Bulk metallic glass matrix composites :

Bulk metallic glass matrix composites : processing, microstructure, and application as a kinetic energy penetrator - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Bulk metallic glass matrix composites : processing, microstructure, and application as a kinetic energy penetrator
Citation
Dandliker, Richard B.
(1998)
Bulk metallic glass matrix composites : processing, microstructure, and application as a kinetic energy penetrator.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/HTJS-N846.
Abstract
NOTE: Text or symbols not renderable in plain ASCII are indicated by [...]. Abstract is included in .pdf document.

The development of alloys with high glass forming ability allows fabrication of bulk samples of amorphous metal. This capability makes these materials available for applications which require significant material thickness in all three dimensions. Superior mechanical properties and advantages in processing make metallic glass a choice candidate as a matrix material for composites.

This study reports techniques for making composites by melt-infiltration casting using the alloy [...] ([...]) as a matrix material. Composite rods 5 cm in length and 7 mm in diameter were made and found to have a nearly fully amorphous matrix; there was less than 3 volume percent crystallized matrix material. The samples were reinforced by continuous metal wires, tungsten powder, or silicon carbide particulate preforms. The most easily processed samples were made with uniaxially aligned tungsten and carbon steel continuous wire reinforcement; the majority of the analysis presented is of these samples. The measured porosity was typically less than 3%. The results also indicate necessary guidelines for developing processing techniques for large scale production, new reinforcement materials, and other metallic glass compositions.

Analysis of the microstructure of the tungsten wire and steel wire reinforced composites was performed by x-ray diffraction, scanning electron microscopy, scanning Auger microscopy, transmission electron microscopy, and energy dispersive x-ray spectroscopy. The most common phase in the crystallized matrix is most likely a Laves phase with the approximate formula [...]. In tungsten-reinforced composites, a crystalline reaction layer 240 nm thick of tungsten nanocrystals in anamorphous matrix formed. In the steel reinforced composites, the reaction layer was primarily composed of a mixed metal carbide, mainly [...].

One promising application of the metallic glass mat composite is as a kinetic energy penetrator material. Ballistic tests show that a composite of 80 volume percent uniaxially aligned tungsten wires and a [...] 1 matrix has self-sharpening behavior, which is a necessary characteristic of superior penetrator materials. Small-scale tests with both aluminum and steel targets show that this composite performs better than tungsten heavy alloys typically used for penetrator applications, and comparably with depleted uranium.
Item Type:
Thesis (Dissertation (Ph.D.))
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Materials Science
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Johnson, William Lewis
Thesis Committee:
Unknown, Unknown
Defense Date:
24 October 1997
Record Number:
CaltechETD:etd-01242008-074925
Persistent URL:
DOI:
10.7907/HTJS-N846
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
322
Collection:
CaltechTHESIS
Deposited By:
Imported from ETD-db
Deposited On:
15 Feb 2008
Last Modified:
21 Dec 2019 02:24
Thesis Files
Preview
PDF (Dandliker_rb_1998.pdf)
- Final Version
See Usage Policy.
53MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
BULK METALLIC GLASS MATRIX
COMPOSITES:
PROCESSING, MICROSTRUCTURE, AND
APPLICATION AS A KINETIC ENERGY
PENETRATOR

Thesis by
Richard B. Dandliker

In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy

California Institute of Technology

Pasadena, California
1998

(Submitted October 24, 1997)

Richard B. Dandliker

ACKNOWLEDGMENTS

None of this work would have been possible without Dr. Bill Johnson, a man
whose intuition and insight is often stupefying to behold. I must thank him for
giving me the freedom to find a path which truly made my research dear to me,
and for his amazingly infectious enthusiasm for science and discovery.

Special thanks also goes to Dale Conner, my collaborator for much of the
work on this project and a good friend. His expertise in mechanics was very
valuable, as were our discussions which helped me focus on what was
important.

Many other members of the Johnson group, both past and present, have
also been a great help. Dr. Atakan Peker's work in developing the first
Vitreloy™ alloys laid the basis for much of this study. Drs. Peker and David S.
Lee showed me how to operate most of the lab equipment and many of the
techniques I was to use later in my graduate career. Dr. Ralf Busch was
always willing to give me a tutorial on thermodynamics and the glass
transition. Drs. Susanne Schneider, Ulli Geyer, and Y. J. Kim brought a wide
range of knowledge and experience to our discussions. Dr. Mo Li was a great
source of stability in the group during his time at Caltech, and his work always
made me grateful for the opportunity to be an experimentalist. The work of Dr.
Xianghong Lin provided essential insights and brought increased understanding
to processing techniques for metallic glass. I would also like to thank Haein
Choi-Yim for many useful discussions about composites and processing. The
rest of the Johnson group, including Steven Glade, Sven Bossuyt, Valerie
Scruggs, Jian Li, Andreas Masuhr, Eric Bakke, Brian Burkholder, Wenshan
Liu, Andy Waniuk, and de facto member Dr. Karina Montilla Edwards, helped
make my tenure at Caltech thoroughly enjoyable.

Carol Garland was invaluable for help and instruction with the TEM, as well
as other microscopy techniques. Professor Brent Fultz instructed me in
electron and x-ray diffraction both in the classroom and the laboratory. Dr.
Channing Ahn happily tolerated my repeated borrowing of his 35 mm camera.
Dr. Jack Worrell at the University of Southern California operated the
scanning Auger microscope and was of great help in the analysis of data from
that instrument. Pam Albertson was always eager to help find creative
solutions to any administrative or budgetary problem. Eve Kiefer and Elaine
Granger were also instrumental in making sure the lab ran smoothly.

This report is based upon work supported in part by the U. S. Army
Research Office under grant number DAAH04-95-1-0233. Robert Dowding
from the Army Research Laboratory at Aberdeen was very helpful in providing
tungsten heavy alloy samples and background data for penetrator testing. I
would also like to thank the National Science Foundation for the fellowship
which supported me through the first three years of graduate school. |

Amorphous Technologies International (Laguna Niguel, California) provided
some samples of metallic glass, and made their facilities and equipment
available to me without hesitation. Dr. Michael Tenhover has always been a
great help and a nearly limitless source of ideas. In addition, I would like to
thank Dr. Tenhover and the Carborundum Company for supplying silicon
carbide fibers and preforms, and performing scanning Auger and secondary ion
mass spectrometry (SIMS) analysis on silicon carbide reinforced samples.

Finally, I would like to thank my parents and my brother for always being
encouraging and supportive of my education and research. And to Mary,
thanks for putting up with me throughout the entire process and for helping me

make it through with my sanity intact.

ABSTRACT

The development of alloys with high glass forming ability allows fabrication
of bulk samples of amorphous metal. This capability makes these materials
available for applications which require significant material thickness in all
three dimensions. Superior mechanical properties and advantages in
processing make metallic glass a choice candidate as a matrix material for
composites.

This study reports techniques for making composites by melt-infiltration
casting using the alloy Zr4, 9Tij3 gCuyg sNi19. pBego.5 (Vitreloy™ 1) as a matrix
material. Composite rods 5 cm in length and 7 mm in diameter were made and
found to have a nearly fully amorphous matrix; there was less than 3 volume
percent crystallized matrix material. The samples were reinforced by
continuous metal wires, tungsten powder, or silicon carbide particulate
preforms. The most easily processed samples were made with uniaxially
aligned tungsten and carbon steel continuous wire reinforcement; the majority
of the analysis presented is of these samples. The measured porosity was
typically less than 3%. The results also indicate necessary guidelines for
developing processing techniques for large scale production, new reinforcement
materials, and other metallic glass compositions.

BeyoZr3TiNiCu. In tungsten-reinforced composites, a crystalline reaction layer

240 nm thick of tungsten nanocrystals in an amorphous matrix formed. In the
steel reinforced composites, the reaction layer was primarily composed of a
mixed metal carbide, mainly ZrC.

One promising application of the metallic glass matrix composite is as a
kinetic energy penetrator material. Ballistic tests show that a composite of 80
volume percent uniaxially aligned tungsten wires and a Vitreloy™ 1 matrix
has self-sharpening behavior, which is a necessary characteristic of superior
penetrator materials. Small-scale tests with both aluminum and steel targets
show that this composite performs better than tungsten heavy alloys typically

used for penetrator applications, and comparably with depleted uranium.

TABLE OF CONTENTS

ACKNOWLEDGMENTS .........ccccccsscssscssssssessscscsesesscsssssesssssssssscsssosesssssessseceeseseseees
ABSTRACT.........scscsssssscscssscsssssssscssscssscscessssssssssssssssscesesssossscesscecesssssssseseseseasaseceseneeees
TABLE OF CONTENTS........cccssssssssssssssssssscsssscsssssssssssscsssecesesscesssscseseseseseseseeeeees
LIST OF FIGURES... cscscssssssssssoscsesesssscsssesssssssessessssssssscsssesssesscessesesesesesesseeeens
LIST OF TABLES uu... ccscssssesssssssssssscsssssssscsssssssssssssscesesssscesssssssssssssssesesessensesees

1.1.1 History of Metallic Glasses ..0....... cc cescsssscsssssssersecessessseeceees

1.1.2 Scientific and Technological Interest in Metallic

GASSES uo... .ccceccceessccsscccsssssccsscccsssccesssececsssccssccecsssossssscsesssccesssccesseasensnes

1.2.1 Homogeneous Nucleation Theory..........cscsssssscseeessseseees
1.2.2 Heterogeneous Nucleation Theory ...........sssscsessseseseeseers

1.3 COMPOSITE THEORY AND TRADITIONAL COMPOSITE

SYSTEMS. .......cccssssscssscssssssesesssecessscsescscecscessessessseeeecessessscessssseaseseessseeseseecees
1.3.1 Classification and Types of Composites...........c:ccsseeseeers
1.3.2 Role of the Composite Interface..............cccsscessssscesseesteseeeess
1.4 METALLIC GLASS AND COMPOSITES .............cessscsssssessererseee
REFERENCES... essssssesssssssssscssssssssecsssessessescscssscssesesssssssessesssnsnsscesoses

CHAPTER 2: PROCESSING A BULK-METALLIC-GLASS MATRIX

COMPOSITE BY MELT INFILTRATION CASTING... .esesssssssssesseeseseees

2.1 TRADITIONAL PROCESSING OF METAL MATRIX

COMPOSITES. ............ssscssscsssssssssssssessseccsescsesesessssesesessssesssesesesesesssneseaeseecsaees

eee

2.1.2 Solid-Matrix Primary Processing ...........ccsscssssssscsssssssseseeees 34

2.1.3 Reactive Primary Processing ...........sssssssssssssssssesssssssscessesers 36
2.2 PROCESSING OF METALLIC GLASS COMPOSITES BY
RAPID QUENCHING 1... sssscssssecesssccessscecsssesessssssesecssscsessecassesesecseneees 36
2.3 EXPERIMENTAL... .ccscsscssssssessscscscssssesesecscesesessssscacsesessscssssersssnsaces 39
2.4 RESULTS... sssssssesssscsssscsesssscssssssesssssecessesesesessesesessssssssscsssesessssessces 44
2.5 DISCUSSION... cscsssesssscssssesecssccsssssessssssocscsesecsceesesssesessssececsesssesseseseees 46
2.6 CONCLUSIONS 00... essessscecssececsssecscsesececsesessesesssssecscssscseseseseseeseseees 49
REFERENCES... ccsssssssssssssccsssssessesesessssescscsecscscsesscessesesssnsassesnescassecseaeaes 51

