Flow Fields and Particle Trajectories in Abrasive Slurry ... · constructive discussions on my...

Flow Fields and Particle Trajectories in Abrasive Slurry-jet Micro-machining of Sintered Ceramics and Metallic-layered

Structures

by

Kavin Kowsari

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Graduate Department of Mechanical and Industrial Engineering

University of Toronto

© Copyright by Kavin Kowsari, 2017

ii

Flow Fields and Particle Trajectories in Abrasive Slurry-jet Micro-machining of Sintered Ceramics and Metallic-layered

Structures

Kavin Kowsari

2017

Ph.D.

Graduate Department of Mechanical and Industrial Engineering University of Toronto

Abstract

The extreme hardness of sintered ceramics makes it difficult to machine them

economically. Abrasive slurry-jet micro-machining (ASJM), in which a target is eroded by the

impingement of a micro-jet of water containing fine abrasive particles, is a low-cost alternative

for micro-machining of sintered ceramic materials without tool wear and thermal damage, and

without the use of patterned masks. The experimental phase of the present research utilized

several model systems that have industrial relevance while incorporating sufficient generality to

illustrate generic characteristics of ASJM of ceramics and metallic-layered structures.

Experiments were complemented by extensive computational slurry-flow modeling to

understand the effects of the ASJM process parameters on the particle trajectories and the

resulting erosion.

iii

In this study for the first time, computational fluid dynamic (CFD) modeling was used to

derive a generalized relation between channel geometry and erosive flow, which was used in an

existing numerical-empirical model to predict the cross-sectional profiles of ASJM micro-

channels in sintered ceramics. The predictions agreed with experimental measurements to within

about 8%.

It was found that cavitation played a significant role in the slurry erosion of curved

features such as the edges of holes and channels. Features with sharper edges, flat bottoms, and

relatively steep sidewalls could be machined by minimizing cavitation through the use of liquids

with low vapor pressure and relatively high viscosity.

The use of ASJM to polish channels was investigated experimentally and with CFD.

Post-blasting channels after their initial machining under typical process conditions reduced the

Rrms roughness to about 23 nm in brittle and ductile targets.

Flat micro-pockets in sintered ceramic substrates containing copper-filled through-holes

were machined using a hybrid AJM (abrasive jet machining)-ASJM methodology, in which AJM

was used to selectively erode the brittle ceramic without etching the ductile copper, followed by

levelling of the protruding copper pillars to the depth of the ceramic using ASJM.

It was demonstrated that electrodeposited copper and nickel-phosphorous layers could be

selectively removed without eroding the underlying ceramic or metallic substrate using over-

lapping ASJM channels. A CFD-aided process design methodology was developed to predict the

ASJM parameters to remove a given coating thickness.

iv

Acknowledgements

I would like to express my sincere appreciation to my advisors, Professor Jan K. Spelt

and Professor Marcello Papini, for their continuous support and guidance throughout the course

of my research. This experience will everlastingly be beneficial to my life. I would also like to

thank my thesis advisory committee, Professor Markus Bussmann and Professor David Sinton

for their constructive suggestions and recommendations in my committee meetings.

Grateful acknowledgement is also paid to Hooman Nouraei for his friendship and

constructive discussions on my research. I am also thankful for my lab colleagues, Reza Haj

Mohammad Jafar, Kamyar Hashemnia, Amir Nourani, Saeed Akbari, Lucas Maciel, Ernst van

Wijk, and Ryan Brown, for maintaining a cozy workplace; as well as Thais R. Dotto, Jonathan

Smith, Zahin Rahman, Eric Chong, Lin Sen Mu, Leonardo de Faria and Qiaozhi Liu who did

great jobs as summer research assistants.

The support of Natural Sciences and Engineering Research Council of Canada (NSERC),

Canada Research Chairs, Nanowave Technologies, and Magna International Inc. is also

acknowledged.

Last and foremost, I have been very fortunate in receiving continuous love, support, and

encouragement from my father, Amir Kowsari, my mother, Mahva Karimi, and my brother,

Kamyar Kowsari. I am indebted to them for enabling me to pursue my goals. This thesis is

dedicated to them.

v

Table of Contents

Abstract..........................................................................................................................................ii

Acknowledgements ...................................................................................................................... iv

Table of Contents .......................................................................................................................... v

List of Tables. ............................................................................................................................... ix

List of Figures ............................................................................................................................... xi

Chapter 1: Introduction ........................................................................................................... 1

1.1. Justification and motivation ......................................................................................... 1

1.2. Objectives .................................................................................................................... 5

1.3. Experimental apparatus ................................................................................................ 6

1.4. Thesis outline ............................................................................................................... 8

1.5. References .................................................................................................................. 10

Chapter 2: CFD-aided Prediction of the Shape of Abrasive Slurry Jet Micro-machined

Channels in Sintered Ceramics.......................................................................... 12

2.1. Introduction ................................................................................................................ 12

2.2. Experiments and flow modeling ................................................................................ 12

2.2.1. Experiments................................................................................................. 15

2.2.2. CFD modeling ............................................................................................. 17

2.3. Results and discussion ............................................................................................... 18

2.3.1. ASJM channels in sintered ceramics – changes in centerline specific

erosion rate and shape with depth ............................................................ 18

2.3.2. ASJM erosion parameters for sintered ceramics ........................................ 25

2.4. Channel profile modeling ......................................................................................... 30

2.4.1. Method I: CFD erosion simulation of each pass ........................................ 30

2.4.2. Method II: CFD with approximate stagnation zone model ........................ 35

2.4.2.1. Method II predictions - 90 machining ........................................ 40

2.4.2.2. Method II - 45 machining ........................................................... 42

2.5. Conclusions ................................................................................................................ 43

2.6. References .................................................................................................................. 45

vi

Chapter 3: The Effects of Fluid Vapor Pressure and Viscosity on the Shapes of Abrasive

Slurry-jet Micro-machined Holes and Channels ............................................. 47

3.1. Introduction ................................................................................................................ 47

3.2. Machining experiments and CFD flow simulations .................................................. 50

3.2.1. Experiments................................................................................................. 50

3.2.2. Ultrasonic apparatus and experiments ....................................................... 52

3.2.3. CFD modeling ............................................................................................. 53


3.3.1. ASJM hole formation mechanism ............................................................... 55

3.3.2. Ultrasonic abrasive cavitation .................................................................... 57

3.3.3. Effects of viscosity and vapor pressure on the shape of ASJM holes in

brittle materials ......................................................................................... 59

3.3.4. Effect of surface roughness ......................................................................... 69

3.3.5. Through-holes in sintered ceramics............................................................ 70

3.3.6. Channels in glass and zirconium tin titanate using an oil-based slurry..... 73

3.3.7. Edge rounding in ductile materials............................................................. 74

3.4. Conclusions ................................................................................................................ 75

3.5. References .................................................................................................................. 77

Chapter 4: Erosive Smoothing of Abrasive Slurry-Jet Micro-machined Channels in

Glass, PMMA, and Sintered ceramics: Experiments and Roughness Model ..

................................................................................................................................80

4.1. Introduction ................................................................................................................ 80

4.2. Experiments and flow modeling ................................................................................ 82

4.2.1. Experiments................................................................................................. 83

4.2.2. CFD modeling ............................................................................................. 85

4.3. Roughness model ....................................................................................................... 87


4.4.1. As-received target surfaces ......................................................................... 90

4.4.2. Mechanism of surface topography evolution .............................................. 92

4.4.3. Roughness of ASJM channels under standard conditions .......................... 94

4.4.3.1. Effect of particle dose .................................................................. 97

4.4.3.2. Effect of particle kinetic energy ................................................... 97

4.4.4. Roughness of post-blasted ASJM channels ............................................... 102

4.4.5. Roughness prediction during post-blasting .............................................. 106

4.5. Conclusions .............................................................................................................. 108

4.6. References ................................................................................................................ 109

vii

Chapter 5: Hybrid Erosive Jet Micro-milling of Sintered Ceramic Wafers With and

Without Copper-filled Through-holes ............................................................ 112

5.1. Introduction .............................................................................................................. 112

5.2. Experiments and flow modeling .............................................................................. 115

5.2.1. Experiments............................................................................................... 115

5.2.2. CFD modeling ........................................................................................... 116

5.3. Results and discussion ............................................................................................. 118

5.3.1. Erosion mechanism ................................................................................... 118

5.3.2. ASJM pockets in sintered ceramics........................................................... 121

5.3.3. AJM pockets in sintered ceramics............................................................. 131

5.3.4. Pockets in sintered aluminum nitride containing copper-filled through-

holes..... ............................................................................................................... 133

5.3.4.1. Application of ASJM .................................................................. 133

5.3.4.2. Hybrid use of AJM and ASJM .................................................... 137 5.4. Conclusions .............................................................................................................. 142

5.5. References ................................................................................................................ 143

Chapter 6: Prediction of the Erosive Footprint in the Abrasive Jet Micro-machining of

Flat and Curved Glass ...................................................................................... 146

6.1. Introduction .............................................................................................................. 146


6.2.1. Jet and footprint measurements ................................................................ 148

6.2.2. CFD modeling ........................................................................................... 151


6.3.1. AJM jet structure ....................................................................................... 154

6.3.2. Erosive footprint prediction for flat surfaces ............................................ 157

6.3.2.1. Experimental validation ............................................................. 163

6.3.3. Erosive footprint prediction for curved surfaces ...................................... 165

6.3.4. Implications for AJM ................................................................................ 170

6.4. Conclusions .............................................................................................................. 171

6.5. References ................................................................................................................ 173

Chapter 7: Selective Removal of Metallic Layers from Sintered Ceramic and Metallic

Substrates Using Abrasive Slurry-jet Micro-machining ............................... 175

7.1. Introduction .............................................................................................................. 175


7.2.1. Target materials ........................................................................................ 177

7.2.2. ASJM apparatus and experiments ............................................................ 179

7.2.3. CFD modeling ........................................................................................... 184

viii


7.3.1. Target erosion characterization - experiment #1 ..................................... 185

7.3.2. Selective removal of copper pillars - experiment #2 .............................. 187

7.3.3. Selective removal of metallic layers using over-lapping channels ........... 194

7.3.3.1. Effect of machining front slope on erosion rate - experiment #3 ....

................................................................................................................ 194

7.3.3.2. Removal of nickel-phosphorous layer from aluminum - experiment

#4 ............................................................................................................ 203

7.3.3.3. Removal of copper layer from aluminum nitride containing

copper-filled through-holes - experiment #4 ......................................... 206

7.3.3.4. Prediction of the layer thickness removed for machined over-

lapping channels – experiment #4 ......................................................... 208

7.4. Conclusions .............................................................................................................. 213

7.5. References ................................................................................................................ 215

Chapter 8: Conclusions and Future Work ......................................................................... 217

8.1. Conclusions .............................................................................................................. 217

8.2. Directions for Future Work ...................................................................................... 223

Thesis References ...................................................................................................................... 224

ix

List of Tables

Table 2.1 Properties of the target materials. ................................................................................. 16

Table 2.2 Standard process parameters. ....................................................................................... 16

Table 2.3 Particle size distribution for 10 μm alumina particles (Comco Inc., Burbank, CA,

USA). Mass flow rate for each fraction based on standard conditions of 1 wt% particle

concentration and 1.67 mL/s slurry flow rate (Table 2.2). ............................................... 18

Table 2.4 Best-fit constants (Eq. (2.3)) for the impact velocity dependence of erosion. ............. 27

Table 2.5 Best-fit coefficients of erosion data using 3th

order polynomial

3 2

3 2 1 0f a x a x a x a .. .................................................................................................. 30

Table 3.1 Properties of the test fluids at 20 C. The vapor pressures of soybean oil and mineral

oil were taken from Ndiaye et al. (2005) [18] and Sigma-Aldrich (St. Louis, MO, USA,

http://www.sigmaaldrich.com), respectively. The values in bold were used as inputs in

the CFD simulations........ ................................................................................................. 51

Table 3.2 Properties of the target materials.. ................................................................................ 52

Table 4.1 Properties of the target materials. The dynamic hardnesses were estimated using the

methodology of Section 4.4.5.. ......................................................................................... 84

Table 4.2 ASJM process parameters used in the two types of experiments: (i) channel machining

over a range of typical conditions, and (ii) channel smoothing using post-blasting.

Standard process conditions shown in bold..... ................................................................. 85

Table 4.3 Input parameters used in the roughness model for three sets of process conditions. The

average centerline particle impact angles were reproduced from the CFD models of

Kowsari et al. (2016b) [24] on flat targets, and the particle properties were obtained from

the manufacturers (Comco Inc., Burbank, CA, USA; Zaozhuang City-hsin Ltd., China). ..

............................................................................................................................................89

Table 4.4 Percentage change in channel centerline Rrms roughness compared to the as-received

surfaces or channel centerline surfaces after machining under typical process conditions.

+ indicates an increase in roughness, - indicates a decrease. .......................................... 106

Table 5.1 Properties of the target materials... ............................................................................. 115

Table 5.2 Standard process parameters...... ................................................................................ 116

x

Table 7.1 Properties of the target materials, obtained from the manufacturer of aluminum nitride,

from ASM (1990) for aluminum, and from Zhaojiang (1999) for copper and nickel-

phosphorous... ................................................................................................................. 169

Table 7.2 Properties of the test fluids at 20 C...... .................................................................... 183

Table 7.3 Process parameters used in experiment #1 to determine dependence of erosion on

velocity and impact angle for copper and nickel-phosphorus. ....................................... 184

Table 7.4 Best-fit constants for Eq. (7.3) giving the dependence of specific erosion rate on the

centerline average particle impact velocity of a water slurry-jet (63-110 m/s) at

perpendicular incidence.. ................................................................................................ 187

Table 7.5 CFD predictions of average particle impact angles at machining front along plane of

symmetry through the centerline (primary footprint) for various β in the 90, 45

forwards, and 45 backward machining orientations using water and soybean oil slurry

jets............... .................................................................................................................... 199

Table 7.6 Percentage change in channel depth at doses of 3.4 g/mm for water and 4.7 g/mm for

soybean oil produced by the leading edge effect in slow, single-pass machined channels

in copper using water and soybean oil slurry jets in the 90, 45 forwards, and 45

backward machining orientations. The symbols (+) and (-) indicate an increase or

decrease in the depth, respectively, relative to channels machined using rapid, multiple

shallow passes at 0.3 g/mm for water and 0.5 g/mm for soybean oil which gave a very

small slope, < 2....... .................................................................................................. 199

xi

List of Figures

Figure 1.1 (a) Schematic of the ASJM apparatus (not to scale), Kowsari et al. (2016) [27]. (b)

Orifice geometry, reproduced from Kowsari et al. (2014) [16]. .................................... 7

Figure 2.1 Nonlinearity of centerline specific erosion rate in deep channels machined in

backward configuration at 45 in aluminum nitride using the standard conditions

(Table 2.2). (a) CFD model of a relatively deep channel (5 passes) showing stagnation

zone and particle flow, (b) Channel depth versus dose (number of machining passes).

(c) Cross-sectional profiles of multi-pass channels. P denotes the number of machining

passes. (d) Profiles of 1-5 pass channels, each normalized by their width and depth.. 20

Figure 2.2 Results for aluminum nitride at perpendicular incidence using the standard conditions

(Table 2.2). (a) Cross-sectional profiles of channels using 1-5 machining passes

(denoted by P). Half of the symmetric profiles shown along with the streamline

bounding the jet. (b) Depth versus particle dose (machining passes) of channels at 90

and 45 jet incidence. (c, d) Static pressure contours and bounding streamlines at the

water-air interface of the flows within a relatively shallow channel (1 pass, (c)) and a

relatively deep channel (5 passes, (d)). (e) Profiles of 1-12 pass channels (denoted by

p), each normalized by their width and depth. (f) Depth versus particle dose

(machining passes), both normalized by those of the first-pass channels at 90 and 45

jet incidence.. ............................................................................................................... 24

Figure 2.3 Domain and boundary conditions of a two-dimensional axisymmetric model of the

90 impingement of a slurry jet on a flat target.. ......................................................... 26

Figure 2.4 Dependence of ASJM specific erosion rate, normalized by the maximum for each

target material (0.04, 0.02, and 0.08 mg/g for aluminum nitride, alumina, and

zirconium tin titanate, respectively), on jet-centerline-averaged particle impact

velocity. Error bars represent ±1 standard deviation for 3 measurements. The lines

show best-fit curves given in Table 2.4.. ..................................................................... 27

Figure 2.5 (a) Centerline trajectories of 10 μm alumina particles in 45 impingement of a jet on a

flat target. (b) ASJM specific erosion rate normalized by that at perpendicular

incidence (0.003, 0.003, and 0.055 mg/g for aluminum nitride, alumina, and zirconium

tin titanate, respectively) versus jet centerline average particle impact angle at a jet

velocity of 89 m/s. Error bars represent ±1 standard deviation for 3 measurements. The

lines show best-fit curves given in Table 2.5.. ............................................................. 29

Figure 2.6 First-pass channel machined in aluminum nitride at perpendicular incidence using the

standard conditions (Table 2.2) (a) Trajectories of 10 μm alumina particles (b) Three-

dimensional erosion map (c) Cross-sectional profiles the first-pass channel and the

second-pass channel predicted using method I. ........................................................... 32

Figure 2.7 Comparison of the measured (solid lines) and predicted (symbols) cross-sectional

profiles of channels machined in aluminum nitride using the standard conditions

(Table 2.2) using COR of 0.2 and 0.8. Half of the symmetric profiles shown. P denotes

the number of passes.. .................................................................................................. 35

xii

Figure 2.8 ASJM channels machined in aluminum nitride at 90 incidence using the standard

conditions (Table 2.2). (a) Cross-sectional profiles and pressure contours (b) Jet-

centerline-averaged impact velocities for 10 μm alumina particles versus filled-region

aspect ratio of multi-pass channels having depth/width aspect ratios of up to about 0.8

in aluminum nitride, alumina, and zirconium tin titanate using the standard conditions

(Table 2.2).. .................................................................................................................. 37

Figure 2.9 Multi-pass channels machined in aluminum nitride at perpendicular incidence using

the standard conditions (Table 2.2). (a) Erosive footprints of three-dimensional erosion

maps. Comparison of two-dimensional erosion patterns, each normalized by their

centerline thickness loss, of (b) the measured and trimmed 1-pass channel in stage I

(Fig. 2.2(a)), and (c) the measured 3-pass channel in stage II of channel formation, and

the trimmed 1-pass channel in stage I. ......................................................................... 40

Figure 2.10 Comparison of predicted (symbols) and measured (solid lines) cross-sectional

channel profiles in (a) aluminum nitride, (b) alumina, and (c) zirconium tin titanate at

perpendicular jet incidence using the standard conditions. Half of the symmetric

profiles shown. P denotes the number of passes. ......................................................... 41

Figure 2.11 Comparison of predicted (symbols) and measured (solid lines) cross-sectional

profiles of channels machined in aluminum nitride with backward 45 passes using the

standard conditions (Table 2.2). Half of the symmetric profiles shown. P denotes the

number of passes. ......................................................................................................... 43

Figure 3.1 Schematic of the setup used in the ultrasonic cavitation experiments.. ..................... 53

Figure 3.2 Domain and boundary conditions of an axisymmetric CFD model of the ASJM flow

within a relatively deep blind hole measured in glass.................................................. 55

Figure 3.3 Scanning electron microscope (SEM) images of cross-section of a hole machined in

glass using the standard conditions for 8 min, and the surface texture in region A (Ra =

0.26 μm), the transition zone, and region B (Ra = 0.33 μm).. ...................................... 56

Figure 3.4 Volume fraction contours of water vapor within a relatively deep hole machined in

glass using a water-particle slurry.. .............................................................................. 57

Figure 3.5 SEM images of slurry cavitation damage on glass created using ASJM (Ra = 0.27 μm)

and an ultrasonic apparatus (Ra = 0.36 μm). Densely and sparsely-impacted surfaces

shown on the left and right, respectively. The roughness values correspond to the

surfaces on the left.. ..................................................................................................... 58

Figure 3.6 Effect of slurry liquid on the profiles of blind holes in glass. (a) Cross-sectional

profiles, and (b) top views of approximately 200 μm-deep blind holes using the liquids

of Table 3.1 and standard conditions. The depths were normalized by the center depth

of each hole. Half of the symmetric holes shown.. ...................................................... 61

Figure 3.7 Velocity contours of the flow fields of: (a) water and (b) the aqueous glycerin

solution (Table 3.1) within a relatively deep holes. Particle rebounds not shown. ...... 63

xiii

Figure 3.8 Half of the symmetrical cross-sectional profiles of holes machined using water and

soybean oil-based slurries in zirconium tin titanate.. ................................................... 64

Figure 3.9 SEM images of cross-section and plan view of blind hole machined using soybean oil

(=45 cP, vp=0.35 kPa) and standard conditions in zirconium tin titanate for 10 min,

and top view of a portion of the edge of a blind hole in glass machined for 2.5 min

using soybean oil.. ........................................................................................................ 65

Figure 3.10 The volume fraction contour of water vapor of a water-particle slurry flow within a

relatively shallow hole initially machined in glass using soybean oil. ........................ 67

Figure 3.11 (a) Cross-sectional profiles of blind holes machined in glass, initially using the

soybean slurry, then continuing to machine using a water-particle slurry. Half of the

symmetric profiles shown. (b) Top view of the rounded hole in (a) that was finished

with the aqueous slurry.. .............................................................................................. 68

Figure 3.12 Rounding radius of curvature as a function of abrasive dose for ASJM holes in glass

machined using a water slurry under the standard conditions.. ................................... 68

Figure 3.13 Pressure contour of the perpendicular impingement of an ASJM jet on a flat plate

having an Ra of approximately 400 nm. Model topography taken from an actual

profilometer scan of the centerline of an ASJM channel machined in glass using

typical process conditions.. .......................................................................................... 70

Figure 3.14 Plan view of the exit of a through-hole machined in zirconium tin titanate without a

backing plate.. .............................................................................................................. 71

Figure 3.15 (a) Section view of a through-hole in a zirconium tin titanate plate machined when

attached to another plate using epoxy. (b) Plan view of the hole exit after second plate

was separated by heating it to 316 C.. ........................................................................ 72

Figure 3.16 Cross-sectional profiles of approximately 50 μm deep channels machined in: (a)

glass and (b) zirconium tin titanate using water and oil-based slurries at 90 and 45

jet incidences. Half of the symmetric profiles shown. The depths were normalized by

the centerline depth of each channel. ........................................................................... 74

Figure 3.17 Half of symmetrical cross-sectional profiles of approximately 100 μm- deep single-

pass channels in copper machined using water and soybean-oil slurries using the

standard conditions. ..................................................................................................... 75

Figure 4.1 Domain and boundary conditions of a three-dimensional CFD model of the ASJM

flow within a channel in PMMA at 45 incidence. The schematic is to scale. ............ 87

Figure 4.2 Schematic of the geometry of a model particle impacting a target.. .......................... 89

Figure 4.3 (a) SEM images of the as-received glass, PMMA, zirconium tin titanate, and

aluminum nitride surface. (b) AFM measurements of as-received PMMA.. ............... 91

xiv

Figure 4.4 SEM images of (a) plastically-deformed craters without cracking in glass, (b) region

A (Fig. 4.1) surfaces of channels machined in glass, PMMA, zirconium tin titanate,

and aluminum nitride using the standard conditions (Table 4.2).. ............................... 94

Figure 4.5 (a) Trajectories of 10 μm alumina particles on the jet centerline plane of the ASJM

flow within a PMMA channel machined at 45 shown in Fig. 4.1. Primary footprint –

region A, secondary footprint – region b. (b) SEM images of regions A and B of a

channel machined at 45 channel in glass.. .................................................................. 96

Figure 4.6 Measured channel centerline Rrms roughness as a function of particle dose using the

range of standard process conditions (Table 4.2). Error bars represent ±1 standard

deviation for 3 areal scans along a single channel.. ..................................................... 97

Figure 4.7 Channel machining using standard conditions of Table 4.2. Measured channel

centerline Rrms roughness and specific erosion rate as a function of average particle

impact kinetic energy of surface-normal velocity component in: (a) glass, (b) PMMA,

(c) zirconium tin titanate, and (d) aluminum nitride. Error bars represent ±1 standard

deviation for 3 areal scans along a single channel.. ..................................................... 99

Figure 4.8 Channel machining using standard conditions of Table 4.2. Measured channel

centerline Rrms roughness and specific erosion rate as a function of average particle

impact angle in (a) glass, (b) PMMA, (c) zirconium tin titanate, and (d) aluminum

nitride. Error bars represent ±1 standard deviation for 3 scans along a single channel..

.................................................................................................................................... 101

Figure 4.9 Measured centerline Rrms roughness of post-blasted channels in zirconium tin titanate

at 15 jet incidence as a function of particle dose using different particles. Stationary

jet. Error bars represent ±1 standard deviation for 3 scans within the footprint.. ...... 103

Figure 4.10 (a) Measured centerline Rrms roughness of post-blasted channels in glass, PMMA,

and zirconium tin titanate as a function of particle dose using 3 μm silicon carbide

particles at 15 jet incidence. Error bars represent ±1 standard deviation for 3

measurements. (b) Plan view SEM images of post-blasted surfaces using the same

conditions in glass, PMMA, and zirconium tin titanate. (c) Isometric AFM view region

A of a channel in zirconium tin titanate post-blasted with 3 μm silicon carbide

particles at a dose of approximately 90 g/mm2.. ........................................................ 105

Figure 4.11 Measured (black) and predicted (gray) channel centerline Rrms roughnesses for the

process conditions II and III (Table 4.3), selected for channel-machining and post-

blasting (peak removal), respectively.. ...................................................................... 108

Figure 5.1 Schematic of (a) cross-sectional profile and (b) plan view of a pocket in an aluminum

nitride wafer containing copper-filled through-holes (vias).. .................................... 113

Figure 5.2 Domains and boundary conditions of (a) a 2D axisymmetric model of the

impingement of a slurry jet on a flat target, and (b) a 3D model for the simulation of

the flow within a channel. Elements not to scale.. ..................................................... 117

xv

Figure 5.3 SEM images of the surfaces of sintered alumina: (a) with exposed grains without any

ASJM (reproduced from CoorsTek (www.coorstek.com) material property catalog),

and (b) showing single-particle-impact sites created by scanning the ASJM jet

containing 0.01 wt% 10 m diameter alumina particles over the target at a scan speed

of 5 mm/s at normal incidence. .................................................................................. 119

Figure 5.4 (a) Schematic of the oblique jet orientation in the machining of an asymmetrical

ASJM channel. (b) Dependence of ASJM normalized erosion rate (erosion rate at a

given angle divided by that at 90) of sintered alumina (Table 5.1) on jet impact angle.

The erosion rate at perpendicular incidence was measured as 0.05 mg/g. Error bars

represent ±1 standard deviation for 3 measurements. ................................................ 120

Figure 5.5 (a) Machining path of slurry jet during the machining of a pocket using the over-

lapping channel method. (b) Isometric views of the surface profiles of pockets in

sintered alumina for 25 and 200 m channel offsets. The in-plane dimensions are to

scale, while the depth is amplified by 15%. ............................................................... 122

Figure 5.6 (a) Measured and predicted cross-sectional profiles of ASJM pockets machined using

the over-lapping channel method in alumina for offsets of 50 and 200 m (b) Pocket

roughness, aR , versus offset in sintered alumina using the standard conditions of

Table 5.2.. ................................................................................................................... 123

Figure 5.7 (a) Surface geometries of a shallow (depth = 25 μm) and relatively deep (depth = 135

μm) channel machined in sintered alumina using standard ASJM conditions. Static

pressure contours for the flows within a shallow channel and a relatively deep channel

(b) without any offset, and (c) with an offset of 150 m. A gage pressure of 0.25 MPa

defined the stagnation zone boundary.. ...................................................................... 127

Figure 5.8 (a) Depth as a function of the number of machining passes of channels machined in

alumina using the standard conditions. (b) Cross-sectional profiles of pockets

machined in alumina after each machining operation using the overlapping channel

method (standard conditions; 150 m offset; 6 overlapped channels per operation).

(c) Pocket depth as a function of the number of machining operations. Error bars

represent ±1 standard deviation for 3 measurements..................................................129

Figure 5.9 Isometric view of the surface geometry of a pocket in alumina using over-lapping

channels each machined at an oblique angle (standard conditions; 5 machining passes

per channel). ............................................................................................................... 130

Figure 5.10 Isometric views of the surface profiles of masked AJM pockets (standard

conditions; 15 s machining time) in (a) sintered alumina and (b) sintered aluminum

nitride... ...................................................................................................................... 132

Figure 5.11 (a) Isometric view of the surface geometry of channels machined using ASJM along

a series of copper-filled through-holes in a matrix of sintered aluminum nitride. #P

denotes the number of machining passes. (b) Dependence of maximum depths of

aluminum nitride and copper-filled through-holes within the channels on the number

of machining passes. Error bars represent ±1 standard deviation for 3 measurements

xvi

taken from three separate filled through-holes along a channel. Standard conditions,

but at 0.1 mm/s scan speed.. ....................................................................................... 134

Figure 5.12 (a) Schematic of the lateral flow exiting the footprint in the perpendicular

impingement of an ASJM jet on a target plate near a copper through-hole. (b)

Deepening filled through-holes due to lateral flow of increasing passes (#P) of jet

between rows of filled through-holes using standard conditions (Table 5.2).. .......... 136

Figure 5.13 Static pressure contours and particle trajectories for targets at 90° and 45° for the

impingement of; (a) an ASJM jet (8 MPa orifice pressure, 10 m alumina), and (b) an

AJM jet (200 kPa orifice pressure, 25 m alumina). A gage pressure of 0.25 MPa

defined the stagnation zone boundary.. ...................................................................... 139

Figure 5.14 Optical profilometer images of the two stages in pocket machining in sintered

aluminum nitride with copper-filled through-holes: (a) after masked AJM and (b) after

masked AJM followed by unmasked ASJM to flatten the copper pillars.. ................ 141

Figure 6.1 Double-pulsed shadowgraphy apparatus.. ................................................................ 150

Figure 6.2 Domains and boundary conditions of: (a) 2D axisymmetric CFD model of the

impingement of an air-particle jet on a flat target and (b) 3D CFD model of the

impingement of an air-particle jet on a curved target.. .............................................. 153

Figure 6.3 AJM jet. (a) Microscope images of the AJM jet. (b) Air and particle velocity

magnitude contours obtained using CFD. .................................................................. 156

Figure 6.4 Radial distribution of particles within the jet: (a) at nozzle exit (b) at 20 mm standoff,

obtained from shadowgraphy. The error bars indicate the standard deviations obtained

from three measurements of approximately 15000 particles each. ............................ 157

Figure 6.5 Impingement of AJM jets on flat targets. Air velocity magnitude contours and 10 μm

diameter particle trajectories for standoff distances of: (a) 5 mm, (b) 10 mm, (c) 20

mm, and (d) 30 mm. (e) Drag energy loss as a function of standoff distance for

particles released at the nozzle centerline obtained from either hapex and Vr or direct

integration of Fd (Eq. (6.2)). (f) CFD prediction of particle drag force versus particle

rebound displacement for different standoffs. (g) Axial velocities of particles released

from a given mesh element at the inlet boundary approximately 100 μm from the

nozzle centerline (13% of the nozzle diameter) at various distances from the target

using the models of Fig. 6.5.. ..................................................................................... 162

Figure 6.6 (a) Predicted (dashed lines) and measured (solid lines) erosive footprint diameter

versus standoff with and without secondary particle impacts. The lines are to guide the

eye only. Error bars represent ±1 standard deviation for 3 measurements. (b)

Schematic representation of intersections of primary and secondary plumes with

successive target planes at standoffs of 10 mm and 30 mm. ψ defines the second-strike

cone angle, and h՛ is the apex height of a corresponding particle after rebound from

the target.. ................................................................................................................... 165

xvii

Figure 6.7 Impingement of AJM jets on curved targets at a standoff of 20 mm. Air velocity

magnitude contours and particle trajectories for rod diameters of: (a) 5 mm, and (b) 3

mm. ............................................................................................................................ 167

Figure 6.8 CFD-obtained normalized erosion maps on a flat target and a 5 mm diameter rod.

Each map was normalized by its maximum specific erosion rate (mass eroded per unit

mass of erodent).. ....................................................................................................... 168

Figure 6.9 A schematic of Weibull-type function describing the shallow eroded profile. The

coordinates (y, x) of a typical point on the profile are shown.....................................169

Figure 6.10 Normalized erosive efficacies and the best fits (Weibull distribution) for flat and

curved (5 mm diameter) glass targets. The abscissa was normalized by standoff

distance and the ordinate was normalized by the depth of the channel centerline.. .. 170

Figure 7.1 Schematic section views through the 3 test specimens. (a) copper-plated aluminum

nitride containing copper-filled through-holes. (b) nickel-phosphorous-plated

aluminum. (c) protrusion formed due to over-filling of through-hole in aluminum

nitride wafer. The dashed regions are those to be removed using ASJM... ............... 178

Figure 7.2 (a) Schematic of the position of the stationary primary jet footprint and the secondary

flow with respect to the copper pillar in experiment #2. Section through the centerline

of the jet. (b)-(c): Domain and boundary conditions of three-dimensional CFD models

of the ASJM flow within channels measured in copper (experiment #3) machined

using a 110 m/s soybean oil slurry-jet scanned at 0.005 mm/s in the (b) 45 forward

and (c) 45 backward configurations. (d) Machining path of the slurry jet in the

overlapping channel-machining method of experiment #4 illustrated for the specimen

of Fig. 7.1(a)... ............................................................................................................ 183

Figure 7.3 ASJM specific erosion rates for copper and nickel-phosphorous, respectively, vs. (a)

jet velocity and centerline average particle impact velocity of a water slurry-jet at

perpendicular incidence, and (b) jet impact angle and actual centerline average particle

impact angle of an 89 m/s water slurry-jet. Experiment #1. Error bars represent ±1

standard deviation for 3 measurements. The lines serve only to guide the eye.. ....... 186

Figure 7.4 Surface topography of un-eroded and eroded copper-filled through-holes subjected to

stationary 15 slurry-jets at 89 m/s in the configuration shown in Fig. 7.2(a)

(experiment #2) using: (a) a water slurry, (b) a soybean oil slurry. (c) Elevation of

copper with respect to the aluminum nitride substrate vs. time of exposure to

stationary 15 slurry-jets of water and soybean oil using the same process conditions

as in (a) and (b).. ........................................................................................................ 188

Figure 7.5 Comparison of boundary layer thickness vs. x (defined in Fig. 7.2(a)) for a 15 jet

impact angle with water (89 m/s) and soybean oil (89 m/s) jets as measured from CFD

and computed using Eq. (7.4). ................................................................................... 189

Figure 7.6 Particle trajectories in vicinity of dimples placed in the secondary flow (x ≈ 1.3 mm,

Fig. 7.2(a)) of a jet having an inclination of 15: (a) water slurry and (b) soybean oil

jets. ............................................................................................................................. 191

xviii

Figure 7.7 Particle trajectories for flow fields over pillars placed about 1.3 mm downstream of a

15 soybean oil jet (Fig. 7.2(a), experiment #2). (a) 10 μm and (b) 15 μm particle

trajectories over a 65 μm high protrusion; (c) 15 μm particle trajectories over a 20 μm

high protrusion.. ......................................................................................................... 193

Figure 7.8 (a) Channel depth vs. dose (g/mm of channel length) of single-pass channels

machined in copper using a 89 m/s water jet in the 90, 45 forward, and 45 backward

orientations (experiment #3). The lines serve only to guide the eye. (b) Side view of

the local machined front geometry of the channel in Fig. 7.2(a). CFD three-

dimensional erosion map of 90 (89 m/s) water slurry-jet on (c) a flat copper target,

and (d) a 117 μm deep channel in

copper..........................................................................................................................197

Figure 7.9 CFD particle trajectories in the primary footprint at the leading edge of single-pass

channels machined in copper using a water slurry-jet scanned at 0.005 mm/s in the (a)

90, (b) 45 forward, and (c) 45 backward orientations. αavg is the average impact

angle along the centerline of the primary footprint. Particle rebounds not shown.

Channel leading edge angle defined in Fig. 7.8(b). ................................................ 198

Figure 7.10 Channel depth vs. dose (g/mm of channel length) of single-pass channels machined

in copper using a soybean oil jet in the 90, 45 forward, and 45 backward

orientations. The lines serve only to guide the eye... ................................................. 201


channels machined in copper using a 110 m/s soybean oil jet scanned at 0.005 mm/s in

the (a) 90, (b) 45 forward, and (c) 45 backward orientations. αavg is the average

impact angle along the centerline of the primary footprint. Particle rebounds not

shown. Channel leading edge angle defined in Fig. 7.8(b)... .................................. 202

Figure 7.12 (a) Surface topography of a nickel-phosphorous layer removed to expose aluminum

substrate. Result of a single machining operations using a 45 water slurry-jet (89 m/s)

scanned at 1.4 mm/s in the configuration of Fig. 7.2(d). Each operation used

overlapping scans offset by 50 μm. The jet was not rotated between passes so the

orientation was alternately forward (Fig. 7.2(b)) and backward (Fig. 7.2(c)) between

passes. (b) Cross-sectional measured profiles along line A-A of the pocket in (a). The

plot shows only a portion of the profiles. A scanning electron microscope image of a

section view of an uneroded specimen of Fig. 7.1(b) is shown on the right... ........... 205

Figure 7.13 (a) Surface topography and (b) a portion of the cross-sectional profiles along line B-

B of a copper layer removed to expose a flat surface of aluminum nitride containing

copper-filled through-holes. Results of 8 and 10 machining operations using a 110 m/s

perpendicular soybean oil jet scanned at 4 mm/s. Each operation used overlapping

scans offset by 50 μm... .............................................................................................. 207

Figure 7.14 CFD three-dimensional erosion map of (a) a 90 (110 m/s) soybean oil jet on

copper, and (b) a 45 (89 m/s) water slurry-jet on nickel-phosphorous. (c) Two-

xix

dimensional representative erosion patterns of the models in (a) and (b). The specific

erosion rates were normalized by the specific erosion rate along line C-C in (a).. ... 210

Figure 7.15 Measured (solid lines) and predicted (dashed lines) cross-sectional channel profiles

of pockets removed within (a) the copper layer of Fig. 7.1(a), machined using 8

operations of over-lapping 4 mm/s channels with a perpendicular soybean oil jet (110

m/s), and (b) the nickel-phosphorous layer of Fig. 7.1(b), machined using 1 operation

of over-lapping 1.4 mm/s channels with a 45 water slurry-jet (89 m/s). The offset was

50 μm in both (a) and (b)... ........................................................................................ 211

1

Chapter 1: Introduction

1.1. Justification and motivation

The fabrication of microfluidic and microelectromechanical systems (MEMS) components is

complicated by their small sizes, complex details, and the need for smooth, defect-free surfaces.

