heeling, T., and Wegener, K., 2018. “The effect of multi-beam strategies on selective lasermelting of stainless steel 316l.” Additive Manufacturing, 22, pp. 334–342 Gr0ay TimesCited:4 Cited References Count:27. 10

[44] Gockel, J., Sheridan, L., Koerper, B., and Whip, B., 2019. “The influence of additivemanufacturing processing parameters on surface roughness and fatigue life.” InternationalJournal of Fatigue, 124, pp. 380–388 Ic4mo Times Cited:30 Cited References Count:27.10, 37

[45] Koepke, J. R., 2019. “The influence of process variables on physical and mechanical prop-erties in laser powder bed fusion.” Thesis. 10

[46] Whip, B., Sheridan, L., and Gockel, J., 2019. “The effect of primary processing parameterson surface roughness in laser powder bed additive manufacturing.” International Journalof Advanced Manufacturing Technology, 103(9-12), pp. 4411–4422 Im5ad Times Cited:11Cited References Count:25. 10, 37, 43

[47] Meier, C., Penny, R. W., Zou, Y., Gibbs, J. S., and Hart, A. J., 2017. “Thermophysical phe-nomena in metal additive manufacturing by selective laser melting: Fundamentals, model-ing, simulation, and experimentation.” Annual Review of Heat Transfer, 20(1), pp. 241–316.10, 11

[48] Tan, J. L., Tang, C., and Wong, C. H., 2018. “A computational study on porosity evolutionin parts produced by selective laser melting.” Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, 49a(8), pp. 3663–3673 Gl1wz Times Cited:5Cited References Count:53. 11, 37

[49] Yadroitsev, I., and Smurov, I., 2011. “Surface morphology in selective laser melting of metalpowders.” Physics Procedia, 12, pp. 264–270. 11, 13

[50] Aversa, A., Lorusso, M., Trevisan, F., Ambrosio, E., Calignano, F., Manfredi, D., Biamino,S., Fino, P., Lombardi, M., and Pavese, M., 2017. “Effect of process and post-processconditions on the mechanical properties of an a357 alloy produced via laser powder bedfusion.” Metals, 7(2), p. 68 Ep3by Times Cited:23 Cited References Count:30. 11

[51] Amirjan, M., and Sakiani, H., 2019. “Effect of scanning strategy and speed on the mi-crostructure and mechanical properties of selective laser melted in718 nickel-based su-peralloy.” The International Journal of Advanced Manufacturing Technology, 103(5-8),pp. 1769–1780 Ik5la Times Cited:1 Cited References Count:22. 11, 37

[52] Choo, H., Sham, K.-L., Bohling, J., Ngo, A., Xiao, X., Ren, Y., Depond, P. J., Matthews,M. J., and Garlea, E., 2019. “Effect of laser power on defect, texture, and microstructure of a

92

laser powder bed fusion processed 316l stainless steel.” Materials & Design, 164, p. 107534Hh0ty Times Cited:19 Cited References Count:50. 11

[53] He, Y., Zhong, M., Beuth, J., and Webler, B., 2020. “A study of microstructure and crackingbehavior of h13 tool steel produced by laser powder bed fusion using single-tracks, multi-track pads, and 3d cubes.” Journal of Materials Processing Technology, 286, p. 116802. 11,37, 38

[54] Gaikwad, A., Giera, B., Guss, G. M., Forien, J.-B., Matthews, M. J., and Rao, P., 2020.“Heterogeneous sensing and scientific machine learning for quality assurance in laser pow-der bed fusion – a single-track study.” Additive Manufacturing, p. 101659. 12, 13, 37

[55] Snyder, J. C., and Thole, K. A., 2020. “Understanding laser powder bed fusion surfaceroughness.” Journal of Manufacturing Science and Engineering, 142(7), pp. 1–37. 12, 13,33, 34, 37, 38, 49

[56] Yadroitsev, I., Gusarov, A., Yadroitsava, I., and Smurov, I., 2010. “Single track formationin selective laser melting of metal powders.” Journal of Materials Processing Technology,210(12), pp. 1624–1631. 12, 38, 39, 71

[57] Gunenthiram, V., Peyre, P., Schneider, M., Dal, M., Coste, F., and Fabbro, R., 2017. “Analy-sis of laser–melt pool–powder bed interaction during the selective laser melting of a stainlesssteel.” Journal of Laser Applications, 29(2), p. 022303. 12, 38

[58] Rubenchik, A. M., King, W. E., and Wu, S. S., 2018. “Scaling laws for the additive man-ufacturing.” Journal of Materials Processing Technology, 257, pp. 234–243 Ge4ch TimesCited:10 Cited References Count:41. 12, 13

[59] Prashanth, K. G., Scudino, S., Maity, T., Das, J., and Eckert, J., 2017. “Is the energy densitya reliable parameter for materials synthesis by selective laser melting?.” Materials ResearchLetters, 5(6), pp. 386–390. 13, 37

[60] Vastola, G., Pei, Q. X., and Zhang, Y. W., 2018. “Predictive model for porosity in powder-bed fusion additive manufacturing at high beam energy regime.” Additive Manufacturing,22, pp. 817–822 Gr0ay Times Cited:3 Cited References Count:45. 13

[61] Carter, L. N., Wang, X., Read, N., Khan, R., Aristizabal, M., Essa, K., and Attallah, M. M.,2016. “Process optimisation of selective laser melting using energy density model for nickelbased superalloys.” Materials Science and Technology, 32(7), pp. 657–661. 13

[62] Wang, D., Liu, Y., Yang, Y., and Xiao, D., 2016. “Theoretical and experimental studyon surface roughness of 316l stainless steel metal parts obtained through selective lasermelting.” Rapid Prototyping Journal, 22(4), pp. 706–716. 13

[63] Francis, Z. R., 2017. “The effects of laser and electron beam spot size in additive manufac-turing processes.” Thesis. 13, 38

[64] Hann, D. B., Iammi, J., and Folkes, J., 2011. “A simple methodology for predicting laser-weld properties from material and laser parameters.” Journal of Physics D: Applied Physics,44(44), p. 445401 842io Times Cited:58 Cited References Count:17. 13

93

[65] Vecchiato, F. L., De Winton, H., Hooper, P. A., and Wenman, M. R., 2020. “Melt poolmicrostructure and morphology from single exposures in laser powder bed fusion of 316lstainless steel.” Additive Manufacturing, 36, p. 101401. 13, 38

[66] Zhang, C. H., Kimura, A., Kasada, R., Jang, J., Kishimoto, H., and Yang, Y. T., 2011. “Char-acterization of the oxide particles in al-added high-cr ods ferritic steels.” Journal of NuclearMaterials, 417(1-3), pp. 221–224 852um Times Cited:38 Cited References Count:14. 13,57, 76

[67] Zhang, L., 2013. “Nucleation, growth, transport, and entrapment of inclusions during steelcasting.” JOM, 65(9), pp. 1138–1144 207sf Times Cited:51 Cited References Count:87. 13,57

[68] Boley, C. D., Khairallah, S. A., and Rubenchik, A. M., 2015. “Calculation of laser absorp-tion by metal powders in additive manufacturing.” Applied Optics, 54(9), pp. 2477–2482Ce2ic Times Cited:66 Cited References Count:14. 13

[69] Mussatto, A., Groarke, R., A-Hameed, A., Ahad, I. U. I., Vijayaraghavan, R. K., O’Neill,A., McNally, P., Delaure, Y., and Brabazon, D., 2019. “Evaluation via powder metallurgy ofnano-reinforced iron powders developed for selective laser melting applications.” Materials& Design, 182, p. 108046 Jb3ll Times Cited:1 Cited References Count:56. 13

[70] Brueckner, F., Schab, J. C., Marquardt, F., Muller, M., Riede, M., Seidel, A., Lopez, E.,and Leyens, C. “Phenomena in multi-material fabrication using laser metal deposition.”In Laser 3D Manufacturing VI 2019, February 5, 2019 - February 7, 2019, Vol. 10909 ofProceedings of SPIE - The International Society for Optical Engineering, SPIE, pp. The So-ciety of Photo–Optical Instrumentation Engineers (SPIE) Compilation and indexing terms,Copyright 2019 Elsevier Inc. 20192607094967 Conventional manufacturing Functional in-tegration Graded structure Laser metal deposition Material characterizations Material com-bination Material compositions Multi materials. 14, 57

[71] Yoozbashizadeh, M., Yavari, P., and Khoshnevis, B., 2018. “Novel method for additive man-ufacturing of metal-matrix composite by thermal decomposition of salts.” Additive Manu-facturing, 24, pp. 173–182 Hg6cl Times Cited:2 Cited References Count:28. 14