CHAPTER 3: MICROSTRUCTURE OF THE BULK-METALLIC-

GLASS MATRIX COMPOSITE .........cssssssssscssssssseescecessesecececseeesesesssesssscsessssseseesees 53
3.1 RESULTS OF MICROSTRUCTURAL ANALYSIS OF
COMPOSITES REINFORCED BY TUNGSTEN AND STEEL........... 53

8.1.1 X-Ray Diffraction... csssscsssesssessssssesesessecesssseccsessesssessescseeees 53
3.1.2 Scanning Electron Microscopy.........csssscsccsssseceesssescssssssseeees 59
3.1.3 Scanning Auger Microscopy..........:sscssccsscssessssssssessssscsscsssscecees 67

3.1.4 Transmission Electron Microscopy and Scanning

Transmission Electron Microscopy........sssssssssssssssesesesesesesessesesesseees 72
3.2 ANALYSIS OF RESULTS..........ccccssssssssssssssssesssssssescscsssesesessseseesesessees 82
3.3 SILICON-CARBIDE AND OTHER REINFORCEMENTS
IN METALLIC-GLASS-MATRIX COMPOSITEG............ccscssssssessssssssees 87
B.4 CONCLUSION... ccesssessessssesssscsecsccessecscsesscsesscsesecsesseseeecsssseesseseeees 93
REFERENCES... ccssssssssssssssesecsssesssccnccsssecussesuesesscsesscssseceessssesscsesssoseaees 98

CHAPTER 4: KINETIC ENERGY PENETRATORS: A NOVEL
APPLICATION OF BULK-METALLIC-GLASS-MATRIX
COMPOSITEG.........ccccscsssssssssssssssssssccuscucsscnecsecucseenssscsessesscsessesecucseesesesaeseceesssaeeseasenenees 99

4,1 TRADITIONAL KINETIC ENERGY PENETRATOR

| SYSTEMS. .......cssssssssssssscessssssscsssecsssecsssecscsessssesscsessssessesssssessssesesesscsucsesececeeses 99
4.1.1 Background of Kinetic Energy Penetrators.............c:esssss0e. 99
4.1.2 Adiabatic Shear of Materials...........ccssssssssssssssssssssssssssesececees 102
4.2 ADIABATIC SHEAR BANDING IN METALLIC GLASS
AND METALLIC GLASS COMPOSITES..........ccccsssssssssssssscsssceseeesecneens 106
4.2.1 Failure Mode of Unreinforced Metallic Glass... 106

4.2.2 Quasi-Static and Low Strain-Rate Testing of

Metallic Glass Matrix Composites.........c.cccccccssscseseeees reateseseessenceeneees 108
4.3 PROCEDURE FOR BALLISTIC TESTING uuu..cccscccseccesescesseseeees 110
4.3.1 Procedure for Reverse Ballistics ........c.cccssssssssseccesescessssesees 112
4.3.2 Procedure for Forward Ballistics ........cccccsssssssssscssecescsssesseees 115
A.A RESULTS wu. sesessessesssesscssssssssssssscssssessssscsscsesssssesscssssesaesascussussecsess 119
4.4.1 Results of Reverse Ballistic Testing........ccccccscccsssseceseceeees 119

4.4.2 Results of Forward Ballistics into Aluminum

Target ....cccsccsccsecsscsssssssssssessssssssssessssesecssssessssesessssececassusacsesscsecaesscaveess 119

4.4.3 Results of Forward Ballistics into Steel Targets ................ 127

4.5 CONCLUSIONS w.csssssssssssssstntstistissisisintiatinintineietntienanansne 133
REFERENCES Wu. cccsscsssssssssssscsssssessscsucsscssssssssscsussuecuscussuscuccucsassusssessecase 137
APPENDIX 1 o....scssssssssssssssssssvsvsssssssssssssesseseseeeceeececcesssssssssssssesssssssssssesssssssssuusuesseseceecees 139

APPENDIK 2.o...ccssssssssssssssssscccssecsssesssuscssssccssscsssscssussssssssecsssscessucesucesereenessasesnecsssecenses 140

Fig. 1.1.

Fig. 1.2.

Fig. 1.3.

Fig. 1.4.

Fig. 1.5.

Fig. 1.6.

Fig. 1.7.

Fig. 2.1.

Fig. 2.2.

Fig. 2.3.

Fig. 2.4.

Fig. 2.5.

LIST OF FIGURES

Schematic of melt spinning apparatus .........sccsssesesssssssssececsssescscsssesess 4
Critical cooling rate, R,, versus the reduced glass transition

temperature, T,,=T,/T,p, for different metals and metallic

Gibbs free energy AG for crystalline embryo as a function of

LACUS. .....s.ssssssscsssceescssscssscscsescscsssssesessscsssesssssesssessaesesessscsssesssessssssssesesesseeevees 12
Schematic of a spherical-cap-shaped crystalline embryo...............006. 16
Spherical cap Geometry. ........ssssssssccsssssssssesssssecsescsesesessesssescscsssessesssessoses 17

Schematic of crack bridging and fiber pull out in a brittle-

matrix brittle-fiber COMpoOSIte............scscsscscssessscsesessecesscsssscssctscsssecesseseees 23

Schematic of toughening mechanisms in a brittle-matrix
ductile-fiber composite. dis the debond length, and R is the

radius Of the fiber............ecssssssssssscssssssessscsesssessesecesescsesssessssesesssseasseacesees 25
Different processing techniques for fabricating metal matrix
COMPOSITES. ........seesessssssesssesssessscsseneeescecssssesssesescecesesesesceesssscscssssesscesesvsoeses 31
Schematic of modified melt-spinner to make continuous-wire
reinforced aMorphouS ribDOs .........c.sssssssesssssssesessesssessssssssssssssscsesseesseses 38
Schematic diagram of apparatus for casting bulk metallic glass
MATIX COMPOSILES.........cccsceceressesesesseseeeseeseces . sesssesscsscesceessesceecsssssassensenseees 41
Graph of time, temperature, and pressure during processing of

the metallic glass matrix composites. Solid line represents
temperature (left axis). Dotted line represents pressure (right

axis). Dashed lines represent melting temperature and glass
transition temperature. .........cccccsesecesesescscsescssssssssscssscscscscecacsssctensessencececs 43

Photograph of metallic glass matrix composite samples ...............0.0. 45

Fig. 3.1.

X-ray diffraction patterns of metallic glass matrix composites.

Cobalt Ko radiation (A=0.1790 nm) was uSed.............cccccccsesccsssecees

Fig. 3.2 X-ray diffraction patterns of unreinforced metallic glass and

metallic-glass-matrix composites with low fiber fractions.
Measured on Inel diffractometer with Co Ka radiation

(A=1.790 A). X-axis converted to interplanar spacing from 20

USING Bragg's law.........sssssssscssssssesssessssessseecssssesecssssssssssssssssssscssessscesscesees

Fig. 3.3 X-ray diffraction pattern of 20 vol% steel music wire composite.

Fig. 3.4.

Fig. 3.5.

Fig. 3.6.

Fig. 3.7.

Fig. 3.8.

Siemens D-500 was used on the same sample for both

PALCEITS.........ccececcscsesccsecssesecsecsesescescessenssssesceessasscssssnsenssssessessessessesesees

SEM micrograph of metallic glass matrix composite.

Reinforced with 80 vol% W wire. Light areas are wires. Image

taken with backscatter detector..........:...ccsccssssssssccsssccsssssesssscessscessesees

SEM micrograph of metallic glass matrix composite.

Reinforced with 20 vol% steel music wire. Dark areas are

wires. Image taken with backscatter detector. ........cccssssesesesseeess

Metallic glass matrix composite. Reinforced with 20 vol% steel
music wire. Round dark areas are wires. Dark faceted regions

are crystallized portions of the matrix. Image taken with

backscatter detector in SEHM..............cccccccsccssssssesscsscsccsssssesssesscsscsscsees

Close up view of metallic glass matrix composite. Large round
dark area is single steel wire. Crystals are visible at some

locations along the wire-matrix interface. Image taken with

backscatter detector in SEM...........c:ccccccsccssssssssssscsssssssessssssscssecseseeeseee

SEM micrograph of metallic glass matrix composite. Lighter

areas are tungsten wires. Dark, faceted crystals are visible at

eee

...60

some locations along the wire-matrix interface. Image taken

with backscatter detector. ...........ccccccsscsssssssccssscsssessssssssscscecscccecssecesssssees 65

Fig. 3.9. SEM micrograph of metallic glass matrix composite reinforced

Fig. 3.10.

Fig. 3.11.

Fig. 3.12.

Fig. 3.13.

Fig. 3.14.

by 40 vol% tungsten wire. Matrix allowed to infiltrate wires for

160 min. Light area is tungsten wire. Light gray area is

matrix, and dark gray area are regions of crystallized matrix.

Image taken in backscatter mode with Philips field-emission
electron gun by Philips representative. ........ccccsssssssssssssccecesssssecescsees 65
Close up view of metallic glass matrix composite. Darker

area on the top is single steel wire. Crystals are visible at

some locations along the wire-matrix interface. Image taken

with backscatter detector in SEM.........ccccccssesssssssssssscscsssssesssseceseaceeees 66
Close up view of metallic glass matrix composite. Darker

area on the left is single steel wire. Crystals are visible at

some locations along the wire-matrix interface. Image taken

with backscatter detector in SEM.........ccccccsssssssssssssssscsssssssssscseseseesecees 66
Backscatter electron image of tungsten/metallic glass

interface in composite. Each white dot and number

corresponds to a point of elemental analysis in Table 3.1 with

AUeYr SPCCtLOSCOPY. .....ssessssssesessosssesssssessesssssscscecscsesecestsesscssscaescnsasseseserees 68
Backscatter electron image of steel/metallic glass interface.

Each white dot and number corresponds to a point of

elemental analysis in Table 3.2 with Auger spectroscopy.............00+ 71
Transmission electron micrographs of interfacial region

between tungsten wire and metallic glass Matrix .......cccsscsssssssssseceeeees 74

Fig. 3.15.

Fig. 3.16.

Fig. 3.17.

Fig. 3.18.

Fig. 3.19.

Fig. 3.20.

Fig. 3.21.

Compositional profile across tungsten wire/matrix interface in
tungsten wire reinforced composite. Measured by EDS in
STEM. Beryllium concentration omitted from analysis (see
COX)... ceccccsesescsecssesesecececssesectsessccsssscusncscsssecesssecssssesesesssesecesessesssesssessseseseves 76
Transmission electron micrograph (bright field) of interface
between steel music wire and amorphous Vitreloy™ 1 matrix.
Amorphous region on right, and steel on left. Diffraction
patterns taken with 0.5 um diameter selected area aperture,
except for pattern from 8 tm into steel, which was taken with
1.5 WM APerture. ......eecscseeesesesssesesesecssesesscsssessssstsessssssssssessesessessssesessenss 77
Transmission electron micrographs of interfacial region in
steel- wire-reinforced metallic glass. (A) bright field and (B)
dark field. Lower right area is amorphous matrix, and upper
left is crystalline Wire...........ccssssssssssssssssscsescsessssssssssssssssesssssssssssesssesseees 79
Transmission electron micrographs in (A) BF and (B) DF of
steel wire/metallic glass interface. Region used for EDS
analysis profile shown IN Fg. 8.19. ....reeccsscccnsessnsecsssesssneessnesssssecseseeeeses 80
Compositional profile across steel wire/matrix interface in
steel wire reinforced composite. Measured by EDS in STEM.
Beryllium and carbon concentrations omitted. Location of
wire/matrix interface at 0 OM X-AXIS. .....cccssessssessssessssessssssssssesesesssseeseosees 81
Carbon compositional map for SiC/Vitreloy™ 1 composite.
Image taken with SAM. Sample infiltrated for 3 hh. w...cscsessescseeeeee 88
Silicon compositional map for cross section of SiC/Vitreloy™
1 composite. Image taken with SAM. Sample infiltrated for 3
Tr, ceesssssesssesssessssscsssccuccssecusessccssesuscsusssscsucssscssscaucsascsucssessessssssussssesssssusssessseseees 88

Fig. 3.22.