These requirements can challenge traditional machining processes such as micro-milling [1],

chemical etching [2], electrical discharge machining (EDM) [3] and laser micro-machining [4]. The

latter process is commonly used for the micro-machining, but requires relatively expensive, high-

frequency lasers to minimize thermal damage and micro-cracking, as explained by Jandeleit et al.

(1998) [5].

Abrasive jet micro-machining (AJM) is an erosion-based material removal process in which

abrasive particles are blasted toward a target in an air jet. It has been used in recent years to make

MEMS devices such as inertial sensors in Blloy et al. (2000) [6], and microfluidic components in

Schlautmann et al. (2001) [7]. Due to the relatively large erosive footprint compared to the nozzle

size, AJM typically involves the use of patterned erosion resistant masks in order to define the micro-

feature edges as explained by Zhang et al. (2005) [8].

Abrasive water jet machining (AWJM) is similar to AJM, except that water is used to

accelerate the abrasive particles instead of air. In both processes, material removal occurs without

thermal damage, allowing the material to be machined without changing its intrinsic properties as

2

explained by Liu (2010) [9]. High-pressure abrasive water jets (150-400 MPa) have been extensively

used for cutting metals, ceramics, polymers, and composite materials as reviewed in detail by

Momber and Kovacevic (1992) [10]. In typical AWJM systems, a target is eroded by the cutting

action of small diameter (70-1500 µm), high velocity (350-1000 m/s) jet of water or water with

abrasive particles. The abrasive media is added to the water through a feed port immediately prior to

the nozzle.

Miller (2004) [11] introduced abrasive slurry-jet micro-machining (ASJM), in which water

and abrasive particles were pre-mixed in a chamber before being pumped through an orifice at

relatively high pressures. The author found similar jet cutting energy densities as AWJM systems,

but at much lower pressures (e.g. 70 MPa). However, the design of the particle storage vessel

resulted in excessive settling and thus inconsistency in the abrasive flux. This problem was resolved

by utilizing a shaker in the low-pressure (2-14 MPa) ASJM apparatus used in the work of Pang et al.

(2010) [12]. Despite the improvement, the machined features suffered from surface waviness caused

by mechanical vibrations. Both of these shortcomings were overcome in the low-pressure (2-8 MPa)

apparatus developed by Nouraei et al. (2014a) [13], who demonstrated the high repeatability of the

system by machining micro-channels and micro-holes in glass with a maximum variation in the

depth and width along a single channel of less than 3%. They concluded that low-pressure ASJM is a

promising technology in the manufacturing of microfluidic and MEMS devices.

Kowsari et al. (2014a) [14] demonstrated the feasibility of ASJM to machine micro-

channels and micro-holes in sintered alumina, and used a basic surface evolution model to predict

the shapes of the profiles. However, that study was limited to feature depths smaller than 50 m so

that the near-flat target geometry had no effect on the slurry flow field. The modeling of deeper,

more practical features must account for changes in the erosive flow.

3

Wang et al. (2009) [15] drilled blind holes in glass using a slurry jet apparatus (6-14 MPa

pressure) and found that the hole openings were rounded into a bell-shaped cross-section. The

authors hypothesized that the edge rounding occurred during the initial stage of machining, but they

did not investigate the mechanism. Moreover, the results were obscured by a high degree of

asymmetry in the holes. Kowsari et al. (2014a) [14] and Kowsari et al. (2014b) [16] used ASJM to

drill sub-millimeter sized blind and through-holes in glass, polymethylmethacrylate (PMMA),

metals, and sintered ceramics. The same group predicted the shapes of ASJM holes in glass using a

computational fluid dynamics (CFD)-aided surface profile model in Nouraei et al. (2014b) [17].

While these models were able to accurately predict the hole depths, there were significant errors in

the predicted opening width due to a progressive edge rounding that was caused by a mechanism

not captured by the CFD erosion model. Liu et al. (2015) [18] used high-pressure AWJM (80-150

MPa) to drill holes in a titanium alloy (Ti-6Al-4V). They noted a ring-shaped zone near the hole

openings in which the surface was rougher, and hypothesized that it might be caused by cavitation.

However, they did not pursue the hypothesis using CFD. Therefore, the erosion mechanism

responsible for edge rounding observed in existing slurry-jet micro-machining studies has not yet

been explained.

The surface finish of micro-features such as channels and holes can affect fluid flow in

microfluidic applications. For example, relatively rough channels can lower the separation

efficiency and increase the solute dispersion as explained by Ghobeity et al. (2012) [19] and

Solignac et al. (2001) [20]. Moreover, Zhao et al. (2003) [21] found that surface roughness

significantly affected micro-scale adhesion contact in MEMS devices. For these reasons,

methodologies for the smoothing of channels have been investigated by Haj Mohammad Jafar et al.

(2013) [22] for air-driven abrasive jet machining (AJM). The authors reduced the channel centerline

4

roughness by up to 60% by ‘post-blasting’; i.e. performing a secondary blast using smaller jet

angles and particle velocities. In another AJM study, Wensink et al. (2002) [23] found that channels

in glass and silicon were smoothed after post-blasting with relatively small particles (3 and 9 μm

diameter alumina) or annealing at 750C, but roughened after treating with hydrofluoric (HF)

etching. However, the erosive footprint in AJM is relatively large (3 mm wide channels) compared

to that in ASJM, requiring patterned masks to reduce the blast footprint to the sub-millimeter range.

Moreover, the higher impact velocities and local impact angles in AJM can lead to greater

roughness in brittle materials. For example, the use of 25 μm diameter alumina particles in AJM

resulted in an average channel centerline roughness, Ra, of approximately 1.6 m in glass, which

was approximately 54% larger than that measured by Haj Mohammad Jafar et al. (2015) [24] for

ASJM using 25 μm particles. These differences can be attributed to differences in damage

mechanisms explained in detail in Section 4.4.2. Post-blasting of ASJM micro-channels may

provide further reductions in roughness, although this has not been investigated in the literature.

The smoothing of sintered ceramics has been demonstrated by, for example Choi et al. (2004) [25],

using chemical mechanical polishing (CMP). However, the relatively large flat CMP pad limited

the process to bulk surface finishing, without the capability to smooth the inner walls of micro-

features.

A useful feature of the ASJM process is its capability to erode brittle and ductile materials at

different rates by controlling the process parameters. The selective removal of metallic layers is of

industrial interest in many applications including as heat sinks for electronic components as well as

enclosures for such components. The use of ASJM as a low-cost and relatively quick alternative for

the selective removal of metallic layers compared to conventional processes such as CMP has not

been attempted.

5

In summary, most previous research on ASJM has been limited to relatively soft ceramics

such as borosilicate glass. The overall objective of the present research was to use a combination of

experiments and modeling to understand and predict the relationships between ASJM operating

conditions and the erosion in a variety of material systems of industrial interest: (i) sintered

ceramics such as alumina, aluminum nitride, zirconium tin titanate; (ii) metals such as

electrodeposited copper and nickel-phosphorous; and (iii) ceramic-metallic and metallic-metallic

composites such as aluminum nitride with metallic-filled through-holes and nickel-phosphorous-

electrodeposited aluminum, respectively.

1.2. Objectives

The thesis research had the following specific objectives:

1. Develop CFD models to obtain the flow fields and particle trajectories in the ASJM of

typical machined features such as holes, channels, and pockets, and then use these flow

fields to optimize the erosion process.

2. Develop a modified version of an existing surface evolution model to enable the prediction

of the cross-sectional profiles of ASJM holes, channels, and pockets.

3. Control the shape and minimize edge rounding near the opening of holes, channels, and

pockets.

4. Investigate maskless smoothing to minimize the surface roughness of channels machined

using ASJM.

5. Explore the possibility of the selective removal of metallic layers covering metallic and

ceramic substrates using ASJM.

6

1.3. Experimental apparatus

The ASJM apparatus used in the present work utilizes an abrasive slurry pump

(LCA/M9/11-DC, LEWA Inc., Leonberg, Germany) and a pulsation damper (FG 44969/01-9,

Flowguard Ltd., Houston, TX, USA) connected to an open reservoir tank (Fig. 1.1(a)), permitting

operation over a relatively wide range of flow rates and pressures. The pre-pressurized pulsation

damper was installed downstream of the pump to reduce pressure and flow rate pulsations to within

±3%.

A sharp sapphire orifice with a diameter of 180 m having a length-to-diameter ratio of 1.67

(KMT Waterjet, KS, USA, Fig. 1.1(b)) produced a turbulent jet having a velocity of 89 m/s for a

typical back-pressure of 4 MPa. The jet had a Reynolds number of 13,350 since its diameter was

measured to be 150 m using a microscope attached to a digital camera (field of view of 3×2 mm).

The diameter was consistent over the standoff distance (20 mm); i.e. the distance between the

orifice and target, which was below the theoretical breakup length, computed to be 36 mm as

explained in Kowsari et al. (2013) [26]. Typical slurries contained water and 1 wt% alumina

abrasive particles having a nominal diameter of 10 m (Comco Inc., CA, USA) premixed in a

reservoir tank using a propeller (Fig. 1.1(a)). Homogeneity was confirmed from jet concentration

measurements over periods of up to 1 hour, thus did not negatively affect the repeatability of the

micro-machined features.

7

(a)

(b)

Figure 1.1 (a) Schematic of the ASJM apparatus (not to scale), Kowsari et al. (2016) [27]. (b)

Orifice geometry, reproduced from Kowsari et al. (2014) [16].

8

1.4. Thesis outline

The first chapter introduces ASJM and reviews the literature, while Chapter 2 presents a

CFD-aided surface profile model developed for ASJM channels in sintered ceramics (objectives 1

and 2). This work has been published as:

K. Kowsari, H. Nouraei, B. Samareh, M. Papini, J.K. Spelt, CFD-aided prediction of the shape of

abrasive slurry-jet micro-machined channels in sintered ceramics, Ceramics International 42

(2016) 7030-7042.

Chapter 3 focuses on the control of the shape of ASJM holes and channels in brittle

materials by minimizing cavitation erosion within the flow field (objectives 1 and 3), and has been

published as:

K. Kowsari, M.H. Amini, M. Papini, J.K. Spelt, The effects of fluid vapor pressure and viscosity on

the shapes of abrasive slurry-jet micro-machined holes and channels, International Journal of

Machine Tools & Manufacture 110 (2016) 80-91.

Chapter 4 explores the roughness of ASJM channels in brittle and ductile materials, and

investigates smoothing by post-blasting (objective 4). This work has been submitted for publication

as:

K. Kowsari, J. Schwartzentruber, J.K. Spelt, M. Papini, Erosive smoothing of abrasive slurry-jet

micro-machined channels in glass, PMMA, and sintered ceramics: experiments and roughness

model, Precision Engineering (2016, submitted).

9

Chapter 5 includes a hybrid AJM-ASJM process developed for micro-milling of ceramic-

metallic composite substrates (objectives 1 and 5), and has been published as:

K. Kowsari, M.R. Sookhaklari, H. Nouraei, M. Papini, J.K. Spelt, Hybrid erosive jet micro-milling

of sintered ceramic wafers with and without copper-filled through-holes, Journal of Materials

Processing Technology 230 (2016) 198-210.

Chapter 6 presents a CFD procedure for the prediction of the erosive footprint size in the

AJM of flat and curved targets (objective 1). This work has been published as:

K. Kowsari, A. Nouhi, V. Hadavi, J.K. Spelt, M. Papini, Prediction of the erosive footprint in the

abrasive jet micro-machining of flat and curved glass, Tribology International 106 (2016) 101-108.

Chapter 7 investigates the selective removal of metallic layers from ceramic and metallic

substrates using ASJM (objectives 1 and 5), and has been submitted for publication as:

K. Kowsari, M. Papini, J.K. Spelt, Selective removal of metallic layers from sintered ceramic and

metallic substrates using abrasive slurry-jet micro-machining, Journal of Manufacturing Processes

(2016, submitted).

Finally, the main conclusions of this dissertation and recommendations for future work are

the subject of Chapter 8.

10

1.5. References

[1] X. Cheng, A. Wang, K. Nakamoto, K. Yamazaki, A study on the micro tooling for micro/nano

milling, International Journal of Advanced Manufacturing Technology 53 (2011) 523-533.

[2] C. Iliescu, B. Chen, F.E.H. Tay, G. Xu, J. Miao, Characterization of deep wet etching of glass,

Proc. of SPIE Vol. 6037 (2003) 60370A-2.

[3] C. T. Yang, S.S. Ho, B.H. Yan, Micro-hole machining of borosilicate glass through electro

mechanical discharge machining, Key Engineering Materials 196 (2001) 149- 166.

[4] H. Ogura, Y. Yoshida, Hole drilling of glass substrates with a CO2 laser, Japanese Journal of

Applied Physics 42 (2003) 2881–2886.

[5] J. Jandeleit,, A. Horn, R. Weichenhain, E.W. Kreutz, R. Poprawe, Fundamental investigations of

micromachining by nano- and picosecond laser radiation, Applied Surface Science 127-129

(1998) 885-891.

[6] E. Blloy, S. Thurre, E. Walchiers, A. Sayah, M.A.M. Gijs, The introduction of powder blasting

for sensor and micro system applications, Sensors and Actuators 84 (2000) 330-337.

[7] S. Schlautmann, H. Wensink, R. Schasfoort, M. Elwenspoek, A. Vandenberg, Powder-blasting

technology as an alternative tool for micro fabrication of capillary electrophoresis chips with

integrated conductivity sensor, Journal of Micro-mechanics and Micro-engineering 11 (2001)

386-389.

[8] L. Zhang, T. Kuriyagawa, Y. Yasutomi, J. Zhao, Investigation into micro abrasive intermittent

jet machining, International Journal of Machine Tools and Manufacture 45 (2005) 873–879.

[9] H. T. Liu, Water jet technology for machining fine features pertaining to micro-machining,

Journal of Manufacturing Processes 12 (2010) 8-18.

[10] A.W. Momber, R. Kovacevic, Principles of Abrasive Water Jet Machining, Springer Verlag

Limited, London (1992).

[11] D.S. Miller, Micro-machining with abrasive water jets, J. of Materials Processing Technology

149 (2004) 37-42.

[12] K.L. Pang, T. Nguyen, J.M. Fan, J. Wang, Machining of micro-channels on brittle glass using an

abrasive slurry jet, Key Eng. Mat. 443 (2010) 639-644.

[13] H. Nouraei, K. Kowsari, J.K. Spelt, M. Papini, Surface evolution models for abrasive slurry jet

micromachining of channels and holes in glass, Wear 309 (2014a) 65-73.

[14] K. Kowsari, H. Nouraei, M. Papini, J.K. Spelt, Surface evolution models for abrasive slurry jet

micro-machining of channels and holes in alumina, Proceedings of the 9th international

conference on micromanufacturing (ICOMM) (2014a).

11

[15] C.Y. Wang, P.X. Yang, J.M. Fan, Y.X. Song, Effect of slurry and nozzle on hole machining of

glass by micro abrasive suspension jets, Key Eng. Mat. 404 (2009) 177-183.

[16] K. Kowsari, H. Nouraei, D.F. James, M. Papini, J.K. Spelt, Abrasive slurry jet micro-machining

of holes in brittle and ductile materials, J. Materials Processing Tech. 214 (2014b) 1909–1920.

[17] H. Nouraei, K. Kowsari, B. Samareh, M. Papini, J.K. Spelt, A combined numerical-analytical

methodology for surface profile prediction of abrasive slurry jet micro-machined holes,

Proceedings of the 10th international conference on micromanufacturing (ICOMM), Milan, Italy

(2014b).

[18] H.X. Liu, Q.M. Shao, C. Kang, C. Gong, Assessment of cavitation and impingement effects of

submerged water jet on Ti alloy surface, Materials Research Innovations 19 (2015) S1-70-74.

[19] A. Ghobeity, H.J. Crabtree, M. Papini, J.K. Spelt, Characterisation and comparison of

microfluidic chips formed using abrasive jet micromachining and wet etching, J.

Micromechanics and Microengineering 22 (2012) 025014.

[20] D. Solignac, A. Sayah, S. Constantin, R. Freitag, M.A.M. Gijs, Powder blasting for the

realization of microchips for bio-analytic applications, Sensors and Actuators A 92 (2001) 388–

393.

[21] Y.P. Zhao, L.S. Wang, T.X. Yu, Mechanics of adhesion in MEMS—a review, J. Adhesion Sci.

Tech. 17 (2003) 519–546.

[22] R. Haj Mohammad Jafar, M. Papini, J.K. Spelt, Simulation of erosive smoothing in the abrasive

jet micro-machining of glass, Mater. Process. Technol. 213 (2013) 2254-2261.

[23] H. Wensink, S. Schlautmann, M.H. Goedbloed, M.C. Elwenspoek, Fine tuning the roughness of

powder blasted surfaces, J. Micromech. Microeng. 12 (2002) 616-620.

[24] R. Haj Mohammad Jafar, H. Nouraei, M. Emamifar, M. Papini, J.K. Spelt, Erosion modeling in

abrasive slurry jet micro-machining of brittle materials, J. Manuf. Proc. 17 (2015) 127–140.

[25] W. Choi, J. Abiade, S. Lee, R.K. Singh, Effects of slurry particles on silicon dioxide CMP, J.

Electrochemical Soc. 151 (8) (2004) G512-G522.

[26] K. Kowsari, D.F. James, M. Papini, J.K. Spelt, The effects of dilute polymer solution elasticity

and viscosity on abrasive slurry jet micro-machining of glass, Wear 309 (2013) 112-119.

[27] K. Kowsari, M.H. Amini, M. Papini, J.K. Spelt, The effects of fluid vapor pressure and viscosity

on the shapes of abrasive slurry-jet micro-machined holes and channels, Int. J. Machine Tools &

Manuf. 110 (2016) 80-91.

12

Chapter 2: CFD-aided Prediction of the Shape of

Abrasive Slurry-jet Micro-machined Channels in

Sintered Ceramics

2.1. Introduction

Abrasive slurry-jet micro-machining (ASJM) has recently been proven by Kowsari et al.

(2016) [1] to be feasible for the micro-milling of difficult-to-machine materials such as fine-grained

sintered ceramics used in microfluidic and microelectromechanical (MEMS) devices. Laser micro-

milling is commonly used for the machining of such materials, but requires relatively expensive,

high-frequency lasers to avoid thermal damage and micro-cracking, as explained by Jandeleit et al.

(1998) [2]. ASJM is a relatively low-cost alternative.

Surface evolution models for predicting the evolving shape of features machined using

abrasive air jet processes were first introduced by Slikkerveer and in’t Veld (1999) [3], and ten

Thije Boonkkamp and Jansen (2002) [4]. Ghobeity et al. (2008) [5] introduced the use of a shallow

"first-pass eroded profile" in these models to capture the erosive efficacy distribution as seen by the

target in abrasive air jet machining (AJM). Nouraei et al. (2014) [6] adapted this model for use in

ASJM, but found that the model could only predict multi-pass channel profiles in glass, where the

channel depths were linearly proportional to the number of machining passes (particle dose). The

model could not predict the profiles of blind holes in glass since the specific erosion rate (mass of

material removed per mass of erodent) decreased with hole depth due to the more confined flow

field that progressively changed throughout the hole formation. Nevertheless, Kowsari et al. (2014)

[7] were able to apply this traditional, first-pass surface evolution model to predict the cross-

sectional profiles of relatively shallow micro-holes and micro-channels (<50 μm deep) in sintered

13

alumina; i.e. cases where the decrease in specific erosion rate could be neglected. In another study,

Kowsari et al. (2016) [1] found through CFD modeling that the flow field changed significantly as

channels in sintered alumina became deeper and evermore "V"-shaped, thereby enlarging the

stagnation zone at the bottom and reducing the impact velocities of the abrasive slurry particles at

perpendicular jet incidence. This introduced a nonlinearity in the specific erosion rate, such that the

depths of channels increased less-than-linearly with increasing particle dose. These authors also

found that the shapes of the channels varied throughout the machining process; i.e. the normalized

depth at any point on the profile was not a unique function of the normalized width (normalized by

the instantaneous centerline depth and channel opening width, respectively). No attempt was made

in Kowsari et al. (2016) [1] to model the specific erosion rate nonlinearity and the change in

channel shape with depth.

Nonlinearity in specific erosion rate with increasing micro-channel depth was also observed

by Haghbin et al. (2015) [8] during the high-pressure abrasive waterjet micro-machining (AWJM)

of high-aspect-ratio channels in 316L stainless steel and 6061-T6 aluminum alloy. In their work, an

existing surface evolution model similar to that used in Kowsari et al. (2014) [7] was corrected

using empirical coefficients to account for the nonlinearity. Billingham et al. (2013) [9] developed a

model to predict the profiles of over-lapping channels made using AWJM in a titanium-based alloy

(Ti6Al4V), but the model was limited to shallow features by specific erosion rate nonlinearities. All

of these existing studies assumed that the erosion pattern produced by the passage of a jet (the

erosive efficacy) could be estimated using the shape of a shallow, “first-pass” profile.

Recently, the present authors developed a novel numerical-empirical methodology to predict

the cross-sectional profiles of deep channels made with ASJM in ductile materials such as

polymethylmethacrylate (PMMA), 6061-T6 aluminum alloy, 316L stainless steel and Ti–6Al–4V

14

titanium alloy [10]. The erosive efficacy of a first-pass channel, obtained using a CFD simulation

rather than from an experimental measurement, was used to capture the progressive widening of

ASJM channels in ductile materials. In these ductile materials the shapes of the channels remained

constant at any depth; i.e. the normalized depth at any point on the profile was a unique function of

the normalized width, and the centerline specific erosion rate remained unchanged with depth.

Therefore, the slurry flow field remained unchanged throughout the machining, in contrast to ASJM

channels in sintered ceramics which are subject to a nonlinearly decreasing specific erosion rate as

the channels become deeper and had different shapes when machined with the jet at 90 incidence.

The present paper describes a novel profile prediction model applicable to the ASJM of

deep channels in brittle, sintered ceramics (aluminum nitride (AlN), alumina (Al₂O₃), and

zirconium tin titanate (Zn-Sn-TiO₂)). The model uses CFD to predict the evolving slurry erosion

patterns on the walls of deepening channels, thereby capturing the changes in the flow field that

caused a nonlinear decrease in the specific erosion rate and a change in the channel shape at

perpendicular jet incidence. Data are also presented for machining at 45 jet incidence where

channel shape did not vary with depth.

15

2.2. Experiments and flow modeling

There were two categories of micro-channel machining experiments: (i) channels machined

under a range of process conditions to examine the variation of both the channel centerline specific

erosion rate and the channel cross-sectional shape with depth (data presented in Section 2.3.1), and

(ii) channels machined to measure the erosion properties of the sintered ceramics (dependence on

particle angle and speed; data presented in Section 2.3.2) for use in the profile models of Section

2.4. The experimental results shown in Section 2.3 are presented with associated CFD modeling to

aid in their interpretation. The experimental procedures and apparatus are explained in Section

2.2.1, while the common features of the CFD modeling used in Sections 2.3 and 2.4 are given in

Section 2.2.2.

2.2.1. Experiments

Channels were machined in aluminum nitride, alumina, and zirconium tin titanate (Table

2.1) by scanning the target with respect to the slurry jet at a known speed using a motorized stage

(KT-LSM100A, Zaber Technologies Inc., Vancouver, BC, Canada). Table 2.2 presents the standard

ASJM conditions used in the channel machining experiments, and the two experiments used to

characterize the fundamental properties of the erosion of each ceramic; i.e. its dependence on jet

angle and particle velocity. A free jet velocity of 89 m/s (pressure of 4 MPa) was maintained in the

channel machining experiments. The aqueous slurry contained 1 wt% alumina abrasive particles

(Comco Inc., Burbank, CA, USA; density 3900 kg/m3; Vickers hardness 16 GPa) having a nominal

diameter of 10 m, and was stirred continuously to maintain homogeneity. Cross-sectional channel

16

profiles were measured using an optical profilometer with a lateral resolution of 426 nm and depth

resolution of 16 nm (ST400, Nanovea Inc., CA, USA).

Table 2.1 Properties of the target materials.

Composition Supplier Dimensions

(mm)

Grain size

(m)

Density

(g/cm³)

Vickers

hardness

(kg/mm²)

Alumina

(Al₂O₃) Superstrate 996, CoorsTek

Inc., Golden, CO, USA 10×10×0.375 < 1 3.88 1800

Aluminum

nitride (AlN)

K170, Toshiba Corp., Minato,

Tokyo, Japan 50×50×0.375 < 1 3.26 1100

Zirconium tin

titanate

(Zn-Sn-TiO₂)

M39, Maruwa, Owariasahi-shi,

Ach, Japan 50×50×0.375 < 5 5.20 950

Table 2.2 Standard process parameters.

Type of experiment

Channel

machining Velocity exponent

Impact angle

function

Pressure (MPa) 4.0 1.2 2.0 3.0 4.0 6.0 8.0 4.0

Slurry flow rate (mL/s) 1.67 1.20 1.34 1.50 1.67 2.00 2.34 1.67

Free jet velocity (m/s) 90 49 63 78 89 110 127 90

Particle concentration (wt%) 1 1.21 1.15 1.08 1.00 0.85 0.70 1

Standoff distance (mm) 20 20 20

Jet traverse speed (mm/s)

Aluminum nitride (AlN) 0.01

Alumina (Al2O3) 0.0025

Zirconium tin titanate

(Zn-Sn-TiO₂)

0.02

Jet incidence (°) 90, 45 90 15, 30, 45, 60, 75, 90

17

2.2.2. CFD modeling

Flow fields, particle trajectories, and erosion patterns of ASJM flows were modeled using

ANSYS Fluent 15.0 (ANSYS Inc., Cecil Township, PA, USA). A multiphase, steady volume-of-

fluid model was used for the water-jet surrounded by air. ANSYS (2015) [11] explained that the κ-

ω shear-stress transport turbulence model, which blends the robust formulation of the κ-ω model in

the near-wall region with the free-stream independence of the κ-ε model, was optimal for highly-

strained flows at much quicker convergence times compared to the standard κ-ε model. The domain

was meshed with quadrilateral 2 μm elements and the simulations converged with residuals of 10-3

.

The water-air interface and centerline flow velocities were within 3% for elements smaller than 5

μm, indicating convergence of the solution to a mesh-independent state. The erodent particles were

injected through the grid elements of the inlet plane at a velocity equal to that of the fluid and were

tracked using the one-way coupling Lagrangian discrete phase model, because their low

concentration eliminated significant particle-particle and particle-flow interactions. The particle

shape factor was 0.76 (area of the sphere having the particle volume divided by the actual particle

surface area), as measured by Dehnadfar et al. (2011) [12]. As described in Nouraei et al. (2016)

[10], the effect of the squeeze film (a thin layer of liquid separating the approaching solid particles

from the target wall) was considered to be insignificant since the range of the relative particle

Reynolds number in ASJM was much larger than the critical range suggested by Clark (2004) [13].

The particle size distribution in manufactured abrasive powders can significantly affect the

resulting erosion pattern due to the size variation in the momentum equilibration number, λ,

(Humphrey, 1990) [14] which reflects the tendency for particles to follow streamlines. For example,

a 2D axisymmetric CFD model of a jet striking a flat plate at normal incidence showed that 2 μm

particles created an erosive footprint of 410 μm in diameter compared with a 220 μm diameter

18

footprint for 20 μm particles. Therefore, particle injection in the present study reflected the actual

particle size distribution (Comco Inc., Burbank, CA, USA). Table 2.3 summarizes the size and mass

flow rate distributions for the 10 μm nominal diameter alumina abrasive and the standard slurry

flow rate of 1.67 mL/s used in the channel machining experiments (Table 2.2).

Table 2.3 Particle size distribution for 10 μm alumina particles (Comco Inc., Burbank, CA, USA).

Mass flow rate for each fraction based on standard conditions of 1 wt% particle concentration and

1.67 mL/s slurry flow rate (Table 2.2).

Particle

diameter (µm)

% volume

in

distribution

Mass flow rate

(mg/s)

2 5.4 0.09

5 6.0 0.10

7 32.9 0.55

10 28.7 0.48

12 13.2 0.22

14 7.8 0.13

15 4.2 0.07

20 1.8 0.03

2.3. Results and discussion

2.3.1. ASJM channels in sintered ceramics – changes in centerline specific

erosion rate and shape with depth

As mentioned in Section 2.1, Kowsari et al. (2016) [1] found that the depths of ASJM multi-

pass channels in sintered alumina at perpendicular incidence increased less-than-linearly with

increasing dose (i.e. number of passes) due to the enlargement of the stagnation zone in the "V"-

19

shaped channels with increasing channel depth. Particles that travelled through the larger stagnation

zone of the relatively deep channels experienced more drag before impact and caused less erosion.

It was of interest in the present study to examine whether this effect could be avoided by machining

with an inclined jet in the backward configuration shown in Fig. 2.1(a), where the jet flows

predominately along the channel after impact. However, Fig. 2.1(b) shows that the channel

centerline depth increased nonlinearly with the number of oblique passes, reflecting a decreasing

specific erosion rate with increasing depth, just as at normal jet incidence.

To better-understand the flow fields involved, shallow (74 μm deep) and relatively deep

(115 μm deep) channels were modeled in CFD with the domains of Fig. 2.1(a). The stagnation zone

was defined as the region having a gage pressure of 0.5 MPa or greater, and its size was defined by

its cross-sectional areas in the plane of symmetry plane of the channel as shown in Fig. 2.1(a). It

was seen that the deeper channel had a stagnation zone area that was about 18% larger than that of

the shallower channel for a 55% increase in depth. Therefore, particles in the deeper channel would

have smaller impact velocities and produce lass erosion than those in shallower channels.

Moreover, Fig. 2.1(c) shows that the channel width did not depend on channel depth and the

number of passes since, as shown in Fig. 2.1(a), the 10 μm alumina particles flowed along the

channel length rather than striking the sidewalls. This in turn caused the channel shape to remain

unchanged as shown in Fig. 2.1(d). In summary, nonlinearity in the specific erosion rate could not

be avoided in the ASJM of channels in sintered ceramics, and must be accounted for in profile

modeling.

20

(a) (b)

(c) (d)

Figure 2.1 Nonlinearity of centerline specific erosion rate in deep channels machined in backward

configuration at 45 in aluminum nitride using the standard conditions (Table 2.2). (a) CFD model

of a relatively deep channel (5 passes) showing stagnation zone and particle flow, (b) Channel

depth versus dose (number of machining passes). (c) Cross-sectional profiles of multi-pass

channels. P denotes the number of machining passes. (d) Profiles of 1-5 pass channels, each

normalized by their width and depth.

21

Figure 2.2(a) shows cross-sectional half-profiles of multi-pass ASJM channels in sintered

aluminum nitride using standard conditions and 90 jet incidence, while Fig. 2.2(b) shows that the

corresponding depth increased nonlinearly with increasing machining passes, reflecting the

expected decrease in the specific erosion rate with increasing depth. Similar results were obtained

for the other sintered ceramics. Figures 2.2(c) and 2.2(d) show the domain and pressure contours for

CFD models of a relatively shallow and deep channel, where the stagnation zone was defined as the

region with a pressure greater than 0.5 MPa. As in Fig. 2.1(a), it was found that the nonlinearity in

specific erosion rate was due to an enlargement in the stagnation zone, which grew by 12% in the 5-

pass channel compared with the first-pass channel.

In contrast to machining with the jet at 45 (Fig. 2.1), a close examination of the profiles of

Fig. 2.2(a) shows that the channel formation occurred in two stages when the jet was at normal

incidence. In stage I, the channel was relatively shallow with a depth/width aspect ratio less than

0.36 (passes 1 and 2), and the channel widened and deepened in region A with each pass due to the

sidewall erosion caused by the lateral flow depicted in Fig. 2.2(a) and 2.2(c). In stage II, Fig. 2.2(a)

shows that the opening width of the channel (in the plane of the target surface) stopped growing

(i.e. region C (Fig. 2.2(a)) did not change with increasing machining passes), and erosion served

only to deepen the channels in region B and hence make the sidewalls steeper. This is explained by

the CFD model of Fig. 2.2(d) which shows that the slurry flow in deeper channels was directed

along the channels and that lateral flow over the upper edge of the sidewalls ceased. This is further

illustrated in Fig. 2.2(e), which shows that the profile of 1-12 pass channels, each normalized by

that channel width and depth, did not conform to a unique shape. Thus the ASJM channel profiles

machined at normal incidence in a sintered ceramic target did not evolve with a constant shape due

22

to the changing flow field in contrast to what was seen in previous studies of ASJM of glass

(Nouraei et al. (2014) [6]) and ductile materials (Nouraei et al. (2016) [10]).

The two-stage channel formation was eliminated when the jet was inclined at 45 (Fig.

2.1(c)), and the cross-sectional profiles developed with a constant channel opening width, as in

region B in stage II of Fig. 2.2(a). This was illustrated in Fig. 2.1(d), which shows that the

normalized profile shape of the 45 multi-pass channels was preserved as the channels deepened.

The widths of these 45 channels were about 7% smaller than the channels machined at 90 in Fig.

2.2(a), consistent with the observations of Nouraei et al. (2016) [15] for the oblique ASJM glass

channels. Another advantage of machining at an oblique jet incidence was that the nonlinearity in

specific erosion rate was less severe as seen in Fig. 2.2(f), in which both the channel depth and dose

were normalized by those of the first-pass channels. Therefore, oblique ASJM would be preferred

for the machining of high aspect-ratio channels since, as shown in Fig. 2.2(f), the channel depth

began to plateau in 90 machining sooner than at 45; i.e. for a given set of process parameters,

machining at 45 yields a higher aspect ratio channel than at 90. However, Fig. 2.2(b) shows that

the specific erosion rate of 45 machining in relatively shallow channels was approximately 55%

lower than that at 90, requiring relatively long machining times. Another drawback of oblique

machining is the requirement to rotate either the target or the jet when machining continuous

channels with segments that change direction at corners and bends. Because of these limitations, the

90 jet incidence is often preferred over the 45 configuration.

23

(a)

(b) (c)

(d)

24

(e)

(f)

Figure 2.2 Results for aluminum nitride at perpendicular incidence using the standard conditions

(Table 2.2). (a) Cross-sectional profiles of channels using 1-5 machining passes (denoted by P).

Half of the symmetric profiles shown along with the streamline bounding the jet. (b) Depth versus

particle dose (machining passes) of channels at 90 and 45 jet incidence. (c, d) Static pressure

contours and bounding streamlines at the water-air interface of the flows within a relatively shallow

channel (1 pass, (c)) and a relatively deep channel (5 passes, (d)). (e) Profiles of 1-12 pass channels

(denoted by p), each normalized by their width and depth. (f) Depth versus particle dose (machining

passes), both normalized by those of the first-pass channels at 90 and 45 jet incidence.

25

2.3.2. ASJM erosion parameters for sintered ceramics

The erosion model provided by ANSYS Fluent 15.0 (2015) [11], defines the rate of surface

erosion, erosionR , in units of kg/m2s as

1

ParticlesNp

erosion

p cell

m ER

A

(2.1)

where P is the abrasive particle mass flow rate and A

cell is the area of a given computational cell on

the target wall. The function E is the specific erosion rate (mass of material removed per mass

of erodent) at particle impact angle, , given by

90E f E (2.2)

where f expresses the dependence of erosion on the particle impact angle, and 90E is the

specific erosion rate at perpendicular incidence and is related to the particle impact velocity, v, by

90

cE Av (2.3)

where A is a constant associated with the material properties and target substrate for a system, and c

is the velocity exponent which expresses the dependence of erosion on the particle impact velocity

(Oka et al., 1997) [16].

The constants A and c were measured by machining relatively shallow blind holes in each of

the three target materials at perpendicular jet incidence at free jet velocities ranging from 49-127

m/s corresponding to pressures of 1.2-8.0 MPa. These jet velocities corresponded to particle impact

velocities of 21-54 m/s along the jet centerline as predicted using the CFD domain shown in Fig.

2.3.

26

Figure 2.3 Domain and boundary conditions of a two-dimensional axisymmetric model of the 90

impingement of a slurry jet on a flat target.

In order to separate the effects of particle dose and particle velocity (controlled via the

pressure), the particle flow rate was held at 16.7 mg/s by adjusting the abrasive concentration in the

tank (Table 2.2). Therefore, for given scan speed and footprint area, the flux striking the target

(kg/m2s) and dose were kept constant.

The values of A and c were obtained by the best fit of Eq. (2.3) to the data as shown in Fig.

2.4, and are given in Table 2.4. The velocity exponents were between 3.5 and 5.5, which is

significantly larger than those typically found in softer targets (e.g. 1.69 in glass in Nouraei et al.

(2014) [6]). Routbort and Scattergood (1992) [17] suggested that values larger than 2 are typical of

sintered ceramics since material removal involves grain removal through intergranular cracking in

contrast to crater formation and micro-chipping in other ceramics such as glass (Nouraei et al.,

2013) [18]. The role of the stagnation zone in decelerating erodent particles is illustrated in Fig. 2.4

by the large differences between the jet velocity far from the target and the actual particle impact

velocities as predicted by the CFD model.

27

Table 2.4 Best-fit constants (Eq. (2.3)) for the impact velocity dependence of erosion.

Material A [(mg/g)×(m/s)-c

] c

Aluminum nitride 2×10-10

5.64

Alumina 2×10-11

5.48

Zirconium tin titanate 3×10-8

3.58

Figure 2.4 Dependence of ASJM specific erosion rate, normalized by the maximum for each target

material (0.04, 0.02, and 0.08 mg/g for aluminum nitride, alumina, and zirconium tin titanate,

respectively), on jet-centerline-averaged particle impact velocity. Error bars represent ±1 standard

deviation for 3 measurements. The lines show best-fit curves given in Table 2.4.