[72] Wei, C., Chueh, Y.-H., Zhang, X., Huang, Y., Chen, Q., and Li, L., 2019. “Easy-to-removecomposite support material and procedure in additive manufacturing of metallic componentsusing multiple material laser-based powder bed fusion.” Journal of Manufacturing Scienceand Engineering, 141(7), p. 1 Ic7cc Times Cited:7 Cited References Count:41. 16, 19

[73] Wei, C., Sun, Z., Chen, Q., Liu, Z., and Li, L., 2019. “Additive manufacturing of horizontaland 3d functionally graded 316l/cu10sn components via multiple material selective lasermelting.” Journal of Manufacturing Science and Engineering-Transactions of the Asme,141(8) Ih0yo Times Cited:0 Cited References Count:44. 16, 19

[74] Chueh, Y.-H., Wei, C., Zhang, X., and Li, L., 2020. “Integrated laser-based powder bedfusion and fused filament fabrication for three-dimensional printing of hybrid metal/polymerobjects.” Additive Manufacturing, 31, p. 100928 Kl9xf Times Cited:5 Cited ReferencesCount:43. 16, 19, 57

94

[75] Wei, C., Gu, H., Zhang, X., Chueh, Y.-H., and Li, L., 2020. “Hybrid ultrasonic and mini-motor vibration-induced irregularly shaped powder delivery for multiple materials additivemanufacturing.” Additive Manufacturing, 33, p. 101138 Lw6sn Times Cited:0 Cited Refer-ences Count:34. 19, 57

[76] Greenwood, T. E., 2020. “Silicone 3d printing processes for fabricating synthetic, self-oscillating vocal fold models.” Thesis. 22

[77] Greenwood, T. E., Hatch, S. E., Colton, M. B., and Thomson, S. L., 2020. “3d printinglow-stiffness silicone within a curable support matrix.” Additive Manufacturing, p. 101681.22

[78] Romero, R. G. T., Colton, M. B., and Thomson, S. L., 2020. “3d-printed synthetic vocalfold models.” Journal of Voice. 22, 23

[79] Colton, T., Liechty, J., McLean, A., and Crane, N. “Influence of drop velocity and dropletspacing on the equilibrium saturation level in binder jetting.” In SFF Symposium. 24, 25,34, 59

[80] Colton, T., and Crane, N. B., 2021. “Influence of droplet velocity, spacing, and inter-arrivaltime on line formation and saturation in binder jet additive manufacturing.” Additive Man-ufacturing, 37, p. 101711. 24, 25, 34, 59

[81] Fan, K. M., Wong, K. W., Cheung, W. L., and Gibson, I., 2007. “Reflectance and transmit-tance of trueformtm powder and its composites to co2 laser.” Rapid Prototyping Journal,13(3), pp. 175–181. 27

[82] Balluffi, R. W., Allen, S. M., and Carter, W. C., 2005. Kinetics of materials. John Wiley &Sons. 30

[83] Weon, B. M., and Je, J. H., 2010. “Capillary force repels coffee-ring effect.” PhysicalReview E, 82(1). 30

[84] Mampallil, D., and Eral, H. B., 2018. “A review on suppression and utilization of the coffee-ring effect.” Advances in Colloid and Interface Science, 252, pp. 38–54. 30

[85] Gu, D., and Shen, Y., 2009. “Balling phenomena in direct laser sintering of stainless steelpowder: Metallurgical mechanisms and control methods.” Materials & Design, 30(8),pp. 2903–2910 467ny Times Cited:227 Cited References Count:33. 33, 34, 37, 38, 39

[86] Li, R., Liu, J., Shi, Y., Wang, L., and Jiang, W., 2012. “Balling behavior of stainless steeland nickel powder during selective laser melting process.” The International Journal ofAdvanced Manufacturing Technology, 59(9-12), pp. 1025–1035 916kv Times Cited:197Cited References Count:18. 33, 34, 37, 39

[87] Colton, T., 2021. “The impact of inkjet parameters and environmental conditions in binderjetting additive manufacturing.” Thesis. 34, 50

[88] Riede, M., Knoll, M., Wilsnack, C., Gruber, S., Cubillo, A. A., Melzer, C., Brandao, A.,Pambaguian, L., Seidel, A., Lopez, E., Brueckner, F., and Leyens, C., 2019. “Material

95

characterization of aisi 316l flexure pivot bearings fabricated by additive manufacturing.”Materials, 12(15) It0zx Times Cited:0 Cited References Count:31. 37, 56

[89] Yadroitsev, I., Krakhmalev, P., Yadroitsava, I., and Du Plessis, A., 2018. “Qualification ofti6al4v eli alloy produced by laser powder bed fusion for biomedical applications.” JOM,70(3), pp. 372–377 Fw4zj Times Cited:17 Cited References Count:10. 37, 56

[90] Pollock, T. M., 2016. “Alloy design for aircraft engines.” Nature Materials, 15(8), pp. 809–815 Pollock, Tresa M eng Research Support, U.S. Gov’t, Non-P.H.S. England Nat Mater.2016 Jul 22;15(8):809-15. doi: 10.1038/nmat4709. 37

[91] Kamath, C., El-Dasher, B., Gallegos, G. F., King, W. E., and Sisto, A., 2014. “Density ofadditively-manufactured, 316l ss parts using laser powder-bed fusion at powers up to 400w.” The International Journal of Advanced Manufacturing Technology, 74(1-4), pp. 65–78An6an Times Cited:140 Cited References Count:24. 37

[92] Narra, S. P., Scime, L., and Beuth, J., 2018. “Integrated control of melt pool geometryand microstructure in laser powder bed fusion of alsi10mg.” Metallurgical and MaterialsTransactions a-Physical Metallurgy and Materials Science, 49a(10), pp. 5097–5106 Gs3fnTimes Cited:0 Cited References Count:30. 37

[93] Tang, M., Pistorius, P. C., and Beuth, J. L., 2017. “Prediction of lack-of-fusion porosity forpowder bed fusion.” Additive Manufacturing, 14, pp. 39–48 Vh8vq Times Cited:84 CitedReferences Count:35. 37

[94] Martin, A. A., Calta, N. P., Khairallah, S. A., Wang, J., Depond, P. J., Fong, A. Y., Thampy,V., Guss, G. M., Kiss, A. M., Stone, K. H., Tassone, C. J., Weker, J. N., Toney, M. F., vanBuuren, T., and Matthews, M. J., 2019. “Dynamics of pore formation during laser powderbed fusion additive manufacturing.” Nature Communications, 10(1), p. 1987 Hv9ws TimesCited:0 Cited References Count:48. 37

[95] Aversa, A., Moshiri, M., Librera, E., Hadi, M., Marchese, G., Manfredi, D., Lorusso, M.,Calignano, F., Biamino, S., Lombardi, M., and Pavese, M., 2018. “Single scan track anal-yses on aluminium based powders.” Journal of Materials Processing Technology, 255,pp. 17–25. 38, 39

[96] King, W. E., Barth, H. D., Castillo, V. M., Gallegos, G. F., Gibbs, J. W., Hahn, D. E., Ka-math, C., and Rubenchik, A. M., 2014. “Observation of keyhole-mode laser melting in laserpowder-bed fusion additive manufacturing.” Journal of Materials Processing Technology,214(12), pp. 2915–2925 Ap7gj Times Cited:298 Cited References Count:19. 38, 70

[97] Kiss, A. M., Fong, A. Y., Calta, N. P., Thampy, V., Martin, A. A., Depond, P. J., Wang, J.,Matthews, M. J., Ott, R. T., Tassone, C. J., Stone, K. H., Kramer, M. J., Van Buuren, A.,Toney, M. F., and Nelson Weker, J., 2019. “Laser-induced keyhole defect dynamics duringmetal additive manufacturing.” Advanced Engineering Materials, 21(10), p. 1900455 Jh7wcTimes Cited:0 Cited References Count:41. 38, 70

[98] Tammas-Williams, S., and Todd, I., 2017. “Design for additive manufacturing with site-specific properties in metals and alloys.” Scripta Materialia, 135, pp. 105–110. 39

96

[99] Marchese, G., Aversa, A., Lorusso, M., Manfredi, D., Calignano, F., Lombardi, M., Bi-amino, S., and Pavese, M., 2018. “Development and characterisation of aluminium matrixnanocomposites alsi10mg/mgal2o4 by laser powder bed fusion.” Metals, 8(3), p. 175 Ga8atTimes Cited:1 Cited References Count:33. 39

[100] Du, Z., Chen, H.-C., Tan, M. J., Bi, G., and Chua, C. K., 2020. “Effect of nal2o3 onthe part density and microstructure during the laser-based powder bed fusion of alsi10mgcomposite.” Rapid Prototyping Journal, 26(4), pp. 727–735. 41, 57, 60