Fig. 3.23.

Fig. 3.24.

Fig. 3.25.

Fig. 3.26.

xiv
Carbon compositional map for cross section of SiC/Vitreloy™

1 composite. Image taken with SAM. Sample infiltrated for

Silicon compositional map for cross section of SiC/Vitreloy™

1 composite. Image taken with SAM. Sample infiltrated for

Beryllium compositional map for cross section of

SiC/Vitreloy™ 1 composite. Image taken with SAM. Sample

Infiltrated for 0.5 bin... ccscsccsscssescsscsccsccescessssesscscessessessscsssescesceessccescece 91

Optical micrograph of SiC preform infiltrated with bulk
metallic glass. Dark regions are SiC, light regions are
AMOFPhous Metal. ........ecssscssssesesessssssssesssssssecsessscsesesesssesesestsessesesesees
SEM micrograph of composite interface. Light area at lower
left is tantalum wire reinforcement, and dark area at upper
right is metallic glass. Image taken with backscatter

AetectOr........ccccssscsssscsssscssscscssescssssssssasecssssscssssessesscsesssssssesssesceceeseccceeees

Fig. 4.1. Computer simulation of the penetration process of a long-rod

kinetic penetrator into steel ArMOY..........c.ccsssssssssessssssccessssseceeeeees

Fig. 4.2. Three types of penetration behavior for various materials that

show (a) self-sharpening, (b) limited self-sharpening and (c) no
self-sharpening (“mushrooming”).........c.cscccscssssssssccssececscscecssecessesens

Fig. 4.3. Photograph of metallic glass (Vitreloy™ 1) cylinder after

compressive failure. Applied stress was along axis of cylinder.

Fig. 4.4 SEM micrograph of fracture surface of metallic glass in

Fig. 4.5. Stress-strain curve of unreinforced metallic glass cylinder

COMPFeEsSSIVe FAIULE. 00.0.0... ecesscesssssscsessssssscscsesescsscssssscscssssesscssscaeseeees

eeecoecce

sessees 91

sesseee 107

Fig. 4.6. SEM micrograph of failure surface of 20 volume percent

tungsten-wire-reinforced Vitreloy™ 1 composite. Sample —

failed on single shear band under dynamic axial compression

on a Hopkinson bar apparatus..........ccsccscssssssssssesesssssssscscsssscscsseceerseseees 111
Fig. 4.7. Schematic of cannon used for ballistic testing...........ccccscsscsssscsosssceeseres 113
Fig. 4.8. Reverse ballistic schematic. ..........ccssssssssssssssssssssesssssssssssssscsssscsesesecececees 114
Fig. 4.9. Forward ballistic schematic.............cccccssssssssssssscsssscessssssssssessecsssecenceseseeces 116
Fig. 4.10. Sabot for forward ballistics.............ccccccssssscsssssscsssssscsscssscsscsssssececsnencsscases 117
Fig. 4.11. Photograph from reverse ballistic testing of tungsten wire

Fig. 4.12.

Fig. 4.13

Fig. 4.14

Fig. 4.15.

Fig. 4.16.

reinforced metallic glass COMPOSItE.........s.ssssssessssessssssssessssesssssereseseeesel ZO
Photograph from reverse ballistic testing of tungsten wire -

reinforced metallic glass COMPOSItE.........c.ssssssssssssssssescsesecssestsesscssseceeess 120
Photograph of composite penetrator embedded in aluminum

target block. Shot in forward ballistic configuration at 605

M/S. L/D=6, P/LHAL.29. oe ecessssssssssssssscsscsncsscescsssssssssssssssesscesscuseuensceecaee 122
Photograph of composite penetrator embedded in aluminum

target block. Shot in forward ballistic configuration at 749

M/S. L/D=6, P/LHAL.21 oo ee esssssesssssssscsecscsscssssscssssssecseasssscseseaneenecuseess 122
Photograph of composite penetrator embedded in aluminum

target block. Shot in forward ballistic configuration at 1032

M/S, L/DH6. P/LH2.1 7. eecssssssessssssesssesseccsecssssssssscsssssscsscessssuseneecuceneesess 124
Photograph of composite penetrator embedded in aluminum

target block. Shot in forward ballistic configuration at 1257

M/S. L/D=6. P/LH2.33..... cc essecssssssscssscssssssssssssssssscsssssscssessecscsatersecaeesaeeass 124

Fig. 4.17. Photograph of WHA (W-Fe; ¢-Ni; 4) penetrator embedded in -

aluminum target block. Shot in forward ballistic configuration

At GOA M/S... sssssesssccsscesescssseeescscscesessessscssssssecesssesesesssesesssesesecsesesees

Fig. 4.18 Penetration efficiency versus impact velocity. Target was

seseeaes 125

aluminum (6061 T651). Penetrators used were flat-ended rods

with L/D=6. Composite was 80 vol% uniaxial tungsten wire in

a metallic glass matrix. WHA was the liquid-sintered tungsten

heavy alloy with composition in weight percent of

Wo3Nis gFe1.4. WHA experiments by S. Yadav...........cscscsseeeees

Fig. 4.19. Photograph of metallic glass matrix composite penetrator in

4130 steel target. V=760 m/s, L/D=8, and P/L = 0.312.............

Fig. 4.20. Photograph of metallic glass matrix composite penetrator in

4130 steel target. V=956 m/s, L/D=8, and P/L = 0.469. ............

Fig. 4.21. Photograph of metallic glass matrix composite penetrator in

4130 steel target. V=1253 m/s, L/D=8, and P/L = 0.812..........
Fig. 4.22. Photograph of WHA (X-27C) penetrator in 4130 steel target.
Fig. 4.23. Photograph of WHA (X-27C) penetrator in 4130 steel target.
V=932 m/s, L/D=8, and P/L = 0.875.......ssssscsscressscssssscssssssssscesseseces
Fig. 4.24. Photograph of WHA (X-27C) penetrator in 4130 steel target.
V=1269 m/s, L/D=8, and P/L = 0.784, wo... esssssessssesssessssessscssessees

Fig. 4.25. Penetration efficiency versus impact velocity. Target was

hardened 4130 steel. Penetrators used were flat-ended rods

seseeees 128

sesseees 128

sessenees 129

seeseees 130

seceesees 131

with L/D=8. Composite was 80 vol% uniaxial tungsten wire in

a metallic glass matrix. WHA was the X-27C liquid-sintered

tungsten heavy alloy with composition in weight percent of

Wo0.73Nig. 55F'€1.97C 09.75. cee ccccecccnceeccencscecesecceceeecseceececcoesccscescecccecooes

sesseeees 132

Fig. 4.26. Penetrator efficiency of metallic glass and WHA penetrators
versus aluminum targets. L/D values shown for the
PCNCCLALOL........ccsssssscscssscecessecsssessesessescesessssescessessessessessecsesosesscessseseeeessaes

Fig. 4.27. Scaled penetrator efficiency of metallic glass composite,

WHA, and depleted uranium penetrators versus steel targets.
Caltech data scaled down so that WHA data from both

sources match. WHA tests performed at Caltech with X-27C
alloy. WHA used by Magness was liquid-sintered 93%

tungsten. DU alloy is U-0.75 Tin... ecessssssssecssssssssesssessscssesssssessssssesees

Table 1.1.

Table 3.1.

Table 3.2.

Table 4.1.

Table 4.2.

LIST OF TABLES

Selected properties of metallic glass and other comparable
engineering Materials. ............sscccssssssscerssssccsssssssenssssssccsesessessesssessssessees 8
Atomic concentrations at different points of interfacial region

of tungsten wire reinforced metallic glass matrix composite.
Location of points is given in Fig. 3.12. Oxygen content was
excluded from analysis (See text). .......sssscsssssscsssresssssscesesseseesssssseeseses 68
Atomic percent concentrations at different points of

interfacial region of steel music wire reinforced metallic glass
matrix composite. Location of points is given in Fig. 3.13............. 71
Comparison of various physical properties of penetrator

MALETIAS. 0.0... ececsssssesecessesesssesecscsecssscsssesssessssesesssessseseseesesesescscscsessssseseees 105

Selected properties of the tungsten heavy alloy X-27C ..........cc0000 118

BCC
BF

DF
DSC
DU
EDS
EDXS
Em; E¢
ESL

Gic

NOTATION AND ABBREVIATIONS

body centered cubic

bright field

specific heat at constant pressure

dark field

differential scanning calorimeter

depleted uranium

energy dispersive spectroscopy

energy dispersive x-ray spectroscopy

elastic modulus of the matrix; fiber/wire

electrostatic levitator

volume fraction of fibers

fracture energy of the interface

Gibbs free energy of formation

free energy of activation for transfer of atoms from the
liquid to the crystal

difference in free energy between the crystal and the liquid

per unit volume

nucleation rate

thermal conductivity

length to diameter ratio

number of equilibrium number of crystalline clusters per
unit volume

number of unassociated molecules per unit volume

number of atoms on the surface of a critical nucleus

penetration depth to length ratio

Vitreloy ™1

WHA
X-27C

a q 3 ¢

radius of the fiber/wire

critical cooling rate

scanning Auger microscope

scanning electron microscopy

scanning transmission electron microscopy

surface area of a critical nucleus

absolute temperature

glass transition temperature

melting point

reduced glass transition temperature (T,/T)

transmission electron microscopy

velocity

metallic glass-forming alloy composition
(Zr41 2Ti13.gCu12.5Ni10.0Bez2.5)

tungsten heavy alloy

designation for tungsten heavy alloy (see Table 4.2)

net rate of transfer of atoms from the liquid to the embryo
per unit area

Poisson's ratio

viscosity

liquid-crystal interfacial energy per unit area

flow stress

strain

strain rate

CHAPTER 1
INTRODUCTION

1.1 BULK GLASS FORMING METALLIC ALLOYS
1.1.1 History of Metallic Glasses

Although glassy solids have been familiar in everyday life for centuries,
glassy metals have been virtually unknown. Silicate and other oxide-based
glasses are found in diverse applications, from window panes to fiber-optic
communication lines. By definition, an amorphous solid lacks long-range
atomic order. Amorphous solids can be ordered within small clusters, but
atoms in crystalline materials are arranged periodically over lengths much
larger than the atomic size. Nearly all metals in common use are composed of
grains, each of which is a region with crystalline ordering of its atoms.

Amorphous structures are rarely found in nature because there is always
a corresponding crystalline phase or mixture of phases with the same overall
composition and a lower free energy. That is, the amorphous state is
thermodynamically unstable or metastable. However, there can be kinetic
constraints so that the transformation to a crystal is very slow. The more
commonly found oxide-based glasses are extremely stable and can be in fact
difficult to crystallize. On the other hand, metals and metal alloys are, as a
whole, very poor glass formers and crystallize easily.