The impact angle function, f , was obtained from specific erosion rate measurements of

shallow blind holes machined at jet impact angles ranging from 15-90 at a pressure of 4 MPa

corresponding to a jet velocity of 89 m/s. The corresponding actual centerline average particle

impact angles were 10-74 for 10 μm diameter alumina particles, as predicted by CFD simulations

using the domains shown in Fig. 2.3 at perpendicular incidence and Fig. 2.5(a) for oblique

incidence. The centerline average impact angles were lower than the global jet incidences since the

28

spreading slurry in the stagnation zone deflects particles to impact at shallower angles. For

example, Fig. 2.5(a) shows that for a 45 jet impact, particles struck at about 34 on average.

Equation (2.2) was fitted to the data using cubic polynomials as shown in Fig. 2.5(b), with the

coefficients given in Table 2.5. The trends of Fig. 2.5(b) were similar to those found by Nouraei et

al. (2014) [6] in the ASJM of glass, showing that all of the sintered ceramic targets behaved in a

brittle manner such that erosion increased monotonically with increasing impact angle. Kowsari et

al. (2016) [1] observed ASJM erodes sintered ceramics through intergranular cracking and grain

dislodgement. Relative to aluminum nitride and alumina, Fig. 2.5(b) shows that zirconium tin

titanate eroded to a larger degree at shallow impact angles.

29

(a)

(b)

Figure 2.5 (a) Centerline trajectories of 10 μm alumina particles in 45 impingement of a jet on a

flat target. (b) ASJM specific erosion rate normalized by that at perpendicular incidence (0.003,

0.003, and 0.055 mg/g for aluminum nitride, alumina, and zirconium tin titanate, respectively)

versus jet centerline average particle impact angle at a jet velocity of 89 m/s. Error bars represent

±1 standard deviation for 3 measurements. The lines show best-fit curves given in Table 2.5.

30

Table 2.5 Best-fit coefficients of erosion data using 3th

order polynomial 3 2

3 2 1 0f a x a x a x a

Coefficients of polynomial fit (R2 ≥ 0.98)

Aluminum nitride Alumina Zirconium tin titanate

a3 -3.00×10

-6 -5.00×10

-6 4.00×10

-6

a2 3.00×10

-4 5.00×10

-4 -7.00×10

-4

a1 9.70×10

-3 7.80×10

-3 4.63×10

-2

a0

-2.95×10-2

2.41×10-2

4.25×10-2

2.4. Channel profile modeling

Two methodologies were developed to predict the channel cross-sectional profiles as a

function of the number of machining passes. Both models captured the nonlinearity of the specific

erosion rate with increasing depth as well as the two stages of profile shape evolution. The first

approach required the simulation of the slurry flow field for every predicted machining pass and

was therefore computationally expensive. The second method simplified the procedure by

generalizing the characteristics of the flow field and using the first-pass profile to predict the

profiles after all subsequent machining passes.

2.4.1. Method I: CFD erosion simulation of each pass

This method of determining profile development requires a new CFD model of the flow

after each machining pass in order to capture the changing flow field. The objective was to predict

the nonlinear growth in channel depth with increasing dose of blasted erodent. The approach of

Nouraei et al. (2016) [10], which used CFD modeling to capture channel widening in ductile

materials, was unable to capture this effect.

31

Obtaining erosion pattern from CFD

Figure 2.6 shows the trajectories of 10 μm alumina particles in the ASJM flow within the

first-pass channel in Fig. 2.2(a). It is seen that the total erosion consists of the contributions of

primary particle impacts directly within the jet footprint as well as secondary impacts on the

sidewalls by particles carried by the return flow toward the surface. Both of these erosive

components can also be seen in the three-dimensional erosion map in Fig. 2.6(b), and caused the

erosive footprint for this channel to be about 18% wider than the 220 μm footprint of the jet on a

flat target. Following the approach of Nouraei et al. (2016) [10], an effective two-dimensional

erosion pattern (i.e. erosion rate versus distance from centerline) resulting from the passage of the

jet footprint can be obtained from the CFD model by summing the total erosion occurring along

parallel scan lines, as shown in Fig. 2.6(b). To investigate the effect of element size on the shape of

the erosion pattern, the grid was refined by a factor of 2 to 1 μm quadrilateral elements, and the data

points in the refined erosion pattern were within 3% of those of Fig. 2.6(b), thus the erosion pattern

was grid-independent.

(a)

32

(b)

(c)

Figure 2.6 First-pass channel machined in aluminum nitride at perpendicular incidence using the

standard conditions (Table 2.2) (a) Trajectories of 10 μm alumina particles (b) Three-dimensional

erosion map (c) Cross-sectional profiles the first-pass channel and the second-pass channel

predicted using method I.

33

Centerline depth prediction

The CFD-predicted centerline specific erosion rate at a jet angle of 90 was 0E =0.63

kg/m2s, which corresponded to a measured first-pass channel depth, 1d 41 μm using the standard

conditions (Table 2.2). Assuming the ratio of specific erosion rate to depth change remains

unchanged, the thickness of the eroded layer on the centerline after the nth

pass is

11

0

n n n

dd d E

E

(2.4)

For example, the CFD model gave an erosion rate of 1 0.60E kg/m2s along the centerline during

the second pass, which leads to a predicted depth d2=80.0 m, which was within 4% of the

measured depth after the second-pass.

Eroded depth off the centerline

For brittle materials, ten Thije Boonkkamp and Slikkerveer (2002) [19] suggested that local

thickness loss occurs normal to the surface, as illustrated in Fig. 2.6(c) for an arbitrary point p(xo,

yo) which shifts to according to the transformation

sinlossdx t (2.5)

coslossdy t (2.6)

where is the angle between the local normal to the surface and the perpendicular coordinate, y,

and is obtained by finding the local slope of the polynomial function, f x , fitted to the channel

profile.

34

Particle rebound model

Figures 2.6(a) and 2.6(b) showed that the total erosive power of ASJM consisted of both

primary and secondary particle impact components. The location of secondary impacts depends on

the coefficient of restitution (CORn and CORt, defined as the ratio of normal and tangential rebound

to the normal and tangential incident velocity, respectively), which is a function of the properties of

both the target and the erodent (e.g. shape, size, hardness). Restitution coefficient values for

alumina abrasives impacting sintered ceramics have not been reported, therefore a range of

coefficients between 0.2 and 0.8 were used to determine the values that best-predicted the channel

profiles in the same way Nouraei et al. (2016) [10] estimated the COR for ductile targets. As shown

in Fig. 2.7, CORn = CORt = 0.2 gave the best fit to the measured profiles in aluminum nitride, with

larger values leading to an over-prediction of the width. This value was within the 0.2-0.5 range of

ratios of total rebound to incident velocity ratio suggested by Slikkerveer and in't Veld (1999) [3]

for similarly-sized alumina particles on glass targets, and was also found to give reasonably

accurate channel profile predictions in alumina and zirconium tin titanate targets. However,

sidewall erosion was seen only in stage I of channel formation as evident from the lack of slurry

flow on the sidewalls in Fig. 2.2(d); therefore the importance of COR was limited to this initial

stage of channel formation in the ASJM of sintered ceramics.

35

Figure 2.7 Comparison of the measured (solid lines) and predicted (symbols) cross-sectional

profiles of channels machined in aluminum nitride using the standard conditions (Table 2.2) using

COR of 0.2 and 0.8. Half of the symmetric profiles shown. P denotes the number of passes.

2.4.2. Method II: CFD with approximate stagnation zone model

This Section presents a novel profile modeling approach that is much faster than method I,

because it eliminates the need for CFD at each pass of the jet by using an approximate model for

the development of the stagnation zone and profile shape.

Approximation of stagnation zone

For high pressure abrasive water jets, Haghbin et al. (2015) [20] hypothesized that the size

and effect of the stagnation zone could be correlated with the aspect ratio, F, of the “filled region”

shown in Fig. 2.8(a) as the projection of the jet diameter on the channel profile; i.e. F = h/b. In the

present work, this dependence was quantified in a similar manner for low pressure slurry jets.

Figure 2.8(a) shows that the 2.62 fold increase in F for the third-pass channel in aluminum nitride

compared to that of the first-pass channel was accompanied by a 7% enlargement in the thickness

of the stagnation zone and an 8% decrease in the average centerline impact velocity (obtained using

CFD). Figure 2.8(b) shows that there was a single, well-defined linear relationship, between F and

36

the predicted average centerline velocity for aluminum nitride, alumina, and zirconium tin titanate

obtained using CFD domains similar to that in 2(c). The relation 90

v F and Eq. (2.3) then allow

for the prediction of the centerline impact velocity as a function of channel depth, and hence the

prediction of the depth after the next machining pass. For example, starting with a flat aluminum

nitride target (c = 5.64, 0 0F ), Fig. 2.8(b) gives the first-pass channel filled aspect ratio

1 0.082F . The centerline depth after the first pass was 1 41d μm (Fig. 2.9(a)), so that the depth

of the second-pass channel can be predicted as

5.64

1902 1 1

090

39.7041 41 74.8

41.08

c

v Fd d d

v F

μm (2.7)

This follows from the depth increment between the first and second passes, d2-d1, being

proportional to E90(1) (erosion rate of flow striking the first-pass channel), and d1 being proportional

to E90(0) (erosion rate of flat target), so that (d2-d1)/d1 is equal to E90(1)/E90(0) or, by Eq. (2.3), the

bracketed term in Eq. (2.7). The error is this prediction was 2.8%, compared to 3.9% using method

I.

37

(a)

(b)

Figure 2.8 ASJM channels machined in aluminum nitride at 90 incidence using the standard

conditions (Table 2.2). (a) Cross-sectional profiles and pressure contours (b) Jet-centerline-

averaged impact velocities for 10 μm alumina particles versus filled-region aspect ratio of multi-

pass channels having depth/width aspect ratios of up to about 0.8 in aluminum nitride, alumina, and

zirconium tin titanate using the standard conditions (Table 2.2).

38

Approximation of profile shape

Regardless of the type of sintered ceramic target, experiments at 90 jet incidence (Fig.

2.2(a)) showed that the opening width of the deepening channels in stage II (beyond an aspect ratio

of 0.36) was approximately equal to the width of the jet footprint on a flat target (220 μm) since

lateral flow toward the sidewalls stopped for greater aspect ratios. This is illustrated in Fig. 2.9(a)

for aluminum nitride at 90 jet incidence. Based on this fundamental flow effect, profile modeling

of channels machined at 90 using method II required trimming the first-pass erosion pattern to

about 220 μm in prediction of second-stage channels. Haghbin et al. (2015) [8] used a similar

approach in the profile modeling of AWJM channels in aluminum by reducing the width of the

erosion pattern to the diameter of the jet in deeper (stage II) channels. However, the significant

divergence of the AWJM jet required the use of an empirically-fitted function to estimate the jet

diameter as a function of channel depth. For the non-divergent ASJM jet, the thickness loss beyond

x = 110 μm was set to be zero regardless of the channel depth as shown in Fig. 2.9(b). This enables

the use of the first-pass erosion pattern to predict all subsequent machining passes, thus minimizing

the number of CFD simulations. For example, Fig. 2.9(c) shows good agreement between the

trimmed erosion pattern of the first-pass channel and measured third-pass channel in aluminum

nitride, each normalized by their respective thickness loss.

As will be discussed in Section 2.4.2.2, machining at a jet angle of 45 eliminated the need

to modify the channel shape from stage I to II, since with opening width remained constant as

shown in Fig. 2.1(c).

39

(a)

40

(b) (c)

Figure 2.9 Multi-pass channels machined in aluminum nitride at perpendicular incidence using the

standard conditions (Table 2.2). (a) Erosive footprints of three-dimensional erosion maps.

Comparison of two-dimensional erosion patterns, each normalized by their centerline thickness

loss, of (b) the measured and trimmed 1-pass channel in stage I (Fig. 2.2(a)), and (c) the measured

3-pass channel in stage II of channel formation, and the trimmed 1-pass channel in stage I.

2.4.2.1. Method II predictions - 90 machining

In summary, method II predicts the profile of subsequent machining passes using only the

erosion pattern of the first-pass channel (obtained using CFD) together with the depth of the next

pass obtained using 90

v F , as was illustrated in Eq. (2.7). Beyond the first machining pass, and

the initial modeling required to produce Fig. 2.8(b), the procedure required no further CFD. Figure

2.10 compares the cross-sectional profiles of measured multi-pass channels with those predictions

using method II for aluminum nitride, alumina, and zirconium tin titanate. It is seen that the depth

predictions were within 8% of those of the measured channels at any distance from the centerline

for aspect ratios of approximately 0.5.

The results indicated that for a given increase in F, the decrease in erosion rate was larger in

aluminum nitride (c = 5.64) than in zirconium tin titanate (c = 3.58), since the degree of erosion was

proportional to particle impact velocity through the velocity exponent (Eq. (2.3)). Hence, the degree

41

of deviation from a linear increase in erosion rate with increasing dose depended not only on

channel shape, but also on the velocity exponent.

(a) (b)

(c)

Figure 2.10 Comparison of predicted (symbols) and measured (solid lines) cross-sectional channel

profiles in (a) aluminum nitride, (b) alumina, and (c) zirconium tin titanate at perpendicular jet

incidence using the standard conditions. Half of the symmetric profiles shown. P denotes the

number of passes.

42

2.4.2.2. Method II - 45 machining

As explained in Section 2.3.1, at 45 jet incidence channels had a constant opening width,

and so method II could be simplified further by eliminating the cross-sectional shape modification.

Following the procedure used to obtain Fig. 2.8(b), a series of CFD models of channels with

increasing depth were used to find the following function giving the decrease in the centerline

specific erosion rate with increasing fill-region aspect ratio, F:

4510.06 50.61v F (2.8)

This relationship was the same for the three sintered ceramics, as was 90

v F in Fig. 2.8(b). It is

seen that the average centerline impact velocity of a 45 jet on a flat plate (F=0) was 50.61 m/s,

which was 23% larger than that for a perpendicular jet. This is explained by the smaller stagnation

zone (about 10% shorter along the jet centerline) and hence particle drag in the 45 case. The

greater effect of the stagnation zone for 90 machining is also evident in the 67% larger slope of

90

v F compared to that of 45

v F ; i.e. the latter was less sensitive to changes in F compared

to 90

v F .

Figure 2.11 compares the predictions of this simplified method II with the measured profiles

in aluminum nitride using 45 backward-only machining passes. The depth predictions were within

6% of those of the measured channels at any distance from the centerline.

43

Figure 2.11 Comparison of predicted (symbols) and measured (solid lines) cross-sectional profiles

of channels machined in aluminum nitride with backward 45 passes using the standard conditions

(Table 2.2). Half of the symmetric profiles shown. P denotes the number of passes.

2.5. Conclusions

ASJM channels in three sintered ceramics (aluminum nitride, alumina, and zirconium tin

titanate) had "V"-shaped profiles, and their depths increased less-than-linearly with increasing dose

of abrasives delivered to the target. CFD simulations showed that the degree of nonlinearity had a

linear relation with the size of the stagnation zone, which could be approximated by the depth/width

aspect ratio of a notional region of a channel that was effectively filled with slurry. The CFD

models also revealed that the channel formation using a perpendicular slurry jet occurred in two

stages defined by a change in profile shape. In the first stage, the sidewalls of shallow channels

(aspect ratios of less than about 0.36) were eroded by the lateral spreading of the slurry flow,

leading to an increase in the channel opening width. In the second stage, the slurry flowed from the

footprint region mainly along the channel length and did not widen the channel opening. Channel

formation using a jet incidence of 45 (forward or backward machining configuration) did not

produce any widening of the channel opening since lateral spreading was reduced and the inclined

44

jet directed the slurry along the channel. This simplified profile modeling since it was only

necessary to account for the decrease in the specific erosion rate with depth.

Two methods were developed to predict the channel cross-sectional profiles as a function of

the number of machining passes or, equivalently, the particle dose delivered to the target. The first

method required a new CFD model of the flow and erosion pattern after each machining pass in

order to capture the changing flow field. The second method predicted the profile of subsequent

machining passes using only the erosion pattern of the first-pass channel together with an

approximate relationship between the particle centerline impact velocity and an estimate of the size

of the stagnation zone obtained using two CFD simulations for all sintered ceramic targets.

Both profile prediction methods required the experimental characterization of the erosion

behavior of the target materials as a function of jet angle and free-stream particle velocity. These

data were then expressed in terms of the actual average impact angles and impact velocities using

three-dimensional CFD models of jet impingement on flat targets. A coefficient of restitution of 0.2

was found to give the most accurate representation of particle second-strike erosion and hence

channel profile shape.

The predictions of both methods were validated by comparing with channel cross-sectional

profiles up to a depth/width aspect ratio of about 0.5. The predicted depths in the three sintered

ceramics were within 8% of those of the measured channels at any distance from the centerline.

45

2.6. References

[1] K. Kowsari, M.R. Sookhaklari, H. Nouraei, M. Papini, J.K. Spelt, Hybrid erosive jet micro-

milling of sintered ceramic wafers with and without copper-filled through-holes, J. Mater.

Process. Technol. 230 (2016) 198-210.

[2] J. Jandeleit, A. Horn, R. Weichenhain, E.W. Kreutz, R. Poprawe, Fundamental investigations of


(1998) 885-891.

[3] P.J. Slikkerveer, F.H. in’t Veld, Model for patterned erosion, Wear 233-235 (1999) 377-386.

[4] J.H.M. ten Thije Boonkkamp, J.K.M. Jansen, An analytical solution for mechanical etching of

glass by powder blasting, J. Engineering Mathematics 43 (2002) 385-399.

[5] A. Ghobeity, T. Krajac, T. Burzynski, M. Papini, J.K. Spelt, Surface evolution models in

abrasive jet micromachining, Wear 264 (2008) 185-198.


micromachining of channels and holes in glass, Wear 309 (2014) 65-73.



conference on micromanufacturing (ICOMM) (2014).

[8] N. Haghbin, J.K. Spelt, M. Papini, Abrasive waterjet micro-machining of channels in metals:

Model to predict high aspect-ratio channel profiles for submerged and unsubmerged

machining, J. Mater. Process. Technol. 222 (2015) 399-409.

[9] J. Billingham, C.B. Miron, D.A. Axinte, M.C. Kong, Mathematical modelling of abrasive

waterjet footprints for arbitrarily moving jets: Part IIOverlapped single and multiple straight

paths, International Journal of Machine Tools & Manufacture 68 (2013) 30-39.

[10] H. Nouraei, K. Kowsari, B. Samareh, J.K. Spelt, M. Papini, Calibrated CFD erosion modeling

of abrasive slurry jet micro-machining of channels in ductile materials, J. Manuf. Proc. 23

(2016) 90-101.

[11] ANSYS Fluent 15.0 Theory guide, ANSYS, Inc., 2015.

[12] D. Dehnadfar, J. Friedman, M. Papini, Laser shadowgraphy measurements of abrasive particle

spatial, size and velocity distributions through micro-masks used in abrasive jet micro-


[13] H.McI. Clark, The influence of the squeeze film in slurry erosion, Wear 256 (2004) 918-926.

46

[14] J. Humphrey, Fundamentals of fluid motion in erosion by solid particle impact, International

Journal of Heat and Fluid Flow 11 (1990) 170-195.

[15] H. Nouraei, K. Kowsari, M. Papini, J.K. Spelt, Operating parameters to minimize feature size

in abrasive slurry jet micro-machining, Precision Engineering 44 (2016) 109-123.

[16] Y.I. Oka, H. Ohnogi, T. Hosokawa, M. Matsumura, The impact angle dependence of erosion

damage caused by solid particle impact, Wear 203-204 (1997) 573-579.

[17] J.L. Routbort, R.O. Scattergood, Solid particle erosion of ceramics and ceramic composites,

Key Eng. Mat. 71 (1992) 23-50.

[18] H. Nouraei, A. Wadoslawsky, M. Papini, J.K. Spelt, Characteristics of abrasive slurry jet

micro-machining: a comparison with abrasive air jet micro-machining, Journal of Materials


[19] J.H.M. ten Thije Boonkkamp, P.J. Slikkerveer, Mathematical modelling of erosion by powder

blasting, Surv. Math. Ind. 10 (2002) 89-105.


Comparison between machining in air and submerged in water, Int. J. Machine Tools &

Manufacture 88 (2015) 108-117.

47

Chapter 3: The Effects of Fluid Vapor Pressure and

Viscosity on the Shapes of Abrasive Slurry-jet Micro-

machined Holes and Channels

3.1. Introduction

Abrasive slurry-jet micro-machined (ASJM) holes, and to a lesser extent, channels

machined in brittle materials using ASJM typically suffer from substantial edge rounding near the

opening. Wang et al. (2009) [1] drilled blind holes in glass using a slurry jet apparatus (6-14 MPa

pressure) and found that the hole openings were rounded into a bell-shaped cross-section. The

authors hypothesized that the edge rounding occurred during the initial stage of machining, but they

did not investigate the mechanism. Moreover, the results were obscured by a high degree of

asymmetry in the holes. Kowsari et al. (2014a) [2] and Kowsari et al. (2014b) [3] used ASJM to

drill sub-millimeter sized blind and through-holes in glass, polymethylmethacrylate (PMMA),

metals, and sintered ceramics without heating the target or changing its properties. The same group

predicted the shapes of ASJM holes in glass using a CFD-aided surface profile model in Nouraei et

al. (2014a) [4]. While these models were able to accurately predict the hole depths, there were

significant errors in the predicted opening width due to a progressive edge rounding that was caused

by a mechanism not captured by the CFD erosion model. Liu et al. (2015) [5] used high-pressure

(80-150 MPa) abrasive waterjet machining (AWJM) to drill holes in Ti-6Al-4V. They noted a ring-

shaped zone near the hole openings in which the surface was rougher, and hypothesized that it

might be caused by cavitation. However, they did not pursue the hypothesis using CFD. Therefore,

48

the erosion mechanism responsible for edge rounding observed in existing slurry-jet micro-

machining studies has not yet been explained.

The role of cavitation in the erosion produced by particle laden slurries has been studied in

the context of hydro-turbines, as explained by Duan and Karelin (2003) [6]. For example, Hengyun

et al. (1986) [7] measured wear rates caused by sand slurries on 18Cr-8Ni steel with and without

cavitation, and found that the combined damage was much more severe than that caused by either

sand erosion or by cavitation alone. Similarly, Sato et al. (1991) [8] studied 8-39 μm alumina

abrasive cavitation on SUS304 stainless steel and pure aluminum targets and found that rougher

surfaces further enhanced cavitation and led to greater mass loss. Toshima et al. (1991) [9]

examined the role of cavitation in silt slurries, observing that particles increased the number of

cavitation sites by 10-15%. This was attributed to gas pockets contained in crevices on the particle

surfaces that could become effective nucleation sites. Gohil and Saini (2014) [10] investigated the

effect of particle properties on cavitation-erosion damage and found that the net erosion depended

strongly on particle composition, size, shape, hardness, and concentration. Borkent et al. (2007)

[11] compared cavitation on glass and several polymers, finding that hydrophobic surfaces

enhanced cavitation. They also found that local pressurization in a flow, such as in the stagnation

zone, can significantly suppress cavitation. Wang et al. (2008) [12] studied cavitation erosion with

100-4000 nm diameter particles and found that surface damage was maximum using 500 nm

particles, and decreased when the particles became large compared to the bubbles. Although the

collapse of cavitation bubbles can cause erosion directly, a synergistic effect can exist with abrasive

erosion when particles are accelerated toward a surface by collapsing cavitation bubbles. Arora et

al. (2004) [13] used a high-speed camera to observe that collapsing 170 μm diameter cavitation

bubbles could accelerate 30-150 μm diameter polystyrene particles to velocities greater than 40 m/s.

49

Using a similar experimental setup, Wagterveld et al. (2011) [14] observed that bubble implosion

tended to fragment calcite crystals before sending them into motion. The authors also explained that

bubble collapse radiated high-pressure waves to neighboring bubbles, leading to a cascade of

cavitation activity. When cavitation bubbles collapse near boundaries or free surfaces, the bubble

interface can be distorted leading to the formation of micro-jets. Li (2006) [15] explained that

particles could be suctioned into the micro-jets and accelerated to impact the bounding surface at

relatively high velocities with their sharpest edge towards the surface. Laguna-Camacho et al.

(2013) [16] conducted cavitation erosion tests on aluminum and steel targets submerged in water

with and without 75 μm silicon carbide particles. The authors found that particles augmented the

cavitation erosion and pitting on the target surfaces, and attributed the additional damage to the

impacts of the particles which were propelled by collapsing bubbles. While these studies

established the role of cavitation-enhanced particle erosion in slurries within conduits, the flow

fields were substantially different than those in the stagnation-point flows in ASJM. Thus, little is

known about the extent and effect of cavitation in ASJM applications.

In summary, the rounding of the edges of holes and channels observed in existing micro-

machining studies with processes similar to ASJM has not yet been explained. Moreover, the role

of cavitation in the stagnation-point flows of slurries in ASJM has not been studied, although it has

been established in the literature on slurry erosion in conduits. The present objective was to

investigate these two phenomena, which will be seen to be related. The approach was to control the

extent of cavitation in ASJM by varying the fluid vapor pressure and viscosity, and to determine

how these properties affected the shapes of holes and channels, and generated the erosion

mechanism responsible for the rounding observed at the edges of these features. These results were

then used to minimize the edge rounding of holes and channels in ASJM.

50

3.2. Machining experiments and CFD flow simulations

Two types of experiments were conducted: (i) micro-sized blind and through-holes, as well

as channels, were machined in borosilicate glass, zirconium tin titanate and copper using slurries

with different fluids, in order to determine the effect of fluid viscosity and vapor pressure on the

hole shape, and (ii) wear scars were produced on borosilicate glass surfaces submerged in a water-

particle slurry using an ultrasonic apparatus to record the surface texture created by cavitation

erosion in the presence of particles.

3.2.1. Experiments

The slurry jets contained 1 wt% of well-suspended alumina particles of 10 μm nominal

diameter (Comco Inc., Burbank, CA, USA; Vickers hardness 16 GPa) in either water, aqueous

sucrose solution, mineral oil, or soybean oil. Table 3.1 gives the fluid viscosities, measured using a

Zahn cup-type viscometer (Boekel Scientific, Feasterville, PA, USA), and their vapor pressures,

either computed using Raoult's law (aqueous sucrose and glycerin solutions), or taken from the

literature.

The holes were machined by the stationary, perpendicular impingement of the slurry jet at a

standoff (distance between orifice plate and target) of 20 mm while controlling the machining

duration using a timed shutter to intercept the jet. These process parameters, referred to below as

the standard conditions, were selected to produce holes with shapes and depth-to-diameter ratios

(0.2-1) similar to those in other ASJM studies such as Kowsari et al. (2014a) [2]. The channels were

machined using standard conditions except that the target plates were scanned below the stationary

jet at 5 μm/s at perpendicular incidence, and at 2 μm/s using a 45 incidence in a forward machining

configuration. Table 3.2 presents the properties of the glass, zirconium tin titanate and copper target

51

materials used in the study. The 400 μm thick copper layer was electrodeposited on sintered

aluminum nitride.

To obtain the cross-sectional profiles of the holes, a negative cast was produced using a

metrology-grade self-curing polymer (Flexbar, Islandia, NY, USA) which was then viewed through

a microscope having a field of view of 2×1.5 mm. To investigate the wall texture of the machined

holes, the samples were sectioned through the hole centerlines, gold-coated, then viewed using a

scanning electron microscope (SU3500, Hitachi, Chiyoda, Tokyo, Japan). Quantitative analyses of

these images were conducted using an image analysis system (Clemex Vision PE, Clemex

Technologies Inc., QC, Canada) and digital software (ImageJ software—http://rsb.info.nih.gov/ij/).

Surface roughness and cross-sectional channel profiles were measured using an optical profilometer

(ST400, Nanovea Inc., CA, USA; lateral resolution 426 nm, depth resolution 16 nm). Centerline

depth measurements were made every 0.1 μm for a length of 2 mm, and the arithmetic mean

roughness, Ra, was measured using a cutoff of 8 μm.

Table 3.1 Properties of the test fluids at 20 C. The vapor pressures of soybean oil and mineral oil

were taken from Ndiaye et al. (2005) [18] and Sigma-Aldrich (St. Louis, MO, USA,

http://www.sigmaaldrich.com), respectively. The values in bold were used as inputs in the CFD

simulations.

Fluid Dynamic viscosity ()

(cP)

Vapor pressure (vp)

(kPa)

Density

(kg/m3)

Water 1 2.34 998

51 wt% aqueous sucrose solution 17±3 2.22 1230

77 wt% aqueous glycerin solution 45±3 1.41 1200

Light mineral oil 17 0.10 850

Soybean oil 45 0.35 917

52



(mm)

Grain size

(m)

Density

(g/cm³)

Vickers

hardness

(kg/mm²)

Borosilicate

glass

Borofloat 33, Schott Inc., NY,

USA 50×50×3 - 2.2 550

Zirconium tin

titanate

(Zn-Sn-TiO₂)


Ach, Japan 50×50×0.375 < 5 5.2 950

Electrodeposited

copper - 50×50×0.4 - 9 150

3.2.2. Ultrasonic apparatus and experiments

As will be discussed in Section 3.3.2, the CFD model indicated that cavitation-enhanced

erosion was likely responsible for the change in surface texture on the rounded portion at the top of

the holes. In order to support this hypothesis, an ultrasonic apparatus similar to the setup of

Laguna-Camacho et al. (2013) [16] was used to generate cavitation erosion in the presence of

particles. It consisted of an ultrasonic transducer (XL-2020 sonicator, Misonic, Farmingdale, NY,

USA) attached to a 6.4 mm-diameter, 110 mm-long stainless steel horn which was submerged in a

slurry containing water and 1 wt% 10 μm alumina particles, as shown in Fig. 3.1. The tip of the

horn was positioned parallel with the target at an elevation of 2 mm, and oscillated at 22.5 kHz with

an amplitude of 60 μm.

53

Figure 3.1 Schematic of the setup used in the ultrasonic cavitation experiments.

3.2.3. CFD modeling

In order to understand the how the flow affected particle trajectories in ASJM, the flow

fields within the machined features were obtained using computational fluid dynamics (CFD)

models. For this purpose it was not necessary to model the actual erosion process, only the flow

fields and particle trajectories. The cross-sectional profiles of the 100-400 μm deep holes were

measured as described in Section 3.2.1 and imported into ANSYS Fluent 15.0 (ANSYS Inc., Cecil

Township, PA, USA) to create axisymmetric domains with the boundary conditions shown in Fig.

3.2. These profiles represented the average smoothed contour of each hole profile. The fluid entered

the domain with a jet velocity of 89 m/s, computed using Bernoulli's equation, over the 150 µm-

diameter jet cross-sectional plane. Particles were injected uniformly across the jet and tracked using

the one-way coupling Lagrangian discrete-phase model. The volume of fluid (VOF) model was

used to simulate the multiphase, steady flow of the primary phase, water, the secondary phase, air,

and the tertiary phase, water vapor. A similar flow field and water-air interface was obtained using

the mixture multi-phase model. The cavitation model of Schnerr and Sauer (2001) [19] was

54

employed to account for the mass transfer between water and its vapor in flow regions having local

pressures lower than the liquid vapor pressure. The surface tensions of water and soybean oil were

set to 72.0 and 31.2 dyn/cm for water and soybean oil, respectively, and the default bubble number

density of 1×1013

was employed. The κ-ω shear-stress turbulent transport (SST) model was used.

The target boundary was treated as a smooth, no-slip wall. The other bounding planes were set as

free boundaries with a pressure outlet condition. The domain containing approximately 250,000

elements was meshed with quadrilateral 1 μm elements, and the simulations converged to

maximum residuals of 10-3

in approximately 3 h on average.

Additional CFD models were made with higher resolution surface profiles that contained the

actual measured surface roughness. The objective was to determine whether flow over local

variations in surface topography on the scale of the surface roughness could contribute to

cavitation. The reference surface was taken as the centerline of a channel machined using a water-

particle slurry under standard conditions and a jet traverse speed of 0.05 mm/s. The surface texture

was measured using the optical profilometer and the data was then imported into ANSYS

Workbench 15.0 (ANSYS Inc., Cecil Township, PA, USA) to create a two-dimensional, three-

phase domain similar to that shown in Fig. 3.2, but with a higher resolution roughness profile on the

surface boundaries. To capture the flow features within the micro-sized peaks and valleys, the

dimensionless wall coordinate, y+, was maintained below 1 using near-wall grid refinement having

a thickness of about 15 μm, where the smallest quadrilateral elements were approximately 200 nm.

55

Figure 3.2 Domain and boundary conditions of an axisymmetric CFD model of the ASJM flow

within a relatively deep blind hole measured in glass.


3.3.1. ASJM hole formation mechanism

Figure 3.3 shows three regions on the cross-section of an ASJM blind hole in glass

machined with the aqueous slurry. There was a distinct change from a relatively smooth texture in

Region A (Fig. 3.3) to a rougher topography in region B that was dominated by distinct craters

characteristic of perpendicular particle impacts in brittle materials (Fig. 3.3). The transition between

these two regions was accompanied by a 27% increase in roughness and significant rounding near

the opening of the hole (Fig. 3.3). The rounding developed steadily for the duration of the

machining time rather than being formed only in the initial stages, as explained by Kowsari et al.

(2014a) [2].

56

Figure 3.3 Scanning electron microscope (SEM) images of cross-section of a hole machined in

glass using the standard conditions for 8 min, and the surface texture in region A (Ra = 0.26 μm),

the transition zone, and region B (Ra = 0.33 μm).

As seen in Fig. 3.4, the CFD model of the slurry flow within a hole predicted that the

pressures would be less than the vapor pressure of water along the entire curvature near the hole

opening of region B. It is hypothesized that the growth and collapse of cavitation bubbles in this

region caused slurry particles to impact and erode the surfaces at approximately normal incidence,

consistent with the observations of Arora et al. (2004) [13], Li (2006) [15] and Laguna-Camacho et

al. (2013) [16].

57

Figure 3.4 Volume fraction contours of water vapor within a relatively deep hole machined in glass

using a water-particle slurry.

3.3.2. Ultrasonic abrasive cavitation

The ultrasonic apparatus described in Section 3.2.2 was used to generate cavitation bubbles

in an aqueous slurry of 1 wt% 10 μm alumina particles near a glass target surface. Figure 3.5

compares densely and sparsely-impacted ASJM surfaces near the opening of the hole in Fig. 3.3

with the glass surface after 30 s of ultrasonic cavitation. In both cases, the craters were typical of

impacts normal to the surface. It is seen that the craters due to ultrasonic cavitation-enhanced

erosion were much larger than those of ASJM cavitation, and appeared similar to the brittle target

impact sites reported by Nouraei et al. (2012) [20] formed by chipping and fracture. In contrast, the

ASJM slurry cavitation appeared to create ductile (plastic) indentations that would indicate much

smaller impact energies. Wensink and Elwenspoek (2002) [21] explained that both damage

mechanisms can occur in glass depending on whether the particle impact energy is greater or less

58

than the ductile-brittle transition threshold of about 17 nJ. This difference was also evident in the

surface roughness measurements since the ultrasonic cavitation surface was about 33% rougher

than the surface created by ASJM cavitation.

Figure 3.5 SEM images of slurry cavitation damage on glass created using ASJM (Ra = 0.27 μm)

and an ultrasonic apparatus (Ra = 0.36 μm). Densely and sparsely-impacted surfaces shown on the

left and right, respectively. The roughness values correspond to the surfaces on the left.

59

3.3.3. Effects of viscosity and vapor pressure on the shape of ASJM holes in brittle materials

As explained in Section 3.3.1, slurry cavitation in ASJM was predicted to occur in region B

(Fig. 3.3) where the flow over the curved surface resulted in a local flow pressure less than the

liquid vapor pressure. It was therefore of interest to investigate whether the degree of cavitation and

edge rounding could be reduced by using slurries with liquids having a lower vapor pressure. Since

the oils that were used for this purpose were also much more viscous than water, it was necessary to

investigate the effects of both liquid viscosity and vapor pressure on the shape of the hole opening

region.

Figure 3.6(a) shows the cross-sectional profiles of ASJM holes machined in glass using

slurries made from the liquids of Table 3.1. A comparison of the profiles for water (=1 cP,

vp=2.34 kPa) and the sucrose solution (=17 cP, vp=2.22 kPa) illustrates that an increase in the

viscosity reduced rounding near the hole opening for approximately the same vapor pressure (just

5% different). Comparing the profiles of the aqueous sucrose solution and the light mineral oil

(=17 cP, vp=0.1 kPa) shows that for a fixed viscosity, the much smaller vapor pressure of the oil

decreased cavitation and produced much less rounding and a smaller hole opening. A similar trend

was seen between the aqueous glycerin solution (=45 cP, vp=1.41 kPa) and soybean oil (=45 cP,

vp=0.35 kPa) which had the same viscosity, but the latter had a much smaller vapor pressure and

therefore less cavitation. Overall, it is evident that both changes in viscosity and vapor pressure can

produce large changes in the profiles of holes in glass, especially in the rounding near the opening.

If the viscosity is fixed, the solution with the smaller vapor pressure produced less cavitation and

hence rounding at the hole opening. For a fixed vapor pressure, the more viscous solution produced

less rounding, presumably because cavitation was reduced.

60

Figures 3.6(a) and 3.6(b) show that the rounded glass surfaces surrounding the holes were

damaged by pitting that was roughly proportional to the degree of expected cavitation in each fluid;

i.e. extensive pitting in the aqueous solutions and much less in the oils since they had very low

vapor pressures. Moreover, the onset of pitting was accompanied by profile rounding as shown in

Fig. 3.6(a). The location and extent of the pitting band was approximately equal to the vapor band

predicted using CFD models similar to that in Fig. 3.4.

61

(a)

(b)

Figure 3.6 Effect of slurry liquid on the profiles of blind holes in glass. (a) Cross-sectional profiles,

and (b) top views of approximately 200 μm-deep blind holes using the liquids of Table 3.1 and

standard conditions. The depths were normalized by the center depth of each hole. Half of the

symmetric holes shown.

To investigate the effect of viscosity, the flow fields of the more viscous fluids were

simulated using CFD. Figures 3.7(a) and 3.7(b) show that the velocity contours of the flow of the

slurry made with the aqueous glycerin solution (=45 cP) within a relatively deep hole in glass

resulted in a boundary layer that was approximately 4.2 times thicker than a purely water-based

62

slurry (=1 cP). It is hypothesized that the decreased flow velocity at the edge of the hole decreased

cavitation. In addition, Hara (1995) [22] found that increased viscosity reduces cavitation by

dampening localized turbulent flow patterns produced by collapsing bubbles, thereby reducing the

local Reynolds number. Moreover, Popinet and Zaleski (2002) [23] explained that increased fluid

viscosity resisted the formation of micro-jets in the collapse of bubbles near a solid boundary.