[101] Scime, L., and Beuth, J., 2019. “Melt pool geometry and morphology variability for theinconel 718 alloy in a laser powder bed fusion additive manufacturing process.” AdditiveManufacturing, 29, p. 100830. 43

[102] Matsunawa, A., and Ohji, T., 1982. “Role of surface tension in fusion welding (part 1) :Hydrostatic effect.”. 45

[103] Ghosh, S. K., and Saha, P., 2011. “Crack and wear behavior of sic particulate reinforcedaluminium based metal matrix composite fabricated by direct metal laser sintering process.”Materials & Design, 32(1), pp. 139–145 655ig Times Cited:82 Cited References Count:31.57

[104] Ghosh, S. K., Bandyopadhyay, K., and Saha, P., 2014. “Development of an in-situ multi-component reinforced al-based metal matrix composite by direct metal laser sintering tech-nique — optimization of process parameters.” Materials Characterization, 93, pp. 68–78Aj7fw Times Cited:15 Cited References Count:37. 57

[105] Aversa, A., Marchese, G., Lorusso, M., Calignano, F., Biamino, S., Ambrosio, E. P., Man-fredi, D., Fino, P., Lombardi, M., and Pavese, M., 2017. “Microstructural and mechanicalcharacterization of aluminum matrix composites produced by laser powder bed fusion.”Advanced Engineering Materials, 19(11), p. 1700180 Fn3wm Times Cited:16 Cited Refer-ences Count:42. 57

[106] Hussain, M., Mandal, V., Kumar, V., Das, A. K., and Ghosh, S. K., 2017. “Development oftin particulates reinforced ss316 based metal matrix composite by direct metal laser sinteringtechnique and its characterization.” Optics & Laser Technology, 97, pp. 46–59 Ff8rf TimesCited:9 Cited References Count:36. 57, 60

[107] Kim, S. H., Shin, G.-H., Kim, B.-K., Kim, K. T., Yang, D.-Y., Aranas, C., Choi, J.-P.,and Yu, J.-H., 2017. “Thermo-mechanical improvement of inconel 718 using ex situ boronnitride-reinforced composites processed by laser powder bed fusion.” Scientific Reports,7(1), p. 14359 Kim, Sang Hoon Shin, Gi-Hun Kim, Byoung-Kee Kim, Kyung Tae Yang,Dong-Yeol Aranas, Clodualdo Jr Choi, Joon-Phil Yu, Ji-Hun eng Research Support, Non-U.S. Gov’t England Sci Rep. 2017 Oct 30;7(1):14359. doi: 10.1038/s41598-017-14713-1.57

[108] Haslam, M. D., and Raeymaekers, B., 2013. “A composite index to quantify dispersion ofcarbon nanotubes in polymer-based composite materials.” Composites Part B: Engineering,55, pp. 16–21. 66, 108

97

[109] Goldstein, J. I., Newbury, D. E., Michael, J. R., Ritchie, N. W., Scott, J. H. J., and Joy, D. C.,2017. Scanning electron microscopy and X-ray microanalysis. Springer. 68

[110] Liu, Y., Yang, Y., Mai, S., Wang, D., and Song, C., 2015. “Investigation into spatter behaviorduring selective laser melting of aisi 316l stainless steel powder.” Materials & Design, 87,pp. 797–806 Cu8tc Times Cited:125 Cited References Count:38. 70

[111] Bidare, P., Bitharas, I., Ward, R. M., Attallah, M. M., and Moore, A. J., 2018. “Fluid andparticle dynamics in laser powder bed fusion.” Acta Materialia, 142, pp. 107–120 Fn1umTimes Cited:35 Cited References Count:33. 70

[112] Hojjatzadeh, S. M. H., Parab, N. D., Guo, Q., Qu, M., Xiong, L., Zhao, C., Escano, L. I.,Fezzaa, K., Everhart, W., Sun, T., and Chen, L., 2020. “Direct observation of pore forma-tion mechanisms during lpbf additive manufacturing process and high energy density laserwelding.” International Journal of Machine Tools and Manufacture, 153, p. 103555 Ls1usTimes Cited:2 Cited References Count:26. 70

[113] Khairallah, S. A., Anderson, A. T., Rubenchik, A., and King, W. E., 2016. “Laser powder-bed fusion additive manufacturing: Physics of complex melt flow and formation mecha-nisms of pores, spatter, and denudation zones.” Acta Materialia, 108, pp. 36–45. 71

[114] Hirata, A., Fujita, T., Wen, Y. R., Schneibel, J. H., Liu, C. T., and Chen, M. W., 2011.“Atomic structure of nanoclusters in oxide-dispersion-strengthened steels.” Nature Materi-als, 10(12), pp. 922–926 Hirata, A Fujita, T Wen, Y R Schneibel, J H Liu, C T Chen, M Weng Letter England Nat Mater. 2011 Oct 23;10(12):922-6. doi: 10.1038/nmat3150. 74

98

APPENDIX A. LACK OF MIXING BETWEEN BUILT MATERIAL AND BUILD PLATE

According to the EDX maps (Figure A.1) and quantitative composition values taken at

different depths into the sample (Figure 2), the Nickel and Chromium (indicative of SS 316L

compared to 1018 mild steel) are uniformly present until a depth of 450-500 µm. This is about

the same depth as the built foundations were designed to be. According to this data, any melt pool

less than about 450 µm deep should be unaffected by the material of the build plate and can be

assumed to be a good representation of a SS 316L environment.

Figure A.1: SEM image (top-left) and EDX maps of Iron (top-right), Nickel (bottom-left), andChromium (bottom-right). Note that the Ni and Cr content drop off dramatically at about 500 µminto the part and Fe content increases substantially in the same region.

99

Figure A.2: Amount of Cr and Ni at depths from surface of built platform. Platform was designedto be 500 µm tall. Dotted lines signify standard acceptable composition range for Cr and Ni in SS316L.

100

APPENDIX B. PARAMETER SETS USED IN CHAPTER ??

Table B.1: Parameter sets used for single-track laser scans. Variable parameters are laser power,laser spot size, and scan speed. Calculated areal energy density is based on the basic

processing parameters.

Position Change from Laser Laser Spot Scan Speed Areal Energy Density

Original Power (W ) Size (µm) (mm/s) P/(vdS) (J/mm2)

1 P-6 190 130 1350 1.083

2 P-5 220 130 1350 1.254

3 P-4 250 130 1350 1.425

4 S+2 370 190 1350 1.442

5 P-1,S+1 340 160 1350 1.574

6 P-3 280 130 1350 1.595

7 S+1 370 160 1350 1.713

8 V+4 370 130 1650 1.725

9 P-2 310 130 1350 1.766

10 V+3 370 130 1575 1.807

11 V+2 370 130 1500 1.897

12 P-1 340 130 1350 1.937

13 V+1 370 130 1425 1.997

14 P-1,V-1 340 130 1275 2.051

15 Original 370 130 1350 2.108

16 P+1,V+1 395 130 1425 2.132

17 V-1 370 130 1275 2.232

18 P+1 395 130 1350 2.251

19 V-2 370 130 1200 2.372

Continued on next page

101

Table B.1 – continued from previous page

Position Change from Laser Laser Spot Scan Speed Areal Energy Density

Original Power (W ) Size (µm) (mm/s) P/(vdS) (J/mm2)

20 P-1,S-1 340 100 1350 2.519

21 V-3 370 130 1125 2.53

22 S-1,V+1 370 100 1425 2.596

23 V-4 370 130 1050 2.711

24 S-1 370 100 1350 2.741

25 V-5 370 130 975 2.919

26 S-1,V-2 370 100 1200 3.083

27 V-6 370 130 900 3.162

28 S-1,V-3 370 100 1125 3.289

29 V-7 370 130 825 3.45

30 P+1,V-7 395 130 825 3.683

31 V-8 370 130 750 3.795

32 S-1,V-5 370 100 975 3.795

33 S-2 370 70 1350 3.915

34 V-9 370 130 675 4.217

35 S-1,V-7 370 100 825 4.485

36 V-10 370 130 600 4.744

37 P+1,V-10 395 130 600 5.064

38 V-11 370 130 525 5.421

39 P+1,V-11 395 130 525 5.788

40 V-12 370 130 450 6.325

41 S-3 370 50 1350 5.481

42 V-13 370 130 375 7.59

43 P+1,V-13 395 130 375 8.103

44 P+1,S-1,V-12 395 100 450 8.778

45 V-14 370 130 300 9.487

Continued on next page

102

Table B.1 – continued from previous page

Position Change from Laser Laser Spot Scan Speed Areal Energy Density

Original Power (W ) Size (µm) (mm/s) P/(vdS) (J/mm2)