Qualitatively, the structure of non-crystalline, or amorphous, solids has
much in common with that of liquids. The primary difference is that the atoms
of liquids are free to move and redistribute themselves on time scales easily
measurable in the laboratory. Amorphous solids can still exhibit flow, but on
much longer time scales. Thus, the division between amorphous solids and

liquids is quite arbitrary, and is often taken to be at a viscosity of 10" poise.*

The most distinctive physical property that defines a glassy solid is a glass
transition. The glass transition can be detected by the rapid increase of the
heat capacity over a small range of temperature. A true second order phase
transition would have a discontinuity in heat capacity at a fixed temperature.
The glass transition does not meet this criteria since it occurs over a range of
temperature, and the onset temperature varies with experimental parameters
such as the heating rate as measured in a differential scanning calorimeter
(DSC). Because most amorphous metallic alloys also exhibit a glass
transition, for simplicity, "amorphous" and "glassy" will be used
interchangeably.

Despite their thermodynamic metastability, amorphous metallic alloys
can be made by a variety of different processes; techniques now exist which
can use gases, liquids, or even crystals as starting material. The first major
breakthrough in metallic glass formation came in 1960 when Klement, Willens,
and Duwez discovered that Au,;Si,, could be made amorphous by rapid cooling,
or quenching, from the liquid state.? The amorphous foil produced by their
technique was so unstable, though, that significant crystallization had
occurred after only 24 h at ambient temperature. In 1983, Schwarz and
Johnson observed the growth of an amorphous layer between two layers of
crystalline metals at elevated temperature.? Other techniques, such as vapor
deposition, electrodeposition, sputtering, plasma spray deposition, and ion
implantation, have also been developed to make thin layers or coatings of
amorphous metals.

This work will focus on glass-forming techniques which quench material
from the molten state. Most metals and metal alloys historically investigated

require cooling rates of 10? K/s or higher to freeze from the melt to the

metastable amorphous state. The high heat transfer rate required limits these
metallic glasses to thin samples produced by such techniques as splat
quenching or melt-spinning.” 4

The process of melt spinning has been widely used because of the relative
ease with which it can be used to make uniform continuous amorphous
ribbons. A schematic of the process is shown in Fig. 1.1. The basic melt-
spinning process involves a heated crucible of molten metal of the desired
composition. The crucible is usually heated with an induction coil. Applied
pressure at the top of the crucible forces a continuous stream of melt through
an orifice and onto a rapidly moving chilled substrate, such as a drum, wheel,
roller, belt, or between twin rollers.* The melt is drawn off and cooled
continuously to form the ribbon.

Fig. 1.2 shows a graph of the reduced glass transition temperature, Tyg,
plotted versus the log of the critical cooling rate, R,, for a range of traditional
glass forming alloys. Tyg is the glass transition temperature, Ty, divided by
the melting point, Tm; R, refers to the minimum cooling rate in degrees per
second required to cool to the glassy state. As the graph shows, as the reduced
glass transition temperature increases, the critical cooling rate required drops
precipitously. The critical cooling rates range from 102 K/s to 101° K/s, with
elemental metals like nickel and tellurium near the highest part of that range.
This graph includes the best glass formers when the article was published in
1976.

Bulk metallic glasses are not limited to these standard methods of
preparation and thin sample geometries. A "bulk" glass can be defined as one
which can be quenched from the melt into a specimen with minimum

dimension on the order

Inert gas ———>

Melt Induction coil

Substrate wheel

Ribbon

Front view Side view

Fig. 1.1. Schematic of melt spinning apparatus. Reproduced from

refs. 6 and 7.

10 — Ni
co)
8 Tee
“a 7 FeP
Sg 6 ® AuGeSi
2 6 eNiP
m ®pip
et @Ge
on @
S -L PdP
FePC ® @PtNiP
4 CuZr
Pdsie ® e
Z NiNb
PdCuSi®
2 =
| | | ] | ] | | |
0 0.2 0.4 0.6 0.8 1.0
T/T

Fig. 1.2. Critical cooling rate, Rc, versus the reduced glass transition
temperature, Tyg=T,/T mn, for different metals and metallic alloys. Reproduced
from ref. 5.

of 1 cm, thus having critical cooling rates of approximately 102 K/s or less.
Clearly, this definition varies with the quenching technique employed. Chen
and Turnbull found in 1969 that amorphous samples of such alloys as Pd-Au-
Si could be made up to 1 mm thick by simply dropping the melt onto a metal
substrate.® In 1982, Drehman, Greer, and Turnbull formed a spheroid of
glassy PdaoNigoP20 after etching the surface heterogeneities away and
thermally cycling the sample to dissolve any crystalline inclusions. The
observed cooling rate for a sample with minor diameter greater than 5 mm was
1.4 K/s; the conclusion was that the critical cooling rate for this specimen was
even less than that of Pd-Au-Si. The reported reduced glass transition
temperature of Pd4pNisgP 29 is 0.67.”

One of the most highly processible of the bulk metallic glasses is
Zr41.2Ti13.gCuy2.5Ni10,0Be22,5, trade name Vitreloy™ 1, which has a critical
cooling rate of about 1 K/s and a reduced glass transition temperature of 0.67
with no special fluxing procedure required.) 11 Rods of this alloy of up to5cm
in diameter have been cast.” In addition to being easily processed, this bulk
metallic glass has many superior mechanical properties and shows promise as
a new engineering material.!* Further research in the area of alloy
development for bulk metallic glasses continues and is yielding new alloy
systems and compositions.1® 1” Nevertheless, Zr41.2Ti13 gCuy sNi10,0Bezo5
remains one of the most robust glass-formers and the most easily processed
by quenching from the melt. In addition, its physical properties are perhaps
the most well characterized of any of the bulk metallic glass alloys. For these

reasons, the studies contained in this work exclusively use this alloy.

1.1.2 Scientific and Technological Interest in Metallic Glasses

The interest in metallic glasses has traditionally been driven by scientific
curiosity. Investigations of these materials have yielded information about
rapid solidification, the nature of the glass transition, metastable phase
equilibria, crystallization, and the effects of atomic symmetry on macroscopic
properties. Newly developed techniques for synthesizing metallic glass allowed
study of a class of material not found in nature.

The many unusual physical, chemical, and mechanical properties of
metallic glasses have led to their use in a number of specialized applications.
The most widely exploited property of amorphous metals to date is the soft
ferromagnetism of certain alloys. The Metglas™ alloys developed by Allied
Signal have been used extensively in such applications as transformer cores.
Mechanical applications have been rare because of limitation on sample
geometry and certain undesirable mechanical properties of amorphous metal
ribbons, such as low fatigue life.1®

Nevertheless, many superior properties of metallic glasses make it an
attractive candidate for mechanical applications. The most notable is the
extremely high strength. Another is the high elastic strain limit, which is
nearly three times that of most crystalline metals. The lack of grain
boundaries and the ease with which an amorphous passive oxide layer is
formed give many metallic glasses good corrosion resistance.

Since good glass forming alloys are usually found near deep eutectics, these
materials can be processed as a liquid at relatively low temperatures, which
can reduce manufacturing costs. Moreover, three factors reduce the amount

of shrinkage in castings. One, the lower casting temperature lowers the total

amount of thermal contraction upon cooling to room temperature. Two,

amorphous alloys usually have low thermal expansion coefficients, and thus

contract proportionally less per degree cooled. Three, there is no volume

contraction from the liquid-to-solid phase transition in metallic glasses

processed by solidification from the melt. These three properties work in

conjunction to yield near net shape castings.

Table 1.1 shows a number of selected properties of the amorphous

Vitreloy™ 1 in comparison with other metals and alloys commonly used for

high strength structural applications.

Vitreloy™ 1 | steel -1080 | tungsten | Ti-6AI-4V | maraging
quenched & cast steel-
tempered C200

Yield Strength | 1.9314 0.97919 0.890! =| 1.4/9
(GPa)

Elastic Strain | 2.0 0.5 0.8 0.7
Limit (%)

Young's 9314 21019 34579 | 11419
Modulus (GPa)

Expansion

(10-6 K-1)

Thermal 0.035 0.38119 =| 1.7420

Conductivity @673 K

(W cm-1K-1)

Melting Point | 9931° 175319 342270

(Tiquidus, K)

Table 1.1. Selected properties of metallic glass and other comparable

engineering materials.

There are a number of promising new applications for metallic glasses,
including sporting goods. Vitreloy™ 1 is currently being used to make golf club
heads under the name LiquidMetal™. The striking face of the club is designed
thinner to take advantage of the large elastic strain limit of the metallic glass.
The result is that proportionally more deformation caused by the collision is in
the club face rather than the golf ball. The elastic losses in the club are
substantially less than those in the ball, so more energy is returned as ball
velocity following the impact.

There are, nevertheless, a number of drawbacks to using metallic glass in
mechanical applications. One of the major problems is that there is no
significant plastic strain before failure. This limits the amount of energy that
the material can absorb and dissipate; also, there is no warning before failure.
This behavior is called "catastrophic" because of its sudden onset. Another
limitation is still the size of the samples produced as determined by the critical
cooling rate of the alloy. Although the minimum dimension of a casting which
can be made has increased dramatically, it still may be insufficient for many
applications. Other drawbacks can be the cost of starting materials,
potentially health hazards from constituents such as beryllium, sensitivity to

contamination, and lack of manufacturing experience with the alloys.

1.2 CRYSTALLIZATION OF METALLIC GLASSES

As mentioned in the previous section, metallic glass is a thermodynamically
metastable or unstable phase. Thus, given the correct conditions, a glass will
spontaneously crystallize. However, it is possible to kinetically constrain
metastable phases so that the time required for transformation is much longer
than is measurable in the laboratory. One excellent example of this

phenomenon is diamond. At room temperature and pressure, graphite is the

thermodynamically stable form of carbon, and diamond is metastable.

Nevertheless, as far as any new fiancée is concerned, "Diamonds are forever."

1.2.1 Homogeneous Nucleation Theory

One of the most important considerations in determining the feasibility of
making a metallic glass composite is that of crystal nucleation and growth.
During composite processing, impurity elements which are not constituents of
the glass forming alloy may dissolve into the melt, elements already present in
the glass may change in concentration, or crystalline debris might act as a
heterogeneous nucleation site.

Turnbull et al.2 2? developed classical nucleation theory to describe the
process of homogeneous nucleation, or nucleation events in the absence of
"structural impurities" to catalyze crystallization. Turnbull's work was an
extension of that done by Volmer and Webber for the condensation of a
vapor.”? We want to consider the Gibbs free energy of formation, AG, for a
small aggregate, or embryo, of crystalline material in a liquid. Although AG is
positive for small crystalline embryos, there is an entropy gain from having
them distributed in the liquid; thus there is some equilibrium concentration of
these embryos.

There are two terms for this energy AG. The first is simply due to the free
energy difference between the crystal and the liquid; this is dependent upon the
temperature. At temperatures above the melting point, the liquid will have a
lower free energy; at temperatures below the melting point, the crystal will
have a lower free energy, assuming it is the equilibrium phase. The second
term arises from the surface tension between the liquid and the crystal. The

surface tension can be thought of as an extra amount of energy needed to

accommodate the mismatch in the atomic arrangements between the two
phases. Thus, as the crystal embryo grows, the interfacial area grows and this

second term in the energy increases. We can write:

AG = <2 AG, +4nro (1.1)

where AG is the free energy of a crystal embryo of radius r, AGy is the
difference in free energy between the crystal and the liquid per unit volume, and
o is the liquid-crystal interfacial energy per unit area. Below the melting point,
AGy will be negative and the two terms will compete as r increases, as shown in
Fig. 1.3. At small r, the interfacial free energy term will dominate, but for large
enough r, the volume term becomes larger and the crystal will grow indefinitely.
At this point, the embryo becomes a nucleus. The radius separating these two
regimes, the critical radius, re, is the point at which the free energy AG reaches

amaximum. Differentiating Eqn. 1.1 and solving, we find a maximum in free

energy at:
=e (1.2)
3(AG,)
when per, =-22 (1.3)
AG,

These values represent the nucleation energy barrier and the critical nucleus
radius of the crystalline embryo, respectively.