Figure 3.6(a) also shows that the shapes of the holes machined with the viscous fluids had flatter

bottoms and more vertical sidewalls. This is explained by the greater ability of the more viscous

fluids to deflect the particles before impact, causing more of them to strike the walls at closer to

normal incidence, thus widening the hole and steepening its sidewalls, as shown in Fig. 3.7(b). The

CFD model in Fig. 3.7(b) indicated that the change in vapor pressure had a negligible effect on the

particle trajectories.

63

(a)

(b)

Figure 3.7 Velocity contours of the flow fields of: (a) water and (b) the aqueous glycerin solution

(Table 3.1) within a relatively deep holes. Particle rebounds not shown.

64

Figure 3.8 shows the profiles of approximately 300 μm-deep blind holes in zirconium tin

titanate made using a water slurry (=1 cP, vp=2.34 kPa) and a soybean oil slurry (=45 cP,

vp=0.35 kPa). In both zirconium tin titanate and glass (Fig. 3.9) the hole openings were sharp

without any of the significant rounding produced by an aqueous slurry as seen in Fig. 3.6(a) and

3.9, and the surfaces in the vicinity of the hole opening did not display the pits seen in Fig. 3.6(b)

that were attributed to cavitation. In particular, comparison with the hole shown in Fig. 3.6(b) that

was produced in glass with the aqueous glycerin solution (=45 cP, vp=1.41 kPa), which had the

same viscosity but a higher vapor pressure, indicates that the lack of pitting in the glass of Fig. 3.9

was due to the decrease in cavitation caused by the lower vapor pressure of the soybean oil.

Figure 3.8 Half of the symmetrical cross-sectional profiles of holes machined using water and

soybean oil-based slurries in zirconium tin titanate.

65

Figure 3.9 SEM images of cross-section and plan view of blind hole machined using soybean oil

(=45 cP, vp=0.35 kPa) and standard conditions in zirconium tin titanate for 10 min, and top view

of a portion of the edge of a blind hole in glass machined for 2.5 min using soybean oil.

To further validate the hypothesis that cavitation was responsible for hole rounding, CFD

models were used to predict whether cavitation would occur if a sharp-edged hole, initially

machined using a soybean oil slurry (=45 cP, vp=0.35 kPa), was subsequently machined using a

water-based slurry (=1 cP, vp=2.34 kPa) where cavitation was much more prevalent because of its

higher vapor pressure, and its lower viscosity which increased the flow velocity (Fig. 3.7). As

66

expected, Fig. 3.10 shows that the aqueous slurry was predicted to cavitate at the sharp edges of the

hole. A similar result was obtained when subsequent machining was performed with the aqueous

glycerin solution slurry (=45 cP, vp=1.41 kPa) which had the same viscosity as the soybean oil,

but a higher vapor pressure and thus a greater tendency to cavitate.

These predictions were tested experimentally by machining a 187 μm deep hole in glass

using soybean oil and then machining again with a water-particle slurry for an additional 1 min.

Figure 3.11(a) shows evidence of cavitation was indeed induced during the second operation, as

evidenced by the significant edge rounding and associated surface damage as seen in Fig. 3.11(b).

These experiments demonstrated that ASJM flow over a sharp edge can initiate cavitation and

rounding of that edge.

The local velocity of a fluid flowing over a solid surface is proportional to the curvature.

Therefore, it is hypothesized that cavitation erosion of the edges of holes and channels decreases

progressively as the opening curvature is decreased by machining. This is consistent with the ASJM

work of Kowsari et al. (2014a) [2] who observed that the rate of edge rounding in holes machined

in glass decreased with increasing abrasive dose, as indicated by the decreasing slope of the curve

in Fig. 3.12.

67

Figure 3.10 The volume fraction contour of water vapor of a water-particle slurry flow within a

relatively shallow hole initially machined in glass using soybean oil.

68

(a) (b)

Figure 3.11 (a) Cross-sectional profiles of blind holes machined in glass, initially using the soybean

slurry, then continuing to machine using a water-particle slurry. Half of the symmetric profiles

shown. (b) Top view of the rounded hole in (a) that was finished with the aqueous slurry.

Figure 3.12 Rounding radius of curvature as a function of abrasive dose for ASJM holes in glass

machined using a water slurry under the standard conditions.

69

3.3.4. Effect of surface roughness

The CFD model of Fig. 3.4 predicted cavitation in the highly-curved surface region at the

edge of a hole. It was therefore of interest to examine whether curvatures found in the much smaller

peaks and valleys of the surface roughness could induce cavitation in a similar way. Figure 3.13

shows the static pressures predicted by a CFD model of a small section of a rough target surface.

Although the high-speed flow over the roughness peaks resulted in local pressure variations, the

pressure was never below the water vapor pressure. This was presumably because the average

profile peak-to-valley distance of about 1 μm was smaller than the 2 μm-thick boundary layer so

that the local flow velocity over the peaks was much smaller than the bulk flow velocity and thus

vapor formation did not occur. Therefore, it can be concluded that cavitation is unlikely to be

induced by surface roughness typically found in features machined using ASJM.

70

Figure 3.13 Pressure contour of the perpendicular impingement of an ASJM jet on a flat plate

having an Ra of approximately 400 nm. Model topography taken from an actual profilometer scan

of the centerline of an ASJM channel machined in glass using typical process conditions.

3.3.5. Through-holes in sintered ceramics

Although through-holes could be made in glass without chipping on the exit side with the

liquid slurries of Table 3.1, consistent with the findings of Kowsari et al. (2014a) [2] for aqueous

slurries, Fig. 3.14 shows that the exits of through-holes machined in zirconium tin titanate were

irregular and wider than the rest of the hole. Such chipping occurs just before the jet breaks through

71

the plate, when the strength of the remaining layer is exceeded by the force of the jet. It is

hypothesized that, in the sintered ceramic, the strength of this remaining ~50 μm thick layer (equal

to the chip thickness) was lower than in glass, leading to the irregular breakage shown in Fig. 3.14.

Figure 3.14 Plan view of the exit of a through-hole machined in zirconium tin titanate without a

backing plate

When a second sintered zirconium tin titanate plate was attached to the bottom using a thin layer of

epoxy adhesive (J-B Weld, Atlanta, GA, USA), Figs. 3.15(a) and 3.15(b) show that chipping was

prevented, but there was leakage as the slurry eroded the epoxy layer. This was similar to the

findings of Kowsari et al. (2014a) [2] who reported leakage at the interface between a glued glass

sheet and a glass target during ASJM hole drilling using an aqueous slurry. For the present

configuration of a sacrificial plate attached to the bottom using an epoxy, it was hypothesized that

the leakage could be minimized by stopping the machining process immediately after piercing the

target plate. This instant could be identified by observing the flow of the return slurry during

experimentation; i.e. the initially sharp cone of the return flow became unstable and fuzzy beyond

72

the depth of the top plate. In summary, ASJM through-holes having sharp openings and relatively

steep sidewalls and exits without any chipping could be machined using an oil based slurry with a

sacrificial plate attached at the bottom using an epoxy adhesive.

(a)

(b)

Figure 3.15 (a) Section view of a through-hole in a zirconium tin titanate plate machined when

attached to another plate using epoxy. (b) Plan view of the hole exit after second plate was

separated by heating it to 316 C.

73

3.3.6. Channels in glass and zirconium tin titanate using an oil-based slurry

Kowsari et al. (2013) [24] and Kowsari et al. (2016) [25] found that ASJM channels

machined using aqueous slurries had round edges in glass and were "V"-shaped in sintered

ceramics. It was therefore of interest to explore the effects of an oil-based slurry on channel shape.

Figure 3.16 compares the cross-sectional profiles of channels machined in glass and zirconium tin

titanate using the soybean oil slurry. In glass, the channels machined using soybean oil slurry had

sharp edges, eliminating the rounding seen at the edge of channels machined using a water-based

slurry. The oil-based channels were also about 20% wider than those made with the aqueous slurry

(Fig. 3.16(a)). The wider channels are explained by the wider erosive footprint of the oil jet brought

about by the increased viscosity as described in Section 3.3.3 (Fig. 3.7(a) and 3.7(b)). Figure 3.16

also shows that the channel width could be reduced by about 5% by machining at 45 jet incidence

rather than at 90. In zirconium tin titanate, Fig. 3.16(b) shows that the use of the oil-based slurry

made the bottoms of the channels flatter. This is explained by the wider footprint of the soybean oil

slurry compared to water as shown in Figs. 3.7(a) and 3.7(b). In both materials, the oil-based slurry

channels were approximately 28% rougher than those machined using the water slurry and

otherwise identical conditions (Ra = 0.5 and 0.3 μm in glass and zirconium tin titanate,

respectively). This is consistent with the greater ability of the water-based slurry to polish the

surface as it flowed along the channel length. In contrast, the relatively thick boundary layer of oil-

based slurries, as explained in Section 3.3.3, reduced the particle velocities near the surface and thus

reduced the smoothing action.

74

(a)

(b)

Figure 3.16 Cross-sectional profiles of approximately 50 μm deep channels machined in: (a) glass

and (b) zirconium tin titanate using water and oil-based slurries at 90 and 45 jet incidences. Half

of the symmetric profiles shown. The depths were normalized by the centerline depth of each

channel.

3.3.7. Edge rounding in ductile materials

In contrast to the edge rounding of glass and zirconium tin titanate using water-based

slurries (Section 3.3.3), Fig. 3.17 shows that ASJM of ductile materials such as copper produced

only a small amount of edge rounding using a water slurry, which was eliminated using a soybean

oil slurry. Kowsari et al. (2014a) [2] also obtained relatively sharp edges using aqueous slurries in

75

the ASJM of holes 316 stainless steel, 110 copper and 6061 aluminum. As explained in Section

3.3.1, imploding cavitation bubbles caused abrasive particles to impact at near-perpendicular

angles. This created more damage in brittle materials, where the maximum erosion occurs at 90,

than in ductile materials where erosion in greatest at about 30. In summary, cavitation-enhanced

slurry erosion has a much smaller effect on edge rounding in ductile materials than on brittle targets

for typical ASJM conditions.

Figure 3.17 Half of symmetrical cross-sectional profiles of approximately 100 μm- deep single-

pass channels in copper machined using water and soybean-oil slurries using the standard

conditions.

3.4. Conclusions

The effects of slurry liquid viscosity and vapor pressure on the shapes of ASJM holes in

brittle materials were investigated. CFD analyses of the flow fields and measurements of the

surface textures within the machined holes indicated that the edge rounding observed in micro-

machined features in ASJM was due to abrasive-enhanced cavitation caused by vapor formation as

76

the high-speed slurry flowed over the edges at the tops of the holes and channels. The collapse of

cavitation bubbles accelerated particles in their vicinity to impact the target at near-perpendicular

incidence and velocities sufficient to cause erosion, thus damaging and rounding the edges. This

was demonstrated by producing comparable damage in glass using an ultrasonic apparatus

immersed in an aqueous slurry. Such cavitation-enhanced erosion has also been observed in slurry

flows within conduits, but this is the first time its role has been established in the stagnation-point

flows typical of slurry-jet micro-machining.

Experimental results showed that reducing the slurry vapor pressure decreased the cavitation

activity, producing holes and channels in glass and zirconium tin titanate with much less rounding

at the top. This effect became more pronounced as the liquid viscosity was increased, since the flow

velocities were reduced and hence the decreases in pressure were smaller. ASJM using slurries of

low-vapor pressure liquids such as mineral oil not only significantly sharpened the hole entrances,

but also produced changes in the local particle impact angle that led to flatter hole bottoms and

steeper sidewalls. Through-holes with sharp entrance and exit holes were machined in glass, and in

sintered zirconium tin titanate with the aid of a sacrificial layer at the hole exit. Edge rounding

caused by cavitation-enhanced slurry erosion was much less pronounced in ductile materials than in

brittle targets due to the difference in their erosion mechanisms.

The effects of surface geometry on abrasive cavitation were also investigated. It was found

that flow of a water-based slurry within an initially sharp hole caused rounding of its opening,

consistent with the cavitation predictions of CFD models. Moreover, CFD predicted that the local

profile peaks and valleys on the scale of the surface roughness for typical ASJM surfaces did not

generate vapor formation. In summary, sharp-edged holes and channels with flat bottoms and

77

relatively steep sidewalls were machined for the first time using ASJM by minimizing cavitation

through the use of liquids with low vapor pressure and relatively high viscosity.

3.5. References




of holes in brittle and ductile materials, J. Materials Processing Tech. 214 (2014a) 1909–1920.



conference on micromanufacturing (ICOMM), Singapore, Singapore (2014b)



Proceedings of the 10th international conference on micromanufacturing (ICOMM), Milan,

Italy (2014a).



[6] C.G. Duan, V.Y. Karelin, Abrasive erosion and corrosion of hydraulic machinery, Imperial

College Press London, 2003.

[7] J. Hengyun, Z. Fengzhen, L. Shiyun, H. Chenzhao, The role of sand particles on the rapid

destruction of the cavitation zone of hydraulic turbines, Wear 112 (1986) 199-205.

[8] J. Sato, K. Usami, T. Okamura, S. Tanaba, Basic studies of coupled damage caused by silt

abrasion and cavitation erosion, Japan Society of Mechanical Engineers 34 (3) (1991) 292-297.

[9] M. Toshima, T. Okamura, J. Satoh, K. Usami, S. Tanabe, Basic study of coupled damage caused

by silt abrasion and cavitation erosion, Japan Society of Mechanical Engineers 57 (1991) 20–

25.

78

[10] P.P. Gohil, R.P. Saini, Coalesced effect of cavitation and silt erosion in hydro turbines — a

review, Renew Sustain Energy Rev 33 (2014) 280–289.

[11] B.M. Borkent, M. Arora, C.D. Ohl, Reproducible cavitation activity in water–particle

suspensions, J. Acoust. Soc. Am. 121 (2007) 1406–1418.

[12] J. Wang, H. Chen, L. Qin, Y. Li, D. Chen, Key roles of micro-particles in water on occurrence

of cavitation-erosion of hydro-machinery, Chin. Sci. Bull. 53 (2008) 1603–1610.

[13] M. Arora, C.D. Ohl, K.A. Mørch, Cavitation inception on microparticles: a self-propelled

particle accelerator, Phys. Rev. Lett. 92 (2004) 174501.

[14] R.M. Wagterveld, L. Boels, M.J. Mayer, G.J. Witkamp, Visualization of acoustic cavitation

effects on suspended calcite crystals, Ultrasonics Sonochem 18 (2011) 216–241.

[15] S. Li, Cavitation enhancement of silt erosion — an envisaged micro model, Wear 260 (2006)

1145–1195.

[16] J.R. Laguna-Camacho, R. Lewis, M. Vite-Torres, J.V. Mendez-Mendez, A study of cavitation

erosion on engineering materials, Wear 301 (2013) 467-476.


micromachining of channels and holes in glass, Wear 309 (2014b) 65-73.

[18] P.M. Ndiaye, F.W. Tavares, I. Dalmolin, C. Dariva, D. Oliviera, J.V. Oliviera, Vapor pressure

data of soybean oil, castor oil, and their fatty acid ethyl ester derivatives, J. Chem. Eng. Data 50

(2005) 330-333.

[19] G.H. Schnerr and J. Sauer, Physical and Numerical Modeling of Unsteady Cavitation

Dynamics, Fourth Intl. Conf. on Multiphase Flow, New Orleans, USA (2001).

[20] H. Nouraei, A. Wodoslawsky, M. Papini, J.K. Spelt, Characteristics of abrasive slurry jet

micro-machining: a comparison with abrasive air jet micro-machining, J. Materials Processing

Tech. 213 (2012) 1711-1724.

[21] H. Wensink, M.C. Elwenspoek, A closer look at the ductile-brittle transition in solid particle

erosion, Wear 253 (2002) 1035-1043.

[22] J. Hara, J. Deshimaru, M. Kasai, Study on cavitation of water soluble hydraulic fluid, Proc. of

the Int. Tribology Conf., Yokohama (2) (1995) 909-914.

79

[23] S. Popinet, S. Zaleski, Bubble collapse near a solid boundary: a numerical study of the

influence of viscosity, J. Fluid. Mech. 464 (2002) 137-163.



[25] K. Kowsari, H. Nouraei, B. Samareh, M. Papini, J.K. Spelt, CFD-aided prediction of the shape

of abrasive slurry jet micro-machined channels in sintered ceramics, Ceramics International 42

(2016) 7030-7042.

80

Chapter 4: Erosive Smoothing of Abrasive Slurry-jet

Micro-machined Channels in Glass, PMMA, and

Sintered ceramics: Experiments and Roughness Model

4.1. Introduction

Similar to abrasive slurry-jet micro-machining (ASJM), fluid jet polishing (FJP) uses a

slurry of small particles to polish surfaces in jets much larger than that of ASJM. For example, Tsai

et al. (2008) [1] used a 4 mm diameter jet and 5 μm silicon carbide particles to polish steel. Peng et

al. (2013) [2] used 100 nm cerium oxide particles in a 30 m/s, 5 mm diameter slurry jet at an impact

angle of 45 to finish the surface of quartz glass to a root mean square (Rrms) roughness of 0.2 nm.

In a similar work, Zhang et al. (2009) [3] used 40 nm silicon oxide particles in a 4 mm diameter,

100 m/s slurry jet to smooth the surface of borosilicate glass to a roughness of about 1 nm. Both of

these authors explained that the dominant material removal mechanism was adhesion and

polymerization reactions between the particles and glass targets forming cerium-oxide-silicon. In

another FJP study, Fang et al. (2006) [4] obtained a smoother final glass surface using 1-2 μm

cerium oxide particles compared to 1-3 μm silicon carbide, but at far longer machining times.

Recently, Wang et al. (2016) [5] used a stationary 500 μm-diameter slurry jet containing

water and 0.7-2.5 μm cerium oxide particles and 0.2 wt% of polyacrylamide (PAM) and found that

the polymeric additive reduced the size of the transition zone between the polished and unpolished

regions, but the study was limited to scars, but not machined features, on glass surfaces and did not

81

consider ductile materials or sintered ceramics. In general, the use of nano-sized particles

necessitate machining durations too long to be practical; e.g. 20 h in Zhang et al. (2009) [6].

Moreover, all of these studies involved jets that were much larger than those in ASJM (e.g. 4 mm in

FJP compared to about 180 μm in ASJM), and therefore the FJP polishing zone cannot be focused

on individual micro-features such as microfluidic channels and holes without eroding neighboring

features.

The smoothing of sintered ceramics has been demonstrated by, for example Choi et al.

(2004) [7], using chemical mechanical polishing (CMP). However, the relatively large flat CMP

pad limited the process to bulk surface finishing, without the capability to smooth the inner walls of

micro-features.

Haj Mohammad Jafar et al. (2015) [8] developed a combined analytical-numerical surface

roughness model for the ASJM of channels in glass. The assumption of brittle erosion and chip

removal due to lateral cracking was appropriate since relatively large (~25 μm) abrasives were

used. However, as explained by Nouraei et al. (2013) [9], typical ASJM abrasives are smaller (< 10

μm) and because of their much lower impact kinetic energies, they tend to erode glass in a ductile

manner, by cutting and ploughing craters. The model of Haj Mohammad Jafar et al. (2015) [8]

neglected this type of erosion.

Existing modeling work in FJP includes that of Cao and Cheung (2014) [10] who used

computational fluid dynamics (CFD) to obtain local particle impact angles and velocities which

were then used in the rigid-plastic model of Papini and Spelt (2000) [11] to predict the shapes of

individual impact craters and hence overall shape and depth of the FJP footprint after many

82

impacts. In a similar FJP study, Li et al. (2010) [12] used CFD to obtain the erosive footprint at an

impact angle of 60, and found that the shape was non-uniform. The authors corrected this by

rotating the oblique jet about the normal of the stagnation point, thereby improving the flatness of

the polished zone. Although these models predicted the shape and size of the erosive footprint in

FJP, they could not predict the local surface roughness within the relatively large (approximately 4

mm) footprint.

In summary, local smoothing of micro-sized channels using abrasive slurry-jets has not been

attempted on any target materials. Larger scale surface smoothing of sintered ceramics has been

limited to CMP. Moreover, there is currently no model that can predict the roughness of ASJM

surfaces due to ductile erosion mechanisms. The present work investigated the effects of process

parameters on the roughness of micro-channels milled using ASJM in glass, PMMA, and two

sintered ceramics: zirconium tin titanate and aluminum nitride. ASJM post-blasting of existing

channels was developed as a means of polishing, and the topography and roughness of these

surfaces was predicted using a numerical-analytical, ductile-regime model. Computational fluid

mechanics was used to predict the particle impact conditions.


Two types of experiments were performed as summarized in Table 4.2: (i) micro-channels

were machined with a typical range of ASJM range of process parameters to examine the variation

in the resulting surface roughness and erosion rate without any post-blasting, and (ii) existing

micro-channels were post-blasted with varying process conditions in order to determine the extent

83

to which the channels could be smoothed. The particle impact conditions were predicted using

computational fluid dynamics of the flow in the channels.

4.2.1. Experiments

Four different slurries were investigated, each containing 1 wt % abrasive: 10 μm alumina

(Comco Inc., Burbank, CA, USA; Vickers hardness 16 GPa), 3 μm alumina (Agsco Corp.,

Wheeling, IL, USA; Vickers hardness 16 GPa), 3 μm silicon carbide (Zaozhuang City-hsin Ltd.,

China; Vickers hardness 20 GPa), and 0.7 μm cerium oxide (Universal Photonics Inc., Hicksville,

NY, USA; Vickers hardness 8 GPa). Channels were machined and post-blasted in glass, PMMA,

zirconium tin titanate, and aluminum nitride (Table 4.1) by scanning the target with respect to the

jet at a known speed using a motorized stage (KT-LSM100A, Zaber Technologies Inc., Vancouver,

BC, Canada). Table 4.2 shows the process parameters used in the channel machining and post-

blasting experiments and shows that the main difference was that the latter employed smaller

particles and more inclined jets.

Surface topographies were measured using atomic force microscopy (AFM5300E, Hitachi,

Tokyo, Japan) in contact mode at a frequency of 0.5 Hz using a cantilevered probe (NCST-10,

NanoWorld Holding AG, Schaffhausen, Switzerland) having a force constant of 7.4 N/m. As

explained by Poon and Bhushan (1995) [14], AFM was selected rather than an optical profilometer,

in order to avoid noise created due to optical diffraction. The centerline roughness of a machined,

unpolished channel was defined as the average root-mean-square roughness (Rrms) of the scans of

three 50×50 μm areas, spaced about 50 μm apart along the channel centerline. Three unpolished

channels were post-blasted using a stationary, inclined jet similar to the glass polishing work of

84

Wang et al. (2016) [5]. The scans of the polished regions in each channel were taken in the central

zone of the stationary footprint.

Table 4.1 Properties of the target materials. The dynamic hardnesses were estimated using the

methodology of Section 4.4.5.


(mm)

Grain

size

(m)

Vickers

hardness

(kg/mm²)

Dynamic

hardness

(kg/mm²)

As-

supplied

Rrms

roughness

(nm)

Borosilicate

glass

Borofloat 33, Schott

Inc., NY, USA 50×50×3 - 550 2600 8

Polymethylme

thacrylate

(PMMA)

Piedmont Plastics Inc.,

ON, Canada 50×50×3 - 17 2750 24

Zirconium tin

titanate

(Zn-Sn-TiO₂)

M39, Maruwa,

Owariasahi-shi, Ach,

Japan

50×50×0.375 < 5 950 2950 395

Aluminum

nitride (AlN)

K170, Toshiba Corp.,

Minato, Tokyo, Japan 50×50×0.375 < 1 1100 - 200

85

Table 4.2 ASJM process parameters used in the two types of experiments: (i) channel machining

over a range of typical conditions, and (ii) channel smoothing using post-blasting. Standard process

conditions shown in bold.

Type of experiment

Channel machining Channel post-blasting

Pressure (MPa) 2 4 6 4

Slurry flow rate (mL/s) 1.34 1.67 2.00 1.67

Free jet velocity (m/s) 63 89 110 89

Particle size and material 10 μm alumina

(Al2O3)

3 μm alumina (Al2O3), 3 μm silicon carbide

(SiC), 0.7 μm cerium oxide (CeO2)

Particle concentration (wt%) 1 1

Standoff distance (mm) 20 20

Jet incidence (°) 30, 45, 90 15

Scan speed (mm/s)

Glass 0.4 0.6 0.8 -

PMMA 1.0 1.4 1.8 -

Zirconium tin titanate 0.05 0.1 0.2 -

Aluminum nitride 0.025 0.05 0.1 -

4.2.2. CFD modeling

The instantaneous flow field and particle trajectories during the machining of the ASJM

channels using a transversely inclined jet (orientation shown in Fig. 4.1) were modeled using

ANSYS Fluent 15.0 (ANSYS Inc., Cecil Township, PA, USA). As will be explained in Section

4.4.3, this machining orientation narrowed the range of angles at which the particles struck the

channel centerline. Although the cross-sectional shape of these channels was asymmetrical at the

edge, the central region was relatively flat as evident in Fig. 4.1. Moreover, the configuration

eliminated the potential secondary erosion on the channel sidewalls in regions outside the primary

footprint resulting from perpendicular machining as explained by Kowsari et al. (2016) [15]. These

effects are demonstrated in Section 4.4.3 using the model geometry of the machining front of a

86

channel in PMMA. It was reasoned that any small differences in the cross-sectional shapes of

channels machined in the other targets would not significantly change the particle impact angles in

region A (Fig. 4.1), so that the conclusions would remain valid for all or the materials.

After machining a channel in PMMA to 80 μm depth using a 0.2 mm/s scan speed and

otherwise standard conditions (Table 4.2), a shutter was used to interrupt the slurry flow. The

instantaneous three-dimensional channel geometry was measured using an optical profilometer with

a lateral resolution of 426 nm (ST400, Nanovea Inc., CA, USA). The surface data was then

imported into ANSYS Workbench 15.0 (ANSYS Inc., Cecil Township, PA, USA) using the three-

dimensional curve tool to obtain the three-dimensional profile shown in Fig. 4.2. The fluid entered

the domain over the 150 µm-diameter jet cross-sectional plane at a velocity of 89 m/s, determined

using Bernoulli's equation. The particles were uniformly injected at the inlet and tracked using the

one-way coupling Lagrangian discrete phase model. The channel and target surfaces were treated as

smooth, no-slip walls. The remaining bounding planes of the domain were treated as pressure

outlets. A multiphase, steady volume-of-fluid model was used for the water-jet surrounded by air.

The κ-ω shear-stress transport turbulence model was used with quadrilateral 2 μm elements and a

convergence residual of 10-3

.

87

Figure 4.1 Domain and boundary conditions of a three-dimensional CFD model of the ASJM flow

within a channel in PMMA at 45 incidence. The schematic is to scale.

4.3. Roughness model

Section 4.4.5 presents predictions of the topographies of surfaces using the ductile-mode

roughness simulation model described in Schwartzentruber et al. (2016) [16]. Briefly, the

methodology first predicts the two-dimensional shape of a crater produced upon the impact of a

single rhomboid particle using the fully-plastic analytical model presented by Papini and Spelt

(1997) [17]. The input parameters include particle properties such as impact velocity (Vi), size (d,

D), angularity (A1, A2), density, impact angle (α), and impact orientation (θ), in addition to the

dynamic hardness of the target material as shown in Fig. 4.2. The particle size distributions were

obtained from the manufacturers (Section 4.2.1), and the particle angularity was defined as the arc-

88

tangent and arc-tangent inverse of the measured average particle aspect ratio. Table 4.3 gives the

input parameters used in the modeling trials. The dynamic hardness of each material was calibrated

based on the experimental results corresponding to Process I in Table 4.3, which was then

subsequently used in predicting the remaining experimental values.

Similar to the procedure followed by Schwartzentruber et al. (2016) [16], the first step was

to generate and then concatenate the craters made by individual particles with randomly varying

particle orientation and angularity within the ranges measured for the abrasive (Table 4.3). These

individual craters were then linked in series to create a 50 µm long surface profile, consistent with

the length of the roughness measurements made using AFM. The program then superimposed

multiple single particle impact profiles about their individual mean lines (the location where the

integral of the profile above the mean line equaled that below it) to generate a multiple impact

profile, until the change in roughness between iterations was less than 0.1 nm. ISO 4288 (1996)

[18] was used to assess the Rrms of the model generated profiles using Mountain 6 surface analysis

software (Digital Surf, France) with an 8 µm cutoff wavelength based on the recommend practice in

ISO 4288 (1996) [18].

89

Figure 4.2 Schematic of the geometry of a model particle impacting a target.

Table 4.3 Input parameters used in the roughness model for three sets of process conditions. The

average centerline particle impact angles were reproduced from the CFD models of Kowsari et al.

(2016) [15] on flat targets, and the particle properties were obtained from the manufacturers

(Comco Inc., Burbank, CA, USA; Zaozhuang City-hsin Ltd., China).

Process conditions and model inputs

I

(machining)

II

(machining)

III

(smoothing)

Particle composition Alumina Alumina Alumina

Particle nominal diameter (μm) 10 10 3

Particle density (g/cm3) 3.95 3.95 3.95

Particle angularity 0.76 0.76 0.76

Jet impact angle (°) 45 30 15

Average centerline particle impact angle

(°) 34 15 6

Jet impact velocity (m/s) 89 89 89

Average centerline particle impact

velocity (m/s) 51 81 88

90


4.4.1. As-received target surfaces

Figure 4.3(a) presents the topography of the as-received targets (Table 4.1). The surface of

the as-received PMMA was rougher (24 nm) than that of glass (8 nm), but smooth enough so that it

was still optically transparent. It contained shallow scratches, as shown using both SEM (Fig.

4.3(a)) and AFM (Fig. 4.3(b)). The pre-existing pits shown in the aluminum nitride and zirconium

tin titanate (Fig. 4.2(a)) were from the manufacturing process as explained by Zhang et al. (1999)

[3], who noted that sintered ceramics such as zirconium tin titanate are prone to the embedment of

abrasives, and thus cannot be further smoothed using conventional processes such as CMP.

91

(a)

(b)

Figure 4.3 (a) SEM images of the as-received glass, PMMA, zirconium tin titanate, and aluminum

nitride surface. (b) AFM measurements of as-received PMMA.

92

4.4.2. Mechanism of surface topography evolution

The material removal mechanisms due to particle impact differ depending on whether the

target is a ductile or brittle. Brittle erosion involves lateral cracks leading to the removal of chips,

while ductile erosion is due to ploughing and cutting in materials capable of significant plastic

deformation. Nevertheless, at sufficiently low kinetic energies, nominally brittle materials can also

undergo some degree of ductile erosion. For example, in the ASJM of borosilicate glass using 25

μm alumina particles, Nouraei et al. (2013) [9] noted four types of impact sites: (i) chip removal

from lateral crack formation, (ii) brittle fracture without chip removal, (iii) scratches resulting from

ductile erosive ploughing, and (iv) plastically-deformed craters without cracking. Wensink and

Elwenspoek (2002) [19] examined the ductile-brittle transitions for a number of ceramics, and

found that the threshold value of the particle kinetic energy perpendicular to the surface (i.e., the

‘normal kinetic energy’) was 17 nJ for borosilicate glass. Using data obtained from CFD

simulations with domains similar to Fig. 4.1, the normal particle kinetic energy was calculated as

0.9 nJ for 10 μm particles using the standard process conditions on glass in Table 4.2; i.e. well

below the ductile-brittle erosion threshold energy. However, Nouraei et al. (2013) [13] explained

that the larger particles in the 10 μm powder distribution produced brittle craters under process

conditions that were similar to the present standard conditions. The authors found the brittle impact

sites to be much larger than those created by ductile erosion and concluded that the brittle erosion

dominated the material removal process, while the ductile mechanism governed the surface

roughness. Therefore, the material removal mechanism in glass was likely a mixture of brittle and

ductile erosion. Figure 4.4 shows images of the ductile glass impact sites; i.e. type (iii) damage -

scratches resulting from ductile erosive ploughing, and type (iv) damage - plastically-deformed

craters without cracking.

93

For particle impacts on ductile materials such as PMMA, the material beneath the impact

site yields in compression, and then laterally displaces toward the surface to form the elongated

"lips" shown in Fig. 4.4(a). These lips are removed from the surrounding material by the action of

successive impacts. Figure 4.4(b) shows the centerline surface texture of a channel in PMMA

machined using a 45 jet and otherwise standard conditions (Table 4.2). The surface contains

relatively deep scratches that were created by the plastic deformation caused by the striking

particles.

In contrast to brittle and ductile targets, the ASJM damage mechanism in sintered ceramics

occurs by grain boundary cracking leading to grain-dislodgement, as explained by Kowsari et al.

(2016) [15]. The machined surface of aluminum nitride in Fig. 4.4(b) showed evidence of removed

grains and plastically-deformed grains. In a softer sintered ceramic such as zirconium tin titanate,

the channel surfaces appeared smoother than channels in aluminum nitride (Fig. 4.4(b)), suggesting

that there was a larger degree of plastic grain deformation. This is consistent with the finding of

Kowsari et al. (2016) [15] that ASJM of zirconium tin titanate produced a mixed ductile and brittle

erosion, with the maximum erosion occurring at perpendicular impact, as is typical of brittle

materials, but with significant erosion also occurring at shallow impact angles, typical of ductile

materials. In contrast, Kowsari et al. (2016) [15] found that sintered aluminum nitride behaved in a

strictly brittle manner using alumina abrasives.

94

(a)

(b)

Figure 4.4 SEM images of (a) plastically-deformed craters without cracking in glass, (b) region A

(Fig. 4.1) surfaces of channels machined in glass, PMMA, zirconium tin titanate, and aluminum

nitride using the standard conditions (Table 4.2).

4.4.3. Roughness of ASJM channels under standard conditions

Haj Mohammad Jafar et al. (2015) [8] found that the roughness of ASJM features in glass

depended on the particle impact kinetic energy transfer perpendicular to the target surface (the

‘normal kinetic energy’). Using CFD, the same authors showed that at perpendicular jet incidence,

particles impacted at angles in the range 30-90. Moreover, Kowsari et al. (2016) [15] showed that

the secondary ASJM flow could erode features away from the primary footprint. To narrow the

broad range of particle impact angles and to eliminate the secondary erosion, the channels in the

95

present work were machined at a transverse jet angle of 45 in the configuration shown in Fig. 4.1.

The CFD results supported this conjecture, showing that under these conditions incident particles

impacted region A in Fig. 4.5(a) (the primary footprint) at approximately 34, while region B

outside the primary footprint was struck with relatively few impacts ranging from 0-30.

Consequently, Fig. 4.5(b) shows that the surface of region B was smoother than that of region A

due to the predominantly shallow impacts of the secondary slurry flow across the channel beyond

the primary footprint of region A.

96

(a)

(b)

Figure 4.5 (a) Trajectories of 10 μm alumina particles on the jet centerline plane of the ASJM flow

within a PMMA channel machined at 45 shown in Fig. 4.1. Primary footprint – region A,

secondary footprint – region b. (b) SEM images of regions A and B of a channel machined at 45

channel in glass.

97

4.4.3.1. Effect of particle dose

The possibility of a transient roughness was investigated by gradually increasing the particle

dose, defined as particle mass delivered over the area of a 1 mm-long channel, in the range 0.1-7.7

g/mm2 by either increasing the number of machining passes, or decreasing the scan speed of the

slurry jet. Figure 4.6 shows that there was no detectable transient roughness in channels in glass,

PMMA, and sintered ceramics machined over the range of typical ASJM process conditions (Table

4.2). Therefore, even at the lowest dose in the present experiments the number of impacts per unit

area (about 4.8×107 impacts/mm

2) was high enough to generate a steady-state roughness.

Figure 4.6 Measured channel centerline Rrms roughness as a function of particle dose using the

range of standard process conditions (Table 4.2). Error bars represent ±1 standard deviation for 3

areal scans along a single channel.

4.4.3.2. Effect of particle kinetic energy

Under conditions of brittle erosion, roughness increases as the particle kinetic energy

associated with the surface-normal velocity component increases, as observed in the ASJM of glass

by Haj Mohammad Jafar et al. (2015) [8]. The role of impact velocity and impact angle were

98

investigated for the present materials over the range of typical ASJM process conditions of Table

4.2.

Impact velocity

The effect of jet impact velocity was investigated by machining approximately 50 μm deep

channels at pressures ranging from 2-6 MPa (jet velocities 63-110 m/s) using a 45 jet in the

configuration shown in Fig. 4.1. Figure 4.7 shows that roughness and specific erosion rate (mass of

material removed per mass of erodent) increased with increasing impact velocity for glass (Rrms:

9%, specific erosion rate: 151%), PMMA (146%, 345%), and zirconium tin titanate (57%, 633%).

Since particles largely deflect before impact in ASJM stagnation flows, the increase in normal

impact velocity was accompanied by an increase in the tangential velocity regardless of the jet

incidence. Therefore, it could not be established if the increasing roughness and specific erosion

rate were due solely to the increasing normal velocity component. The trends were however

consistent with those found in the air-driven work of Haj Mohammad Jafar et al. (2013) [20] where

the normal velocity was varied independently of the tangential component. Moreover, Hasem

(2013) [21] found that surface roughness of glass did not depend on the tangential velocity

component. In contrast, for the ductile PMMA, it was hypothesized that the increased tangential

velocity did play a role in creating larger craters, and producing a larger increase in roughness

compared to glass. Aluminum nitride behaved differently, with the roughness remaining constant.

This was due to is erosion mechanism (Section 4.4.2) where particle impacts induced intergranular

cracks and tended to remove entire, discrete grains rather than parts of grains, as was seen in Fig.

4.4(b). Of the four materials, the largest increase in roughness was in PMMA (246%) due to its

relatively low resistance to plastic deformation.

99

(a) (b)

(c) (d)

Figure 4.7 Channel machining using standard conditions of Table 4.2. Measured channel centerline

Rrms roughness and specific erosion rate as a function of average particle impact kinetic energy of

surface-normal velocity component in: (a) glass, (b) PMMA, (c) zirconium tin titanate, and (d)

aluminum nitride. Error bars represent ±1 standard deviation for 3 areal scans along a single

channel.