46 S-1,V-13 370 100 375 9.867

47 P+1,S-1,V-13 395 100 375 10.533

48 V-15 370 130 225 12.65

49 P+1,V-15 395 130 225 13.504

50 S-1,V-15 370 100 225 16.444

51 P+1,S-1,V-15 395 100 225 17.556

52 V-16 370 130 150 18.974

53 P+1,V-16 395 130 150 20.256

54 S-1,V-16 370 100 150 24.667

103

APPENDIX C. DERIVATION OF ZIRCONIA FRACTION VALUE FROM ZIRCONIUMFRACTION VALUE FOR A GIVEN PIXEL

Each pixel of data recorded using EDX reports the amount of Zirconium measured com-

pared to the measured amounts of the main elemental components of SS 316L (Fe, Ni, Cr, Mn,

and Mo) as a mass fraction, fZr. This section explains the steps to calculate the mass fraction of

zirconia (FZrO2) in the volume represented by a given pixel based on that initial mass fraction of

zirconium ( fZr). That mass fraction of zirconia is defined as

FZrO2 =mZrO2

mtot(C.1)

where mZrO2 is the mass of zirconia in a pixel volume and mtot is the total mass of that same pixel

volume. The mtot of the pixel volume is further defined as

mtot = mZrO2 +mSS (C.2)

where mSS is the mass of stainless steel in the pixel volume. Equation C.2 assumes that the total

mass of the pixel is made up solely of zirconia and stainless steel. This can then be used in Equation

C.1 to get

FZrO2 =mZrO2

mZrO2 +mSS. (C.3)

The ratio of the mass of zirconium in the pixel volume to the mass of zirconia in the same

volume can be stated asmZr

mZrO2

=MZr

MZrO2

(C.4)

with mZr being the mass of zirconium in the pixel volume, MZr being the molar mass of zirconium,

and MZrO2 being the molar mass of zirconia (ZrO2). This comparison is based on the assumption

that all of the zirconium detected is in the form of zirconia and that the oxygen present in the

104

zirconia is not part of the initial EDX measurements. The mZrO2 can then be isolated to obtain

mZrO2 = mZrMZrO2

MZr. (C.5)

Similarly, the ratio of the mass of zirconium in the pixel volume to the mass of stainless steel in

the pixel volume can be expressed asmZr

mSS=

fZr

fSS(C.6)

where fSS is the mass fraction of stainless steel measured using EDX. mSS can then be isolated as

mSS = mZrfSS

fZr. (C.7)

Since the mass fractions ( fSS and fZr) make up the entirety of the measured composition, fSS can

be expressed as

fSS = 1− fZr (C.8)

Equations C.5 and C.7 can be used in Equation C.3 to get

FZrO2 =

(mZr

MZrO2MZr

)(

mZrMZrO2

MZr

)+(

mZrfSSfZr

) (C.9)

which can be simplified to

FZrO2 =MZrO2 fZr

MZrO2 fZr +MZr fSS. (C.10)

Equation C.8 can then be used in Equation C.10 to obtain

FZrO2 =MZrO2 fZr

MZrO2 fZr +MZr (1− fZr). (C.11)

Equation C.11 yields the value for the fraction of the mass represented by the pixel that is made up

of zirconia as opposed to stainless steel. This is done for each pixel of EDX data.

105

APPENDIX D. DERIVATION OF PREDICTED MELT POOL CONCENTRATIONS

The predicted mass fraction of zirconia in the melt pool (PZrO2) can be calculated by

PZrO2 =mZrO2

mMP(D.1)

where mZrO2 is the total mass of zirconia in the melt pool and mMP is the total mass of the melt

pool. The mass of the zirconia can be obtained from

mZrO2 = ϕZrO2(wMPl) (D.2)

where ϕZrO2 is the area density of zirconia deposited (known), wMP is the measured width of the

melt pool, and l is a given length of the melt pool for which the volume and zirconia content are

consistent. The total mass of the melt pool, mMP, from Equation D.1 can be calculated as

mMP = mZrO2 +mSS (D.3)

where mSS is the mass of stainless steel in the melt pool, which can be calculated from

mSS =VSS ∗ρSS (D.4)

where ρSS is the density of stainless steel and VSS is the volume of stainless steel in the melt pool.

This volume can be defined as

VSS =VMP−VZrO2 (D.5)

with VMP being the volume of the melt pool and VZrO2 being the volume of zirconia in the melt

pool. The volume of the melt pool can be calculated with

VMP = AMPl (D.6)

106

where AMP is the measured cross-sectional area of the melt pool, and the volume of the zirconia

can be calculated using

VZrO2 =mZrO2

ρZrO2

(D.7)

where ρZrO2 is the density of zirconia. Equations D.7 and D.6 can be used in Equation D.5 to get

VSS = AMPl− mZrO2

ρZrO2

(D.8)

where Equation D.2 can be input to get

VSS = AMPl− ϕZrO2(wMPl)ρZrO2

(D.9)

This can go back into Equation D.4 to get

mSS = ρSS

(AMPl− ϕZrO2(wMPl)

ρZrO2

)(D.10)

Equations D.10 and D.2 can both go into Equation D.3 to make

mMP = ϕZrO2(wMPl)+ρSS

(AMPl− ϕZrO2(wMPl)

ρZrO2

)(D.11)

which can be put into Equation D.1 along with Equation D.2 to get

PZrO2 =ϕZrO2(wMPl)

ϕZrO2(wMPl)+ρSS

(AMPl− ϕZrO2(wMPl)

ρZrO2

) (D.12)

which can be simplified to

PZrO2 =ϕZrO2wMP

ϕZrO2wMP +ρSS

ρZrO2(AMPρZrO2−ϕZrO2wMP)

(D.13)

and further simplified to

PZrO2 =ϕZrO2wMPρZrO2

ϕZrO2wMPρZrO2 +ρSS (AMPρZrO2−ϕZrO2wMP)(D.14)

107

APPENDIX E. IMPLEMENTATION OF BOX SEPARATION FOR DINDEX CALCU-LATIONS

The dIndex is a metric that quantifies the dispersion of the particles throughout an area.

It uses a quadrat method to compare the density of the included particles over different sections

throughout the region of interest. Haslam and Raeymaekers [108] claimed that the quadrat to

particle size ratio must be over 1000 to produce reliable results. The zirconia particles used in

Chapter 5 are around 100 nm and are generally spherical. The pixel side length of the EDX data is

about 2.3 µm. Comparatively, the pixel area is about 681 times the size of the cross-sectional area

of a zirconia particle. However, the area of a quadrat including 4 pixels (2 pixels square) is about

2725 times the size of the zirconia particle cross section. To show the way this difference affects

the resulting dIndex values, values calculated using quadrat sizes of single pixels and 2x2 pixels

are both represented in Chapter 5.

For both the single-pixel and 2x2 pixel quadrat sizes, only pixels that were inside the melt

pool were used for the dIndex calculations. With single-pixel quadrats this is fairly straightforward.

However, if a 2x2 pixel quadrat lays on the edge of the melt pool, it may have some pixels in the

melt pool and some that are not included. In these cases, the overall mass fraction for that quadrat

is calculated using only the pixel values that reside inside the melt pool. These quadrat values are

then weighted when being used to calculate the final s(b) and b values by the fraction of pixels

in the quadrat that are included in the calculations. Calculations were made using MATLAB as

shown in Appendix G.

108

APPENDIX F. CUMULATIVE DISTRIBUTION GRAPHS

The figures in this section of cumulative distribution graphs for individual parameter sets

compared over multiple input dopant amounts and one graph of a single dopant amount compared

over various processing parameter sets. Each figure contains both a full view of the data as well as

a zoomed-in view focused on the area immediately surrounding the overall calculated noise level

from the EDX analysis method.

109

AverageNoise

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 7

0 wt%

4.8 wt%

9.1 wt%

17.2 wt%

AverageNoise

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

0 3 6 9 12

15

18

21

24

27

30

33

36

39

42

45

48

51

54

57

60

63

66

69

72

75

78

81

84

87

90

93

96

99

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 7

0 wt%

4.8 wt%

9.1 wt%

17.2 wt%

Figure F.1: Cumulative distributions of zirconia content in the melt pool by pixel value in wt%.Represented samples were created using parameter set 7 (P = 370 W , v = 675 mm/s, and dS= 130 µm). Four unique samples are measured for each of the four input dopant levels. Thisillustrates the relative contribution to the overall zirconia content amount by pixels of a givenvalue. For most of the samples, the majority of the zirconia is contained in pixels with large valuesindicating grouping or agglomeration of the zirconia and poor overall dispersion. Top: Full viewof cumulative distribtuions. Bottom: Zoomed-in view of region where the multiple distributionsbegin to separate.