The density of embryos of critical radius follows a Boltzmann distribution:
AG,

n,=ne (1.4)

where n is the number of equilibrium crystalline clusters per unit volume, ng is

Interfacial free energy “

‘ AG

Volume free energy}

Fig. 1.3. Gibbs free energy AG for crystalline embryo as a function of radius.

the number of unassociated molecules per unit volume, k is Boltzmann's
constant, and T is the absolute temperature. If we assume that when a
crystal grows to the critical nuclei size it is removed from the equilibrium
considerations, then the nucleation rate I is governed by the rate at
which smaller embryos can grow to the critical size. Then,

I= ZS,n, (1.5)
where Z is the net rate of transfer of atoms from the liquid to the embryo per
unit area, S, is the surface area of a critical nucleus, and n, is defined by Eqn.

1.4. This becomes:
(1.6)

I=K exp| ASG,

where AGza is free energy of activation for transfer of atoms from the liquid to
the crystal. Ky, is given by:
K,=n* [°% mer) (1.7)

where n* is the number of atoms on the surface of a critical nucleus, ais a
_constant depending on the shape of the nucleus, n is the number of atoms per

unit volume of the liquid, and h is Planck's constant.

Turnbull also did an analysis of glass-forming materials, and made the
further assumptions that the average jump time of the molecules in the
interfacial region is proportional to the viscosity, and that there is no difference

in the heat capacity between the liquid and the crystal. This gives:

k ba’B
[=—t —_-—__ .
n os ee 18)

where ky is a constant, 7 is the viscosity, b is a shape constant, T,=T/Typ,

AT,=(Tm-T)/Tm, and a and B are dimensionless constants defined by:

N*V40
a= ms

L (1.9)

_L
RT,,

and | B (1.10)

N is Avogadro's number, L is the molar latent heat of melting, Vm is the molar
volume of the crystal, and R is the molar gas constant. The viscosity can be

modeled by the Vogel-Fulcher relation:

n= Aens|

where A, B, and To are constants.

Eqn. 1.8 shows the two competing forces to homogeneously nucleate
crystals in an undercooled melt. When undercooling is small, there is little
thermodynamic driving force to crystallize. AT is small, and the exponential
term keeps the nucleation rate low. As the undercooling increases, the
nucleation rate increases until the increasing viscosity begins to dominate.
Physically, the mobility of atoms is drastically reduced, making crystallization
more difficult. Thus, it is over the range in temperature below the melting
point and above the glass transition temperature that crystallization of a
glass-former is possible. This is frequently stated in terms of the reduced glass
transition temperature Tyg, which is T,/T;,. Turnbull notes that there is a
strong dependence upon Tyg for glass-forming ability; as Trg increases to
approach 2/3, it becomes quite easy to cool the material to a glass, provided

that only homogeneous nucleation is considered.

1.2.2 Heterogeneous Nucleation Theory
In practice, very rarely is a melt totally homogeneous; there can be free
surfaces, internal surfaces, container walls, or suspended crystals which can

catalyze crystallization by acting as heterogeneous nucleation sites. It is

extremely difficult to eliminate all of these effects in an experiment, so
heterogeneous nucleation often dominates homogeneous nucleation in physical
experiments.

Heterogeneous nucleation can be thought of as a crystalline embryo in a
supercooled liquid growing on an impurity crystal. The shape of the embryo will
approximate a sphere in order to minimize the surface area per unit volume
between the embryo and the liquid; it can take the form of a spherical cap.

This is shown schematically in Fig. 1.4. Assuming that the interface is stable
and in equilibrium, the forces from the surface tensions balance where the

three interfaces meet:22

Os, = Osc + Op COS O (1.12)
where C refers to the crystal embryo, L to the liquid, S to the catalytic
substrate, and 0 is the angle of contact between the crystalline embryo and
the substrate, also known as the wetting angle.

Clearly, there are only certain conditions for which Eqn. 1.12 is satisfied.
In particular, since -1 > cos@ < 1, the parameter m is defined as :
Os, — Osc

m= (1.13)
Orc

For a stable interface, the condition -1 contact occurs; it is a lower energy state for the crystalline embryo to be
completely within the liquid. This then degenerates to the case of homogeneous
nucleation. If m > 1, then there is complete contact; there is a driving force for
the crystal to spread completely across the surface of the substrate. A
spherical cap forms for wetting angle between 0° and 90°; that is, O

For a spherical cap, the volume is given by (see Fig. 1.5)":

Crystal embryo @

Fig. 1.4. Schematic of a spherical-cap-shaped crystalline embryo
forming in a supercooled liquid on a catalytic substrate.

Reproduced from ref. 24.

Fig. 1.5. Spherical cap geometry.

V. =%nh?(3r-h)
= Kn(r—rcos 6) [3r—(r—rcos 6)] (1.14)
= %n(1-cos@)[2 +cos 6]r?
and the surface area representing the interface between the liquid and the
embryo is:
A, = 2m(1—cos 6)r? (1.15)
The Gibbs free energy change associated with the formation of a spherical cap-

shaped crystalline embryo on a substrate is given by:

AG,

hetero

=2n(1-cos 6)r’o,, + m(rsin 0) (Osc — Os) (1.16)
+4(1—cos6)’(2 + cos 6)r°AG,

The first term represents the new interfacial energy between the liquid and the

crystalline embryo, the second term is the energy associated with replacing the

interface between substrate and liquid by one between substrate and crystal,

and the third term is the energy of converting the volume of liquid to crystal.

This can be simplified using Eqn. 1.12 and our result from homogeneous

nucleation theory (Eqn. 1.1):

_ (1—cos6)"(2 +.cos 6)
AG petero ~ 4 AG (1. 17 )

= f(@)-AG

recalling that AG is the energy change for an embryo according to
homogeneous nucleation theory. Differentiating with respect to r, we find that
AGhetero has a maximum at the same point as AG. Thus, the radius

of a critical nucleus is the same for heterogeneous and for homogeneous

nucleation, but the energy of formation is reduced by the factor f(6).

Referring back to Eqn. 1.6 and the theory of homogeneous nucleation, we
see that since this factor of f(6) is between 0 and 1, it increases the rate of

heterogeneous nucleation as given by:

= K, exo|- a exa|- ie | (1.18)

hetero

1.8 COMPOSITE THEORY AND TRADITIONAL COMPOSITE
SYSTEMS
1.3.1 Classification and Types of Composites

A composite can be defined as having a heterogeneous structure composed
of two or more distinct components which are bonded together to achieve a
definite goal for a specific purpose.”° Composite materials have been in
frequent use for many centuries and often provide better properties than any
of the constituent materials would alone. Brick made from mud and straw,
plywood, and reinforced concrete are all composites engineered to give certain
properties. Extensive research continues to give new types of composites,
better processing techniques for fabrication, and improved characterization of
existing materials.

Although certain high-performance composites can have many superior
properties, no single material will have all of the best properties. Some of the
most common properties desired are strength, fracture toughness, ductility,
fatigue resistance, oxidation resistance, and low cost. There are inevitable
trade-offs, so the material must be chosen to have the best characteristics for

the most important properties, and minimal degradation of the others.

One way to categorize different high performance composites is based on
the matrix material used. There are three basic matrices: polymer, ceramic,
and metal; reinforcements are used from all three of these classes as well.
Composite materials can also be divided into five different groups depending
upon the geometry of the reinforcement phase: particles, fibers, flakes,
skeletal, and laminar. Particulate reinforced composites are usually reinforced
with a discontinuous, roughly spherical second phase giving an isotropic final
material. Fibers can vary widely in size and aspect ratio, in addition to being
either randomly oriented or aligned. Flakes would be considered discontinuous
reinforcements with one dimension significantly less than the other two.
Skeletal composites are composed of a continuous porous matrix which is then
filled by a second material. Since both the starting material and the filling
material are continuous, it is unclear which to consider the matrix and which
the reinforcement. Layers of different materials can be stacked one on top of
another to make a laminar composite.”°

Combinations of these different reinforcement geometries are possible as
well. Fiber reinforced composites can use short fibers to yield an isotropic
material, aligned short fibers, uniaxially aligned continuous fibers, or fibers
woven in two dimensions to give sheets, which can then be stacked and made
into a laminar composite. No one configuration is ideal for all applications; the
requirements of a given task must be considered in determining both the
materials and the reinforcement geometry to use.

Techniques of fabrication of composites can also vary widely. For metal
matrix composites, three basic techniques exist: powder metallurgy, ingot
metallurgy, and mechanical alloying. Powder metallurgy begins with the

matrix in powder form; it is then incorporated with the reinforcement, and then

consolidated in some fashion at elevated temperature and pressure. The
consolidated composite can then be shaped by either extrusion or forging.

Ingot metallurgy involves mixing the reinforcement with the molten matrix;
this mixture can then be cast directly. Also, vacuum infiltration can be used to
make metal-matrix composites. In this technique, the reinforcing fibers are
placed in a die, a vacuum is applied to one end, and molten matrix is supplied at
the other. The vacuum draws the matrix into the mold and infiltrates the
reinforcement simultaneously. Finally, the mechanical alloying technique can
be used to form an alloy matrix with discontinuous reinforcement starting with
the elemental metal powders and the reinforcement. These are all
mechanically alloyed in an apparatus such as a high energy ball mill. The
powder from this procedure can then be consolidated in a similar fashion as in
the powder metallurgy approach.?’

There are also numerous types of processing techniques for other types of
composites. The most pertinent one to this study is injection molding, which is
primarily used with thermoplastic matrix materials. In this technique, the
polymer is heated, the softened polymer is forced into a cooled mold by a high
pressure ram, and the part is allowed to cool and ejected from the mold. Die
casting is a similar process used for making metal parts. Typically, since
metal is involved, a higher temperature is required for die casting, and die

casting is rarely used for processing composites.

1.3.2 Role of the Composite Interface
The work presented in this thesis will focus primarily on reinforcement with

continuous uniaxially aligned fiber and wire reinforcements. Thus the

background discussion will focus on these types of composites in traditional
systems.

There are a number of advantages given to a composite material by the
reinforcement by fibers. The properties of interest are usually the ones which
the matrix material is most clearly lacking for a given application. For
example, graphite-fiber-reinforced epoxy is a commonly used composite
system because the graphite fibers increase the epoxy's poor strength and
modulus. Conversely, ceramic and other brittle matrix composites are often
designed to improve other properties such as fracture toughness and strain-to-
failure of the matrix, since strength and modulus are generally properties
which need little improvement.

As discussed previously, toughness and strain-to-failure are properties
which would be beneficial to improve in metallic glasses; in brittle matrix
composite systems, these properties can be imparted by crack bridging by the
fibers. An essential component of this mechanism is control of the interface
properties. When a composite is loaded and reaches a stress at which a crack
begins to propagate through the matrix, the interface between the matrix and
the fiber can serve to lower the stress intensity experienced by the fiber and
allow the matrix crack to bypass it. Instead of the matrix crack continuing
directly through the fiber, a debonding crack is propagated along the interface.
(See Fig. 1.6.) The condition for this to occur has been calculated to be that the
ratio of interface fracture energy to fiber fracture energy is less than ~1/4.”8 If
this condition is not met, the interface can still transfer enough stress to crack
the fibers, rather than fail itself.