Impact angle

Figure 4.8 shows that, for a given particle velocity, rougher channels were machined at

higher jet angles in glass, PMMA, and zirconium tin titanate. As was also seen in Fig. 4.7,

increasing the impact angle resulted in larger normal impact forces that led to the removal of larger

ductile wear scars and hence increased roughness. Conversely, as explained by Finnie (1960) [22],

100

the greater tangential impact forces of relatively shallow impacts enhanced the removal of crater

lips, thus smoothing the surface. As in Fig. 4.7, although the erosion rate increased with increasing

impact angle in aluminum nitride, the roughness did not change since this material was eroded by

the removal of discrete grains rather than by plastic deformation (Section 4.4.2). Figure 4.8 also

shows that glass and aluminum nitride behaved in a brittle manner; i.e. the specific erosion rate was

maximum at perpendicular jet incidence, since the brittle erosion governed the erosion process in

glass as explained in Section 4.4.2. The maximum erosion in PMMA occurred at 45 jet incidence,

typical of ductile targets, while zirconium tin titanate eroded in a mixed mode of both brittle and

ductile mechanisms where the erosion rate was maximum at 90, but showed a smaller decrease at

shallower angles.

In summary, over the range of typical ASJM process conditions, decreases in normal

particle kinetic energy achieved by lowering the particle impact angle, velocity, or size could be

exploited to machine channels that were approximately 35% smoother than the roughest channels

machined at the largest normal kinetic energy. This roughness reduction, however, also resulted in

an average decrease in the specific erosion rate of 64%.

101

(a) (b)

(c) (d)

Figure 4.8 Channel machining using standard conditions of Table 4.2. Measured channel centerline

Rrms roughness and specific erosion rate as a function of average particle impact angle in (a) glass,

(b) PMMA, (c) zirconium tin titanate, and (d) aluminum nitride. Error bars represent ±1 standard

deviation for 3 scans along a single channel.

102

4.4.4. Roughness of post-blasted ASJM channels

As explained in Section 4.4.3, the roughness of the machined channels could be reduced by

35% on average over the present range of typical ASJM process conditions, but this also decreased

the erosion rate by about 64%, thereby necessitating unrealistically long machining times to reach

practical channel depths. Therefore, the present work investigated a hybrid process of relatively

rapid channel machining to a desired depth using the standard ASJM conditions of Table 4.2,

followed by post-blasting with much smaller normal kinetic energies to smooth the channel using

process parameters suited to peak removal. These post-blasting experiments were conducted on

reference channels machined in glass, PMMA, zirconium tin titanate, and aluminum nitride under

the standard process conditions of Table 4.2 (in bold) that produced Rrms roughnesses of 245, 127,

141, and 390 nm, respectively.

Figure 4.9 shows that the roughness of the post-blasted channel in zirconium tin titanate

showed a marked transient behavior, decreasing with increasing post-blast particle dose, defined as

particle mass delivered by the jet by the resulting channel area. The AFM scans showed that there

was a gradual flattening of the profile peaks with increasing dose up to a steady-state, dose-

independent roughness. This is in contrast to Fig. 4.6 which showed that no transient was apparent

in the roughness with increasing dose under typical ASJM conditions. The same steady-state post-

blasted roughness of ~23 nm was obtained using both 3 μm alumina and 3 μm silicon carbide, but

the machining time (dose) to achieve a given roughness was about 50% higher using alumina. The

accelerated peak-removal observed with silicon carbide was probably due to its higher Vickers

hardness; approximately 25% larger than that of alumina, thus removing more material under

identical ASJM conditions. In the present study, post-blasting with 0.7 μm cerium oxide required

103

machining times too long to be practical (approximately 1 h/mm2), consistent with the longer

machining times using cerium oxide instead of silicon carbide in the FJP study of Fang et al. (2006)

[4] on glass.

Figure 4.9 Measured centerline Rrms roughness of post-blasted channels in zirconium tin titanate at

15 jet incidence as a function of particle dose using different particles. Stationary jet. Error bars

represent ±1 standard deviation for 3 scans within the footprint.

Since the use of 3 μm silicon carbide resulted in the shortest machining time while giving

the same steady-state roughness, it was used to post-blast reference channels in glass, PMMA, and

aluminum nitride. Figure 4.10(a) shows that a transient roughness was evident during the post-

blasting of all the target materials, and indicates that the steady-state roughness was reached more

quickly in zirconium tin titanate and PMMA surfaces than in glass, since the former materials

exhibited more plastic deformation than the latter, as explained in Section 4.4.2. Figure 4.10(a) also

shows that the steady-state roughnesses were approximately equal for glass, PMMA, and zirconium

tin titanate, presumably due to the large difference between the hardness of silicon carbide and the

104

targets. As discussed in Section 4.4.2, the material removal mechanism in aluminum nitride using

alumina particles was grain removal without significant grain damage. However, Fig. 4.10(a) shows

that the harder silicon carbide particles could smooth aluminum nitride, but the process of

roughness-profile peak removal was much slower than with the other three target materials because

of the much higher starting roughness. Longer machining times were avoided to minimize the wear

of the pump valve seats that was detected with the use of silicon carbide particles. Figure 4.10(b)

shows the post-blasted surfaces of the target materials at steady state. The effect of the polishing is

evident by comparing with the initial surfaces shown in Fig. 4.4(b). The aluminum nitride displayed

a large decrease in the number of pits, consistent with the hypothesis that the silicon carbide

particles under the post-blasting conditions were able to plastically erode individual grains of

aluminum nitride, in contrast to the removal of entire, discrete grains evident during ASJM under

typical conditions (Fig. 4.4(b)).

105

(a)

(b)

(c)

Figure 4.10 (a) Measured centerline Rrms roughness of post-blasted channels in glass, PMMA, and

zirconium tin titanate as a function of particle dose using 3 μm silicon carbide particles at 15 jet

incidence. Error bars represent ±1 standard deviation for 3 measurements. (b) Plan view SEM

images of post-blasted surfaces using the same conditions in glass, PMMA, and zirconium tin

titanate. (c) Isometric AFM view region A of a channel in zirconium tin titanate post-blasted with 3

μm silicon carbide particles at a dose of approximately 90 g/mm2.

106

An energy-dispersive X-ray spectroscopy (EDS) analysis of the target surfaces showed that

there was no particle embedment with the slurries of either the 10 μm alumina or the 3 μm silicon

carbide particles. It is believed that the abrasives were washed away after impact by the relatively

high-speed fluid flow.

Table 4.4 summarizes the achievable percentage reductions in channel Rrms roughness

compared to the starting roughnesses of either the as-received material or the channels after

machining using typical ASJM process conditions.

Table 4.4 Percentage change in channel centerline Rrms roughness compared to the as-received

surfaces or channel centerline surfaces after machining under typical process conditions. + indicates

an increase in roughness, - indicates a decrease.

Target material

Percent change in Rrms roughness with respect to:

as-received

surface

centerline surface of channel

machined under typical ASJM

conditions

Glass +65% -90%

PMMA -0.4% -81%

Zirconium tin titanate -94% -83%

Aluminum nitride -15% -56%

4.4.5. Roughness prediction during post-blasting

As discussed in Section 4.4.2, the material removal mechanism during the ASJM post-

blasting in glass, PMMA, and zirconium tin titanate involved plastic deformation. The ductile

roughness model described in Section 4.3 was used to predict the surface roughness of the post-

blasted ASJM channels. The dynamic hardness of the target materials was estimated by treating it

107

as an adjustable parameter in a roughness simulation using process condition set I in Table 4.3. The

value of dynamic hardness that gave the best match to the measured roughness for each target

material under condition set I is given in Table 4.1. These best-fit values were then used to predict

the surface roughness using process condition sets II and III (Table 4.3). The main differences

between the calibration condition set (I) and sets II and III were the particle size, impact angle and

impact velocity.

Figure 4.11 shows that the channel roughnesses were predicted with an average error of

about 12% for glass, PMMA, and zirconium tin titanate. The relatively high accuracy of the model

can be attributed in part to using calibrated dynamic hardness values for these target materials. The

relatively small variation in these values (2600, 2750 and 2950 MPa, for glass, PMMA and

zirconium, respectively; Table 4.1) was consistent with the close similarity of the steady-state Rrms

values as seen in Fig. 4.10(a). The calibrated dynamic hardness value of PMMA (2600 MPa) was

much larger than that measured by Getu et al. (2012) [23] (970 MPa) using the coefficient of

restitution of impacting 50 μm-diameter spherical stainless steel particles. It is believed that this

was due to the difference in the impact conditions between traditional dynamic hardness testing and

the ductile erosion conditions for the very small particles used in the present post-blasting

experiments.

The model was limited to predicting the steady-state surface roughness because of the way in which

it superimposed simulated surface profiles until a constant Rrms was achieved.

108

Figure 4.11 Measured (black) and predicted (gray) channel centerline Rrms roughnesses for the

process conditions II and III (Table 4.3), selected for channel-machining and post-blasting (peak

removal), respectively.

4.5. Conclusions

The potential of abrasive slurry-jet micro-machining (ASJM) as a means of polishing

surfaces was investigated by measuring changes in the centerline roughness of micro-channels

machined in borosilicate glass, polymethylmethacrylate (PMMA), sintered zirconium tin titanate,

and sintered aluminum nitride. The principal parameters were the abrasive particle dose (i.e.

number of machining passes or jet traverse speed), particle material, diameter, and kinetic energy

(impact velocity or impact angle). It was found that for typical ASJM conditions, ductile plastic

deformation was the dominant erosion mode, even in glass since the particle kinetic energies were

below the theoretical transition energy required for fracture. Slower particle impacts at shallower

angles using smaller particles could produce approximately 35% smoother channels compared to

the roughest channels machined in glass, PMMA, and zirconium tin titanate at the largest normal

particle kinetic energy, but at 64% lower etch rates on average. At conditions optimized to obtain

the smoothest surfaces, machining of channels of practical depths would require relatively long

109

machining times, therefore the post-blasting of channels machined under typical parameters was

explored as a means of polishing. Under post-blasting conditions (89 m/s jet velocity, 15 jet

inclination, 3 μm silicon carbide particles), channels in glass, PMMA, zirconium tin titanate, and

aluminum nitride were smoothed to root-mean-square (Rrms) roughnesses of 23, 23, 19, and 170 nm.

These surfaces were smoother than the as-received surfaces for PMMA (0.4% smoother), zirconium

tin titanate (94%), and aluminum nitride (15%), but rougher for glass (65%). An existing ductile-

regime surface roughness simulation model could predict the steady-state roughness of the ASJM

surfaces with an average error of 12%.

4.6. References

[1] F.C. Tsai, B.H. Yan, C.Y. Kuan, F.Y. Huang, A Taguchi and experimental investigation into the

optimal processing conditions for the abrasive jet polishing of SKD61 mold steel, Int. J. Machine

Tools Manuf. 48 (2008) 932–945.

[2] W. Peng, C. Guan, S. Li, Material removal mode affected by the particle size in fluid jet

polishing, Applied Optics 52 (33) (2013) 7927-7933.

[3] L. Zhang, S. Raghavan, M. Weling, Minimization of chemical-mechanical planarization (CMP)

defects and post-CMP cleaning, J. Vacuum Science Tech. B 17 (1999) 2248-2255.

[4] H. Fang, P. Guo, J. Yu, Surface roughness and material removal in fluid jet polishing, Appl.

Optics 45 (17) (2006) 4012-4019.

[5] R.J. Wang, C.Y. Wang, W. Wen, J. Wang, Experimental study on a micro-abrasive slurry jet for

glass polishing, Int. J. Adv. Manuf. Technol. (2016) DOI 10.1007/s00170-016-9109-z.

[6] F. Zhang, X. Song, Y. Zhang, D. Luan, Figuring of an ultra-smooth surface in nanoparticle

colloid jet machining, J. Micromech. Microeng. 19 (2009) 054009.

110





[9] H. Nouraei, A. Wadoslawsky, M. Papini, J.K. Spelt, Characteristics of abrasive slurry jet micro-

machining: a comparison with abrasive air jet micro-machining, Journal of Materials Proc. Tech.

213 (2013) 1711-1724.

[10] Z. Cao, C.F. Cheung, Theoretical modelling and analysis of the material removal

characteristics in fluid jet polishing, Int. J. Mech. Sci. 89 (2014) 158–166.

[11] M. Papini, J.K. Spelt, Impact of rigid angular particles with fully-plastic targets part II:

parametric study of erosion phenomena, Int. J. Mech. Sci. 42 (2000) 1007–1025.

[12] Z. Li, S. Li, Y. Dai, X. Peng, Optimization and application of influence function in abrasive jet




[14] C. Y. Poon, B. Bhushan, Comparison of surface roughness measurements by stylus profiler,

AFM and non-contact optical profiler, Wear 1190 (1995) 76-88.

[15] K. Kowsari, H. Nouraei, , B. Samareh, M. Papini, J.K. Spelt, CFD-aided prediction of the

shape of abrasive slurry jet micro-machined channels in sintered ceramics. Ceramics Int. 42

(2016) 7030-7042.

[16] J. Schwartzentruber, M. Papini, J.K. Spelt, Prediction of Surface Roughness in Abrasive

Waterjet Trimming of Fiber-Reinforced Polymer Composites, Int. J. Machine Tools Manuf.

(2016, submitted).

[17] M. Papini and J.K. Spelt, Organic coating removal by particle impact, Wear 213 (1997) 185-

199.

[18] ISO 4288: Geometrical Product Specifications (GPS)—Surface texture: profile method—rules

and procedures for the assessment of surface texture (1996).


erosion, Wear 253 (2002) 1035-1043.

[20] R. Haj Mohammad Jafar, M. Papini, J.K. Spelt, Numerical simulation of surface roughness and

erosion rate of abrasive jet micro-machined channels, Wear 303 (2013) 302–312.

111

[21] Hasem, Md. A., The effect of tangential velocity component in abrasive jet micro-machining

of borosilicate glass, Ryerson University Ph.D. Thesis (2013).

[22] I. Finnie, Erosion of surfaces by solid particles, Wear 3 (2) (1960) 87-103.

[23] H. Getu, J.K. Spelt, M. Papini, Conditions leading to the embedding of angular and spherical

particles during the solid particle erosion of polymers, Wear 292-293 (2012) 159-168.

112

Chapter 5: Hybrid Erosive Jet Micro-milling of

Sintered Ceramic Wafers With and Without Copper-filled

Through-holes

5.1. Introduction

Kowsari et al. (2014) [1] demonstrated the feasibility of abrasive slurry-jet micro-machining

(ASJM) to machine micro-channels and micro-holes in sintered alumina, and used a basic surface

evolution model to predict the shapes of the profiles. However, that study was limited to feature

depths smaller than 50 m so that the near-flat target geometry had no effect on the slurry flow

field. The modeling of deeper features must account for changes in the erosive flow.

Machining of micro-pockets in ceramics containing metallic through-holes (Fig. 5.1) is of

interest in industrial applications such as the packaging of hybrid microwave integrated circuits

(HMIC) involving high-power, low-noise amplifiers operating at high frequencies; i.e. 3-30 MHz,

as explained by Khalil et al. (2009) [2]. The use of relatively thick (i.e. 375-675 m) sintered

ceramic wafers such as alumina, aluminum nitride, and zirconium tin titanate as substrates can

result in an electronically optimal circuit. However, the relatively low thermal conductivity of

alumina for instance (26.9 W/m·k) can cause an active device such as a field effect transistor to

overheat if placed on such a substrate. Temperatures can be reduced using copper-filled through-

holes (vias) to conduct heat through a wafer to an attached heat sink as illustrated in Fig. 5.1.

113

(a)

(b)

Figure 5.1 Schematic of (a) cross-sectional profile and (b) plan view of a pocket in an aluminum

nitride wafer containing copper-filled through-holes (vias).

The devices are soldered to the bottom of the pockets, making thermal and electrical

connection with the copper-filled through-holes or vias. Laser micro-milling is commonly used for

such applications, but, as explained by Jandeleit et al. (1998) [3], avoiding thermal damage and

micro-cracking require the use of relatively expensive, high-frequency lasers. The aim of the

present work was to investigate the feasibility of ASJM as a low-cost alternative to micro-mill

pockets into composite substrates of sintered ceramic wafers containing metallic-filled through

holes.

The only existing investigation of the ASJM of pockets was conducted by Tamannaee et al.

(2016) [4] who used overlapping adjacent nozzle passes to create 800 m wide flat-bottomed

pockets in ductile polymers. The effect of overlapping eroded footprints was also considered in the

fluid jet shaping and polishing of optical glass (e.g. Fähnle et al. (1998) [5], Booji et al. (2004) [6]

and (2002) [7], and Fang et al. (2006) [8]), and the ion beam milling of optical glass by Shanbhag et

al. (2000) [9]; however, these applications considered only the removal of a few microns from

114

rotating glass targets. Of related interest is the work of Ghobeity et al. (2008a) [10] who used

abrasive air jet micro-machining (AJM) to machine flat pockets in brittle glass using over-lapping

machined channels; however, the minimum pocket width was about 10 mm because of the

relatively large air jet footprint. The much larger divergence of the AJM jet relative to an ASJM jet

usually requires the use of patterned masks to reduce the size of the blast zone to the micron range.

For example, Park et al. (2005) [11] used an ultraviolet (UV) hardening polyurethane mask to

produce pockets as small as 50 m wide in metals. More recently, Billingham et al. (2013) [12]

used a high-pressure (413.7 MPa) abrasive water jet machine (AWJM) with a 1 mm diameter

nozzle and 180-300 m garnet particles to machine pockets into a titanium-based alloy (Ti6Al4V)

using over-lapping channels. They developed a model to predict the pocket profiles, but its

applicability was limited to very shallow pockets because of non-linearities in the local erosion rate

and fluid flow field brought about by the steepening sidewalls.

In summary, although ASJM has the potential to be a relatively inexpensive technology for

milling flat pockets into sintered ceramics, little is known about the associated material removal

mechanisms in such materials. Relevant milling studies of the past are limited to high-pressure

AWJM and AJM of glass and metal targets. This paper presents an experimental study of the

effects of process parameters on the shape and roughness of micro-pockets machined in sintered

ceramics with and without copper-filled through-holes using, for the first time, a hybrid AJM and

ASJM methodology. The observed trends were explained using CFD modeling of the slurry flow

fields.

115


5.2.1. Experiments

Table 5.1 gives the properties of the sintered ceramic target materials used in the experiments.

In addition, tests were conducted on aluminum nitride wafers (Table 5.1) which also contained

copper-filled 180 μm diameter through-holes in the configuration shown in Fig. 5.1.



(mm)

Grain size

(m)

Density

(g/cm³)

Vickers

hardness

(kg/mm²)

Alumina

(Al₂O₃) Superstrate 996, CoorsTek

Inc., Golden, CO, USA 10×10×0.375 < 1 3.88 1800

Aluminum

nitride (AlN)

K170, Toshiba Corp., Minato,

Tokyo, Japan 50×50×0.375 < 1 3.26 1100

Zirconium tin

titanate

(Zn-Sn-TiO₂)


Ach, Japan 50×50×0.375 < 5 5.20 950

The AJM experiments were conducted using an AccuFlo abrasive blaster (Comco Inc.,

Burbank, CA, USA), described in detail in Dehnadfar et al. (2011) [13]. The nozzle had a diameter

of 1700 m and a length-to-diameter ratio of about 22. In both ASJM and AJM, the targets were

clamped to computer-controlled linear stages (accuracy 15 μm) at a standoff of 20 mm.

Table 5.2 outlines the standard machining conditions in the ASJM and AJM experiments, which

were selected to obtain typical feature depths and widths.

116

Table 5.2 Standard process parameters.

Process Abrasives Particle flow rate

(g/min)

Pressure

(MPa)

Standoff

(mm)

Scan speed

(mm/s)

ASJM 10 μm alumina 1.4 8.0 20 0.05

AJM 25 μm alumina 8.0 0.2 20 0 (stationary jet)

5.2.2. CFD modeling

CFD models in ANSYS Fluent 14.0 (ANSYS Inc., Cecil Township, PA, USA) were used to

predict the abrasive particle trajectories in both ASJM and AJM. This included models of the

impingement of air-in-air jets AJM and slurry-in-air jets in ASJM on flat plates at perpendicular

and 45° incidence, as well as the slurry jet flow within shallow and relatively deep channels. At

perpendicular incidence, 2D axisymmetric domains were used with the boundary conditions shown

in Fig. 5.2(a), while 3D models, similar to that shown in Fig. 5.2(b), were used at oblique incidence

or for flows within channels. The jet velocities used at the inlet boundaries were obtained from

experimental measurements; i.e. 150 and 126 m/s for AJM and ASJM jets, respectively. The

domains were meshed with quadrilateral elements having widths of approximately 1 m. To

capture the shear flow in the viscous sub-layer, the dimensionless wall coordinate, y+, was

maintained below 1 using near-wall grid refinement, as shown in Fig. 5.2(a).

117

(a)

(b)

Figure 5.2 Domains and boundary conditions of (a) a 2D axisymmetric model of the impingement

of a slurry jet on a flat target, and (b) a 3D model for the simulation of the flow within a channel.

Elements not to scale.

118

For the multi-phase simulations of ASJM, the volume of fluid (VOF) model was employed

to simulate the flow of the primary phase, water, and the surrounding secondary phase, air. Particles

were injected and tracked using the discrete phase model (DPM) in which the particle shape factor

was set to be 0.76 for the 10 m alumina abrasives, as measured by Dehnadfar et al. (2011) [13]. As

suggested by Murakami (1993) [14], the standard к-є turbulence model provided only moderate

agreement in the prediction of flow fields near a stagnation point due to the high strain rates.

Therefore, turbulence was modeled using the Reynolds stress model (RSM) which was considered

suitable for stagnation point flows in the work of Gnanavelu et al. (2011) [15]. The simulations

were made to converge to a maximum residual of 10-3

, as suggested by Tu et al. (2008) [16] and

ANSYS (2009) [17].


5.3.1. Erosion mechanism

Hockin et al. (1995) [18] explained that the erosion mechanism in sintered ceramics such as

alumina involves cracking along grain boundaries (Fig. 5.3(a)), leading to grain dislodgement.

Missing grains can be seen in Fig. 5.3(b), which shows a scanning electron microscope (SEM)

image of the post-blasted surface of sintered alumina.

119

(a) (b)

Figure 5.3 SEM images of the surfaces of sintered alumina: (a) with exposed grains without any

ASJM (reproduced from CoorsTek (www.coorstek.com) material property catalog), and (b)

showing single-particle-impact sites created by scanning the ASJM jet containing 0.01 wt% 10 m

diameter alumina particles over the target at a scan speed of 5 mm/s at normal incidence.

The specific erosion rate (mass of removed material per dry mass of abrasive particles) of

the alumina target subjected to ASJM was measured as a function of jet impact angle from 15-90°

in the configuration shown in Fig. 5.4(a) using shallow machined channels in the same manner as

Ghobeity et al. (2008b) [19]. The results, summarized in Fig. 5.4(b), indicate that the alumina

eroded in a typically brittle manner, similar to the ASJM of borosilicate glass as reported by

Nouraei et al. (2014) [20]. However, whereas glass erosion was by cracking and chipping, the

brittle behavior in sintered alumina was brought about by a tendency for normal impacts to cause a

greater degree of grain removal. The fitted curve in Fig. 5.4(b) was drawn to fit through the origin,

since there was no evidence of an erosion threshold for sintered grain removal.

120

(a) (b)

Figure 5.4 (a) Schematic of the oblique jet orientation in the machining of an asymmetrical ASJM

channel. (b) Dependence of ASJM normalized erosion rate (erosion rate at a given angle divided by

that at 90) of sintered alumina (Table 5.1) on jet impact angle. The erosion rate at perpendicular

incidence was measured as 0.05 mg/g. Error bars represent ±1 standard deviation for 3

measurements.

The percentage of incident particles that caused grain dislodgement was measured by

scanning a dilute jet of 0.1 wt% alumina particles at a relatively fast speed of 6 mm/s to ensure that

the impact sites were sufficiently scattered to enable them to be counted. A relatively small number

of pre-existing pits on the as-received wafers, presumably removed during the manufacturing

process, were identified on SEM images of the region to be scanned and were subtracted from the

SEM of the eroded surface using the digital analysis software (ImageJ—http://rsb.info.nih.gov/ij/).

It was found that 1786 impact pits were created by ASJM on an area of 3.5×1.0 mm, of which 189

were pre-existing. Assuming spherical particles of average size, approximately 45000 particles

impacted the surface under the given experimental conditions, meaning that only 4% of the

impacting particles caused grain dislodgement.

121

5.3.2. ASJM pockets in sintered ceramics

Planar areas in alumina were machined using over-lapping channels with offsets of 25 m

and 200 m (Fig. 5.5) using the standard conditions (Table 5.2). The profiles of Fig. 5.5(b) were

obtained using an optical profilometer (ST400, Nanovea Inc., Irvine, CA) having a depth resolution

of 10 nm. As expected, the smaller offset resulted in a flatter pocket floor.

122

(a)

(b)

Figure 5.5 (a) Machining path of slurry jet during the machining of a pocket using the over-lapping

channel method. (b) Isometric views of the surface profiles of pockets in sintered alumina for 25

and 200 m channel offsets. The in-plane dimensions are to scale, while the depth is amplified by

15%.

The effect of offset on pocket roughness was investigated in greater detail by varying the

offset from 25 m to 250 m (16%-167% of the jet diameter) while selecting the total number of

adjacent traverses in order to machine pockets having widths of approximately 2200 m. The

resulting profiles in Fig. 5.6(a) were measured along the A-A cross-section shown in Fig. 5.5(b). To

123

quantify the roughness of the bottom surfaces of these pockets; i.e. section B-B (Fig. 5.6(a)), a fast

Fourier transform (FFT) (MATLAB, Version 7.12.0, Mathworks, Natick, MA, USA) was first used

to convert the spacial domain of the profile to a frequency domain to determine the wavelength

having the largest amplitude, consistent with the methodology used by Jafar et al. (2013) [21]. This

frequency was then input as the cutoff in a high-pass filter applied according to the ISO 11562

(1996) [22] standard to separate the roughness from the waviness. The filtered profile was then used

to obtain the arithmetic average roughness, aR , using the ISO 4288 (1996) [23] standard with a

cutoff of 0.25 μm as suggested by Jafar et al. (2013) [21]. Figure 5.6(b) presents pocket roughness

versus offset and shows that the aR of the flattest ASJM pocket in alumina was about 0.4 m when

the offset was set to 25 m.

(a) (b)

Figure 5.6 (a) Measured and predicted cross-sectional profiles of ASJM pockets machined using

the over-lapping channel method in alumina for offsets of 50 and 200 m (b) Pocket roughness, aR ,

versus offset in sintered alumina using the standard conditions of Table 5.2.

124

Figure 5.6(a) also shows the profiles of the 50 and 200 m offset pockets predicted by the

superposition procedure used in Tamannaee et al. (2016) [4] and Ghobeity et al. (2008a) [10], in

which the measured profile of a single channel was summed with the adjacent profile, one offset

away, repeating for the total number of machining passes. The waviness of the 200 μm offset

pocket can be explained by the relatively large offset/channel width ratio of 0.69 (channel width

was 290 μm). It is seen that the predicted and measured values were in good agreement; however,

it was found that this method gave accurate results only in cases where the individual channel

depths were shallow enough (<50 m) to result in sidewall angles less than about 25° from the

horizontal. Under these conditions the flow field was similar to that on a flat plate.

For deeper pockets, such as that shown in Fig. 5.7(a), the eroded shape produced significant

changes in the slurry flow, and thus the resulting erosion pattern. Deeper pockets caused an

enlargement of the stagnation zone in the jet footprint as illustrated in the 3D CFD models of Figs.

5.7(b) and 5.7(c) using domains similar to that in Fig. 5.2(b). The shallow (25 m) and relatively

deep (135 m) channels correspond to those made using the standard conditions with 1 and 10

machining passes, respectively. Figure 5.7(b) shows that the height of the stagnation zone increased

by a factor of 1.8 to about 360 m in the deeper channel. Particles passing through this larger

stagnation had much lower impact velocities, thus causing less erosion. Therefore, because of this

depth-dependent flow field effect, the depths of ASJM channels increased less than linearly with

the number of machining passes (Fig. 5.8(a)), thus invalidating the linear superposition when the

over-lapping channels of a single operation were relatively deep. The decelerating effect of the

stagnation zone is similar to the observation of Haghbin et al. (2015) [24] that the depths of

channels machined using abrasive water jets were proportional to the depth-to-width aspect ratio of

the region within the channel that was effectively filled with slurry during machining. This filled

125

portion of the channel served to decelerate incident abrasive particles, thereby decreasing the rate of

erosion.

When adjacent channels were machined using an offset of 150 m, Fig. 5.7(c) shows that

the enlargement of the stagnation zone was marginal, but that the steeper sidewall of the deep

channel significantly decreased the particle impact angles, which, according to Fig. 5.4(b), led to a

reduced erosion rate.

126

(a)

(b)

127

(c)

Figure 5.7 (a) Surface geometries of a shallow (depth = 25 μm) and relatively deep (depth = 135

μm) channel machined in sintered alumina using standard ASJM conditions. Static pressure

contours for the flows within a shallow channel and a relatively deep channel (b) without any

offset, and (c) with an offset of 150 m. A gage pressure of 0.25 MPa defined the stagnation zone

boundary.

128

The non-linear effects brought about by the larger aspect ratio of the filled region of deep

channels in sintered ceramics as shown in Fig. 5.8(a) can be eliminated by machining relatively

shallow over-lapping channels to form a shallow pocket, then repeating this operation with

additional shallow over-lapping channels until the desired pocket depth has been reached. Figure

5.8(b) presents the cross-sectional profiles of multi-operation pockets, and Fig. 5.8(c) illustrates the

linear increase in the depths of these pockets with increasing machining operations. It is also seen

that increasing the number of operations had little effect on the width or roughness of the pocket

floor.

129

(a)

(b) (c)

Figure 5.8 (a) Depth as a function of the number of machining passes of channels machined in

alumina using the standard conditions. (b) Cross-sectional profiles of pockets machined in alumina

after each machining operation using the overlapping channel method (standard conditions; 150 m

offset; 6 overlapped channels per operation). (c) Pocket depth as a function of the number of

machining operations. Error bars represent ±1 standard deviation for 3 measurements.

Since the maximum sidewall angle of ASJM pockets machined using the overlapping

channel method was less than 15° at a depth of about 50 m, it was of interest to investigate

methods to steepen the sidewalls. Tamannaee et al. (2016) [4] found that sidewalls of pockets in

ductile materials could be steepened by using a compound inclination of the jet as it machined the

entire pocket; i.e. the jet axis was both 45 from the target plane and 45 from the axis of the

machining pass. However, in the present work with a brittle ceramic, it was found that 60°

130

sidewalls could be machined at a depth of about 70 m as shown in Fig. 5.9 by machining the entire

pocket using the asymmetrical channel configuration of Fig. 5.4(a) at an oblique jet impingement of

45. When machining pockets in ductile materials, Tamannaee et al. (2016) [4] found that this

orientation created a secondary slurry flow that accentuated the waviness of the floor of the pocket.

This difference in behavior was due to the low erosion rate of brittle materials at relatively small

impact angles where ductile materials experience high erosion, as explained by Oka et al. (1997)

[25].

Figure 5.9 Isometric view of the surface geometry of a pocket in alumina using over-lapping

channels each machined at an oblique angle (standard conditions; 5 machining passes per channel).

131

5.3.3. AJM pockets in sintered ceramics

The relatively small divergence of the ASJM jet required machining of multiple over-

lapping channels to erode planar areas having the desired widths of about 500 m (Fig. 5.1(a)). The

relatively large ~4 mm footprint of the AJM jet at a standoff of 20 mm made it possible to machine

larger areas using a stationary jet and a rectangular mask to define the pocket boundaries and

provide an adequate sidewall angle of 12 for the present application. Steeper sidewall angles could

have been machined using the approach of Ghobeity et al. (2008a) [10]. Figures 5.10(a) and

6.10(b) shows the surface geometries of pockets machined in sintered alumina and sintered

aluminum nitride using AJM with a 0.9 mm thick copper-tungsten mask. It is seen that for the given

parameters, alumina was approximately 6.25 times more erosion resistant than aluminum nitride,

but the pocket in alumina was roughly twice as wavy, presumably due to the differences in grain

size and/or sinter bond strength. In summary, flat pockets can be machined in sintered ceramics

using masked AJM.

132

(a)

(b)

Figure 5.10 Isometric views of the surface profiles of masked AJM pockets (standard conditions;

15 s machining time) in (a) sintered alumina and (b) sintered aluminum nitride.

133

5.3.4. Pockets in sintered aluminum nitride containing copper-filled through-holes

5.3.4.1. Application of ASJM

As discussed in Section 5.1, the machining of flat pockets in aluminum nitride wafers

containing copper-filled through-holes is of interest in a variety of microelectronics applications.

The use of ASJM to create pockets in sintered aluminum nitride containing copper-filled through-

holes was investigated by first machining single channels at perpendicular incidence over a series of

filled through-holes using 2, 4, and 6 passes. Figure 5.11 shows that the copper through-holes were

preferentially etched within each channel, becoming relatively deeper with increasing numbers of

passes (Fig. 5.11(b)).

134

(a) (b)

Figure 5.11 (a) Isometric view of the surface geometry of channels machined using ASJM along a

series of copper-filled through-holes in a matrix of sintered aluminum nitride. #P denotes the

number of machining passes. (b) Dependence of maximum depths of aluminum nitride and copper-

filled through-holes within the channels on the number of machining passes. Error bars represent ±1

standard deviation for 3 measurements taken from three separate filled through-holes along a

channel. Standard conditions, but at 0.1 mm/s scan speed.

Figure 5.11(a) also revealed that the erosion of the copper through-hole was highly non-

uniform, with much more erosion occurring in the center of each filled through-hole. This was

probably a consequence of property gradients that can be created in filled through-holes as the

copper plating deposition proceeds inward from the walls and ends of the hole, as explained by

Dixit and Miao (2006) [26] and Dow et al. (2008) [27]. This was supported by micro-hardness

measurements on the present holes which showed that the Vickers hardness in the central portion of

each hole was 70 kg/mm2 (average of 5 filled through-holes with a standard deviation of 23),

whereas it was 135 kg/mm2 (average of 5 filled through-holes with a standard deviation of 7) near

the perimeter. To minimize edge effects in these hardness measurements, the distance between the

centers of the peripheral indentations and the surrounding ceramic was kept greater than the

indentation size (35 m) as suggested by Pollock et al. (1986) [28]. An attempt was made to

135

eliminate the hardness non-uniformity within a filled through-hole by annealing at 600° C for 6 h

followed by a slow cooling in the closed oven. Although this failed to significantly change the

pattern of hardness variation within a filled through-hole, it did decrease the degree of additional

erosion in the center as well as the hardness variability seen from one hole to the next.

The preferential erosion of the copper-filled through-holes compared to the surrounding

ceramic was attributed to the action of the slurry flow as it spread from the footprint, as illustrated

in Fig. 5.12(a). Within much of the footprint away from the jet axis, and especially in the regions

scoured by the spreading flow, the ductile copper was eroded relatively quickly by the small impact

angle of the abrasive particles. This is illustrated in Fig. 5.12(b), which shows that filled through-

holes were preferentially eroded by the spreading flow even when the jet passed between rows of

holes such that the footprint (approximately 200 m diameter) remained largely over the ceramic.

136

(a)

(b)

Figure 5.12 (a) Schematic of the lateral flow exiting the footprint in the perpendicular impingement

of an ASJM jet on a target plate near a copper through-hole. (b) Deepening filled through-holes due

to lateral flow of increasing passes (#P) of jet between rows of filled through-holes using standard

conditions (Table 5.2).

137

An attempt was made to reduce the difference in erosion between the copper and aluminum

nitride by increasing the jet traverse speed (decreasing particle dose) when machining the copper, so

that the erosion matched that of aluminum nitride. For a scan speed of 0.05 mm/s over the

aluminum nitride, the required traverse speed over each copper through-hole was determined to be

0.09 mm/s using the measured erosion rate difference (factor of 1.7) in aluminum nitride and

copper. It was found that this procedure did indeed equalize the initial erosion of the aluminum

nitride and copper through-holes, but that the effect began to disappear beyond a pocket depth of

about 25 m, with the copper through-holes again becoming progressively deeper than the

surrounding aluminum nitride.

5.3.4.2. Hybrid use of AJM and ASJM

The combined use of AJM and ASJM in milling flat pockets in aluminum nitride wafers

containing copper through-holes was motivated by the differences in local particle impact angles

between the processes. The erosion of brittle and ductile materials depends strongly on the local

particle impact angle, which is governed by the local flow field close to the target surface. To

illustrate the effect of flow field on particle impact angle, the impingement of both ASJM and AJM

jets containing 10 and 25 μm-diameter alumina particles, respectively, on targets at 90° and 45°

incidence was simulated using CFD. Figure 5.13(a) shows that the relatively large stagnation zone

in the ASJM footprint region caused the water to spread laterally over the surface. As a result, the

impact angles were about 90° at the center of the perpendicular jet and approximately 45° near the

edges of the footprint (overall average of about 50°), consistent with the findings of Haj

Mohammad Jafar et al. (2014) [29]. For the jet inclined at 45°, the average particle impact angle

was approximately 35°. Thus the average local impact angles at global impact angles of 90° and

138

45° were quite close, differing by only 15°, which helps explain why the ASJM erosion rate of

copper was significantly larger than that of aluminum nitride for both 90° and 45° jet angles

(Section 5.3.4.1).

Figure 5.13(b), on the other hand, shows that there was virtually no stagnation zone for AJM

at a standoff of 10 mm, resulting in a very abrupt change in the air streamlines near the surface.

Moreover, since the fluid viscosity was very small in AJM compared to that in ASJM, the particle

momentum equilibration number, (calculated using the ASJM jet and AJM nozzle diameters, with

the respective jet velocities) was approximately 36 times larger in AJM so that the particles in the

air stream were much less likely to follow the streamlines, but rather to continue on their original

trajectories and strike the surface. Therefore, the average particle impact angles for AJM nozzle

angles of 90° and 45° were about 85° and 45°, respectively. This is consistent with the AJM local

impact angle measurements of Dehnadfar et al. (2011) [13]. These higher local impact angles, and

the angular dependence of erosion in brittle and ductile materials (Oka et al., 1997 [25]) explain

why AJM produced much greater relative erosion of the brittle ceramic compared with the ductile

copper for air-jet impact angles between 45° and 90°.