110

AverageNoise

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 8

0 wt%

4.8 wt%

9.1 wt%

17.2 wt%

AverageNoise

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 8

0 wt%

4.8 wt%

9.1 wt%

17.2 wt%

Figure F.2: Cumulative distributions of zirconia content in the melt pool by pixel value in wt%.Represented samples were created using parameter set 8 (P = 370 W , v = 525 mm/s, and dS= 130 µm). Four unique samples are measured for each of the four input dopant levels. Thisillustrates the relative contribution to the overall zirconia content amount by pixels of a givenvalue. For most of the samples, the majority of the zirconia is contained in pixels with large valuesindicating grouping or agglomeration of the zirconia and poor overall dispersion. Top: Full viewof cumulative distribtuions. Bottom: Zoomed-in view of region where the multiple distributionsbegin to separate.

111

AverageNoise

0

1

2

3

4

5

6

7

8

0 3 6 9

12

15

18

21

24

27

30

33

36

39

42

45

48

51

54

57

60

63

66

69

72

75

78

81

84

87

90

93

96

99

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 9

0 wt%

4.8 wt%

9.1 wt%

17.2 wt%

AverageNoise

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 9

0 wt%

4.8 wt%

9.1 wt%

17.2 wt%

Figure F.3: Cumulative distributions of zirconia content in the melt pool by pixel value in wt%.Represented samples were created using parameter set 9 (P = 370 W , v = 1350 mm/s, and dS= 50 µm). Four unique samples are measured for each of the four input dopant levels. Thisillustrates the relative contribution to the overall zirconia content amount by pixels of a givenvalue. For most of the samples, the majority of the zirconia is contained in pixels with large valuesindicating grouping or agglomeration of the zirconia and poor overall dispersion. Top: Full viewof cumulative distribtuions. Bottom: Zoomed-in view of region where the multiple distributionsbegin to separate.

112

AverageNoise

0

1

2

3

4

5

6

0 3 6 9

12

15

18

21

24

27

30

33

36

39

42

45

48

51

54

57

60

63

66

69

72

75

78

81

84

87

90

93

96

99

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 10

0B-10a

1B-10a

2B-10a

4B-10a

AverageNoise

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 10

0B-10a

1B-10a

2B-10a

4B-10a

Figure F.4: Cumulative distributions of zirconia content in the melt pool by pixel value in wt%.Represented samples were created using parameter set 10 (P = 370 W , v = 300 mm/s, and dS= 130 µm). Four unique samples are measured for each of the four input dopant levels. Thisillustrates the relative contribution to the overall zirconia content amount by pixels of a givenvalue. For most of the samples, the majority of the zirconia is contained in pixels with large valuesindicating grouping or agglomeration of the zirconia and poor overall dispersion. Top: Full viewof cumulative distribtuions. Bottom: Zoomed-in view of region where the multiple distributionsbegin to separate.

113

AverageNoise

0

1

2

3

4

5

6

0 3 6 9

12

15

18

21

24

27

30

33

36

39

42

45

48

51

54

57

60

63

66

69

72

75

78

81

84

87

90

93

96

99

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 11

0B-11a

1B-11a

2B-11a

4B-11a

AverageNoise

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

Parameter Set 11

0B-11a

1B-11a

2B-11a

4B-11a

Figure F.5: Cumulative distributions of zirconia content in the melt pool by pixel value in wt%.Represented samples were created using parameter set 11 (P = 370 W , v = 225 mm/s, and dS= 130 µm). Four unique samples are measured for each of the four input dopant levels. Thisillustrates the relative contribution to the overall zirconia content amount by pixels of a givenvalue. For most of the samples, the majority of the zirconia is contained in pixels with large valuesindicating grouping or agglomeration of the zirconia and poor overall dispersion. Top: Full viewof cumulative distribtuions. Bottom: Zoomed-in view of region where the multiple distributionsbegin to separate.

114

AverageNoise

0

0.5

1

1.5

2

2.5

3

3.5

0 3 6 9 12

15

18

21

24

27

30

33

36

39

42

45

48

51

54

57

60

63

66

69

72

75

78

81

84

87

90

93

96

99

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

4.8 wt% Zirconia Input

Parameter Set 7

Parameter Set 8

Parameter Set 9

Parameter Set 10

Parameter Set 11

AverageNoise

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

0 3 6 91

21

51

82

12

42

73

03

33

63

94

24

54

85

15

45

76

06

36

66

97

27

57

88

18

48

79

09

39

69

9

Cu

mu

la�

ve w

t% Z

irco

nia

in M

elt

Poo

l

Pixel Value (wt% Zirconia)

4.8 wt% Zirconia Input

Parameter Set 7

Parameter Set 8

Parameter Set 9

Parameter Set 10

Parameter Set 11

Figure F.6: Cumulative distributions of zirconia content in the melt pool by pixel value in wt%.Represented samples were created using various parameter sets with 4.8 wt% input zirconia. Fourunique samples are measured for each of the four input dopant levels. This illustrates the relativecontribution to the overall zirconia content amount by pixels of a given value. For most of thesamples, the majority of the zirconia is contained in pixels with large values indicating groupingor agglomeration of the zirconia and poor overall dispersion. Top: Full view of cumulative distrib-tuions. Bottom: Zoomed-in view of region where the multiple distributions begin to separate.

115

APPENDIX G. MATLAB SCRIPT FOR ZIRCONIA CONTENT AND DISPERSIONCALCULATIONS BASED ON EDX DATA

% Attempt to measure average wt% zirconia over melt pool area using EDX,

% SEM, and etched melt pool images.

% Asks for user input of all files and max value of Zr EDX reading.

% Adjusts images to be comparable in terms of size and orientation.

% User traces melt pool area on transformed version of etched image to

% create mask of melt pool area.

% Mask is used to selectively average the values of the Zr EDX map in the

% melt pool area.

clc; clear all; close all;

% choose file in correct folder (size1, orient1, zoom1)

[filename, pthname] = uigetfile({'*.*','All Files (*.*)'},...

'Select file in next sample folder');

if isequal(filename,0) | | isequal(pthname,0) % check if user pressed cancel

disp('User pressed cancel')

return

end

zzzz = length(pthname);

aaaa = zzzz-2;

bbbb = zzzz-1;

SP = pthname(aaaa:bbbb);

TTT = table();

numba = 0;

startnum = 7;

for num = startnum:11

116

if num >= 10

NUM = sprintf('%u',num);

else

NUM = sprintf('0%u',num);

end

for track = 1:2

if track == 1

TR = 'a';

else

TR = 'b';

end

%% IMPORT images and data

-------------------------------------------------

label = sprintf('%s-%s%s',SP,NUM,TR);

filename = sprintf('%s(1) WeightPct Zr L map.tif',label);

fullFileName = fullfile(pthname,filename);

if ~isfile(fullFileName)

continue

end

% prompt = sprintf('Analyze sample %s? (y/n): ',label);

% sig = input(prompt,'s');

% if sig == 'n'

% continue

% end

numba = numba+1;

fprintf('Sample: %s\r',label)

EDX = imread(fullFileName);

Sample = label;

Samp = filename(1:2);

Param = filename(4:5);

MeltPool = filename(6);

[rows1,cols1,numColors1] = size(EDX);

figure(numba);

117

subplot(3,5,1);

imshow(EDX);

axis on; %

show axis markers (pixels)

caption = sprintf('EDX map of %s', Sample);

title(caption);

set(gcf,'Units','Normalized','WindowState','maximized'); %

enlarge to fullscreen

set(gcf,'Name',['Melt Pool # ',Sample],'NumberTitle','Off'); %

name on title bar

drawnow;

hp = impixelinfo(); %

see values by mousing over image

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

imshow(EDX); %

display original image

axis on; %

show axis markers (pixels)

caption = sprintf('EDX map of %s', Sample);

title(caption);

set(gcf,'Units','Normalized','WindowState','maximized'); %

enlarge to fullscreen

drawnow;

hp = impixelinfo(); %

see values by mousing over image

% EDX map value limits (highest value = x wt%)

EDXmax = 100;% input('What is the max wt% value in the EDX map? ');

filename = sprintf('%s(1) WeightPct Grey map.tif',label);

fullFileName = fullfile(pthname,filename);