The toughening effect arises from the pull-out of fibers from the matrix.

Most theories of composite toughening include an initial debonding shear stress

Debonded interface with
frictional sliding Bonded fiber

4 |
a |
a L“] a
hee +] d Lt a :
4 a d 5
1g yp 4 ar
E a M a a :
c a d 5 : >
: I [ we
1 AN .
IN Crack tip

Fiber break and frictional pullout

Fig. 1.6. Schematic of crack bridging and fiber pull out in a brittle-matrix
brittle-fiber composite. Reproduced from ref. 30.

as well as a frictional sliding stress for the debonded fibers. Most of the energy
absorption comes from the work done in opposing this frictional force. In
addition to this mechanism, toughening can be accomplished by the work done
of plastic deformation of ductile fibers or wires. (See Fig. 1.7.)

The technique of fabrication of the composite is important in determining its
final properties. One reason for this is that processing can drastically affect
the interface between the matrix and the reinforcement. The desired
properties of the interface depend on the type of composite system; systems
can be classified as having either a brittle or ductile matrix. In ductile matrix
systems, fiber failure precedes matrix failure, and in a brittle matrix system,
matrix cracking occurs before or simultaneously with fiber failure. Similarly,
the interface between matrix and fiber can be characterized as either "strong"
or "weak."29 As discussed above, the interface must be sufficiently weak to
blunt the crack and encourage fiber pull out, but strong enough to transfer the

load effectively from the matrix to the fibers.

1.4 METALLIC GLASS AND COMPOSITES

Because of its high strength and high strain limit, metallic glass is a natural
candidate for use as a reinforcement in composite materials. There have been
a number of attempts to produce composites with amorphous ribbon
reinforcements. Strife and Prewo*! fabricated and mechanically tested an
amorphous-metal-ribbon-reinforced resin-matrix composite. It was found to
have high strength, both longitudinally and transversely, and good fracture
toughness. However, it was not competitive in specific modulus and fatigue
resistance with other resin-matrix composites. As a result, they concluded it

would not be suitable to replace most other high performance composites.

Debonded interface Bonded fiber
\ |

Crack bridging-
plastic deformation

Matrix

Fig. 1.7. Schematic of toughening mechanisms in a brittle-matrix ductile-
fiber composite. d is the debond length, and R is the radius of the fiber.
Reproduced from ref. 29.

Nevertheless, a number of other studies have investigated the properties of

polymer and glass ceramic matrices reinforced by amorphous metal ribbons.3*

There are a number of clear processing advantages in making a composite
material with a metallic glass matrix. One is that since the metallic glass-
forming alloys usually are at deep eutectic compositions®®, they have low
melting points, considerably lower that those of their constitutive elements.
Accordingly, these alloys can be used in liquid-phase processing at lower
temperatures; this lowers the cost of processing, decreases the need for
specialized equipment, and reduces the interfacial reaction rate between
matrix and reinforcement. Further, there is more potential flexibility in
processing since many of the good glass forming alloys are stable against
crystallization above their glass transition temperature as an undercooled
liquid. This characteristic allows for the possibility of processing at even lower
temperatures slightly above the glass transition, although higher pressures
would be required due to the higher viscosity. Also, upon cooling from a liquid to
a glass, good glass forming alloys have minimal shrinkage, as previously
discussed. This results in a composite with near net shape, as well as lower
differential thermal stresses between the metallic glass and reinforcement.

The past work performed on metallic glass matrix composites is reviewed in
the beginning of Chapter 2, which describes the technique developed for

processing a bulk metallic glass matrix composite.

10.

11.

12.

13.

REFERENCES

. T.R. Anantharaman, in Metallic Glasses: Production, Properties, and
Applications (edited by T. R. Anantharaman), Transtec, Switzerland
(1984).

. W. Klement IV, R. H. Willens and P. Duwez, Nature 187, 869-870 (1960).
. R.B. Schwarz and W. L. Johnson, Phys. Rev. Lett 51, 5, 415-418 (1988).

. H.H. Liebermann, in Amorphous Metallic Alloys (edited by F. E.
Luborsky), 26-41, Butterworths, London (1983).

. H.A. Davies, Phys. Chem. Glasses 17, 5, 159-173 (1976).

. X. Lin, Bulk Glass Formation and Crystallization of Zr-Ti Based Alloys,
Ph.D. thesis, California Institute of Technology (1997).

. R.W. Cahn, in Physical Metallurgy (edited by R. W. Cahn and Haasen), ,
1779, North-Holland Physics Publishing, Amsterdam (1983).

. H.S. Chen and D. Turnbull, Acta Metall. 17, 1021 (1969).

A. J. Drehman, A. L. Greer and D. Turnbull, Appl. Phys. Lett. 41, 8, 716-
717 (1982).

A. Peker and W. L. Johnson, Appl. Phys. Lett. 63, 17, 2342-2344 (1998).

Y. J. Kim, R. Busch, W. L. Johnson, A. J. Rulison and W. K. Rhim, Appl.
Phys. Lett. 68, 8, 1057-1059 (1996).

W. L. Johnson, Curr. Opinion in Solid State & Mater. Sci. 1, 3, 383-386
(1996).

W. L. Johnson, Materials Science Forum Proc. ISMANAM-95 (edited by
Robert Schultz), 225-227, 35-50, Transtec, Switzerland (1996).

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

28
H. A. Bruck, T. Christman, A. J. Rosakis and W. L. Johnson, Scripta
Metall. Mater. 30, 4, 429-434 (1994).

H. A. Bruck, A. J. Rosakis and W. L. Johnson, J. Mater. Res. 11, 2, 503-
511 (1996).

T. Zhang, A. Inoue and T. Masumoto, Mater. Trans. JIM 32, 11, 1005
(1991).

X. Lin and W. L. Johnson, J. Appl. Phys. 78, 11, 6514-6519 (1995).

D. Raskin and C. H. Smith, in Amorphous Metallic Alloys (edited by F. E.
Luborsky), 381-400, Butterworths, London (1983).

Metals Handbook, American Society for Metals, Metals Park, OH (1985).
Handbook of Chemistry and Physics, CRC Press, Boca Raton (1991).

D. Turnbull, Solid State Physics (edited by F. Seitz and D. Turnbull), 226-
306, Academic Press, New York (1956).

J. H. Hollomon and D. Turnbull, Prog Metal Phys 4, 333-388 (1953).
M. Volmer and A. Webber, Z Phys Chem 119, 277 (1925).
B. Chalmers, Principles of Solidification, Wiley & Sons, New York (1964).

K. G. Kreider, Composite Materials: 4 , 1, Academic Press, New York,
(1974).

Composite Materials Handbook, McGraw-Hill, New York (1984).

T. S. Srivatsan, T.S. Sudarshan and E. J. Lavernia, Prog. Mater. Sci. 39,
317-409 (1995).

M. Y. He and J. W. Hutchinson, Int. J. Solids Struct. 25, 1053 (1989).

29.

30.

31.

32.

33.

34.

35.

36.

A. G. Evans, Mater. Sci. Eng. A 148, 63-76 (1991).
J. W. Hutchinson and H. M. Jensen, Mech. mater. 9, 139-163 (1990).
J. R. Strife and K. M. Prewo, J. Mater. Sci. 17, 359-368 (1982).

R. U. Vaidya and K. N. Subramanian, J Mater Sci L 9, 12, 1397-1399
(1990).

R. U. Vaidya, C. Norris and K. N. Subramanian, J. Mater. Sci. 27, 18,
4957-4960 (1992).

R. U. Vaidya and K. N. Subramanian, J. Mater. Sci. 26, 23, 6453-6457
(1991).

R. U. Vaidya, K. K. Chawla and K. N. Subramanian, J. Mater. Sci. 29, 7,
1719-1723 (1994).

H. A. Davies, in Amorphous Metallic Alloys (edited by F. E. Luborsky), 8-
25, Butterworths, London (1983).

CHAPTER 2
PROCESSING A BULK-METALLIC-GLASS MATRIX
COMPOSITE BY MELT INFILTRATION CASTING

2.1 TRADITIONAL PROCESSING OF METAL MATRIX COMPOSITES

A number of techniques are currently used for the fabrication of metal
matrix composites. The type of matrix, reinforcement, and target application
will all influence the technique for processing the material. Reinforcement size
and morphology are influential parameters in determining which fabrication
processes are possible and practical. Processing must consolidate and bond
the reinforcement and matrix in the desired configuration without excessive
reaction between the two, which can degrade the overall properties of the
composite.

Fig. 2.1 shows a number of different processing techniques for the
fabrication of metal-matrix composites. Processing can be divided up into two
stages: primary, for incorporation of the reinforcement into the matrix, and
secondary, which may be required for consolidation, shaping, or fiber alignment.
Some primary processing techniques may provide satisfactory properties upon
incorporation and not require secondary processing.

Processing techniques which are most applicable to the composite system
studied in this work will be emphasized. For this reason, traditional secondary

processing techniques will not be examined in depth.

2.1.1 Liquid-Matrix Primary Processing
There are a number of techniques for composite fabrication in which the

reinforcement is initially incorporated with a molten matrix. This type of

Processing of Metal Matrix

Composites
Primary
Liquid (melt) Non-Liquid Reactive
Processing (solid) Processing Processing
¢ Incorporation
Solidification
- Die Casting
- wet or dry
-S - followed by secondary . .
Casting! Infiltration compaction stage Exothermic Reaction

- Melt Atomization (Osprey
Deposition Process)
Stacked Laminates

- Th aS)
ermas 2Pray - primarily used with foils

- bonded during
¢ Incorporation consolidation
before forming
Physical Vapor
- Slurry casting Deposition (PVD)

- Semi-solid (rheocasting
and compocasting)

Secondary

- extrusion / drawing

- forging

- rolling

- Hot Isostatic Pressing (HIPing)

- High energy, high rate processing

- Superplastic forming

- dynamic powder compaction
(shock wave consolidation)

Fig. 2.1. Different processing techniques for fabricating metal matrix

composites.

processing is advantageous because this usually results in a strong bond and
intimate contact between matrix and reinforcement. Its primary drawback is
that the rate of chemical reaction at temperatures required to keep the matrix
molten is high; excessive reaction can lead to a brittle interfacial layer,
degradation of the reinforcement, and contamination of the matrix. Due to the
high reaction rates, this type of processing often requires that infiltration be
accomplished in relatively short times and at high pressures.

One example of liquid-matrix primary processing is casting, which involves
forcing molten material into a mold with back pressure, often supplied by a
hydraulic ram. During squeeze casting, infiltration is usually done more slowly
and at higher pressures than die casting, and typically the ram continues to
move during solidification. This continuing movement during solidification
refines the microstructure by deforming the growing dendritic array and
compensates for the freezing contraction by continuing to feed in new
material.1 This technique can be used to make metal matrix composites either
by casting a mixture of molten matrix and reinforcement, or by forcing the
matrix into a reinforcement array, such as a fiber preform. This technique has
the capability to yield low porosity and near net shape pieces; there is usually
little need for any secondary processing. Die casting can provide a smaller
grained microstructure due to higher cooling rates, but there can be problems
with porosity and achieving complete infiltration.