139

(a)

(b)

Figure 5.13 Static pressure contours and particle trajectories for targets at 90° and 45° for the

impingement of; (a) an ASJM jet (8 MPa orifice pressure, 10 m alumina), and (b) an AJM jet

(200 kPa orifice pressure, 25 m alumina). A gage pressure of 0.25 MPa defined the stagnation

zone boundary.

A series of experiments were performed to evaluate whether the differences in local particle

impact angles in ASJM and AJM could be exploited to machine pockets of uniform depth in the

aluminum nitride containing copper-filled through-holes. The first step was to use the stationary 90

AJM nozzle to erode the pocket area (standard conditions; 20 s exposure) through a 500 m-wide

copper-tungsten mask of the type described in Section 5.3.3. Figure 5.14(a) shows that this

produced a flat aluminum nitride surface at the pocket floor depth of 60 m with a sidewall angle of

140

57°. The copper through-holes were largely un-eroded and remained as pillars of about 60 m

height. The second step was to remove the copper pillars by scanning the 90 ASJM jet four times

over each row of filled through-holes using 4 MPa pressure and 0.4 mm/s scan speed and otherwise

standard conditions (Table 5.2). Figure 5.14(b) shows that this produced a uniformly flat pocket,

since the ASJM under these conditions (50% reduction in pressure and 800% increase in scan speed

compared to the standard conditions of Table 5.2 and Fig. 5.11) essentially eroded only the copper

with negligible additional erosion of the aluminum nitride. Furthermore, this lower pressure and

higher scan speed did not cause erosion in the neighboring row of through-holes, unlike the

standard ASJM conditions (Table 5.2) as seen in Fig. 5.12.

In summary, combining AJM and ASJM allowed for selective erosion of adjacent brittle and

ductile materials in each step. AJM is characterized by local particle impact angles that are close to

the nozzle angle, and can thus be close to 90. Moreover, particle impact velocities in AJM are

much higher than in ASJM, so that relatively high erosion occurs in brittle materials, even at lower

angles. For this reason, AJM alone could not be used to erode the ceramic at 90 and then only the

copper at some smaller nozzle angle. ASJM in contrast, is characterized by both smaller local

particle impact angles at all nozzle angles, and by smaller impact velocities, making it possible

erode ductile materials at low impact angles while not eroding adjacent brittle material.

The machining protocol to create pockets in such a composite material should first establish

the AJM exposure time for a stationary jet at perpendicular incidence to propagate the ceramic

pocket floor to its final depth. The remaining copper pillars can then be leveled by scanning a

perpendicular ASJM jet over the pillars at a relatively fast speed to minimize the degree of lateral

flow.

141

(a)

(b)

Figure 5.14 Optical profilometer images of the two stages in pocket machining in sintered

aluminum nitride with copper-filled through-holes: (a) after masked AJM and (b) after masked

AJM followed by unmasked ASJM to flatten the copper pillars.

142

5.4. Conclusions

The material removal mechanism in the ASJM of sintered ceramics such as alumina,

aluminum nitride, and zirconium tin titanate was found to involve grain dislodgement brought about

by particle impact. The lower limit on the machined surface roughness was determined by the size

of these sintered grains which tended to be removed intact. The dependence of the erosion rate on

global impact angle was typical of brittle materials.

Using an over-lapping channel methodology in ASJM, pockets with a roughness, Ra, of

about 0.4 m were machined in alumina. The pocket shape could be predicted using the same

superposition method as Tamannaee et al. (2016) [4] for machining operations in which each pass

removed less than about 50 m from the floor of the pocket. Deeper passes changed the local

geometry of the machining front under the jet footprint and caused the stagnation zone to enlarge,

thereby decreasing the erosion rate. It was also demonstrated that pockets of similar size and

roughness could be machined in sintered ceramics using masked AJM. Finally, it was demonstrated

that 60 m deep flat pockets having a sidewall angle of 57° from the horizontal could be milled in

aluminum nitride wafers containing 180 m-diameter copper through-holes using a hybrid AJM-

ASJM methodology. AJM was used first to selectively erode the brittle ceramic without eroding the

ductile copper through-holes. ASJM was used in a second step to selectively erode the copper

pillars remaining from the first step while leaving the surrounding ceramic essentially intact.

143

5.5. References




[2] I. Khalil, M. Rudolph, A. Leiro, M. Neumann, W. Heinrich, High power, high linearity and low-

noise hybrid RF amplifiers based on GaN HEMTs, Microwave Conference 1 (2009) 16-18.



(1998) 885-891.

[4] N. Tamannaee, J.K. Spelt, M. Papini, Abrasive slurry jet micro-machining of edges, planar areas

and transitional slopes in a talc-filled co-polymer, Precision Engineering 43 (2016) 52-62.

[5] O.W. Fähnle, H. van Brug, H.J. Frankena, Fluid jet polishing of optical surfaces, Applied Optics

37 (28) (1998) 6771–6773.

[6] S.M. Booij, O.W. Fähnle, J.J. Braat, Shaping with fluid jet polishing by footprint optimization,

Applied Optics 43 (1) (2004) 67–69.

[7] S. Booij, I. Partosoebroto, J.J. Braat, H. van Brug, Computational model for prediction of

shaping with FJP and experimental validation, Proceedings of Optical Fabrication and Testing

Conference, Tucson, USA (2002) 52-54.

[8] H. Fang, P. Guo, J. Yu, Dwell function algorithm in fluid jet polishing, Applied Optics 45 (18)

(2006) 4291–4296.

[9] P.M. Shanbhag, M.R. Feinberg, G. Sandri, M.N. Horenstein, T.G. Bifano, Ion-beam machining

of millimeter scale optics, Applied Optics 39 (4) (2000) 599–611.

[10] A. Ghobeity, J.K. Spelt, M. Papini, Abrasive jet micro-machining of planar areas and

transitional slopes, J. Micromech. Microeng. 18 (2008a) 055014.

[11] D.S. Park, M.W. Cho, T.I. Seo, Mechanical etching of micro pockets by powder blasting, Int.

J. of Adv. Manuf. Technol. 25 (2005) 1098-1104.




144




[14] S. Murakami, Comparison of various turbulence models applied to a bluff body, Journal of

Wind Engineering and Industrial Aerodynamics, 46-47 (1993) 21-36.

[15] A. Gnanavelu, N. Kapur, A. Neville, J.F. Flores, N. Ghorbani, A numerical investigation of a

geometry independent integrated method to predict erosion rates in slurry erosion, Wear 271

(2011) 712-719.

[16] J. Tu, H.G. Yeoh, C. Liu, Computational Fluid Dynamics: A Practical Approach, Butterworth-

Heinemann, Oxford, UK, 2008.

[17] ANSYS Fluent 12.0, 2009. Theory guide. ANSYS, Inc.

[18] H.K. Hockin, K. Xu, S. Jahanmir, Microfracture and material removal in scratching of

alumina, Journal of materials science 30 (1995) 2235-2247.


abrasive jet micromachining, Wear 264 (2008b) 185-198.



[21] R. Haj Mohammad Jafar, J.K. Spelt, M. Papini, Surface roughness and erosion rate of abrasive

jet micro-machined channels: Experiments and analytical model, Wear 303 (2013) 138-145.

[22] ISO 11562:1996 Geometrical Product Specifications (GPS) -- Surface texture: Profile

method—Metrological characteristics of phase correct filters.

[23] ISO 4288:1996 Geometrical Product Specifications (GPS) -- Surface Texture: Profile

Method—Rules and Procedures for the Assessment of Surface Texture.


comparison between machining in air and submerged in water, Int. J. Machine Tools and

Manufacture 88 (2015) 108-117.



[26] P. Dixit, J. Miao, Aspect-ratio-dependent copper electodeposition technique for very high

aspect-ratio through-hole plating, Journal of the Electrochemical Society 153 (6) (2006) G552-

G559.

145

[27] W. Dow, H. Chen, M. Yen, W. Chen, K. Hsu, P. Chuang, H. Ishizuka, N. Sakagawa, R.

Kimizuka, Through-hole filling by copper electroplating, Journal of the electrochemical society

155 (12) (2008) D750-D757.

[28] H.M. Pollock, D. Maugis, M. Barquins, Characterization of submicrometre surface layers by

indentation. Microindentation Techniques in Material Science and Engineering, Balu, P.J. and

Lawn, B.R., Eds., ASTM (American Society for Testing and Materials) STP 889, Philadelphia,

USA (1986) 47-71.

[29] R. Haj Mohammad Jafar, H. Nouraei, M. Emamifar, M. Papini, J.K. Spelt, Erosion modeling

in abrasive slurry jet micro-machining of brittle materials, Journal of Manufacturing Processes

17 (2014) 127-140.

146

Chapter 6: Prediction of the Erosive Footprint in the

Abrasive Jet Micro-machining of Flat and Curved Glass

6.1. Introduction

As demonstrated by, for example, Solignac et al. (2001) [1] and Liu et al. (2003) [2],

abrasive jet micro-machining (AJM) can be used to machine micro-features in brittle and ductile

materials. In AJM, a small, high-speed air jet is used to accelerate fine abrasive particles which are

made to impact and erode the target material. The topography of the eroded surface depends

strongly on the distribution of the velocity and number density of the impacting particles, termed

the erosive efficacy within the footprint of the jet on the surface, as defined by Ghobeity et al.

(2008) [3]. AJM erosive footprints have thus far been inferred from the profile of shallow eroded

scars on flat targets. For example, Ghobeity et al. (2008) [3] found that at a typical standoff distance

of 20 mm between the nozzle exit plane and target, the footprint diameter was approximately 3

times wider than the jet diameter, and resulted in a roughly V-shaped eroded topography in glass,

indicative of a maximum erosive efficacy near the jet centerline. Using an analytical model,

Ghobeity et al. (2009) [4] demonstrated that the shape and depth of machined micro-channel

profiles made through an erosion-resistant mask were affected by the abrasive particle size

distribution and the width of the mask opening. Dehnadfar et al. (2012) [5] implemented a pulsed

laser shadowgraphy method to measure the abrasive particle size and velocity distribution in both a

free jet and through a mask opening. The shadowgraphy measurements were in good agreement

with the analytical model of Ghobeity et al. (2009) [4].

147

Shipway (1997) [6] measured the depth profiles of wear scars to investigate the effect of

particle divergence in an abrasive jet plume for relatively large nozzles (4.93 mm diameter). He

found that the distribution of particle trajectories followed a gamma distribution, and observed that

the local impact angle due to the plume divergence needed to be taken into account to obtain an

accurate prediction of the erosion on a flat surface. Using a particle capturing technique, Burzynski

and Papini (2011) [7] found that the spatial distribution of abrasive particles within a micro-

abrasive jet produced by nozzles having diameters between 460 μm and 1.5 μm followed a Weibull

distribution. Mansouri et al. (2015) [8] modeled abrasive jet flows using computational fluid

dynamics (CFD) and showed secondary impacts of the particles after rebounding from a flat target,

However, that study focused on sand blasting, which involved much larger length scales (7 mm

nozzle diameter and particle sizes of 150-300 μm) than those used in AJM (460-760 µm nozzle

diameters of particle sizes of 10-25 µm). None of these earlier studies quantified or discussed the

effect of particle second strikes on the footprint size.

More recently, Qi et al. (2016a) [9] used CFD to model the flow field and particle

trajectories in ultrasonic vibration-assisted abrasive slurry-jet micro-machining of glass and found

an increase in the erosion rate due to target vibration. Moreover, Qi et al. (2016b) [10] and Kowsari

et al. (2016a) [11] obtained the erosive footprints from CFD for use in surface profile models.

However, the trajectories and erosive patterns caused by the second strikes in these water slurries

were significantly different from those in air-driven jets due to the large difference in the viscosities

of water and air. The only AJM study involving CFD models of particle secondary strikes was by

Nouhi et al. (2016) [12]. While studying the effect of the variation in the local nozzle standoff

distance and divergence angle of particle trajectories in the jet plume on the erosion of cylinders,

they found that the apparent erosive footprint size changed depending on the surface curvature. The

148

results of their preliminary CFD study revealed that this change in footprint size was due to

differences in particle second strike locations brought about by the target curvature. That

observation provided the motivation for the present work, which considers the effect of second

strike on footprint size in detail.

In summary, the effect of secondary particle impacts in AJM remains largely unexplored,

and the present work aims to characterize and numerically predict the AJM footprint at various

standoffs using CFD models.


An AccuFlo AF10 micro-abrasive blaster (Comco Inc., Burbank, CA, USA), described in

detail in Dehnadfar et al. (2012) [5] was used in all the experiments. The air pressure upstream of

the 760 m inner-diameter nozzle (length-to-diameter ratio of ~ 6.6) was 200 kPa, and aluminum

oxide (Al2O3) powder (Comco Inc., Burbank, CA, USA) having a nominal diameter of 10 m (log-

normal distribution with a standard deviation of 3.31 µm) was used in all the experiments. The

powder mass flow rate was 2.7 g/min.

6.2.1. Jet and footprint measurements

As will be shown in Section 6.3.2, the jet footprint could be viewed as the superposition of a

primary particle plume originating from the nozzle, and a secondary plume consisting of particles

that rebounded from the surface and struck a second time. The jet divergence was measured using a

digital camera attached to a microscope having a field of view of 3×2 mm. The diameters of the net

149

footprints (including both primary and secondary particle impacts) were inferred by using an optical

profilometer (NANOVEA ST400 Micro Photonics Inc., Irvine, CA, USA, depth resolution of 25

nm; lateral resolution of 0.1 μm), to measure the shallow eroded profile resulting from jet

impingement on flat 100×50×3 mm thick glass (Borofloat, Swift Glass Co. Inc., Elmira, NY, USA)

targets. Experiments were performed at standoffs (distance between nozzle exit and target) of 5, 10,

20, and 30 mm at a perpendicular, stationary jet incidence. A shallow eroded profile on a 5 mm

diameter glass rod (Borofloat, Swift Glass Co. Inc., Elmira, NY, USA) at a 10 mm standoff distance

under similar blasting conditions was measured previously in Nouhi et al. (2016) [12].

To determine the footprint diameter resulting from only the primary plume originating at the

nozzle, the jet was made to impact 100×100×0.1 mm thick sheets of multi-purpose paper (Canon

Canada Inc., Mississauga, ON, Canada) such that the impacting particles pierced the paper, but did

not rebound. A small amount of tension was applied to the paper to prevent it bending due to the

particle impacts. The size of the primary plume was measured using a microscope with a field of

view of 6×4 mm.

The divergence of the abrasive particles in the air jet was studied further using double-

pulsed laser shadowgraphy as explained in detail in Dehnadfar et al. (2012) [5] and Hadavi et al.

(2015) [13]. Briefly, a double-pulsed frequency-double Nd: YAG (neodymium:yttrium aluminum

garnet) laser, capable of generating a maximum of 0.3 J/pulse pair at a frequency of 1000 Hz, was

coupled with a high efficiency diffuser (Item No.: 1108417, Lavision, Gmbh, Goettingen,

Germany). The laser with diffuser was positioned directly opposite a high speed CCD camera

(Imager Pro PlusX, Lavision GmbH, Goettingen, Germany) with a high magnification zoom lens

(Navitar zoom 12x, Navitar Inc., Rochester, New York, USA) so that the axis of the diffuser and

150

lens of the CCD camera were aligned. The abrasive jet flowed in a chamber between the diffuser

and the lens of the CCD camera as shown in Fig. 6.1.

Figure 6.1 Double-pulsed shadowgraphy apparatus.

The particle velocity distribution was measured from the image pairs using Davis Software

(Lavision GmbH, Goettingen, Germany). It was found that a pulse duration of 1 ns and time

intervals of 1-3 µs between the two pulses were suitable to capture the particle spatial distribution

and measure the particle velocities for the given conditions. The radial and axial particle velocities

at the nozzle exit were measured and set as input parameters for CFD modeling (Section 6.3). The

coordinates of the particles at 0 mm (nozzle exit) and 20 mm away from the nozzle were recorded

and used to determine the radial distribution of particles within the jet.

151

6.2.2. CFD modeling

CFD models in ANSYS Fluent 15.0 (ANSYS Inc., Cecil Township, PA, USA) were used

to predict the abrasive particle trajectories for an air-particle jet surrounded by stationary air

impinging a non-deforming target at perpendicular incidence. Figure 6.2 shows the two-

dimensional axisymmetric and three-dimensional domains used to model the impingement of the air

jet on flat targets and curved rods having diameters between 3 and 5 mm (about 4-6.5 times the jet

diameter). The realizable κ-ε turbulence model was used to model the fluid, following the AJM

modeling work of Li et al. (2014) [14]. For a given simulation, the same flow field was obtained

using the κ-ω shear-stress turbulent transport (SST) model, but at slightly shorter convergence

times. Li et al. (2014) [14] used two-way coupling and modeled particle-particle collisions for

relatively large 27 μm diameter particles. However, Crowe et al. (2012) [15] explained that particle

volume fractions smaller than 0.001 can be treated with one-way coupling. Therefore, for the

present volume fraction of 5.4×10-8

particle-particle interaction was assumed negligible, and one-

way coupling was used. The models converged with residuals below 10-3

. The maximum Mach

number was about 0.57, thus fluid compressibility was assumed to be negligible.

The 10 m particles described in Section 6.2 were given a shape factor of 0.76 as measured

by Dehnadfar et al. (2012) [5], and were uniformly injected using the same discrete phase injection

ratio settings described in Kowsari et al. (2016a) [11] through the inlet. The particles were assigned

initial axial and radial velocities of 195 m/s and 0-3 m/s, respectively, to match those measured

using shadowgraphy (Section 6.3.1). The particles were tracked using the Lagrangian discrete phase

model. The flat target boundary was treated as a smooth, no-slip wall, and the other boundaries of

the domain were treated as free with a pressure outlet condition. The surface roughness of as-

152

received glass (Rrms = 8 nm) was assumed to have a negligible effect on the rebound particle

trajectories since the particles had much larger diameters than the local surface peaks and valleys.

Although the mesh was refined near the target, the single-phase, air domains surrounding the jet

were meshed using mostly 10 μm quadrilateral elements. The impact velocities (both primary and

secondary) of a particle released 100 μm from the jet centerline differed by 6% for domains meshed

with 20 μm and 10 μm elements. The difference in velocity for domains between 5 and 10 μm,

however, was only 1%, indicating convergence of the solution to a mesh-independent state. To

capture the shear flow in the viscous sub-layer near the targets, the dimensionless wall coordinate,

y+, was maintained below unity using near-wall grid refinement, as shown in Fig. 6.2. Both normal

and tangential restitution coefficients were set to 0.2 as suggested by Slikkerveer and in't Veld

(1999) [16] for similarly sized Al2O3 particles impacting glass.

153

(a)

(b)

Figure 6.2 Domains and boundary conditions of: (a) 2D axisymmetric CFD model of the

impingement of an air-particle jet on a flat target and (b) 3D CFD model of the impingement of an

air-particle jet on a curved target.

154


6.3.1. AJM jet structure

Figure 6.3(a) shows that the particles exited the nozzle with a relatively small divergence

angle. This is also evident in Fig. 6.4(a) which shows the particle distribution obtained by analyzing

shadowgraphic images of the abrasive jet plume in a region between the nozzle exit and 5 mm

downstream of it. These data imply that the jet plume diameter at the nozzle exit was about 1 mm

((Fig. 6.4(a)), and that the divergence angle was about 1.5 from the jet axis. There was good

agreement between the size of the primary plume measured on paper after 30 s exposure to the

abrasive jet (2.8 mm diameter at 20 mm standoff) and the one obtained from shadowgraphy (3.1

mm at 20 mm standoff). As explained in Section 6.2.2, the axial and radial particle velocity

components obtained from these shadowgraphy measurements were used in the CFD model to

produce the same divergence, as shown in Fig. 6.3(b). The relatively small divergences in the

present work, measured both optically and numerically, are consistent with the findings of Shipway

and Hutchings (1993) [17] for nozzles having smooth inner walls so that particle scattering due to

wall collisions was relatively small.

The results of the shadowgraphy experiments in Fig. 6.4(b), showed that approximately

90% of the particles were within 1.55 mm of the centerline, at a standoff distance of 20 mm. The

relatively small dose of particles outside this region did not significantly affect the erosive footprint,

in part because they had relatively small velocities. This was confirmed by conducting experiments

on paper targets at a standoff of 20 mm while varying the exposure time. It was found that the size

of the footprint remained constant for up to 30 s, but increased by about 20% after 5 min. In

155

summary, approximately 90% of the particles were found in an approximately conical primary

plume of about 3 mm diameter at a 20 mm standoff. The stray particles outside of this primary

plume did not significantly affect the erosive footprint, since they produced appreciable erosion

only after relatively long exposure times (~30 s). In comparison, when machining micro-channels a

given point on a target is exposed to the jet for about 5 s at a typical scan speed of 0.5 mm/s.

156

(a)

(b)

Figure 6.3 AJM jet. (a) Microscope images of the AJM jet. (b) Air and particle velocity magnitude

contours obtained using CFD.

157

(a)

(b)

Figure 6.4 Radial distribution of particles within the jet: (a) at nozzle exit (b) at 20 mm standoff,

obtained from shadowgraphy. The error bars indicate the standard deviations obtained from three

measurements of approximately 15000 particles each.

6.3.2. Erosive footprint prediction for flat surfaces

Figure 6.5(a) shows the CFD-predicted flow field and particle trajectories near a target at a

standoff of 5 mm. It is seen that particles within the primary plume travel in straight lines from the

nozzle until they impact the target, and did not follow the curvature in the air streamlines near the

158

surface. As explained by Kowsari et al. (2016b) [18] for a similar AJM jet, the momentum

equilibration number, , was approximately 900 as calculated using

2

18

p p jet

n

d v

d

(6.1)

where p is the particle density,

pd , is the particle diameter, jetv is the jet velocity, , is the

dynamic viscosity of air, and nd is the nozzle diameter. As explained by Humphrey (1990) [19],

values of >>1 indicate that the particles in the air jet are unlikely to follow the fluid streamlines.

Upon initial impact, Fig. 6.5(a) shows that, depending on their radial position, the particles

rebounded to various heights to impact a second time farther away at angles between about 30-70°

to the surface. Figures 6.5(a)-6.5(d) show that the rebound height increased with increasing standoff

distance despite the decrease in the incident velocity at larger standoffs. This was because the air

velocities at the maximum rebound heights also decreased with increasing standoff, thus the

particles experienced less resistance in their rebound trajectories and rebounded to greater heights.

This was further-investigated using an energy balance on a rebounding particle described as

2

02

apexh

rapex d

mVmgh F dx (6.2)

where m is particle mass, rV is rebound velocity, g is the gravitational acceleration, apexh is the

maximum rebound height, dF is the particle drag force, and x is the vertical distance from the

surface. In order to estimate rV , a coefficient of restitution of 0.2 was assumed, although this choice

was not critical since the main objective was a relative comparison of the effect of standoff

distance. Using the CFD predictions of the incident velocity and apexh , Fig. 6.5(e) shows that the

particle drag loss computed using the left-hand side of Eq. (6.2) did indeed decrease with increasing

159

standoff distance, explaining why apexh increased. Figure 6.5(e) also shows that these drag loss

values were consistent with those obtained from the integration of the particle drag force, dF ,

shown in Fig. 6.5(f), as a function of x as calculated using the CFD model. The drag coefficient was

obtained from Haider and Levenspiel (1989) [20] as

2 31

4

Re241 Re

Re Re

b

d

bC b

b

(6.3)

where the coefficients b1-b4 are functions of the ratio of the surface area of an equivalent sphere to

the actual particle surface area, which was set to 0.76 for the 10 μm nominal diameter alumina

abrasives as measured by Dehnadfar et al. (2012) [5].

It is seen in Fig. 6.5 that the higher rebounds at larger standoffs caused the second-strikes to

occur farther away from the centerline which enlarged the erosive footprint. Moreover, the average

second-strike impact angles increased with increasing standoff even as the footprint became larger;

i.e. 72, 77, 82, and 83 for 5 mm, 10 mm, 20 mm, and 30 mm standoffs, respectively, as evident

in Fig. 6.5(a), 6.5(b), 6.5(c), and 6.5(d). This was because the stagnation zone and the associated

lateral flow of air was smaller at larger standoff distances so that second-strike particles

experienced less deflection immediately before impact. For example, Fig. 6.5(a) shows that the

lateral air velocity near the surface in the y-direction caused the second-strike particles to deflect

and impact at shallower angles. This lateral flow did not significantly enlarge the footprint on its

own.

160

(a)

(b)

161

(c)

(d)

162

(e)

(f)

(g)

Figure 6.5 Impingement of AJM jets on flat targets. Air velocity magnitude contours and 10 μm

diameter particle trajectories for standoff distances of: (a) 5 mm, (b) 10 mm, (c) 20 mm, and (d) 30

mm. (e) Drag energy loss as a function of standoff distance for particles released at the nozzle

centerline obtained from either hapex and Vr or direct integration of Fd (Eq. (6.2)). (f) CFD prediction

of particle drag force versus particle rebound displacement for different standoffs. (g) Axial

velocities of particles released from a given mesh element at the inlet boundary approximately 100

μm from the nozzle centerline (13% of the nozzle diameter) at various distances from the target

using the models of Fig. 6.5.

163

The centerline rebound height at the 20 mm standoff was 5.1 times that in the 5 mm case. At

20 mm standoff, the 42 m/s flow velocity at the apex peak of a rebounding centerline particle was

sufficient to re-accelerate the particles to impact the target a second time with a velocity of

approximately 18 m/s. Figure 6.5(g) shows the axial velocity of particles released from a mesh

element about 100 μm from the jet centerline (13% of the nozzle diameter) for the 5-30 mm

standoff CFD simulations. It is seen that both the primary impact velocity, the maximum rebound

velocity (negative), and the second-strike velocity all decreased with increasing standoff distance.

This same trend was evident at all distances from the nozzle centerline. For a given standoff

distance, the maximum variation in the first and second strike velocities between particles released

at different nozzle radii was about 20%.

Wensink and Elwenspoek (2002) [21] explained that the ductile-brittle transition occurred at

17 nJ for borosilicate glass. Although the secondary impact kinetic energies (0.1-0.6 nJ) are lower

than this threshold value for brittle cracking, the values are sufficient to cause ductile erosion in the

glass targets as in the work of Nouraei et al. (2012) [22] in which the impact velocities of similarly-

sized alumina particles ranged between 20-60 m/s corresponding to kinetic energies of 0.4-3.7 nJ.

6.3.2.1. Experimental validation

The CFD simulations revealed that the total erosion at all standoffs consisted of the

contributions of the first strikes near the center of the jet and the second impacts in an outer ring as

shown in Fig. 6.5. The first-strike footprints enlarged with increasing standoff due to the divergence

of the primary plume, as observed in Fig. 6.6(a). The second-strike footprints also grew with

increasing standoff since the drag loss of rebounding particles was greater at smaller standoffs (Fig.

164

6.5(e)). These predictions were compared to the experimentally-measured AJM footprints on glass

at standoffs between 5-30 mm. Figure 6.6(a) shows good agreement between the predicted and

measured results for the net footprints including both the first and second strikes. As described in

Section 6.2.2, the agreement between the predicted and measured results demonstrated that fluid

compressibility had a negligible effect on the particle trajectories. The results suggested that the

effective AJM footprint is characterized by the superposition of two cone-shaped plumes for first

and second strike erosion, as illustrated schematically in Fig. 6.6(b). While the divergence angle of

the primary plume, α, is governed only by the jet divergence, the second-strike angle, ψ, defining

the boundary of the second-strike plume, depends on the rebound conditions. For example, it is

hypothesized that ψ increases with increasing coefficient of restitution, and thus the numerical jet

footprint prediction methodology could serve as a tool to predict restitution coefficients for various

abrasive-target combinations.

165

(a)

(b)

Figure 6.6 (a) Predicted (dashed lines) and measured (solid lines) erosive footprint diameter versus

standoff with and without secondary particle impacts. The lines are to guide the eye only. Error bars

represent ±1 standard deviation for 3 measurements. (b) Schematic representation of intersections

of primary and secondary plumes with successive target planes at standoffs of 10 mm and 30 mm. ψ

defines the second-strike cone angle, and h՛ is the apex height of a corresponding particle after

rebound from the target.

6.3.3. Erosive footprint prediction for curved surfaces

Nouhi et al. (2016) [12] showed that the erosive efficacy inferred from the measurement of

an eroded footprint on a flat surface could not be used to predict the footprint on a curved surface.

It was therefore of interest to determine whether the CFD model could be used to do this. The

166

erosive footprints of an AJM jet from a 460 μm diameter nozzle (200 kPa, 10 m aluminum oxide

particles, as in Nouhi et al. (2016) [12]) on 3 and 5 mm rods were predicted using computational

domains similar to that shown in Fig. 6.1(b). Figure 6.7 presents the CFD-predicted air velocity

magnitude contours and particle trajectories at a standoff of 10 mm. The curvature of the targets

caused the first strike impact angles to vary, which significantly affected the degree of lateral

rebound leading to second strikes. For a given particle within the air jet, the local normal of the first

strike, θ, was larger for a 5 mm-diameter rod than for the 3 mm rod, φ, thereby widening the net

erosive footprint with increasing target curvature. However, despite the footprint enlargement, the

number of rebounds without second strikes also increased since a larger number of particles

deflected beyond the edges of the rod, as seen in Fig. 6.7(b), thus reducing the dose of secondary

impacts. In summary, differences in target curvature can strongly affect the particle impact

trajectories in both the primary and secondary plumes.

167

(a)

(b)

Figure 6.7 Impingement of AJM jets on curved targets at a standoff of 20 mm. Air velocity

magnitude contours and particle trajectories for rod diameters of: (a) 5 mm, and (b) 3 mm.

The net effect of the differences in particle trajectories brought about by target curvature

was determined by predicting the distribution of erosive efficacies on a flat glass surface and a 5

mm glass rod using erosion maps produced by CFD as described in Kowsari et al. (2016a) [11].

Briefly, the measured dependence of erosion on particle impact angle and impact velocity were

168

defined in the erosion model of ANSYS Fluent to obtain the three-dimensional erosion maps shown

in Fig. 6.8. These maps reflected the net erosion produced by both the primary and secondary

plumes. The maps were then converted to the two-dimensional erosion patterns across a machined

channel that would result from a nozzle scan by summing the erosion rates along lines parallel to

the scan direction across the footprint.

Figure 6.8 CFD-obtained normalized erosion maps on a flat target and a 5 mm diameter rod. Each

map was normalized by its maximum specific erosion rate (mass eroded per unit mass of erodent).

169

The resulting erosive efficacy distribution was then fit to a Weibull-type function

(6.4)

where y is the transverse coordinate along the channel width, x is the vertical coordinate measured

from the nozzle tip to the target surface (Fig. 6.9) and β is an effective nozzle focus coefficient that

reflects both first and second strikes.

Figure 6.9 A schematic of Weibull-type function describing the shallow eroded profile. The

coordinates (y, x) of a typical point on the profile are shown.

The β values for both the flat and curved surface cases were then inferred from curve-fitting a

Weibull distribution to the normalized erosion patterns to obtain Fig. 6.10. Although the curves of

Fig. 6.10 were obtained from the superposition of two different plumes, one due to primary and the

other due to secondary impact, their sum created a single smooth erosive efficacy curve. The

predicted values of β, 31 on flat target and 24 on 5 mm diameter glass rod, were in good agreement

(~10% difference) with the measured ones, 28 and 22 on flat and rod targets, respectively, given by

Nouhi et al. (2016) [12].

170

Figure 6.10 Normalized erosive efficacies and the best fits (Weibull distribution) for flat and

curved (5 mm diameter) glass targets. The abscissa was normalized by standoff distance and the

ordinate was normalized by the depth of the channel centerline.

6.3.4. Implications for AJM

Ghobeity et al. (2008) [3] and Getu et al. (2008) [23] showed that a shallow channel profile

which was machined by a scanning nozzle on flat target at 90˚ incidence could be used to

characterize the erosive efficacy delivered to both ductile and brittle surfaces. They also showed

that when implemented in an appropriate surface evolution model, the erosive efficacy inferred

from the shallow profile could be used to successfully predict the shape of micro-channels

machined using AJM. On flat surfaces, the erosive efficacy determined from a shallow profile

includes both the primary and secondary particle strikes on the flat surface, and therefore the

Weibull distribution obtained in this manner can be used to predict the evolution of machined

surface profiles. This footprint on flat targets can be viewed as being at the intersection of a single

effective particle cone and the target (Fig. 6.6(b)). However, the results of the present work confirm

the hypothesis of Nouhi et al. (2016) [12] that erosion due to second strikes is more pronounced

171

when the initial target surface curvature is higher, consequently making the shallow channels

machined on flat targets inappropriate for directly characterizing the erosive efficacy on rods or

other curved targets, where the effective cone and footprint will be a function of curvature.

In general, the effective value of β for a given target curvature can be determined either by

CFD modeling of the erosion maps as in Section 6.3.3, or an effective value can be obtained by

adjusting its value in the surface evolution model to fit the measured shallow first-pass profile of a

channel machined on the curved target (Nouhi et al. (2016) [12]). Measurements and modeling have

shown that β is sensitive to the target curvature and must be adjusted if, for example, a rotating rod

is being machined using AJM as a lathe. For instance, according to Nouhi et al. (2016) [12], β had

to be decreased about 9%, from 22 to 20, when the rod diameter was decreased (curvature

increased) from 5 mm to 3 mm. For relatively deep channels in curved targets, the increasing slope

of the local surface geometry with increasing channel depth would likely alter the erosive pattern.

Therefore, additional CFD modeling would be required beyond the first pass, analogous to what

was done in Kowsari et al. (2016a) [11] for abrasive slurry-jet micro-machining.

6.4. Conclusions

A computational fluid dynamics (CFD)-aided procedure was presented for the prediction of

the erosive footprints resulting from abrasive jet machining (AJM) of both flat and curved targets.

The divergence of an AJM jet was measured using laser-pulsed shadowgraphy and by blasting

holes through paper. Using these results together with CFD models, it was found that the net

erosive efficacy footprint on a surface was the result of the superposition of two approximately

conical erodent plumes; a primary one leading to first strikes and a secondary one reflecting second

172

particle impacts. CFD modeling showed that approximately 90% of the particles travelled within

the primary plume, with the remaining 10% at the periphery being so sparse that they did not affect

the footprint. On flat targets, the particle incident velocities, the air velocities, and the rebound

particle drag losses were found to decrease with increasing standoff distance. These effects caused

an increase in the particle rebound heights after their first strike and a broadening of their

trajectories such that the net footprint of first- and second-strike particles was enlarged, but the

average impact angles decreased with increasing standoff. The predicted kinetic energies of

particles striking a second time were large enough to erode glass targets.

The erosive footprint was also found to depend on target curvature, because the local slope

changed the angle at which the particles rebounded, thus changing the distribution of second strikes

to the surface. The presented methodology provided fundamental understanding of air-driven

particle erosive footprints that is needed in modeling of curved surfaces. In such cases, the footprint

size would depend on the local surface slope that changes with increasing feature depth, thereby

requiring further CFD modeling beyond those for shallow features, analogous to the approach taken

in Kowsari et al. (2016a) [11] for abrasive slurry-jet micro-machining.

173

6.5. References

[1] D. Solignac, A. Sayah, S. Constantin, R. Freitag, M.A.M. Giijs, Powder blasting for realization

of microchips for bio-analytic applications Sens. Actuators, A92 (2001) 388–393.

[2] C. Liu, J. Chen, J. Engel, J. Zou, X. Wang, Z. Fan, K. Ryu, K. Shaikh, D. Bullen, Polymer

micromachining and applications in sensors, microfluidics, and nanotechnology The 226th

National Meeting of the American Chemical Society (ACS), New York, NY, 11–17 September

(2003).



[4] A. Ghobeity, D. Ciampini, and M. Papini, An analytical model of the effect of particle size

distribution on the surface profile evolution in abrasive jet micromachining, Journal of

Materials Processing Technology, Vol. 209, Issue 20 (2009) 6067-6077.

[5] D. Dehnadfar, J. Friedman, and M. Papini, Laser shadowgraphy measurements of abrasive

particle spatial, size and velocity distributions through micro-masks used in abrasive jet micro-

machining, Journal of Materials Processing Technology, 212 (1) (2012) 137–149.

[6] P.H. Shipway, The effect of plume divergence on the spatial distribution and magnitude of wear

in gas-blast erosion, Wear 205 (1997)169–77.

[7] T. Burzynski and M. Papini, Measurement of the particle spatial and velocity distributions in

micro-abrasive jets, Measurement Science and Technology, 22 (2011) 025104.

[8] A. Mansouri, M. Mahdavi, S.A. Shirazi, B.S. McLaury, Investigating the effect of sand

concentration on erosion rate in slurry flows, Proceedings of the 2015 NACE Corrosion

Conference and Expo. Paper No. 6130.

[9] H. Qi, D. Wen, C. Lu, G. Li, Numerical and experimental study on ultrasonic vibration-assisted

micro-channelling of glasses using an abrasive slurry jet, Int. J. Mech. Sci., 110 (2016a) 94-107.

[10] H. Qi, D. Wen, Q. Yuan, L. Zhang, Z. Chen, Numerical investigation on particle impact

erosion in ultrasonic-assisted abrasive slurry jet micro-machining of glasses, Powder

Technology (2016b, in press).


of abrasive slurry jet micro-machined channels in sintered ceramics, Ceramics Int'l 42 (2016a)

7030-7042.

[12] A. Nouhi, K. Kowsari, J.K. Spelt, M. Papini, Abrasive jet machining of channels on highly-

curved glass and PMMA surfaces, Wear 356-357 (2016) 30-39.

174

[13] V. Hadavi, B. Michaelsen, M. Papini, Measurements and modeling of instantaneous particle

orientation within abrasive air jets and implications for particle embedding, Wear 336–337

(2015) 9–20.

[14] H. Li, A. Lee, J. Fan, G.H. Yeoh, J. Wang, On DEM–CFD study of the dynamic characteristics

of high speed micro-abrasive air jet, Powder Technology 267 (2014) 1611-1179.

[15] C.T. Crowe, J.D. Schwarzkopf, M. Sommerfeld, Y. Tsuji, Multiphase flows with droplets and

particles – second edition, Taylor & Francis Group, LLC – CRC Press (2012), pp. 17–34.

[16] P.J. Slikkerveer, F.H. in't Veld, Model for patterned erosion, Wear 233 (1999) 377-386.

[17] P.H. Shipway, I.M. Hutchings, Influence of nozzle roughness on conditions in a gas-blast

erosion rig, Wear 162 (1993) 148-158.