SEM1 = imread(fullFileName);

subplot(3,5,2);

imshow(SEM1); %

display original image

118

axis on; %

show axis markers (pixels)

caption = sprintf('SEM image of EDX area');

title(caption);

drawnow;

hp = impixelinfo(); %

see values by mousing over image

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

imshow(SEM1); %

display original image

axis on; %

show axis markers (pixels)

caption = sprintf('SEM image of EDX area');

title(caption);

drawnow;

hp = impixelinfo(); %

see values by mousing over image

filename = sprintf('%s(1) WeightPct RefGrey.tif',label);

fullFileName = fullfile(pthname,filename);

SEM2 = imread(fullFileName);

subplot(3,5,6);

imshow(SEM2); %

display original image

axis on; %

show axis markers (pixels)

caption = sprintf('Full size SEM image');

title(caption);

drawnow;

hp = impixelinfo(); %

see values by mousing over image

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

imshow(SEM2); %

display original image

119

axis on; %

show axis markers (pixels)

caption = sprintf('Full size SEM image');

title(caption);

drawnow;

hp = impixelinfo(); %

see values by mousing over image

SEM2 = imresize(SEM2,[rows1 cols1]);

filename = sprintf('%s etch.jpg',label);

fullFileName = fullfile(pthname,filename);

Etched = imread(fullFileName);

[rows2,cols2,numColors2] = size(Etched);

if numColors2 > 1

Etchgray = rgb2gray(Etched);

else

Etchgray = Etched;

end

subplot(3,5,7);

imshow(Etchgray); %

display original image

axis on; %

show axis markers (pixels)

caption = sprintf('Etched sample');

title(caption);

drawnow;

hp = impixelinfo();

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

imshow(Etchgray); %

display original image

axis on; %

show axis markers (pixels)

caption = sprintf('Etched sample');

title(caption);

120

drawnow;

hp = impixelinfo();

%% Alignment and Melt Pool Mask

UseOld = 'n';

filename = sprintf('%s points.mat',label);

fullFileName = fullfile(pthname,filename);

if isfile(fullFileName)

% UseOld = input('Would you like to use saved transform and mask?

(y/n): ','s');

UseOld = 'y';

end

if UseOld == 'y'

original = SEM2;

distorted = Etchgray;

load(fullFileName,'movingPoints','fixedPoints');

filename = sprintf('%s recovered.tif',label);

fullFileName = fullfile(pthname,filename);

recovered = imread(fullFileName);

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

showMatchedFeatures(distorted,original,movingPoints,fixedPoints);

title('Matching points');

subplot(3,5,11);

showMatchedFeatures(distorted,original,movingPoints,fixedPoints);

match = getframe;

Match = match.cdata;

title('Matching points');

MPimage = imfuse(original,recovered);

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

imshow(MPimage);

title('Overlayed Etched/SEM Images');

subplot(3,5,12);

imshow(MPimage);

title('Overlayed Etched/SEM Images');

121

Etch2 = recovered;

filename = sprintf('%s MeltPool.tif',label);

fullFileName = fullfile(pthname,filename);

MeltPool = imread(fullFileName);

MeltPoolLine = bwboundaries(MeltPool);

xy = MeltPoolLine{1};

x = xy(:,2);

y = xy(:,1);

figure(numba);

subplot(3,5,3);

imshow(MPimage);

hold on;

plot(x,y,'LineWidth',1);

title('Reoriented Etched Image');

EtchJustMP = MPimage;

EtchJustMP(~MeltPool) = 0;

subplot(3,5,4);

imshow(EtchJustMP);

title('Etched Melt Pool');

hold off;

SEMJustMP = SEM2;

SEMJustMP(~MeltPool) = 0;

subplot(3,5,5);

imshow(SEMJustMP);

title('SEM Melt Pool')

EDXonSEMjustMP = EDX;

EDXonSEMjustMP(~MeltPool) = 0;

subplot(3,5,[8,9,10,13,14,15]);

imshow(EDXonSEMjustMP);

122

title('Zirconia in Melt Pool')

else

% MATCH and TRANSFORM optical image to orientation/size of SEM

images -----

signal = 'n';

while signal ~= 'y'

% Manual Point Generation

% (https://www.mathworks.com/help/images/find-image-rotation-

and-scale.html?searchHighlight=find%20image%20rotation%20

and%20scale%20using%20manual%20feature%20matching&s tid=

srchtitle)

original = SEM2;

distorted = Etchgray;

% fprintf('\rPick > 2 matching points on the images, then

close tool.')

% fprintf('\rSelect points in pairs (1 on each image before

moving to next pair).\r')

[movingPoints,fixedPoints] = cpselect(distorted,original,'Wait

',true);

tform = fitgeotrans(movingPoints,fixedPoints,'

nonreflectivesimilarity');

% tform = fitgeotrans(movingPoints,fixedPoints,'affine');

tformInv = invert(tform);

Tinv = tformInv.T;

ss = Tinv(2,1);

sc = Tinv(1,1);

scale recovered = sqrt(ss*ss + sc*sc);

theta recovered = atan2(ss,sc)*180/pi;

Roriginal = imref2d(size(original));

recovered = imwarp(distorted,tform,'OutputView',Roriginal);

filename = sprintf('%s points.mat',label);

fullFileName = fullfile(pthname,filename);

save(fullFileName,'movingPoints','fixedPoints');

123

imwrite(recovered,sprintf('%s\\%s recovered.tif',pthname,label

))

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

showMatchedFeatures(distorted,original,movingPoints,

fixedPoints);

title('Matching points');

subplot(3,5,11);

showMatchedFeatures(distorted,original,movingPoints,

fixedPoints);

match = getframe;

Match = match.cdata;

title('Matching points');

MPimage = imfuse(original,recovered);

subplot(3,5,[3,4,5,8,9,10,13,14,15]);

imshow(MPimage);

title('Overlayed Etched/SEM Images');

subplot(3,5,12);

imshow(MPimage);

title('Overlayed Etched/SEM Images');

Etch2 = recovered;

signal = input('Are the overlayed images aligned correctly? (y

/n): ','s');

end

% CREATE MASK of melt pool area at s1,o1,z1

-------------------------------

% (https://www.mathworks.com/matlabcentral/answers/38547-masking-

out-image-area-using-binary-mask)

while signal ~= 'n'

figure(numba+1);

imshow(MPimage);

set(gcf,'Units','Normalized','WindowState','maximized');

% enlarge to fullscreen

124

title('Outline melt pool. Double click to close loop. Any key

to continue.')

Q = drawfreehand('Multiclick',true);

signal = input('Would you like to redo the outline? (y/n): ','

s');

end

MeltPool = Q.createMask();

imwrite(MeltPool,sprintf('%s\\%s MeltPool.tif',pthname,label))

MeltPoolLine = bwboundaries(MeltPool);

xy = MeltPoolLine{1};

x = xy(:,2);

y = xy(:,1);

figure(numba);

subplot(3,5,3);

imshow(MPimage);

hold on;

plot(x,y,'LineWidth',1);

title('Reoriented Etched Image');

EtchJustMP = MPimage;

EtchJustMP(~MeltPool) = 0;

subplot(3,5,4);

imshow(EtchJustMP);

title('Etched Melt Pool');

hold off;

SEMJustMP = SEM2;

SEMJustMP(~MeltPool) = 0;

subplot(3,5,5);

imshow(SEMJustMP);

title('SEM Melt Pool')

EDXonSEMjustMP = EDX;

EDXonSEMjustMP(~MeltPool) = 0;

125

subplot(3,5,[8,9,10,13,14,15]);

imshow(EDXonSEMjustMP);

title('Zirconia in Melt Pool')

% signal = 'y';

% while signal ~= 'n'

% figure(numba+1);

% imshow(MPimage);

% set(gcf,'Units','Normalized','WindowState','maximized');

% enlarge to fullscreen

% title('Designate noise reference area. Double click to close

loop. Any key to continue.')

% QQ = drawfreehand('Multiclick',true);

% signal = input('Would you like to redo the outline? (y/n):

','s');

% end

% NoiseRef = QQ.createMask();

% imwrite(NoiseRef,sprintf('%s\\%s NoiseRef.tif',pthname,label))

end

%% Calculations

% CALCULATE average wt% of zirconia in melt pool

-------------------------

M Zr = 91.224;% g/mol

M ZrO2 = 123.218;% g/mol

M SS = 56.235;% g/mol

% convert image scale to weight fraction of zirconium

EDX Zr = double(EDX)*EDXmax/254/100;

% convert weight fraction Zr to weight fraction ZrO2

EDX O2 = zeros(rows1,cols1);

for j = 1:rows1

for k = 1:cols1

wt = EDX Zr(j,k);

if wt == 1

EDX O2(j,k) = 1;

126

else

EDX O2(j,k) = ((wt/(1-wt))*(M ZrO2/M Zr))/...