Melt atomization creates bulk material by blowing drops of molten metal
onto a substrate. The metal is melted in a furnace, released in a stream, and
atomized by a jet of cold gas. This process was developed in the late 1970s and
1980s by Osprey Ltd., and has become know as the Osprey process. Feed

rates of molten material in the range of kilograms per second are possible with

this process. This can be used to make metal matrix composites by
introducing a stream of reinforcement particles along with the atomized metal.
This allows minimal contact time and interfacial reaction between the melt
and the reinforcement. It is an economical process because it makes little
waste material, and use of an inert gas jet results in minimal oxide
contamination. One drawback of this process is that in practice the volume
fraction of reinforcement is limited to about 25 percent. Other disadvantages
of this technique are frequent inhomogeneities in reinforcement distribution
and significant porosity. Further consolidation through secondary processing is
usually required.) 2

Thermal spray techniques are similar to melt atomization because they
also use flowing gas to project molten drops of material towards a cool
substrate. However, the thermal spray techniques melt the material by
introducing it, either in powder or wire form, into a combustion flame or plasma
arc. Like the Osprey process, thermal spray deposition can be used to make
metal matrix composites by introducing a stream of reinforcement particles
along with the matrix material. This family of processes usually results in
lower deposition rates than melt atomization, on the order of grams per second.
Drop velocities are usually higher, so quench rates can be 104 degrees per
second, fast enough to be considered rapid solidification. In fact, this technique
is very similar to the gun quenching technique used by Duwez to rapidly quench
Au7s5Sig5 and make the first metallic glass.? In addition, porosities from
samples made with this technique can be below 1%.

Composites can also be fabricated using techniques which mix the

reinforcement together with the molten matrix prior to forming. Slurry casting

is perhaps the simplest method of making composites; the reinforcement is

added to the molten matrix, and the mixture is cast and allowed to cool. The
advantage of this technique is mainly its simplicity: no specialized equipment is
required, and it can be easily adapted to continuous processing. Also, there is
sufficient contact time for good wetting to occur. There are three primary
problems with this technique: difficulty in casting, inhomogeneities, and
excessive interfacial reactions. When solid particles or short fibers are added
to the melt, the viscosity increases dramatically, and thus flow of the mixture
becomes restricted. This can be ameliorated by agitating or stirring the melt
to keep the particles in suspension. Semi-solid casting involves similar
processing, but the melt is allowed to partially solidify during agitation. This
process, known as rheocasting, encourages formation of spherical crystals
rather than dendrites in the matrix, and keeps the viscosity low. The stirring
_also prevents settling and agglomeration of the particles. With the
introduction of ceramic particles to the melt, the process has been called

compocasting.) 2

2.1.2 Solid-Matrix Primary Processing

There are also a number of techniques suitable for making metal matrix
composites by avoiding having the matrix in the liquid state. The most
common form of solid primary processing is mixing the reinforcement together
with the matrix in powder form. With this technique, the secondary processing
to compact and bond the powder together is essential. The powder can be dry
or suspended in a fluid. When the powder is a suspension, the compaction
steps also serve to evaporate and remove the liquid carrier. This process is a
good one for making composites with all ranges of matrix to reinforcement

ratios. It can be difficult to achieve homogeneous reinforcement distribution,

though, particularly with short fibers or whiskers, which tend to clump and
tangle.

The initial form of the matrix can also be thin foils instead of powders.
Stacking these foils of matrix material alternately with fibers or another form
of reinforcement is the primary step; then, compacting this layered structure
at elevated temperature allows plastic flow of the foil material around the
reinforcement. The different components are joined by diffusion bonding, since
the temperatures used are generally below the melting point of the matrix.
This is a good process for making continuous fiber composites, and interfacial
reactions can be better controlled because of the lower temperatures than
would be required for liquid-phase processing. Still, the stacking process can be
cumbersome and slow; it is also difficult to make samples with high
reinforcement fractions and good fiber distribution. Also, only certain metals,
such as titanium, which can dissolve their own oxides at processing
temperatures are well-suited to diffusion bonding. Otherwise, the native oxide
layer can create problems in joining the foil layers together.!

Physical vapor deposition (PVD) can also be used to join reinforcement and
matrix, and is neither truly liquid nor solid matrix processing. This is usually
done with monofilament reinforcements; the matrix material is deposited, often
by evaporation, directly onto a single wire. The wire is then bundled up and
suitably arranged, and compacted with a secondary technique such as hot
isostatic pressing. This technique provides good control of fiber volume fraction
through control of the layer thickness, and allows a wide variety of matrix

compositions to be used.?

2.1.3 Reactive Primary Processing

Composites need not be made as two distinct phases mixed together; they
can also be made by inducing some sort of reaction, driven by either chemical
or mechanical energy, in a monolithic material to form multiple phases.
Directionally solidified eutectics have a lamellar-type structure of two different
phases. Materials can be engineered so that at some elevated temperature, a
spontaneous reaction occurs which produces a second phase such as a ceramic

or an oxide which acts as a reinforcement.

2.2 PROCESSING OF METALLIC GLASS COMPOSITES BY RAPID
QUENCHING

Metal-matrix composites reinforced by metallic glass have also been
investigated, as reported by Cytron.* The investigator vacuum hot-pressed
amorphous NigoNba4o ribbons sandwiched between two disks of superplastic
aluminum. The materials were chosen so that the aluminum disks could be
heated to their superplastic zone below the crystallization temperature of the
amorphous ribbons. Good bonding was achieved between the reinforcement
and the matrix, and the possibility of using such a technique for making this
type of composite was supported.

Amorphous metallic alloys have also been frequently used as a matrix for
particulate reinforced composites, and to a lesser extent for fiber reinforced
composites as well. Usually the melt spinning technique is used to make
composites in ribbon form. This was the subject of a number of U.S. patents
filed in the early 1980s. Narasimham of Allied Corporation patented the
technique of incorporating particulate matter into the melt, and then

quenching the mixture by melt spinning.> Kimura, Cunningham, and Ast first

reported mechanical tests and microstructural characterization of such a
composite.® They added tungsten carbide particles to NizgSijoBip during the
melt-spinning process. Ast later patented the technique of blowing particles
into the melt between the crucible and the chilled roller, claiming that this
improvement allowed better particle distribution while minimizing contact time
and chemical reaction between the particles and the melt.” This technique has
also been used to make amorphous metal ribbons reinforced with
discontinuous fibers, although the resulting fiber distribution was poor; most of
the fibers were incorporated on the ribbon surfaces, probably due to poor
wetting between the fiber and the melt.8

In addition to being used to fabricate particulate reinforced metallic glass
ribbons, melt-spinning process has been used to make continuous lengths of
fiber reinforced composites, as described in a patent filed in 1972 by Williford
and Pilger.? The inventors described introducing an arrangement of fibers into
contact with the meniscus of molten metal protruding from the crucible orifice.
The fibers were then drawn out along with the melt onto the moving substrate;
the combination was chilled to form the composite. Composites of this type
were made successfully, but the best results were ribbons reinforced by one or
two tungsten wires.® 1° A schematic of the apparatus used by Nussbaum and
Ast is shown in Fig. 2.2. Two pulleys are used, one to feed the wire to the melt-
spinning apparatus, the other to take up the composite at the other end. The
feed pulley (the "Rear pulley" in Fig. 2.2) was also fitted with aerodynamic drag
vanes, which were used in conjunction with a nitrogen jet to regulate the
tension. The front pulley was controlled by a variable speed electric motor

taken from a Dremel tool.

Inert gas ———>

Induction coil

Melt a

Wire in
ribbon

o _, /
—_»>

Feed wire

Inert gas
—————

Rear pulley

OOO

——_

Front pulley

Guide tube

Substrate wheel

Fig. 2.2. Schematic of modified melt-spinner to make continuous-wire

reinforced amorphous ribbons. Reproduced from ref. 10.

2.3 EXPERIMENTAL

This section describes a technique for successfully using a bulk glass-
forming alloy as a matrix material for composites. We used exclusively the
alloy Zr41.2Tiz3,gCui2.5Nii0,0Beo2.5 (trade name Vitreloy™ 1) developed by
Peker and Johnson"! as the matrix material. The low critical cooling rate of
this alloy allows amorphous metal matrix composites to be made in bulk form.
First, ingots of this alloy were prepared by melting together the constitutive
elements in an induction furnace under a titanium-gettered argon atmosphere.
The starting metals were high-purity (99.5% metals basis or better) research
grade material. The metals were alloyed on a copper "boat," which is a 2 cm
diameter copper tube with large indentations on one side. The sample sits atop
the tube in the indentations and can then be inserted into a horizontal
induction coil. The copper boat is cooled with a constant water flow inside the
tube to prevent it from alloying with the sample. Because the cooling rate
provided by contact with the copper boat is sufficiently high, ingots produced
by this technique are mostly amorphous.

Composites were fabricated with a wide variety of reinforcements using
the technique described below. Continuous ceramic fibers, such as silicon
carbide and carbon, and continuous metal wires, such as tungsten, carbon
steel, stainless steel, molybdenum, tantalum, nickel, copper, and titanium,
were used as reinforcement. Particulate reinforced composites were made with
both loose tungsten particles and sintered silicon carbide (SiC) particulate
preforms. The preform was provided by the Carborundum Company.

The reinforcements used in the majority of samples were 254 um diameter
tungsten wire or 254 um diameter high-carbon (1080) steel music wire. The

tungsten wire was obtained from Thermionics Products Company (North

Plainfield, NJ). Both types of wire were straightened and cut into 5 cm lengths.
To avoid clustering of wires and to ensure good distribution, the wires were
bowed slightly in samples with lower fiber fractions. The tungsten wire was
degreased by ultrasonic cleaning in a bath of acetone followed by the same
procedure in a bath of ethanol. The steel wire was cleaned in a solution of 50%
phosphoric acid and 50% water at room temperature for 3 min. This solution
removes the oxide layer and deposits a thin phosphate layer.

Composite specimens were cast in the apparatus shown by the schematic
in Fig. 2.3. The reinforcement material was placed in the sealed end of a 7 mm
inner diameter quartz-glass tube. The tube was necked about 1 cm above the
reinforcement, and then ingots of the matrix material were placed in the tube
above the neck. The constriction minimizes premature contact and thus
excessive reaction between the melt and the reinforcement. The open end of
the quartz tube was clamped to a flexible hose connected to a three-way
switching valve; the tube could thus be evacuated with a roughing pump or
pressurized with argon. Prior to heating, the tube was evacuated and then
flushed with argon gas. This cycle was repeated several times to remove any
residual oxygen. The tube was left under vacuum on the last cycle to minimize
trapped gas in the composite sample to be formed. An additional processing
step was required for casting the steel wire composites. The wires were held at
973 K for 2 h under vacuum to remove any hydrogen absorbed by the steel
during the acid etch; this step will hereafter be referred to as the wire bakeout.

The sample tube was heated in a resistive tube furnace with temperature
feedback control. The initial heating stage was at 1228 K + 20 K, well above
the liquidus temperature (993 K)1! of the glass-forming alloy. This initial

heating stage dissolves residual oxides and other impurity phases which

quartz tube
4%
clamp — Wy ingot of
matrix material
three-way 4 fe) o7
switching valve
O |] O
O O reinforcement
O O-T bundle
to to O | | O
vacuum argon O O
pump supply O O
O O

furnace

Fig. 2.3. Schematic diagram of apparatus for casting bulk metallic glass

matrix composites.

degrade the glass forming ability of the alloy.12 The error in temperature
comes primarily from variations in the temperature profile in the furnace. The
sample was held at this temperature for 15 min. The temperature was then
lowered to 1078 K+ 5 K and allowed to stabilize. When the furnace reached
this target temperature, a positive pressure of 207 kPa of argon gas was
applied above the melt. These conditions were held for 30 min to allow
infiltration of the molten matrix material into the reinforcement. Then the
sample was quickly removed from the furnace and quenched in brine (8 wt%
NaCl/H20 solution). In an attempt to improve the mechanical properties of
the steel, some of the steel wire reinforced samples were tempered at 588 K for
2 h following infiltration and quenching. A graph of the time, temperature, and
pressure during processing is shown in Fig. 2.4.