Process. Technol. 23 (2016b), 190-210.



[20] A. Haider, O. Levenspiel Drag Coefficient and Terminal Velocity of Spherical and

Nonspherical Particles, Powder Technology 58 (1989) 63-70.


erosion. Wear 253, (2002) 1035-1043.


micro-machining: a comparison with abrasive air jet micro-machining, J. Mater. Process.

Technol. 213 (2012) 1711–1724.

[23] H. Getu , A. Ghobeity, J.K Spelt, M. Papini, Abrasive jet micromachining of acrylic and

polycarbonate polymers at oblique angles of attack, Wear, Volume 265, Issues 5-6, (2008), 888-

901.

http://www.sciencedirect.com/science/journal/00431648/336/supp/C

175

Chapter 7: Selective Removal of Metallic Layers from

Sintered Ceramic and Metallic Substrates Using

Abrasive Slurry-jet Micro-machining

7.1. Introduction

Existing studies related to the abrasive erosion of coatings have been designed to assess the

wear resistance of the coatings using slurry or air jets. For example, Iwai et al. (2006) [1] eroded

approximately 2 μm thick titanium nitride coatings on high speed steel (HSS) substrates using a

slurry jet with 1.2 μm alumina from a 3×3 mm square nozzle at about 100 m/s. They found that the

erosion behavior changed with the coating ductility. In a similar study, Hawthorne et al. (1999) [2]

subjected high velocity oxy-fuel (HVOF) sprayed ceramic and metallic coatings to 35-200 μm

diameter alumina abrasives carried by either a 15 m/s slurry jet or an 84 m/s air jet. They found the

specific erosion rate using the air jet was three orders of magnitude greater than using the slurry jet

because of the much higher particle impact velocities in the air jet. Other related studies include

those of Wood (1999) [3], Santa et al. (2009) [4], and Sugiyama et al. (2005) [5], but none

considered the controlled removal of metallic layers from a substrate.

Tamannaee et al. (2016) [6] and Kowsari et al. (2016a) [7] used repeated adjacent passes of

the abrasive slurry-jet micro-machining (ASJM) jet to mill planar areas (pockets) in talc-filled

thermoplastic olefin (TPO) and sintered alumina, respectively. Billingham et al. (2013) [8] used a

high-pressure (414 MPa) abrasive water jet machine (AWJM) with a 1 mm diameter nozzle and

176

180-300 m garnet particles to machine pockets into a titanium-based alloy (Ti6Al4V) using over-

lapping channels. All of these authors predicted the shape of the machined pockets using a

superposition model in which the cross-sectional profile of a single-pass channel, obtained

experimentally, was summed, taking into account the overlap of adjacent passes of the scanning jet.

These studies limited the material removal per pass so that the shape of the single-pass channel

continued to be representative of the erosion pattern for each pass of the nozzle. Prediction of the

single-pass erosion pattern from first principles (e.g. computational fluid dynamics (CFD)) for use

in the superposition model has not yet been attempted.

In the above-mentioned work with sintered ceramics containing copper-filled through-holes,

Kowsari et al. (2016a) [7] distinguished the slurry erosion in the direct footprint of a 150 μm

diameter, 89 m/s ASJM aqueous jet, blasted at perpendicular incidence, from the predominately

ductile erosion that occurred in the secondary slurry flow over the target surface at shallow particle

incidence. The latter preferentially eroded the ductile copper-filled holes within the ceramic

substrate producing unwanted dimples in the finished surface.

Kowsari et al. (2016b) [9] found that the use of viscous fluids such as soybean oil could

enlarge the boundary layer thickness of the jet over the target causing a reduced flow velocity near

the opening of holes made with ASJM. This decreased the pressure drop at the hole edge thereby

reducing the generation of cavitation bubbles. The authors also found that particles deflected to a

much higher degree within the stagnation zone of the relatively viscous soybean oil slurry jet while

machining holes, but the effect of fluid viscosities much greater than that of water on the erosion

rate in ASJM channel machining remains unexplored.

177

Nouraei et al. (2016) [10] machined channels in brittle and ductile materials using ASJM,

and found that the slope of the leading edge during channel machining decreased the effective

erosion rate in glass while increasing it in PMMA. They hypothesized that a steeper leading edge

decreased the local impact angles thereby reducing the erosion rate in brittle glass (maximum

erosion at perpendicular incidence) and increasing it in ductile PMMA (maximum erosion at

approximately 45). However, Nouraei et al. (2016) [10] did not examine this hypothesis with CFD,

and did not study the role of fluid viscosity in controlling this leading-edge effect on erosion rate.

The present objective was to explore the use of ASJM to selectively remove uniform copper

or nickel-phosphorous layers and copper protrusions from sintered ceramic and metallic substrates

without eroding the underlying material. Experiments were complemented by extensive

computational slurry-flow modeling to understand the effects of the ASJM process parameters on

the particle trajectories, the boundary layer thickness and the resulting erosion.


7.2.1. Target materials

The experiments involved 3 configurations of 4 materials as described in Table 7.1: (i)

copper-plated aluminum nitride wafers containing copper-filled through-holes (Fig. 7.1a); (ii)

nickel-phosphorous-plated aluminum substrates (Fig. 7.1(b)); and (iii) aluminum nitride containing

copper pillars protruding from over-filled through-holes (Fig. 7.1(c)). The objective was to use

ASJM to remove the copper or the nickel-phosphorus material within the regions indicated by the

dashed lines, while leaving the substrates intact. Upon selective removal of these regions,

178

specimens such as those in Figs. 7.1(a) and 7.1(c) find application as heat sinks for electronic

components, while the configuration Fig. 7.1(b) is typically used to fabricate enclosures for such

components.

(a)

(b)

(c)

Figure 7.1 Schematic section views through the 3 test specimens. (a) copper-plated aluminum

nitride containing copper-filled through-holes. (b) nickel-phosphorous-plated aluminum. (c)

protrusion formed due to over-filling of through-hole in aluminum nitride wafer. The dashed

regions are those to be removed using ASJM.

179

Table 7.1 Properties of the target materials, obtained from the manufacturer of aluminum nitride,

from ASM (1990) [11] for aluminum, and from Zhaojiang (1999) [12] for copper and nickel-

phosphorous.


(mm)

Grain size

(m)

Density

(g/cm³)

Vickers

hardness

(kgf/mm²)

Aluminum nitride

(AlN)

K170, Toshiba

Corp., Minato,

Tokyo, Japan

50×50×0.375 < 1 3.26 1100

Electrodeposited copper

(Cu) -

50×50×0.014

50×50×0.400 - 8.96 83

6061-T6 aluminum (Al) - 85×25×1 - 2.71 112

Electrodeposited nickel-

phosphorous (Ni-P) - 85×25×0.014 - 8.00 255

7.2.2. ASJM apparatus and experiments

A 180 μm sharp sapphire orifice was used to produce 150 m diameter water and soybean

oil jets (Table 7.2) having velocities of 89-110 m/s, computed using Bernoulli's equation for

pressures of 4-6 MPa. Both types of slurry jets contained 1 wt% of well-suspended alumina

particles of 10 μm nominal diameter (Comco Inc., Burbank, CA, USA; Vickers hardness 16 GPa).

The standoff distance between the orifice plate and the target was set to 20 mm in all experiments,

and the channels were machined by scanning the target using a computer-controlled two-

dimensional linear stage (Zaber Technologies Inc., Vancouver, BC, Canada).

There were four types of experiments using the three specimens of Fig. 7.1. In all cases, the

process conditions were selected to produce specific erosion rates in sintered ceramic and metallic

targets similar to those found in the ASJM work of Kowsari et al. (2016a) [7].

180

Experiment #1. The specific erosion rate (mass of material removed per mass of erodent) of ductile

materials depends on both the particle impact angle and the particle velocity as explained by Oka et

al. (1997) [13]. These functions were required as inputs in the CFD models to obtain the erosion

patterns in these materials. They were measured on the 400 μm thick copper and nickel-phosphorus

layers of Figs. 7.1(a) and 7.1(b) using the volumes of relatively shallow (10-50 μm) blind holes

machined at various particle velocities and angles of attack. In the absence of chemical effects,

these erosion rate functions are independent of the fluid carrying the abrasive particles. The water

slurry-jet was used in the present experiments, because its much thinner boundary layer simplified

the determination of the local average particle impact velocities and angles using CFD models as

explained below.

The global jet inclination angles and free-stream velocities ranged from 15-90 and 63-110 m/s,

respectively. Because the spreading slurry in the stagnation zone deflected incoming particles, the

centerline average particle impact angles, calculated using CFD models in Kowsari et al. (2016c)

[14], were significantly lower than the global jet inclination. The conversion from global jet angle

to average particle impact angle was necessary to obtain the impact angle function, described in

Section 7.3.1, which was used as an input in the CFD models. The experiments to determine the

erosion rate-impact velocity dependence were conducted at perpendicular incidence, and the

particle concentration was adjusted according to Table 7.3 to ensure a constant particle dose (kg/m2)

delivered to the target regardless of the jet velocity.

It was unnecessary to characterize the aluminum nitride and aluminum substrates of Figs.

7.1(a) and 7.1(b) in this manner since the objective was to remove the copper or nickel-

phosphorous layers while leaving these substrates intact.

181

Experiment #2. Stationary water and soybean oil jets at 89 m/s were directed at 15 incidence

toward the protruding copper-filled through-holes of the specimen shown in Fig. 7.1(c). The jet was

approximately 1 mm away from the copper pillar to be flattened as shown in Fig. 7.2(a) so that the

slurry flow was essentially parallel to the flat aluminum nitride surface; i.e. the copper pillar being

machined was eroded by the secondary slurry flow, outside the primary footprint of the jet. The

objective was to level the protrusion leaving a flat surface of aluminum nitride and copper. The jet

was repositioned for each pillar being flattened as shown in Fig. 7.2(a).

Experiment #3. Channels were machined in the 400 μm thick copper layer of Fig. 7.1(a) by

scanning water (89 m/s) and soybean oil (110 m/s) slurry jets at speeds of 0.005-4 mm/s in three jet

configurations: (i) 90, (ii) 45 forward (Fig. 7.2(b)), and (iii) 45 backward (Fig. 7.2(c)). These

experiments were conducted to quantify the effect of the local machining front geometry on the

erosion produced by the slurry jet.

Experiment #4. Partially overlapping parallel channels were machined using water (89 m/s) and

soybean oil (110 m/s) jets that were scanned repeatedly over the targets of Figs. 7.1(a) and 7.1(b).

Figure 7.2(d) illustrates this for the specimen of Fig. 7.1(a). The objective was to remove the copper

and nickel-phosphorus layers, both 14 μm thick, indicated by the dashed boxes in Figs. 7.1(a) and

7.1(b) leaving flat substrates with little or no removal of the underlying copper, aluminum or

aluminum nitride. It will be seen in Section 7.3.3 that the path of the scanned jet (Fig. 7.2(d)) was

essentially arbitrary and did not have to align with the rows of copper-filled through-holes of the

specimen in Fig. 7.1(a), as long as soybean oil was used to prevent dimple formation in the through-

holes. Cross-sectional and areal surface profiles were measured using an optical profilometer

(ST400, Nanovea Inc., CA, USA; lateral resolution 426 nm, depth resolution 16 nm), where depth

measurements were made every 10 μm in each of the scanning directions.

182

(a)

(b)

(c)

183

(d)

Figure 7.2 (a) Schematic of the position of the stationary primary jet footprint and the secondary

flow with respect to the copper pillar in experiment #2. Section through the centerline of the jet.

(b)-(c): Domain and boundary conditions of three-dimensional CFD models of the ASJM flow

within channels measured in copper (experiment #3) machined using a 110 m/s soybean oil slurry-

jet scanned at 0.005 mm/s in the (b) 45 forward and (c) 45 backward configurations. (d)

Machining path of the slurry jet in the overlapping channel-machining method of experiment #4

illustrated for the specimen of Fig. 7.1(a).

Table 7.2 Properties of the test fluids at 20 C.

Fluid Dynamic viscosity (cP) Density (kg/m3)

Water 1 998

Soybean oil 45 917

184

Table 7.3 Process parameters used in experiment #1 to determine dependence of erosion on

velocity and impact angle for copper and nickel-phosphorus.

Type of experiment

Velocity exponent Impact angle function

Pressure (MPa) 2 4 6 4

Water slurry-jet flow rate (mL/s) 1.34 1.67 2.00 1.67

Free jet velocity (m/s) 63 89 110 90

Particle concentration (wt%) 1.15 1.00 0.85 1

Standoff distance (mm) 20 20

Jet incidence (°) 90 15, 30, 45, 60, 75, 90

7.2.3. CFD modeling

The flow fields and particle trajectories of the water and soybean-oil slurry jets were

obtained using CFD models that were constructed using the actual surface profiles obtained using

the optical profilometer and imported into ANSYS Fluent 15.0 (ANSYS Inc., Cecil Township, PA,

USA) to produce two-dimensional planar and three-dimensional domains with the boundary

conditions shown in Fig. 7.2(a). The fluid entered the domain with the known jet velocities (Section

7.2.2) over the 150 μm diameter inlet. The target was modeled as a smooth, no-slip wall, and the

other bounding planes were treated as free boundaries with a pressure outlet condition. Particles

were uniformly injected at the free jet velocities (Table 7.3) across the jet and tracked using the

Lagrangian discrete-phase model. The injections contained particle diameters and volume fractions

that reflected the actual particle size distribution (Comco Inc., Burbank, CA, USA), given in

Kowsari et al. (2016c) [14]. The fluid properties in Table 7.2 were used as model inputs. The

volume of fluid (VOF) model was used to simulate the multiphase, steady flow of the primary

phase, water or soybean oil, the secondary phase, air. Turbulence was modeled using the κ-ω shear-

185

stress turbulent transport (SST) model. The domains were meshed with approximately 1 μm

quadrilateral elements and the simulations converged to maximum residuals of 10-3

.


7.3.1. Target erosion characterization - experiment #1

The erosion model provided by ANSYS Fluent 15.0 (2015) [15], defines the rate of surface

erosion, erosionR , in units of kg/m2

s as

1

ParticlesNp

erosion

p cell

m ER

A

(7.1)

where P is the abrasive particle mass flow rate and A

cell is the area of a given computational cell on

the target. For brittle and ductile materials, Oka et al. (1997) [13] explained that the function E

is the specific erosion rate (mass of material removed per mass of erodent) at particle impact angle,

, given by

90E f E (7.2)

where f defines the dependence of erosion on the particle impact angle, and 90E is the specific

erosion rate at perpendicular incidence and is related to the particle impact velocity, v, by

90

cE Av (7.3)

where A is a system constant that depends on the material properties and target substrate, and c is

the velocity exponent which expresses the dependence of erosion on the particle impact velocity.

Figure 7.3(a) shows how the specific erosion rate at normal incidence, measured from the

volumes of shallow blind holes (approximately 10 μm deep), varied with water slurry-jet velocities

186

of 63-110 m/s. These jet velocities corresponded to particle impact velocities of 27-47 m/s along the

jet centerline (v in Eq. (7.3)) as measured for the present conditions by Kowsari et al. (2016c) [14]

using CFD models of impinging ASJM jets on flat targets. Table 7.4 gives the best-fit values of Eq.

(7.3) to obtain the constants A and c, used as inputs in the CFD models.

Figure 7.3(b) shows how the measured specific erosion rate depended on the water slurry-jet

impact angle for copper and nickel-phosphorous targets. Also shown is the corresponding average

centerline particle impact angle (α in Eq. (7.2)) determined along a two-dimensional centerline

section through the stagnation zone (primary footprint in Fig. 7.2(a)). It is seen that these targets

behaved in a typical ductile manner where the maximum erosion occurred approximately at jet

incidences of 30-45.

(a) (b)

Figure 7.3 ASJM specific erosion rates for copper and nickel-phosphorous, respectively, vs. (a) jet

velocity and centerline average particle impact velocity of a water slurry-jet at perpendicular

incidence, and (b) jet impact angle and actual centerline average particle impact angle of an 89 m/s

water slurry-jet. Experiment #1. Error bars represent ±1 standard deviation for 3 measurements. The

lines serve only to guide the eye.

187

Table 7.4 Best-fit constants for Eq. (7.3) giving the dependence of specific erosion rate on the

centerline average particle impact velocity of a water slurry-jet (63-110 m/s) at perpendicular

incidence.

Material

A

((mg/g)×(m/s)-c

) c

Copper 6×10-6

2.52

Nickel-phosphorous 1×10-6

2.65

7.3.2. Selective removal of copper pillars - experiment #2

An individual 5 μm high protruding copper pillar of the specimen of Fig. 7.1(c) was

removed after 20 s of exposure to the stationary 15 water slurry-jet (89 m/s) in the configuration

shown in Fig. 7.2(a). However, the profilometer scan of Fig. 7.4(a) showed that unwanted 8 μm

deep dimples were created by this secondary aqueous slurry flow. The depth of the dimple was

found to increase to 16 μm after a 30 s exposure (Fig. 7.4(c)). This demonstrated that the near-wall

particles had enough energy to erode the copper-filled through-holes relatively quickly even though

the aqueous slurry flow at a jet angle of 15 was essentially parallel to the aluminum nitride.

Figure 7.4(b) shows that using soybean oil instead of water under otherwise identical

conditions (i.e. 20 s exposure, 89 m/s jet, 15 jet incidence in the configuration of Fig. 7.2(a)), the

copper pillar was removed to the level of the surrounding aluminum nitride without creating a

dimple, leaving the desired flat surface as illustrated in Fig. 7.4(c). Subsequent experiments showed

that, even after 1 min of exposure to the soybean oil jet, the copper surface of the through-hole

remained flat (Fig. 7.4(c)). Therefore, copper pillars could be removed leaving a flat surface of

aluminum nitride and copper-filled through-holes (surface A in Fig. 7.1(a)) using a stationary

188

soybean oil jet placed approximately 1 mm upstream of the through-holes so that the copper pillars

were eroded by the secondary slurry flow.

(a) (b)

(c)

Figure 7.4 Surface topography of un-eroded and eroded copper-filled through-holes subjected to

stationary 15 slurry-jets at 89 m/s in the configuration shown in Fig. 7.2(a) (experiment #2) using:

(a) a water slurry, (b) a soybean oil slurry. (c) Elevation of copper with respect to the aluminum

nitride substrate vs. time of exposure to stationary 15 slurry-jets of water and soybean oil using the

same process conditions as in (a) and (b).

189

The difference in the behavior of the water and soybean oil slurry jets was attributed to the

differences in the boundary layer thickness that results from the differences in viscosity. Figure 7.5

shows that the CFD-predicted boundary layer thickness for both water and soybean oil was within

about 11% of that computed using the analytical relation given by Schlichting and Gersten (2004)

[16] as

0.5

15x v x U

(7.4)

where is the boundary layer thickness, defined as the distance from the target where the flow

velocity was equal to the jet centerline velocity (89 m/s), v is the kinematic viscosity of the fluid,

x is the distance from a leading edge where the boundary layer begins to develop (Fig. 7.2(a)), and

U is the bulk flow velocity.

Figure 7.5 Comparison of boundary layer thickness vs. x (defined in Fig. 7.2(a)) for a 15 jet

impact angle with water (89 m/s) and soybean oil (89 m/s) jets as measured from CFD and

computed using Eq. (7.4).

190

Figure 7.6 displays the flow field and particle trajectories predicted by CFD in the

machining configuration of Fig. 7.2(a) over an 8 μm deep dimple representing that created in the

copper of the through-hole by the water slurry. It shows that the boundary layer of the secondary

soybean oil flow over the dimple (Fig. 7.2(a)) was approximately 6.4 times thicker than that of

water (18 μm). Figure 7.6 also displays the trajectories of 10 and 15 μm particles, initially released

along the jet centerline (Fig. 7.2(a)) then carried downstream to a dimple. It is seen that particles of

both sizes entered the dimple in the case of the water slurry flow (Fig. 7.6(a)), impacting the dimple

surface at a predicted velocity of 33 m/s. In contrast, due to the thicker boundary layer compared to

water, the closest 10 μm diameter particle in the soybean oil (Fig. 7.6(b)) was approximately 13 μm

above the surface and so could not erode the copper.

191

(a)

(b)

Figure 7.6 Particle trajectories in vicinity of dimples placed in the secondary flow (x ≈ 1.3 mm, Fig.

7.2(a)) of a jet having an inclination of 15: (a) water slurry and (b) soybean oil jets.

192

The ability of the soybean oil jet to flatten protruding copper pillars (experiment #2) was

explained using CFD models of the flow over 20-65 μm high pillars representing over-filled copper

through-holes. Although the relatively thick soybean oil boundary layer served as a protective layer

to minimize dimpling of the copper through-holes, Fig. 7.7 shows that 10 and 15 μm diameter

particles could impact protruding pillars at approximately 23 m/s, consistent with the experiment in

Fig. 7.4(b), in which 5 μm high pillars were flattened using a soybean oil jet. This is explained by

the much thinner initial boundary layer that formed on the leading front of the rounded pillar in Fig.

7.7(a). In summary, dimpling of copper-filled through-holes caused by the secondary flow of water

jets in the machining configuration of Fig. 7.2(a) could be minimized using soybean oil, while not

hindering the ability of the same flow from removing protruding copper features.

193

(a)

(b)

(c)

Figure 7.7 Particle trajectories for flow fields over pillars placed about 1.3 mm downstream of a

15 soybean oil jet (Fig. 7.2(a), experiment #2). (a) 10 μm and (b) 15 μm particle trajectories over a

65 μm high protrusion; (c) 15 μm particle trajectories over a 20 μm high protrusion.

194

7.3.3. Selective removal of metallic layers using over-lapping channels

7.3.3.1. Effect of machining front slope on erosion rate - experiment #3

Figure 7.8(a) shows how the depth changes with the particle dose delivered to the target per

unit length of channels in the copper layer of Fig. 7.1(a) in the 90°, 45° forward (Fig. 7.2(b)), and

45° backward (Fig. 7.2(c)) machining orientations using an 89 m/s water slurry-jet. Experiments in

both the forward and backward orientations were required because of the potential differences in

local impact angles on the machining fronts. The scan speeds were 0.005-0.5 mm/s using the slurry

flow rates in Table 7.3 so that the particle dose delivered to the target in a single machining pass

produced either shallow (<20 μm deep) or relatively deep (>70 μm) channels. In this way the

relation between the erosion rate and the leading edge slope, β (Fig. 7.8(b)), governed by the depth

of each machining pass, was determined as a function of the local impact angle. The slope angle β,

was determined from linear fits to profile points measured every 10 μm along the curved front A-B

in Fig. 7.8(b) (Section 7.2.2), beginning at the point where the best-fit slope of 5 consecutive points

deviated 10% from the horizontal.

For the channels machined in copper using the 90 water slurry-jet, Fig. 7.8(a) shows that

the leading edge slope, β, increased from approximately 2 to 17 as the scan speed decreased from

0.5 to 0.005 mm/s, thereby increasing the particle dose impacting the surface per unit length of

channel. The dashed lines show the predicted depths of channels that would result from machining

using multiple high-speed passes of jets with angles of 45 and 90 such that each pass produced a

shallow channel with β < 2. In other words, these predicted shallow-channel trend lines assume

that the erosion rates remained constant at their initial first-pass values. However, the data of Fig.

7.8(a) shows that the etch rate did not remain constant as the scan speed slowed and the dose

195

increased; i.e., at the highest dose of 3.3 g/mm, the depth of the relatively deep channel machined at

90 (slow scan speed producing β = 17) was about 17% larger than that predicted for multiple

passes with β < 2 (dashed line for 90). This increase in machining efficiency at slow jet scan

speeds can be explained by comparing the CFD-obtained three-dimensional erosion maps for 90

water slurry-jets impinging a shallow channel (β = 2, Fig. 7.8(c)), and a deep channel (β = 17,

Fig. 7.8(d)). The erosion rate along the centerline of the shallow channel was predicted to be about

23% greater than that of the deep one, in relatively in good agreement with the measured 17%. The

difference in erosion rate was due to the differences in local impact angles along the machining

front that occurred for shallow and deep channels. The CFD model predicted that the average

particle impact angles measured along the plane of symmetry through the centerline of the primary

footprint decreased from 74 to 60 (Table 7.5), as the channel leading-edge slope β increased from

2 to 17 (Fig. 7.9(a)) due to the lower scan speed. As illustrated in Fig. 7.3(b), this decrease in the

local impact angle increased the erosion rate in copper by approximately 20%.

In the case of shallow channels producing very small β, as shown in Fig. 7.8(a), there was

no difference between forward and backward machining. Figure 7.8(a) also shows that, in contrast

to the 90 water slurry-jet, a β of 25 in the 45 forward orientation decreased the channel depth at a

dose of 3.3 g/mm by 34%. This can be explained by noting the increase of the centerline average

particle impact angles from 34 to 47, evident in the CFD model of Fig. 7.9(b) and reported in

Table 7.5. This increase in the leading edge impact angle decreased the CFD-predicted erosion rate

at the centerline of the relatively deep channel (β = 25) by approximately 25% compared to the

centerline erosion rate of a 45 water slurry-jet on a flat copper target (Table 7.6). As above, this

was a consequence of the erosion dependence on the impact angle as given in Fig. 7.3(b).

196

The 45 backward orientation generated a β of 14 (Fig. 7.8(a)) which increased the channel

depth for a given dose of 3.3 g/mm by 78% relative to the channel machined using multiple rapid

scans indicated by the dashed line for 45 scans. This was due to the decrease in the centerline

average particle impact angle against the leading edge of the machined channel from 34 to 21

(Fig. 7.9(c) and Table 7.5), thereby increasing the CFD-obtained centerline erosion rate by 90%

compared to that on a flat target (Table 7.6). In summary, machining channels with the water slurry

at a slow scan speed in the 45 backward orientation generated smaller local particle impact angles

at the relatively steep machined front which led to the greatest erosion efficiency. These results,

summarized in Table 7.6, are consistent with those of Nouraei et al. (2016) [10] who also found that

the depth of ASJM channels in brittle and ductile materials increased linearly with dose as long as

the leading edge slope, β, was less than about 2 in each pass of the jet.

(a) (b)

197

(c)

(d)

Figure 7.8 (a) Channel depth vs. dose (g/mm of channel length) of single-pass channels machined

in copper using a 89 m/s water jet in the 90, 45 forward, and 45 backward orientations

(experiment #3). The lines serve only to guide the eye. (b) Side view of the local machined front

geometry of the channel in Fig. 7.2(a). CFD three-dimensional erosion map of 90 (89 m/s) water

slurry-jet on (c) a flat copper target, and (d) a 117 μm deep channel in copper.

198

(a)

(b)

(c)


channels machined in copper using a water slurry-jet scanned at 0.005 mm/s in the (a) 90, (b) 45

forward, and (c) 45 backward orientations. αavg is the average impact angle along the centerline of

the primary footprint. Particle rebounds not shown. Channel leading edge angle defined in Fig.

7.8(b).

199

Table 7.5 CFD predictions of average particle impact angles at machining front along plane of

symmetry through the centerline (primary footprint) for various β in the 90, 45 forwards, and 45

backward machining orientations using water and soybean oil slurry jets.

Fluid

Water Soybean oil

Particle impact angle

Machining orientation

Range

(°)

Average

(°)

Range

(°)

Average

(°)

90° Shallow (β ≈ 2°) 57-90 74 8-90 30

Deep (β ≈16°) 40-80 60 15-79 53

45° forward Shallow (β ≈ 2°) 27-45 34 2-70 26

Deep (β ≈ 23°) 36-69 47 2-90 55

45° backward Shallow (β ≈ 2°) 27-45 34 2-70 26

Deep (β ≈ 12°) 13-32 21 1-33 14

Table 7.6 Percentage change in channel depth at doses of 3.4 g/mm for water and 4.7 g/mm for

soybean oil produced by the leading edge effect in slow, single-pass machined channels in copper

using water and soybean oil slurry jets in the 90, 45 forwards, and 45 backward machining

orientations. The symbols (+) and (-) indicate an increase or decrease in the depth, respectively,

relative to channels machined using rapid, multiple shallow passes at 0.3 g/mm for water and 0.5

g/mm for soybean oil which gave a very small slope, < 2.

Machining orientation

Fluid

Water Soybean oil

Measured CFD Measured CFD

90° 17%+ 23%+ 28%- 18%-

45° forward 34%- 25%- 17%- 11%-

45° backward 78%+ 90%+ 57%- 47%-

200

Figure 7.10 shows that the use of a soybean-oil slurry jet produced completely different

behavior than that seen with the water slurry. With soybean oil, the greatest channel depth for a

given dose occurred in the 90° and 45° forward orientations. But in all cases, the machining with

single slow scans which produced steep leading edge fronts was counter-productive, and decreased

the channel depth relative to that obtained using multiple higher-speed scans where β remained less

than about 2.

Table 7.5 shows that in the 90 machining orientation, an increase in β from 2 to 15

increased the centerline average impact angle from 30 to 53 (Fig. 7.11(a)), corresponding to a

18% lower CFD-predicted centerline erosion rate for the deeper channel than that of a flat target

representing the shallow channel. Very similar behavior was found in the 45 forward orientation

where an increase in β from 2 to 20 (Table 7.5) increased the centerline average particle impact

angle from 26 to 55 (Fig. 7.11(b)), and reduced the CFD-predicted centerline erosion rate by 11%

relative to that in shallow channels. In the 45 backward orientation, Table 7.5 shows that an

increase in β from 2 to 10 decreased the CFD-predicted centerline average impact angle from 26

to 14 (Fig. 7.11(c)), which in turn decreased the centerline erosion rate by 47% compared to that in

shallow channels.

Overall, for the water slurry-jet, the 45° backward machining orientation at a slow scan

speed produced the deepest channel for a given particle dose. For the soybean oil slurry jet, it was

most efficient to machine multiple shallow channels (keeping β < 2) using a 90 jet. These

conditions caused the local particle impact angles at the leading edge of the machined front to

approach the angle at which erosion was maximized in copper (about 30).

201

Figure 7.10 Channel depth vs. dose (g/mm of channel length) of single-pass channels machined in

copper using a soybean oil jet in the 90, 45 forward, and 45 backward orientations. The lines

serve only to guide the eye.

202

(a)

(b)

(c)


channels machined in copper using a 110 m/s soybean oil jet scanned at 0.005 mm/s in the (a) 90,

(b) 45 forward, and (c) 45 backward orientations. αavg is the average impact angle along the

centerline of the primary footprint. Particle rebounds not shown. Channel leading edge angle

defined in Fig. 7.8(b).

203

7.3.3.2. Removal of nickel-phosphorous layer from aluminum - experiment #4

Figure 7.12(a) shows the measured topography of a pocket milled through the nickel-

phosphorous layer (Fig. 7.1(b)) using a single operation of over-lapping channels with a 45 water

slurry-jet (89 m/s) being scanned at a relatively high speed of 1.4 mm/s with an offset between

adjacent passes of 50 μm in the configuration shown in Fig. 7.12(a) (experiment #4 illustrated in

Fig. 7.2(d)). Under these conditions each machining pass would produce a 3.5 μm deep channel if

they were widely spaced, and so the leading edge slope was small (β < 2). But since the 50 μm

overlap was just 77% of the primary footprint diameter, a single machining operation removed the

entire 14 μm thick nickel-phosphorous layer to expose bare aluminum. This was confirmed by

energy-dispersive X-ray spectroscopy (EDS) which showed that there was less than 1 wt% of nickel

or phosphorous in the machined area. The operation was performed using the water slurry-jet

instead of the soybean oil jet since this aluminum substrate did not contain copper-filled through-

holes prone to dimpling.

For convenience, the pocket of Fig. 7.12(a) was machined using shallow passes in

alternating 45 forward and backward orientations in order to eliminate the need to rotate the jet or

target. Since the leading edge slope was small, the erosion rate was the same in both orientations at

these high scan speeds (Fig. 7.8(a)). There was no advantage to using lower scan speeds, since they

would produce steeper leading edge fronts which would increase the erosion rate in the backward

orientation, but decrease it in the forward orientation (Fig. 7.8(a)).

Kowsari et al. (2016a) [7] found that the waviness of the surface milled using the same

overlapping ASJM channel machining as in Fig. 7.2(d) depended on the offset of each pass of the

204

jet. For example, for channels having a β of approximately 10, an offset 167% larger than the jet

diameter resulted in a wavy profile having an arithmetic average roughness, Ra, of about 6 μm,

whereas the waviness remained constant at approximately 0.4 μm for offsets smaller than about

33% of the jet diameter corresponding to an offset of 50 μm for the present jet. The machining

experiments of Fig. 7.12 were therefore performed with an offset of 50 μm to minimize the

waviness of the exposed aluminum surface.

205

(a)

(b)

Figure 7.12 (a) Surface topography of a nickel-phosphorous layer removed to expose aluminum

substrate. Result of a single machining operations using a 45 water slurry-jet (89 m/s) scanned at

1.4 mm/s in the configuration of Fig. 7.2(d). Each operation used overlapping scans offset by 50

μm. The jet was not rotated between passes so the orientation was alternately forward (Fig. 7.2(b))

and backward (Fig. 7.2(c)) between passes. (b) Cross-sectional measured profiles along line A-A of

the pocket in (a). The plot shows only a portion of the profiles. A scanning electron microscope

image of a section view of an uneroded specimen of Fig. 7.1(b) is shown on the right.

206

7.3.3.3. Removal of copper layer from aluminum nitride containing copper-

filled through-holes - experiment #4

Figure 7.13(a) shows the measured surface topography of a pocket machined through the 14

µm thick copper layer (Fig. 7.1(a)) using a 90 soybean oil jet (110 m/s). The thicker boundary

layer of soybean oil eliminated the erosive effect of the secondary slurry flow on the copper-filled

through-holes outside the primary jet footprint (Section 7.3.2). As shown in Fig. 7.4, a water slurry

would have produced dimpling in the copper through-holes as the aluminum nitride became

exposed. The result was a flat surface of aluminum nitride and copper-filled through-holes (surface

A in Fig. 7.1(a)), over which the maximum elevation difference between the copper and aluminum

nitride was less than 2 µm (Fig. 7.13(a)). Moreover, the jet was oriented at 90 to take advantage of

the 18% larger depth compared to a soybean oil jet at 45 at a dose of 0.5 g/mm (Fig. 7.10). Each

operation removed slightly less than 2 μm of copper, so 8 operations were used in total to remove

the copper layer. The jet was repeatedly scanned back and forth in the configuration of Fig. 7.2(d)

at a relatively high scan speed of 4 mm/s and offset by 50 μm after each machining pass. The very

small leading edge slope, β, eliminated the reduction in channel depth for a given dose due to the

leading edge effect explained in Section 7.3.3.1, and thus ensured optimal machining efficiency for

soybean oil. An EDS analysis of surface C in Fig. 7.13(a) revealed that there was approximately 16

wt% copper on the exposed aluminum nitride surface after the 8 operations. It is hypothesized that

this residual copper remained intact in the surface troughs of the relatively rough, unpolished

sintered aluminum nitride, measured to have an Ra of 182 nm by Kowsari et al. (2016d) [17].

Figure 7.13(b) shows that two additional operations served to dimple the copper through-

holes, but did not erode the aluminum nitride since it was about 13 times harder than copper. Since

207

it was shown in Section 7.3.2 that the secondary flow of the soybean oil jet could not create dimples

in the copper, it can be concluded that this additional erosion was due to the primary footprint as the

jet passed over the filled through-holes as shown in the configuration of Fig. 7.2(d).

(a)

(b)

Figure 7.13 (a) Surface topography and (b) a portion of the cross-sectional profiles along line B-B

of a copper layer removed to expose a flat surface of aluminum nitride containing copper-filled

through-holes. Results of 8 and 10 machining operations using a 110 m/s perpendicular soybean oil

jet scanned at 4 mm/s. Each operation used overlapping scans offset by 50 μm.

208

7.3.3.4. Prediction of the layer thickness removed for machined over-lapping channels – experiment #4

Obtaining the erosion pattern from CFD

A methodology was developed to predict the ASJM process conditions required to remove a

nickel-phosphorus layer using a water slurry (as in Section 7.3.3.2) and a copper layer using a

soybean oil slurry (as in Section 7.3.3.3) based on the CFD-predicted cross-sectional shape of a

single-pass channel, similar to the approach described in Section 7.1 used by Kowsari et al. (2016c)

[14] for water slurry-jets on sintered ceramic targets.

The impact angle functions and velocity exponents for nickel-phosphorous and copper

(Section 7.3.1) were used as inputs in the CFD models of 45 water and 90 soybean oil slurry jets

on flat targets (very small ) to obtain three-dimensional erosion maps such as that shown in Figs.

7.14(a) for nickel-phosphorous and 7.14(b) for copper. Such erosion patterns represent the erosive

footprint within a relatively shallow channel in which no secondary impacts occur on the sidewalls.

Therefore, they can be used to predict the cross-sectional shape of shallow channels in these

materials.

Following the procedure of Kowsari et al. (2016c) [14], the three-dimensional erosion maps

were reduced to two-dimensional representative erosion patterns for copper and nickel-

phosphorous, shown in Fig. 7.14(c), by summing the erosion rates along lines parallel to C-C (Figs.

7.14(a) and 7.14(b)). The erosion patterns shown in Fig. 7.14(c) are essentially half of the

symmetric cross-sectional profiles of shallow, single-pass channels, normalized by their centerline

depth. The next step in the procedure was to calibrate the normalized erosion patterns in Fig.

7.14(b) by a constant obtained from an experimental calibration pass; e.g. a 1.4 mm/s pass with a

209

45 water slurry-jet (89 m/s) was about 3.5 μm deep in nickel-phosphorous. These predicted single-

pass channel profiles were then used in the superposition model of Tamannaee et al. (2016) [6].

210

(a)

(b)

(c)

Figure 7.14 CFD three-dimensional erosion map of (a) a 90 (110 m/s) soybean oil jet on copper,

and (b) a 45 (89 m/s) water slurry-jet on nickel-phosphorous. (c) Two-dimensional representative

erosion patterns of the models in (a) and (b). The specific erosion rates were normalized by the

specific erosion rate along line C-C in (a).

211

Prediction of layer thickness removed

Figure 7.15 compares the measured cross-sectional profiles of removed pockets in layers of

nickel-phosphorous (Fig. 7.15(a)) and copper (Fig. 7.15(b)) to the profiles predicted using the

above model. The removed layer thickness was predicted with a maximum error of about 10% and

showed good agreement at both the bottom and the sloped edges of the machined regions.