((wt/(1-wt))*(M ZrO2/M Zr)+1);

end

end

end

%%

sizer = 15;

mult = 64;

getem = (2.5*mult+1);%/mult;

numers = 191;

EDXavg7 = filter2(fspecial('average',sizer),EDX O2*mult)*100;

figure(101);

A1sauce = imshow(EDXavg7);

set(gcf,'Units','Normalized','WindowState','maximized'); %

enlarge to fullscreen

colors = jet(numers);

colors = [zeros(64,3);colors];

colors(1:64,:) = 1;

colors(length(colors)+1,:) = 0.875;

C1 = colormap(gca,colors);

caxis([0 4.015625]); colorbar;%mult+numers+1

hold on;

plot(x,y,'m','LineWidth',2);

Newbie = ind2rgb(uint16(EDXavg7),C1);

Newbie1 = Newbie;

for i = 1:length(x)

xer = x(i);

yer = y(i);

Newbie(yer,xer,:) = [1 0 1];

Newbie1(yer,xer,:) = [0 0 0];

end

figure(102);

imshow(Newbie1);

127

set(gcf,'Units','Normalized','WindowState','maximized'); %

enlarge to fullscreen

Title = sprintf('Averaging=%ux%u px',sizer,sizer);% 1.0 < Colors < 2

.5wt%%;

title(Title);

A2sauce = gca;

%%

% separate pixels in MP from those out of MP

MPvals0 = [];

% MPvals5 = [];

% MPvals10 = [];

% MPvals50 = [];

MPvalsFull = zeros(rows1,cols1);

MParray = cell(1,101);

for j = 1:rows1

for k = 1:cols1

if MeltPool(j,k) == 1

MPvals0 = [MPvals0,EDX O2(j,k)]; % create vector for calcs

MPvalsFull(j,k) = EDX O2(j,k); % create array for visual

for jkjk = 1:101

stopnum = (jkjk)/100;

if EDX O2(j,k) < stopnum

MPvals = MParray{jkjk};

MPvals = [MPvals,EDX O2(j,k)];

MParray{jkjk} = MPvals;

end

end

end

end

end

EDXwidth = 592;% um frame width for 350x @ 15

kV

PixSize = EDXwidth/cols1;% um

L = length(double(MPvals0));% 1 # of pixels in melt pool

128

Width = max(sum(MeltPool))*PixSize;% um width is column of MP w/

most pixels

PixArea = PixSizeˆ2;% umˆ2

MParea = L*PixArea;% umˆ2

High = sum(MPvals0>.5);

% Low = sum(MPvals0<.5);

% Avg = mean(MPvals0);

% Med = median(MPvals0);

% SD = std(MPvals0);

% Up1 = Avg+SD;

% Dn1 = Avg-SD;

% Middy = sum(MPvals0>Dn1 & MPvals0<Up1);

Covar = cov(MPvals0);

Covar2 = cov(MPvals0*100);

% Predicted Dopant Content

% mA dop dep = str2double(SP(1))*10/(10ˆ6);% g/mm2

if SP(1) == '0'

mA dop = 0;

elseif SP(1) == '1'

mA dop = 10.113*10ˆ(-6);% g/mm2

elseif SP(1) == '2'

mA dop = 20.124*10ˆ(-6);

elseif SP(1) == '4'

mA dop = 41.577*10ˆ(-6);

end

w mp = Width/1000;% um -> mm

p SS = .00799;% g/mmˆ3

p ZrO2 = .00568;% g/mmˆ3

p Zr = .00649;% g/mmˆ3

z = .05;% mm

PF = .5;% packing fraction

A mp = MParea/(10ˆ6);% mmˆ2

% F predicted = (mA dop dep*w mp)/((mA dop dep*w mp)+(p SS*A mp));

mA SS = z*PF*p SS; % mm*1*g/mmˆ3 = g/mm2

f ZrO2 = mA dop/(mA dop+mA SS); % (g/mm2)/(g/mm2+g/mm2) = 1

129

A dep = (mA dop*w mp/p ZrO2)+(w mp*z*PF); % ((g/mm2)*mm/(g/mm3))+(mm*

mm*1) = mm2

VF dep MP = A dep/A mp; % mm2/mm2 = 1

P ZrO2 = f ZrO2*VF dep MP; % 1*1

M z = mA dop*w mp;

V z = M z/p ZrO2;

V mp = A mp;

V s = V mp-V z;

M s = V s*p SS;

P 2 = M z/(M z+M s);

% Measured/Calculated Dopant Content

m p = (MPvals0*p ZrO2)+((1-MPvals0)*p SS);

m Z = MPvals0.*m p;

mm p = sum(m p);

mm Z = sum(m Z);

F Z = mm Z/mm p;

for iii = 1:101

MPV = 0;

MPV = MParray{iii};

m px = (MPV*p ZrO2)+((1-MPV)*p SS);

m ZrO2 = MPV.*m px;

mm Za(iii) = sum(m ZrO2);

mm px(iii) = sum(m px);

F ZrO2(iii) = sum(m ZrO2)/sum(m px);

RRR(iii) = mm Za(iii)/mm p*100;

end

% F5 = F ZrO2(6);

% F10 = F ZrO2(11);

% F50 = F ZrO2(51);

% if F Z > .018

% noise = 1-(.018/F Z);

% else

% noise = 0;

130

% end

% diff = 1;

% iii = 0;

% while diff > 0

% iii = iii+1;

% diff = RRR(iii)-noise;

% end

% nhigh = RRR(iii-1);

% nlow = RRR(iii);

% ndiff = nhigh-nlow;

% nshort = nhigh-noise;

% nfrac = nshort/ndiff;

% nxval = iii-2+nfrac;

% NXval = sprintf('%.3f',nxval);

figure(49); plot([0:100],RRR);

yline(1.8);

% yline(noise); text(70,noise+.03,'Noise'); xline(nxval); text(nxval

+2,.15,NXval);

title('Fraction of measured zirconia in pixels > X wt%');

% xlim([0 100]); ylim([0 1]);

xlabel('Using pixels > X wt%'); ylabel('Fraction of measured zirconia'

);

Fraccers = gcf;

figure(50); Y = cdfplot(MPvals0);

title('Cumulative Distribution of pixel values');

xlabel('x = Pixel values in wt% zirconia');

ylabel('Cumulative fraction of pixels w/ given value');

CDF = gcf;

Yper = Y.YData;

Xper = Y.XData;

% for 95%

diff = 1;

iii = 1;

while diff > 0

131

iii = iii+1;

diff = .95-Yper(iii);

end

Ydiff = Yper(iii)-.95;

Ygap = Yper(iii)-Yper(iii-1);

Yfrac = Ydiff/Ygap;

Xgap = Xper(iii)-Xper(iii-1);

Xadd = Yfrac*Xgap;

Xcut95 = Xadd+Xper(iii-1);

% for 99%

diff = 1;

iii = 1;

while diff > 0

iii = iii+1;

diff = .99-Yper(iii);

end

Ydiff = Yper(iii)-.99;

Ygap = Yper(iii)-Yper(iii-1);

Yfrac = Ydiff/Ygap;

Xgap = Xper(iii)-Xper(iii-1);

Xadd = Yfrac*Xgap;

Xcut99 = Xadd+Xper(iii-1);

xline(Xcut95,'--',"95%",'LabelVerticalAlignment','middle','

LabelOrientation','horizontal');

% xline(Xcut99,'--',"99%",'LabelVerticalAlignment','middle','

LabelOrientation','horizontal');

figure(numba)

subplot(3,5,8);

imshow(EDXonSEMjustMP);

title('Zirconia in Melt Pool')

subplot(3,5,[13,14,15]);

histogram(MPvals0,'BinWidth',.01)

set(gca,'Yscale','log')

title('# of Pixels with Given Value')

ylim([.9 10000])

132

subplot(3,5,[9,10]);

imshow(uint8(MPvalsFull*100));

colors = jet(101);

colors(1:ceil(7.6),:) = 1;

C95 = colormap(subplot(3,5,[9,10]),colors);

colors(1:ceil(11.6),:) = 1;

C99 = colormap(subplot(3,5,[9,10]),colors);

caxis([0 100]); colorbar;

CC = uint8(MPvalsFull*100);

color = ind2rgb(CC,C95);

Color = color.*repmat(MeltPool,[1,1,3]);

title('Zirconia in Melt Pool');

% %%

% figure(1);

% aiai = 0;

% % for aia = 1:2:23

% % aiai = aiai+1;

% % subplot(3,4,aiai);

% CC = uint8(MPvalsFull*100);

% imshow(CC); set(gcf,'Units','Normalized','WindowState','

maximized');

% colors = jet(101);