(a)

(b)

Figure 7.15 Measured (solid lines) and predicted (dashed lines) cross-sectional channel profiles of

pockets removed within (a) the copper layer of Fig. 7.1(a), machined using 8 operations of over-

lapping 4 mm/s channels with a perpendicular soybean oil jet (110 m/s), and (b) the nickel-

phosphorous layer of Fig. 7.1(b), machined using 1 operation of over-lapping 1.4 mm/s channels

with a 45 water slurry-jet (89 m/s). The offset was 50 μm in both (a) and (b).

212

An empirical relation between the machining parameters and the depth removed by the

machining operation can be expressed in terms of the depth of a single-pass channel of depth

1

1channel p sd C m v (7.5)

where the particle dose is equal to the mass flow rate of the particles (g/s), pm , divided by the scan

speed (mm/s), vs, and C1 (mm2/g) is an erosion constant obtained from the measured first-pass vs.

depth relation (see below). The depth of the layer, dlayer, removed using an overlapping channel

operation is then

2layer channeld C d (7.6)

where C2 is a measured dimensionless constant that gives dlayer resulting from overlapping channels,

each having a depth of dchannel for a given offset. For instance, pm was computed to be 0.0167 g/s

for the 89 m/s water slurry-jet using the process conditions in Table 7.3. Moreover, a single-pass

channel in nickel-phosphorous at vs = 1.4 mm/s had a depth of about 3.5 μm, where dlayer was about

14 μm after one operation at the optimal offset of 50 μm. Therefore, the constants were

1

1 0.3244p s channelC m v d mm2/g and ,

1

2 3.6110layer channelC d d and combining Eqs.

(7.5) and (7.6) yields the scan speed (mm/s)

11.1714s p layerv m d (nickel-phosphorous) (7.7)

Following the same procedure, the scan speed required to remove a copper layer of

thickness of dlayer using a 110 m/s soybean oil jet at perpendicular incidence is given by

10.2162s p layerv m d (copper) (7.8)

213

Equations (7.7) and (7.8) are valid only of the operation is comprised of relatively shallow

channels (β less than about 2) so that the change in the erosion rate due to a relatively steep

leading-edge slope can be neglected (Section 7.3.3.1). The modeling of deeper overlapping

channels would require a methodology that accounts for the changes in erosion due to relatively

large β. However, as explained in Section 7.3.3.1, for a given particle dose using the soybean oil

slurry, the use of multiple shallow passes was preferred to a smaller number of deep passes, because

the material removal efficiency was improved. Similarly, when machining in the 45 forward

orientation using the water slurry, the leading edge effect was found to be detrimental to the erosion

rate so that again multiple shallow passes were preferred to a smaller number of deep passes. The

additional efficiency in material removal that resulted when using a small number of deeper passes

for the case of a water slurry in the 45 backward orientation did not justify the use of a modeling

methodology that would require additional CFD simulations. In this case, an approach similar to

that of Kowsari et al. (2016c) [14] for the ASJM of multi-pass channels in sintered ceramics can be

used.

7.4. Conclusions

Metallic layers were selectively removed without eroding the underlying substrates using

over-lapping ASJM channels. The use of soybean oil eliminated the unwanted dimple formation of

metallic-filled through-holes caused by the secondary slurry flow, consistent with the CFD models

which showed that the much larger soybean oil boundary layer compared to that of water reduced

the near-wall particle velocities and thus minimized the erosion they cause.

214

The selective removal of a copper layer from an aluminum-nitride substrate containing

copper-filled through holes was most efficiently achieved using a soybean oil slurry jet with

machining multiple shallow channels using a 90 jet. These conditions also caused the local particle

impact angles at the leading edge of the machined front to approach the angle at which erosion was

maximized in copper (about 30). A nickel-phosphorous layer was removed most efficiently from

an aluminum substrate using the same machining configuration, but with a 45 water slurry-jet

where the machining direction was alternately forward and backward between adjacent passes. In

both cases, the jet was displaced by an offset of 50 μm to minimize the waviness of the resulting

surface.

A model to predict the depth of the metallic layer removed in a machining operation was

developed using an existing model of the erosion produced by adjacent nozzle passes in ductile

materials. CFD models were used to obtain the single-pass channel erosion pattern in each material.

This required the measurement of the dependence of the specific erosion rate on both the particle

impact velocity and the particle impact angle for the copper and nickel-phosphorous layers. This

model predicted the depth of the metal layers removed in a machining operation with a maximum

error of about 10%.

215

7.5. References

[1] Y. Iwai, T. Miyajima, T. Honda, T. Matsubara, K. Kanda, S. Hogmark, Evaluation of erosive

wear resistance of TiN coatings by a slurry jet impact test, Wear 261 (2006) 112-118.

[2] H.M. Hawthorne, B. Arsenault, J.P. Immarigeon, J.G. Legoux, V.R. Parameswaran,

Comparison of slurry and dry erosion behaviour of some HVOF thermal sprayed coatings, Wear

225-229 (1999) 825-834.

[3] R.J.K. Wood, The sand erosion performance of coatings, Mat and Design 20 (1999) 179-191.

[4] J.F. Santa, L.A. Espitia, J.A. Blanco, S.A. Romo, A. Toro, Slurry and cavitation erosion

resistance of thermal spray coatings, Wear 267 (2009) 160-167.

[5] K. Sugiyama, S. Nakahama, S. Hattori, K. Nakano, Slurry wear and cavitation erosion of

thermal-sprayed cermets, Wear 258 (2005) 768-775.


and transitional slopes in a talc-filled co-polymer, Precision Eng 43 (2016) 52-62.


milling of sintered ceramic wafers with and without copper-filled through-holes, J Mat Proc

Tech 230 (2016a) 198-210.



paths, Int J Machine Tools Manuf 68 (2013) 30-39.


on the shapes of abrasive slurry-jet micro-machined holes and channels, J Mach Tools Manuf

110 (2016b) 80-91.


in abrasive slurry jet micro-machining, Precision Eng 44 (2016) 109-123.

[11] ASM Int. Handbook Committee, ASM Handbook, Volume 02 - Properties and Selection:

Nonferrous Alloys and Special-Purpose Materials, ASM International (1990).

[12] L. Zhaojiang, Hardness characteristics in electrodeposited copper/nickel multilayer systems.

University of Windsor M.Sc. Thesis (1999).



216


of abrasive slurry jet micro-machined channels in sintered ceramics, Ceramics Int 42 (2016c)

7030-7042.

[15] ANSYS Fluent 15.0 Theory guide, ANSYS, Inc; (2015)

[16] H. Schlichting, K. Gersten. Boundary-Layer Theory. Springer; New York (2004)

[17] K. Kowsari, J. Schwartzentruber, J.K. Spelt, M. Papini. Erosive smoothing of abrasive slurry-

jet micro-machined channels in glass, PMMA, and sintered ceramics: experiments and

roughness model. Precision Eng (2016d, submitted).

217

Chapter 8: Conclusions and Future Work

8.1. Conclusions

The main conclusions of the research presented in Chapter 2-7 are summarized below.

(i) ASJM channel profile modeling in sintered ceramics (Ch. 2) - objectives 1 and 2

ASJM channels in three sintered ceramics (aluminum nitride, alumina, and

zirconium tin titanate) had "V"-shaped profiles, and their depths increased less-than-

linearly with increasing dose of abrasives delivered to the target.

The CFD models revealed that the channel formation using a perpendicular slurry

jet occurred in two stages defined by a change in profile shape. In the first stage, the

sidewalls of shallow channels (aspect ratios of less than about 0.36) were eroded by

the lateral spreading of the slurry flow, leading to an increase in the channel opening

width. In the second stage, the slurry flowed from the footprint region mainly along

the channel length and did not widen the channel opening.

Channel formation using a jet incidence of 45 (forward or backward machining

configuration) did not produce any widening of the channel opening compared to

90 machining since lateral spreading was reduced and the inclined jet directed the

slurry along the channel.

Two methods were developed to predict the channel cross-sectional profiles as a

function of the number of machining passes or, equivalently, the particle dose

delivered to the target. The first method required a new CFD model of the flow and

218

erosion pattern after each machining pass in order to capture the changing flow field.

The second method predicted the profile of subsequent machining passes using only

the erosion pattern of the first-pass channel together with an approximate

relationship between the particle centerline impact velocity and an estimate of the

size of the stagnation zone obtained using two CFD simulations for all sintered

ceramic targets.

The predictions of both methods were validated by comparing with channel cross-

sectional profiles up to a depth/width aspect ratio of about 0.5. The predicted depths

in the three sintered ceramics were within 8% of those of the measured channels at

any distance from the centerline.

(ii) Shape-control of ASJM holes and channels in brittle and ductile materials (Ch. 3) -

objectives 1 and 3

CFD analyses of the flow fields and measurements of the surface textures within the

machined holes indicated that the edge rounding observed in micro-machined

features in ASJM was due to abrasive-enhanced cavitation caused by vapor

formation as the high-speed slurry flowed over the edges at the tops of the holes and

channels.

The collapse of cavitation bubbles accelerated particles in their vicinity to impact the

target at near-perpendicular incidence and velocities sufficient to cause erosion, thus

damaging and rounding the edges. This was demonstrated by producing comparable

damage in glass using an ultrasonic apparatus immersed in an aqueous slurry.

219

Experimental results showed that reducing the slurry vapor pressure decreased the

cavitation activity, producing holes and channels in glass and zirconium tin titanate

with much less rounding at the top. This effect became more pronounced as the

liquid viscosity was increased, since the flow velocities were reduced and hence the

decreases in pressure were smaller.

ASJM using slurries of low-vapor pressure liquids such as mineral oil not only

significantly sharpened the hole entrances, but also produced changes in the local

particle impact angle that led to flatter hole bottoms and steeper sidewalls.

Through-holes with sharp entrance and exit holes were machined in glass, and in

sintered zirconium tin titanate with the aid of a sacrificial layer at the hole exit.

Edge rounding caused by cavitation-enhanced slurry erosion was much less

pronounced in ductile materials than in brittle targets due to the difference in their

erosion mechanisms.

(iii) Smoothing of ASJM channels in brittle and ductile materials (Ch. 4) - objective 4

For typical ASJM conditions, ductile plastic deformation was the dominant erosion

mode, even in glass since the particle kinetic energies were below the theoretical

transition energy required for fracture.

Slower particle impacts at shallower angles using smaller particles could produce

approximately 35% smoother channels compared to the roughest channels machined

220

in glass, PMMA, and zirconium tin titanate at the largest normal particle kinetic

energy, but at 64% lower etch rates on average.

At conditions optimized to obtain the smoothest surfaces, machining of channels of

practical depths would require relatively long machining times, therefore the post-

blasting of channels machined under typical parameters was explored as a means of

polishing. Under post-blasting conditions (89 m/s water slurry-jet velocity, 15 jet

inclination, 3 μm silicon carbide particles), channels in glass, PMMA, zirconium tin

titanate, and aluminum nitride were smoothed to root-mean-square (Rrms)

roughnesses of 23, 23, 19, and 170 nm. These surfaces were smoother than the as-

received surfaces for PMMA (0.4% smoother), zirconium tin titanate (94%), and

aluminum nitride (15%), but rougher for glass (65%).

An existing ductile-regime surface roughness simulation model could predict the

steady-state roughness of the ASJM surfaces with an average error of 12%.

(iv) Hybrid AJM-ASJM micro-milling of sintered ceramic wafers with and without filled

through-holes (Ch. 5) - objectives 1 and 5

Using an over-lapping channel methodology in ASJM, pockets with a roughness, Ra,

of about 0.4 m were machined in alumina. The pocket shape could be predicted a

superposition method in which each machining operation removed less than about

50 m from the floor of the pocket.

221

Flat pockets having sidewall angles of 57° from the horizontal were milled in

aluminum nitride wafers containing 180 m-diameter copper through-holes using a

hybrid AJM-ASJM methodology. AJM was used first to selectively erode the brittle

ceramic without eroding the ductile copper through-holes. ASJM was used in a

second step to selectively erode the copper pillars remaining from the first step while

leaving the surrounding ceramic essentially intact.

(v) Prediction of AJM erosive footprint (Ch. 6) - objective 1

The divergence of an AJM jet was measured using laser-pulsed shadowgraphy and

by blasting holes through paper. Using these results together with CFD models, it

was found that the net erosive efficacy footprint on a surface was the result of the

superposition of two approximately conical erodent plumes; a primary one leading to

first strikes and a secondary one reflecting second particle impacts.

On flat targets, the particle incident velocities, the air velocities, and the rebound

particle drag losses were found to decrease with increasing standoff distance.

The predicted kinetic energies of particles striking a second time were large enough

to erode glass targets.

The erosive footprint was also found to depend on target curvature, because the local

slope changed the angle at which the particles rebounded, thus changing the

distribution of second strikes to the surface.

222

(vi) Selective removal of metallic layers from sintered ceramic and metallic substrates

(Ch. 7) - objectives 1 and 5

Metallic layers were selectively removed without eroding the underlying substrates

using over-lapping ASJM channels. The use of soybean oil eliminated the unwanted

dimple formation of metallic-filled through-holes caused by the secondary slurry

flow, consistent with the CFD models which showed that the much larger soybean

oil boundary layer compared to that of water reduced the near-wall particle velocities

and thus minimized the erosion they cause.

The selective removal of a copper layer from an aluminum-nitride substrate

containing copper-filled through holes was most efficiently achieved using a

soybean oil slurry jet with machining multiple shallow channels using a 90 jet. A

nickel-phosphorous layer was removed most efficiently from an aluminum substrate

using the same machining configuration, but with a 45 water slurry-jet where the

machining direction was alternately forward and backward between adjacent passes.

A model to predict the depth of the metallic layer removed in a machining operation

was developed using an existing model of the erosion produced by adjacent nozzle

passes in ductile materials. This model predicted the depth of the metal layers

removed in a machining operation with a maximum error of about 10%.

223

8.2. Directions for Future Work

The following topics may prove to be fruitful areas for future research.

(i) Abrasive slurry-jet micro-machining of highly-curved surfaces such as rods of brittle and

ductile materials using a lathe apparatus. Micro-machining of curved specimens finds

industrial applications in optical and biomedical equipment such as metal or polymer stents.

(ii) Control of boundary layer thickness using magnetorheological (MR) fluids. This would

allow for the continuous control of slurry viscosity and boundary layer thickness using

adjustable magnetic fields without the need to change the test fluid.

(iii) Investigation of abrasive slurry-jet micro-machining combined with chemical polishing

using corrosive fluids. The cutting action of abrasive particles carried by a corrosive fluid

could expose un-corroded material to produce a synergetic erosive effect, reducing polishing

times.

224

Thesis References

Chapter 1

[1] X. Cheng, A. Wang, K. Nakamoto, K. Yamazaki, A study on the micro tooling for micro/nano

milling, International Journal of Advanced Manufacturing Technology 53 (2011) 523-533.

[2] C. Iliescu, B. Chen, F.E.H. Tay, G. Xu, J. Miao, Characterization of deep wet etching of glass,

Proc. of SPIE Vol. 6037 (2003) 60370A-2.

[3] C. T. Yang, S.S. Ho, B.H. Yan, Micro-hole machining of borosilicate glass through electro

mechanical discharge machining, Key Engineering Materials 196 (2001) 149- 166.

[4] H. Ogura, Y. Yoshida, Hole drilling of glass substrates with a CO2 laser, Japanese Journal of

Applied Physics 42 (2003) 2881–2886.

[5] J. Jandeleit,, A. Horn, R. Weichenhain, E.W. Kreutz, R. Poprawe, Fundamental investigations of


(1998) 885-891.

[6] E. Blloy, S. Thurre, E. Walchiers, A. Sayah, M.A.M. Gijs, The introduction of powder blasting

for sensor and micro system applications, Sensors and Actuators 84 (2000) 330-337.

[7] S. Schlautmann, H. Wensink, R. Schasfoort, M. Elwenspoek, A. Vandenberg, Powder-blasting

technology as an alternative tool for micro fabrication of capillary electrophoresis chips with

integrated conductivity sensor, Journal of Micro-mechanics and Micro-engineering 11 (2001)

386-389.

[8] L. Zhang, T. Kuriyagawa, Y. Yasutomi, J. Zhao, Investigation into micro abrasive intermittent

jet machining, International Journal of Machine Tools and Manufacture 45 (2005) 873–879.

[9] H. T. Liu, Water jet technology for machining fine features pertaining to micro-machining,

Journal of Manufacturing Processes 12 (2010) 8-18.

[10] A.W. Momber, R. Kovacevic, Principles of Abrasive Water Jet Machining, Springer Verlag

Limited, London (1992).

[11] D.S. Miller, Micro-machining with abrasive water jets, J. of Materials Processing Technology

149 (2004) 37-42.

[12] K.L. Pang, T. Nguyen, J.M. Fan, J. Wang, Machining of micro-channels on brittle glass using an

abrasive slurry jet, Key Eng. Mat. 443 (2010) 639-644.


micromachining of channels and holes in glass, Wear 309 (2014a) 65-73.

225



conference on micromanufacturing (ICOMM) (2014a).




of holes in brittle and ductile materials, J. Materials Processing Tech. 214 (2014b) 1909–1920.



Proceedings of the 10th international conference on micromanufacturing (ICOMM), Milan, Italy

(2014b).



[19] A. Ghobeity, H.J. Crabtree, M. Papini, J.K. Spelt, Characterisation and comparison of

microfluidic chips formed using abrasive jet micromachining and wet etching, J.

Micromechanics and Microengineering 22 (2012) 025014.

[20] D. Solignac, A. Sayah, S. Constantin, R. Freitag, M.A.M. Gijs, Powder blasting for the

realization of microchips for bio-analytic applications, Sensors and Actuators A 92 (2001) 388–

393.

[21] Y.P. Zhao, L.S. Wang, T.X. Yu, Mechanics of adhesion in MEMS—a review, J. Adhesion Sci.

Tech. 17 (2003) 519–546.

[22] R. Haj Mohammad Jafar, M. Papini, J.K. Spelt, Simulation of erosive smoothing in the abrasive

jet micro-machining of glass, Mater. Process. Technol. 213 (2013) 2254-2261.

[23] H. Wensink, S. Schlautmann, M.H. Goedbloed, M.C. Elwenspoek, Fine tuning the roughness of

powder blasted surfaces, J. Micromech. Microeng. 12 (2002) 616-620.








on the shapes of abrasive slurry-jet micro-machined holes and channels, Int. J. Machine Tools &

Manuf. 110 (2016) 80-91.

226

Chapter 2



Process. Technol. 230 (2016) 198-210.



(1998) 885-891.

[3] P.J. Slikkerveer, F.H. in’t Veld, Model for patterned erosion, Wear 233-235 (1999) 377-386.

[4] J.H.M. ten Thije Boonkkamp, J.K.M. Jansen, An analytical solution for mechanical etching of

glass by powder blasting, J. Engineering Mathematics 43 (2002) 385-399.









Model to predict high aspect-ratio channel profiles for submerged and unsubmerged





[10] H. Nouraei, K. Kowsari, B. Samareh, J.K. Spelt, M. Papini, Calibrated CFD erosion modeling

of abrasive slurry jet micro-machining of channels in ductile materials, J. Manuf. Proc. 23

(2016) 90-101.

[11] ANSYS Fluent 15.0 Theory guide, ANSYS, Inc., 2015.

227




[13] H.McI. Clark, The influence of the squeeze film in slurry erosion, Wear 256 (2004) 918-926.




in abrasive slurry jet micro-machining, Precision Engineering 44 (2016) 109-123.



[17] J.L. Routbort, R.O. Scattergood, Solid particle erosion of ceramics and ceramic composites,

Key Eng. Mat. 71 (1992) 23-50.

[18] H. Nouraei, A. Wadoslawsky, M. Papini, J.K. Spelt, Characteristics of abrasive slurry jet

micro-machining: a comparison with abrasive air jet micro-machining, Journal of Materials


[19] J.H.M. ten Thije Boonkkamp, P.J. Slikkerveer, Mathematical modelling of erosion by powder

blasting, Surv. Math. Ind. 10 (2002) 89-105.


Comparison between machining in air and submerged in water, Int. J. Machine Tools &

Manufacture 88 (2015) 108-117.

228

Chapter 3




of holes in brittle and ductile materials, J. Materials Processing Tech. 214 (2014a) 1909–1920.



conference on micromanufacturing (ICOMM), Singapore, Singapore (2014b)



Proceedings of the 10th international conference on micromanufacturing (ICOMM), Milan,

Italy (2014a).



[6] C.G. Duan, V.Y. Karelin, Abrasive erosion and corrosion of hydraulic machinery, Imperial

College Press London, 2003.

[7] J. Hengyun, Z. Fengzhen, L. Shiyun, H. Chenzhao, The role of sand particles on the rapid

destruction of the cavitation zone of hydraulic turbines, Wear 112 (1986) 199-205.

[8] J. Sato, K. Usami, T. Okamura, S. Tanaba, Basic studies of coupled damage caused by silt

abrasion and cavitation erosion, Japan Society of Mechanical Engineers 34 (3) (1991) 292-297.

[9] M. Toshima, T. Okamura, J. Satoh, K. Usami, S. Tanabe, Basic study of coupled damage caused

by silt abrasion and cavitation erosion, Japan Society of Mechanical Engineers 57 (1991) 20–

25.

[10] P.P. Gohil, R.P. Saini, Coalesced effect of cavitation and silt erosion in hydro turbines — a

review, Renew Sustain Energy Rev 33 (2014) 280–289.

[11] B.M. Borkent, M. Arora, C.D. Ohl, Reproducible cavitation activity in water–particle

suspensions, J. Acoust. Soc. Am. 121 (2007) 1406–1418.

229

[12] J. Wang, H. Chen, L. Qin, Y. Li, D. Chen, Key roles of micro-particles in water on occurrence

of cavitation-erosion of hydro-machinery, Chin. Sci. Bull. 53 (2008) 1603–1610.

[13] M. Arora, C.D. Ohl, K.A. Mørch, Cavitation inception on microparticles: a self-propelled

particle accelerator, Phys. Rev. Lett. 92 (2004) 174501.

[14] R.M. Wagterveld, L. Boels, M.J. Mayer, G.J. Witkamp, Visualization of acoustic cavitation

effects on suspended calcite crystals, Ultrasonics Sonochem 18 (2011) 216–241.

[15] S. Li, Cavitation enhancement of silt erosion — an envisaged micro model, Wear 260 (2006)

1145–1195.

[16] J.R. Laguna-Camacho, R. Lewis, M. Vite-Torres, J.V. Mendez-Mendez, A study of cavitation

erosion on engineering materials, Wear 301 (2013) 467-476.


micromachining of channels and holes in glass, Wear 309 (2014b) 65-73.

[18] P.M. Ndiaye, F.W. Tavares, I. Dalmolin, C. Dariva, D. Oliviera, J.V. Oliviera, Vapor pressure

data of soybean oil, castor oil, and their fatty acid ethyl ester derivatives, J. Chem. Eng. Data 50

(2005) 330-333.

[19] G.H. Schnerr and J. Sauer, Physical and Numerical Modeling of Unsteady Cavitation

Dynamics, Fourth Intl. Conf. on Multiphase Flow, New Orleans, USA (2001).


micro-machining: a comparison with abrasive air jet micro-machining, J. Materials Processing

Tech. 213 (2012) 1711-1724.


erosion, Wear 253 (2002) 1035-1043.

[22] J. Hara, J. Deshimaru, M. Kasai, Study on cavitation of water soluble hydraulic fluid, Proc. of

the Int. Tribology Conf., Yokohama (2) (1995) 909-914.

[23] S. Popinet, S. Zaleski, Bubble collapse near a solid boundary: a numerical study of the

influence of viscosity, J. Fluid. Mech. 464 (2002) 137-163.



230


of abrasive slurry jet micro-machined channels in sintered ceramics, Ceramics International 42

(2016) 7030-7042.

Chapter 4

[1] F.C. Tsai, B.H. Yan, C.Y. Kuan, F.Y. Huang, A Taguchi and experimental investigation into the

optimal processing conditions for the abrasive jet polishing of SKD61 mold steel, Int. J. Machine

Tools Manuf. 48 (2008) 932–945.

[2] W. Peng, C. Guan, S. Li, Material removal mode affected by the particle size in fluid jet


[3] L. Zhang, S. Raghavan, M. Weling, Minimization of chemical-mechanical planarization (CMP)

defects and post-CMP cleaning, J. Vacuum Science Tech. B 17 (1999) 2248-2255.

[4] H. Fang, P. Guo, J. Yu, Surface roughness and material removal in fluid jet polishing, Appl.

Optics 45 (17) (2006) 4012-4019.

[5] R.J. Wang, C.Y. Wang, W. Wen, J. Wang, Experimental study on a micro-abrasive slurry jet for

glass polishing, Int. J. Adv. Manuf. Technol. (2016) DOI 10.1007/s00170-016-9109-z.

[6] F. Zhang, X. Song, Y. Zhang, D. Luan, Figuring of an ultra-smooth surface in nanoparticle

colloid jet machining, J. Micromech. Microeng. 19 (2009) 054009.





[9] H. Nouraei, A. Wadoslawsky, M. Papini, J.K. Spelt, Characteristics of abrasive slurry jet micro-

machining: a comparison with abrasive air jet micro-machining, Journal of Materials Proc. Tech.

213 (2013) 1711-1724.

[10] Z. Cao, C.F. Cheung, Theoretical modelling and analysis of the material removal

characteristics in fluid jet polishing, Int. J. Mech. Sci. 89 (2014) 158–166.

[11] M. Papini, J.K. Spelt, Impact of rigid angular particles with fully-plastic targets part II:

parametric study of erosion phenomena, Int. J. Mech. Sci. 42 (2000) 1007–1025.

231

[12] Z. Li, S. Li, Y. Dai, X. Peng, Optimization and application of influence function in abrasive jet




[14] C. Y. Poon, B. Bhushan, Comparison of surface roughness measurements by stylus profiler,

AFM and non-contact optical profiler, Wear 1190 (1995) 76-88.

[15] K. Kowsari, H. Nouraei, , B. Samareh, M. Papini, J.K. Spelt, CFD-aided prediction of the

shape of abrasive slurry jet micro-machined channels in sintered ceramics. Ceramics Int. 42

(2016) 7030-7042.

[16] J. Schwartzentruber, M. Papini, J.K. Spelt, Prediction of Surface Roughness in Abrasive

Waterjet Trimming of Fiber-Reinforced Polymer Composites, Int. J. Machine Tools Manuf.

(2016, submitted).

[17] M. Papini and J.K. Spelt, Organic coating removal by particle impact, Wear 213 (1997) 185-

199.

[18] ISO 4288: Geometrical Product Specifications (GPS)—Surface texture: profile method—rules

and procedures for the assessment of surface texture (1996).


erosion, Wear 253 (2002) 1035-1043.

[20] R. Haj Mohammad Jafar, M. Papini, J.K. Spelt, Numerical simulation of surface roughness and

erosion rate of abrasive jet micro-machined channels, Wear 303 (2013) 302–312.

[21] Hasem, Md. A., The effect of tangential velocity component in abrasive jet micro-machining

of borosilicate glass, Ryerson University Ph.D. Thesis (2013).

[22] I. Finnie, Erosion of surfaces by solid particles, Wear 3 (2) (1960) 87-103.

[23] H. Getu, J.K. Spelt, M. Papini, Conditions leading to the embedding of angular and spherical

particles during the solid particle erosion of polymers, Wear 292-293 (2012) 159-168.

Chapter 5

232




[2] I. Khalil, M. Rudolph, A. Leiro, M. Neumann, W. Heinrich, High power, high linearity and low-

noise hybrid RF amplifiers based on GaN HEMTs, Microwave Conference 1 (2009) 16-18.



(1998) 885-891.


and transitional slopes in a talc-filled co-polymer, Precision Engineering 43 (2016) 52-62.

[5] O.W. Fähnle, H. van Brug, H.J. Frankena, Fluid jet polishing of optical surfaces, Applied Optics

37 (28) (1998) 6771–6773.

[6] S.M. Booij, O.W. Fähnle, J.J. Braat, Shaping with fluid jet polishing by footprint optimization,

Applied Optics 43 (1) (2004) 67–69.

[7] S. Booij, I. Partosoebroto, J.J. Braat, H. van Brug, Computational model for prediction of

shaping with FJP and experimental validation, Proceedings of Optical Fabrication and Testing

Conference, Tucson, USA (2002) 52-54.

[8] H. Fang, P. Guo, J. Yu, Dwell function algorithm in fluid jet polishing, Applied Optics 45 (18)

(2006) 4291–4296.

[9] P.M. Shanbhag, M.R. Feinberg, G. Sandri, M.N. Horenstein, T.G. Bifano, Ion-beam machining

of millimeter scale optics, Applied Optics 39 (4) (2000) 599–611.

[10] A. Ghobeity, J.K. Spelt, M. Papini, Abrasive jet micro-machining of planar areas and

transitional slopes, J. Micromech. Microeng. 18 (2008a) 055014.

[11] D.S. Park, M.W. Cho, T.I. Seo, Mechanical etching of micro pockets by powder blasting, Int.

J. of Adv. Manuf. Technol. 25 (2005) 1098-1104.







233

[14] S. Murakami, Comparison of various turbulence models applied to a bluff body, Journal of

Wind Engineering and Industrial Aerodynamics, 46-47 (1993) 21-36.

[15] A. Gnanavelu, N. Kapur, A. Neville, J.F. Flores, N. Ghorbani, A numerical investigation of a

geometry independent integrated method to predict erosion rates in slurry erosion, Wear 271

(2011) 712-719.

[16] J. Tu, H.G. Yeoh, C. Liu, Computational Fluid Dynamics: A Practical Approach, Butterworth-

Heinemann, Oxford, UK, 2008.

[17] ANSYS Fluent 12.0, 2009. Theory guide. ANSYS, Inc.

[18] H.K. Hockin, K. Xu, S. Jahanmir, Microfracture and material removal in scratching of

alumina, Journal of materials science 30 (1995) 2235-2247.


abrasive jet micromachining, Wear 264 (2008b) 185-198.



[21] R. Haj Mohammad Jafar, J.K. Spelt, M. Papini, Surface roughness and erosion rate of abrasive

jet micro-machined channels: Experiments and analytical model, Wear 303 (2013) 138-145.

[22] ISO 11562:1996 Geometrical Product Specifications (GPS) -- Surface texture: Profile

method—Metrological characteristics of phase correct filters.

[23] ISO 4288:1996 Geometrical Product Specifications (GPS) -- Surface Texture: Profile

Method—Rules and Procedures for the Assessment of Surface Texture.


comparison between machining in air and submerged in water, Int. J. Machine Tools and

Manufacture 88 (2015) 108-117.



[26] P. Dixit, J. Miao, Aspect-ratio-dependent copper electodeposition technique for very high

aspect-ratio through-hole plating, Journal of the Electrochemical Society 153 (6) (2006) G552-

G559.

[27] W. Dow, H. Chen, M. Yen, W. Chen, K. Hsu, P. Chuang, H. Ishizuka, N. Sakagawa, R.

Kimizuka, Through-hole filling by copper electroplating, Journal of the electrochemical society

155 (12) (2008) D750-D757.

234

[28] H.M. Pollock, D. Maugis, M. Barquins, Characterization of submicrometre surface layers by

indentation. Microindentation Techniques in Material Science and Engineering, Balu, P.J. and

Lawn, B.R., Eds., ASTM (American Society for Testing and Materials) STP 889, Philadelphia,

USA (1986) 47-71.

[29] R. Haj Mohammad Jafar, H. Nouraei, M. Emamifar, M. Papini, J.K. Spelt, Erosion modeling

in abrasive slurry jet micro-machining of brittle materials, Journal of Manufacturing Processes

17 (2014) 127-140.

Chapter 6

[1] D. Solignac, A. Sayah, S. Constantin, R. Freitag, M.A.M. Giijs, Powder blasting for realization

of microchips for bio-analytic applications Sens. Actuators, A92 (2001) 388–393.

[2] C. Liu, J. Chen, J. Engel, J. Zou, X. Wang, Z. Fan, K. Ryu, K. Shaikh, D. Bullen, Polymer

micromachining and applications in sensors, microfluidics, and nanotechnology The 226th

National Meeting of the American Chemical Society (ACS), New York, NY, 11–17 September

(2003).



[4] A. Ghobeity, D. Ciampini, and M. Papini, An analytical model of the effect of particle size

distribution on the surface profile evolution in abrasive jet micromachining, Journal of

Materials Processing Technology, Vol. 209, Issue 20 (2009) 6067-6077.

[5] D. Dehnadfar, J. Friedman, and M. Papini, Laser shadowgraphy measurements of abrasive

particle spatial, size and velocity distributions through micro-masks used in abrasive jet micro-

machining, Journal of Materials Processing Technology, 212 (1) (2012) 137–149.

[6] P.H. Shipway, The effect of plume divergence on the spatial distribution and magnitude of wear

in gas-blast erosion, Wear 205 (1997)169–77.

[7] T. Burzynski and M. Papini, Measurement of the particle spatial and velocity distributions in

micro-abrasive jets, Measurement Science and Technology, 22 (2011) 025104.

[8] A. Mansouri, M. Mahdavi, S.A. Shirazi, B.S. McLaury, Investigating the effect of sand

concentration on erosion rate in slurry flows, Proceedings of the 2015 NACE Corrosion

Conference and Expo. Paper No. 6130.

235

[9] H. Qi, D. Wen, C. Lu, G. Li, Numerical and experimental study on ultrasonic vibration-assisted

micro-channelling of glasses using an abrasive slurry jet, Int. J. Mech. Sci., 110 (2016a) 94-107.

[10] H. Qi, D. Wen, Q. Yuan, L. Zhang, Z. Chen, Numerical investigation on particle impact

erosion in ultrasonic-assisted abrasive slurry jet micro-machining of glasses, Powder

Technology (2016b, in press).


of abrasive slurry jet micro-machined channels in sintered ceramics, Ceramics Int'l 42 (2016a)

7030-7042.

[12] A. Nouhi, K. Kowsari, J.K. Spelt, M. Papini, Abrasive jet machining of channels on highly-

curved glass and PMMA surfaces, Wear 356-357 (2016) 30-39.

[13] V. Hadavi, B. Michaelsen, M. Papini, Measurements and modeling of instantaneous particle

orientation within abrasive air jets and implications for particle embedding, Wear 336–337

(2015) 9–20.

[14] H. Li, A. Lee, J. Fan, G.H. Yeoh, J. Wang, On DEM–CFD study of the dynamic characteristics

of high speed micro-abrasive air jet, Powder Technology 267 (2014) 1611-1179.

[15] C.T. Crowe, J.D. Schwarzkopf, M. Sommerfeld, Y. Tsuji, Multiphase flows with droplets and

particles – second edition, Taylor & Francis Group, LLC – CRC Press (2012), pp. 17–34.

[16] P.J. Slikkerveer, F.H. in't Veld, Model for patterned erosion, Wear 233 (1999) 377-386.

[17] P.H. Shipway, I.M. Hutchings, Influence of nozzle roughness on conditions in a gas-blast

erosion rig, Wear 162 (1993) 148-158.



Process. Technol. 23 (2016b), 190-210.



[20] A. Haider, O. Levenspiel Drag Coefficient and Terminal Velocity of Spherical and

Nonspherical Particles, Powder Technology 58 (1989) 63-70.


erosion. Wear 253, (2002) 1035-1043.


micro-machining: a comparison with abrasive air jet micro-machining, J. Mater. Process.

Technol. 213 (2012) 1711–1724.

http://www.sciencedirect.com/science/journal/00431648/336/supp/C

236

[23] H. Getu , A. Ghobeity, J.K Spelt, M. Papini, Abrasive jet micromachining of acrylic and

polycarbonate polymers at oblique angles of attack, Wear, Volume 265, Issues 5-6, (2008), 888-

901.

Chapter 7

[1] Y. Iwai, T. Miyajima, T. Honda, T. Matsubara, K. Kanda, S. Hogmark, Evaluation of erosive

wear resistance of TiN coatings by a slurry jet impact test, Wear 261 (2006) 112-118.

[2] H.M. Hawthorne, B. Arsenault, J.P. Immarigeon, J.G. Legoux, V.R. Parameswaran,

Comparison of slurry and dry erosion behaviour of some HVOF thermal sprayed coatings, Wear

225-229 (1999) 825-834.

[3] R.J.K. Wood, The sand erosion performance of coatings, Mat and Design 20 (1999) 179-191.

[4] J.F. Santa, L.A. Espitia, J.A. Blanco, S.A. Romo, A. Toro, Slurry and cavitation erosion

resistance of thermal spray coatings, Wear 267 (2009) 160-167.

[5] K. Sugiyama, S. Nakahama, S. Hattori, K. Nakano, Slurry wear and cavitation erosion of

thermal-sprayed cermets, Wear 258 (2005) 768-775.


and transitional slopes in a talc-filled co-polymer, Precision Eng 43 (2016) 52-62.


milling of sintered ceramic wafers with and without copper-filled through-holes, J Mat Proc

Tech 230 (2016a) 198-210.



paths, Int J Machine Tools Manuf 68 (2013) 30-39.


on the shapes of abrasive slurry-jet micro-machined holes and channels, J Mach Tools Manuf

110 (2016b) 80-91.


in abrasive slurry jet micro-machining, Precision Eng 44 (2016) 109-123.

[11] ASM Int. Handbook Committee, ASM Handbook, Volume 02 - Properties and Selection:

Nonferrous Alloys and Special-Purpose Materials, ASM International (1990).

237

[12] L. Zhaojiang, Hardness characteristics in electrodeposited copper/nickel multilayer systems.

University of Windsor M.Sc. Thesis (1999).




of abrasive slurry jet micro-machined channels in sintered ceramics, Ceramics Int 42 (2016c)

7030-7042.

[15] ANSYS Fluent 15.0 Theory guide, ANSYS, Inc; (2015)

[16] H. Schlichting, K. Gersten. Boundary-Layer Theory. Springer; New York (2004)

[17] K. Kowsari, J. Schwartzentruber, J.K. Spelt, M. Papini. Erosive smoothing of abrasive slurry-

jet micro-machined channels in glass, PMMA, and sintered ceramics: experiments and

roughness model. Precision Eng (2016d, submitted).

Flow Fields and Particle Trajectories in Abrasive Slurry ... · constructive discussions on my...

Documents

Transcript of Flow Fields and Particle Trajectories in Abrasive Slurry ... · constructive discussions on my...