% colors(1:ceil(7.6),:) = 1;

% C = colormap(figure(1),colors);

% caxis([0 100]); colorbar;

% Color = ind2rgb(CC,C);

% % imshow(Color);

% title("Pixels > 7 wt%");

% % end

% %%

% CALCULATE dispersion

% X = 1;

% i = 1;

133

xnum = 0;

ynum = 2;

dIndex = 0;

% Xs = [4:1:256];

% for xbox = Xs

% while i > .00000005

% while ynum(X) > 1

for X = [1,2]

% X = X+1;

xbox = X;

ybox = xbox;% box size

xnum(X) = ceil(cols1/xbox); ynum(X) = ceil(rows1/ybox);% # of

boxes

% if xnum(X) ~= xnum(X-1)

% xchange(X) = xbox;

% else

% xchange(X) = 0;

% end

% if ynum(X) ~= ynum(X-1)

% ychange(X) = ybox;

% else

% ychange(X) = 0;

% end

q ZrO2 = zeros(rows1,xnum(X));

q px = zeros(rows1,xnum(X));

m ZrO2 = zeros(rows1,cols1);

m px = zeros(rows1,cols1);

v = zeros(rows1,xnum(X));

a = 0;

for m = 1:rows1% exclude non-melt pool

values

for n = 1:cols1% shorten rows by

combining px

p = floor((n-1)/xbox)+1;% p=1 for n=1:16, p=2

for n=17:32, etc

134

if MeltPool(m,n) == 1% only for px in melt

pool area

a = a+1;

m px(m,n) = (EDX O2(m,n)*p ZrO2)+((1-EDX O2(m,n))*p SS

);

m ZrO2(m,n) = EDX O2(m,n)*m px(m,n);

q ZrO2(m,p) = q ZrO2(m,p)+m ZrO2(m,n);

q px(m,p) = q px(m,p)+m px(m,n);

v(m,p) = v(m,p)+1;% # of px included /16

end

end

end

[rows3,cols3] = size(q px);% 170 rows x 16

columns

r ZrO2 = zeros(ynum(X),xnum(X));

r px = zeros(ynum(X),xnum(X));

w = zeros(ynum(X),xnum(X));

for n = 1:cols3% shorten cols by

combining rbx

for m = 1:rows3% bx = 17x16 px

p = floor((m-1)/ybox)+1;% p=1 for m=1:17, p=2

for m=18:34, etc

r ZrO2(p,n) = r ZrO2(p,n)+q ZrO2(m,n);

r px(p,n) = r px(p,n)+q px(m,n);

% r(p,n) = r(p,n)+q(m,n);% box = sum of 17 rows

(by 16 columns)

w(p,n) = w(p,n)+v(m,n);% # of px values

included in box

end

end

a = 0;

R = 0;

W = 0;

R1 = zeros(ynum(X),xnum(X));

W1 = zeros(ynum(X),xnum(X));

135

for m = 1:ynum(X)% vector form of

boxes & weights

for n = 1:xnum(X)

if w(m,n) ~= 0

a = a+1;

R(a) = r ZrO2(m,n)/r px(m,n);% wt% ZrO2 for each box

R1(m,n) = r ZrO2(m,n)/r px(m,n);

% R wt(a) = w(m,n);% # of pixels included

for each box

W(a) = w(m,n)/max(w,[],'all');% frac of possible

px included

W1(m,n) = w(m,n)/max(w,[],'all');% weight = 0:1

% R(a) = r(m,n)/w(m,n);% avg = sum/# px

included

% R1(m,n) = r(m,n)/w(m,n);

else

% R(a) = r(m,n);

% R1(m,n) = r(m,n);

end

end

end

Wsum = sum(W,'all');

SDbx = std(R,W);% weighted SD of box

intensity values

Ubx = sum(R.*W)/Wsum;% weighted avg of box

intensity values

% Ubx2 = sum(r,'all')/sum(w,'all');

% Upx = mean(MPvals);% check Ubx; average

of px in MP

% Uwt = Ubx*EDXmax/254/100;

% SDwt = SDbx*EDXmax/254/100;

boxnum = sum(W,'all');

SDmax = sqrt((.5ˆ2)*boxnum/(boxnum-1));

136

dIndex(X) = .5*(1-(SDbx/SDmax)+(Ubx)/max(R));

if X > 1

% i = abs(dIndex(X)-dIndex(X-1);

end

end

% figure(5); plot(dIndex,'b');

% xlabel('box size (n x n)'); ylabel('dIndex value');

% for liner = 2:X

% if xchange(liner) ~= 0

% xline(xchange(liner),'--c');

% end

% if ychange(liner) ~= 0

% xline(ychange(liner),'--r');

% end

% end

Dindex1 = dIndex(1);

Dindex2 = dIndex(2);

% Dindex4 = dIndex(4);

% Dindex8 = dIndex(8);

% Dindex16 = dIndex(16);

% Dindex32 = dIndex(32);

%% Record it all

Label{numba,1} = label;

ZrO2Frac(numba,1) = F Z*100;

Predicted(numba,1) = P 2*100;

DIndex1(numba,1) = Dindex1;

DIndex2(numba,1) = Dindex2;

% DIndex4(numba,1) = Dindex4;

% DIndex8(numba,1) = Dindex8;

% DIndex16(numba,1) = Dindex16;

% DIndex32(numba,1) = Dindex32;

% NoiseCutoff(numba,1) = nxval;

width(numba,1) = w mp;

137

area(numba,1) = A mp;

high(numba,1) = High/L*100;

% low(numba,1) = Low/L*100;

% average(numba,1) = Avg*100;

% med(numba,1) = Med*100;

% stdev(numba,1) = SD*100;

% middle68(numba,1) = Middy/L*100;

covariance(numba,1) = Covar*100;

covariance2(numba,1) = Covar2;

cutoff95(numba,1) = Xcut95;

cutoff99(numba,1) = Xcut99;

VarNames = {'Label','ZrFrac','Adjusted','Predicted','%Incorporated',

...

'dIndex1','dIndex2',...%'NoiseCutoff',...%'dIndex4','dIndex8','

dIndex16','dIndex32',...

'Width','W (um)','Area','A (umˆ2)','% @ 50+wt%',...%'Low','Average

','Median',...'Standard Deviation',...

'Covariance','Covar2','95%','99%'};

T = table(Label(numba,1),ZrO2Frac(numba,1),0,Predicted(numba,1),0,...

DIndex1(numba,1),DIndex2(numba,1),...%NoiseCutoff(numba,1),...%

DIndex4(numba,1),DIndex8(numba,1),DIndex16(numba,1),DIndex32(

numba,1),...

width(numba,1),0,area(numba,1),0,high(numba,1),...%low(numba,1),

average(numba,1),med(numba,1),stdev(numba,1),...

covariance(numba,1),covariance2(numba,1),cutoff95(numba,1),

cutoff99(numba,1),...

'VariableNames',VarNames);

newfolder = [pthname,label];

if ~isfolder(newfolder)

mkdir(newfolder);

end

imwrite(Match,sprintf('%s\\%s Match.jpg',newfolder,label))

imwrite(MPimage,sprintf('%s\\%s overlay.tif',newfolder,label))

imwrite(EtchJustMP,sprintf('%s\\%s etchMP.tif',newfolder,label))

138

imwrite(SEMJustMP,sprintf('%s\\%s semMP.tif',newfolder,label))

imwrite(EDXonSEMjustMP,sprintf('%s\\%s edxMP.tif',newfolder,label))

imwrite(Newbie1,sprintf('%s\\%s averaged.tif',newfolder,label))

imwrite(CC,C95,sprintf('%s\\%s colormap95.tif',newfolder,label),'

WriteMode','overwrite')

imwrite(CC,C99,sprintf('%s\\%s colormap99.tif',newfolder,label),'

WriteMode','overwrite')

saveas(CDF,sprintf('%s\\%s CDF.png',newfolder,label))

saveas(Fraccers,sprintf('%s\\%s Noise.png',newfolder,label))

plotname = convertCharsToStrings(label);

TT = table(RRR.','VariableNames',plotname);

TTT = [TTT TT];

if numba == 1 % export excel file with columns: 1=Label, 2=ZrFrac,

3=DIndex

writetable(T,sprintf('%s\\%s values.csv',pthname,SP),...

'WriteVariableNames',true);

else

writetable(T,sprintf('%s\\%s values.csv',pthname,SP),...

'WriteMode','Append','WriteVariableNames',false);

end

end

writetable(TTT,sprintf('%s\\%s plotvals.csv',pthname,SP),...

'WriteVariableNames',true);

end

close all

139

”When you come to last page, close the book.”

- Ancient Chinese Proverb