Porosity was estimated by two methods. The first finds the apparent
porosity by analysis of a micrograph of a cross-sectioned sample. The bulk
sample porosity can then be extrapolated by assuming homogeneity
throughout the sample. This technique has been used for other materials as
described in an ASTM standard.18

The second method of determining porosity combines analysis of the cross
section with Archimedes' principle. As a consequence of the processing
procedure, the volume fraction of reinforcement cannot be inferred from the
relative amounts of the starting materials. There is a substantial reaction
layer between the as-processed composite and the quartz tube, as well asa
layer of quartz which strongly adheres to this reaction layer. The quartz layer
and the reaction layer must be ground off mechanically; unavoidably, some of
the reinforcement and matrix gets removed as well. As a result, we cannot

make a precise determination of the volume fractions from the amounts of the

J preheat h
900- quenc
1 infiltrate \
800- T ;
S) jm ~
2 700- g
S 600-
2 GOCCSCReeeeeeaeeoeseees Geeeeseecssfacee ears wn
jrfc cic opoccc rocco -4
300- zi 3:
Paceossocs OP Devecuecns™ Seseceseeee “RHeeenceehaeseascanecoeeceaCeeUeeoRee 0
200 TT. TTT T TT |
-40 -30 -20- -10 0 10 20 30 40 50 60
Time (min)

Fig. 2.4. Graph of time, temperature, and pressure during processing of the
metallic glass matrix composites. Solid line represents temperature (left axis).
Dotted line represents pressure (right axis). Dashed lines represent melting

temperature and glass transition temperature.

starting materials.

The volume fraction of wires was computed by analyzing a backscatter
SEM micrograph of the polished cross section of the sample. The relative area
covered by the matrix and continuous reinforcement then gives the volume
fraction of wires directly. Archimedes' principle was used to find the overall
density of the sample. Then, having both the density of the sample and the

volume fraction, the porosity could be calculated.

2.4 RESULTS

The quality of the composite samples produced by this technique varied
with the reinforcement. At least one sample with each type of reinforcement
listed in the previous section was made. The samples were judged on the
extent of infiltration and the percentage of matrix which was amorphous. On
this basis, tungsten and carbon steel continuous wires were chosen for more
extensive study; a number of samples of each with nominal reinforcement
volume fractions of 20, 40, 60 and 80 percent were made. Fig. 2.5 shows a
photograph of typical samples produced. The samples shown from top to
bottom are: a steel-reinforced sample as cast, a tungsten reinforced sample
after grinding off the surface layer and being cut to length, a tungsten-wire
reinforced sample as cast.

The porosity of two samples were measured: one with nominally 60 vol %
tungsten wire reinforcement and the other 80 vol %. Upon evaluation of SEM
micrographs of polished cross sections, the apparent porosity of the
composites was found to be less than 1%. By calculating the volume fraction
and using Archimedes' principle as described in the previous section, we

determined a porosity of 3% + 2%.

Fig. 2.5. Photograph of metallic glass matrix composite samples. From
top to bottom: an as-cast steel-wire-reinforced sample, tungsten wire
reinforced sample after grinding and cutting, and an as-cast tungsten-wire-

reinforced sample.

The quality of the tungsten-reinforced composites was very consistent;
more complications arose in making the steel wire reinforced composites. Prior
to using the bakeout step, there was evidence that gases were released from
the steel wires during processing. The samples with nominal 20 and 40 vol %
steel wire were made before the problem was found, and thus without the
bakeout procedure. To avoid possible trapped gas porosity and hydrogen
embrittlement of the steel, the bakeout procedure was added prior to

processing the composite samples in the 60 and 80 vol % steel samples.

2.5 DISCUSSION

Different reinforcements were chosen for initial trials for a number of
different reasons. Properties under consideration included high melting point,
low reactivity, ductility, toughness, strength, elastic modulus, availability,
price, or some combination thereof. The reinforcement systems which worked
exceptionally well initially were chosen for further studies.

To take advantage of the mechanical properties of the metallic glass in the
composite, it is important to avoid crystallization of brittle intermetallics.14
Clearly, for quenching a sample from the melt, a low critical cooling rate is
advantageous and allows larger samples to be easily processed and fabricated.
Vitreloy™ 1 has one of the lowest reported critical cooling rates of any metallic
glass forming alloy. In addition, the low liquidus temperature of the alloy allows
processing of the molten alloy at a relatively low temperature, and,
accordingly, results in a minimal reaction between the matrix and
reinforcement.

The extremely high melting point of tungsten (3680 K) is consistent with

both a small thermal effect on the microstructure of the wire during processing

and limited reactivity with the melt. Although the microstructure of steel is
not as stable at the processing temperatures, the glass is tolerant of small
additions of iron}, which partially explains the ease with which steel-reinforced
composites are made.

Nevertheless, addition of reinforcements to a bulk metallic glass can allow
samples to be made larger than with the glass alone. The amorphous matrix
material has a thermal conductivity of about 0.0385 W cm-1K-1 at room
temperature. We would expect a somewhat lower thermal conductivity for the
undercooled melt, which is the material of interest in determining critical
cooling rates. Tungsten and carbon steel have thermal conductivities of 1.74
W cm-!K-1 and 1.0 W cm-!K-! respectively16, and remain solid during the
processing. The addition of higher thermally conductive materials in the
composite allows heat to be removed from the composite more efficiently than
in the unreinforced amorphous alloy, thus allowing the critical cooling rate to be
reached in larger samples.

During the initial stage of processing, prior to infiltration, the metallic glass
is preheated to the minimum temperature required to remelt any residual
crystalline particles present in the starting ingots. Lin et al. found in another
bulk glass-former that preheating a few hundred degrees above the melting
temperature is necessary to achieve maximum undercooling.!2 The conclusion
from that study is that crystallization of the melt is controlled by nucleation of
oxide particles; the preheat must exceed the liquidus temperature of the oxides
to get maximum undercooling. We expect a similar phenomenon in the Zr-Ti-
Cu-Ni-Be system. This study provided the motivation for the initial preheating

step discussed earlier. Ongoing research in this area, including preliminary

results from the TEMPUS facility aboard the space shuttle flight MSL-1,
indicate that this is indeed an important consideration.

Under the given processing conditions, unreinforced metallic glass samples
are fully amorphous under x-ray diffraction analysis. At higher temperatures,
the viscosity of the melt drops and the likelihood of exposing the reinforcement
to the melt during the preheat step increases. At the temperature of the
preheat step, we found significantly more reaction between the reinforcement
and matrix. In addition, even if the neck in the quartz tube succeeds in
preventing contact between the melt and the reinforcement, there is reaction
between the melt and the quartz. Titanium, zirconium and beryllium all form
very stable oxides; all three are more thermodynamically stable than silicon
dioxide per mole of oxygen. Thus, we expect the quartz to be reduced by
contact with the melt. This reaction is obvious in the final sample and is
responsible for the observed interlayer between the composite and the quartz.
This is unavoidable to some extent with this technique, but because we
suspect that oxygen is detrimental to the glass-forming ability of this alloy, we
try to minimize this silica reduction reaction.

We chose the processing conditions to minimize the total reaction between
the reinforcement and the matrix. In some cases, it is possible that some
reaction is desirable for optimal interfacial characteristics. For insufficient
interfacial reactions, reaction time can simply be extended. Far more common
is the problem of excess reaction between matrix and reinforcement.

Conditions for the final stage of processing were chosen to allow sufficient
time for full infiltration of the reinforcement by the melt with minimal reaction.
Because of its higher viscosity, the glass-forming alloy takes longer to infiltrate

the reinforcement than conventional metal matrix materials do. At the

temperature of this processing step, the viscosity of the molten alloy is about 4
Pa:s.1” By comparison, aluminum at its melting point is about 1 mPa:s.!8
Lower viscosities at higher temperatures would allow shorter infiltration times,
but with increased reaction between the matrix and reinforcement. We found
that there was more reaction between matrix and reinforcement at higher
temperatures, despite shorter processing times.

Although the two samples measured have low porosity, we suspect this
value may not be representative of all samples, particularly those of lower
volume fractions. The porosity of samples with lower volume fractions of
reinforcement is also more difficult to measure with the techniques listed in
this chapter. In these samples, the wires have a greater off-axis
misalignment; the assumption of homogeneity in the cross section of the
composite rod becomes a less valid one. Also, porosity did not generally seem
to be well distributed throughout the samples. Often, there would be a
relatively large void present which became apparent after mechanical testing,

or else there would seem to be no voids at all.

2.6 CONCLUSIONS

Composites with a bulk metallic-glass matrix can successfully be made by
slow melt infiltration of the reinforcement. Under the right processing
conditions, very little reaction occurs between the matrix and reinforcement,
and the matrix freezes to an amorphous structure. Many reinforcement
materials and geometries can be used successfully. The best results were with
uniaxial tungsten and steel wires, and silicon carbide particulate preforms. The
matrix material used was Zr41 2Ti13,.gCure,5Ni10,0Be22.5, but this technique

should be applicable with other bulk metallic glass alloys.

Further work can improve the composite in a number of ways. Simple
changes could easily be made, such as processing with other types of tubes
which would not react and introduce silicon and oxygen into the melt.
Preliminary work has shown that stainless steel tubes can be used for
processing in this manner. Materials other than quartz may be cheaper, more
easily handled without breakage, and even reusable. Also, different container
materials might allow other bulk metallic glasses to be used, in particular, ones
without beryllium. Beryllium oxide is toxic as an airborne particulate;
beryllium-free alloys would alleviate related safety concerns and eliminate the
need for special handling procedures during processing. Most of the other good
glass forming alloys do not wet quartz during processing; as a result, they tend
to not maintain good thermal contact with the container during the water
quench and cannot achieve the critical cooling rate for bulk samples.

Another obvious path of further work is to experiment with other
reinforcements in an effort to achieve specific properties in the composite.
This could include materials such as carbon fibers for low density and high
stiffness, or tantalum and tungsten-rhenium for more plastic strain. Also,
particulate, multiaxial continuous fibers and multiaxial discontinuous fiber
reinforcements have the potential to have properties better suited to certain

applications.

REFERENCES

TW. Clyne and P. J. Withers, An introduction to metal matrix composites,
Cambridge University Press, Cambridge (1993).

. T.S. Srivatsan, T. S. Sudarshan and E. J. Lavernia, Prog. Mater. Sci. 39,
317-409 (1995).

. P. Duwez, Trans ASM 60, 607-633 (1967).

. S.J. Cytron, J. Mater. Sci. Let. 1, 211-213 (1982).

10.

11.

12.

13.

M. Narasimhan, US Patent 4,330,027 (1982).

H. Kimura, B. Cunningham and D. G. Ast, Proc. 4th Int. Conf. on Rapidly
Quenched Metals (edited by T. Masumoto and K. Suzuki), 1885-1388,
Japan Instit. Metals, Sendai (1982).

D. G. Ast, US Patent 4,523,625 (1985).

P. G. Zielinski and D. G. Ast, MRS Symposia Proc.: Rapidly Solidified
Metastable Materials (edited by B. H. Kear and B. C. Giessen), 189-
195, Elsevier, New York (1984).

J. F. Williford and J. P. Pilger, US Patent 3,776,297 (1973).

G. Nussbaum and D. G. Ast, J. Mater. Sci. 22, 23-26 (1987).

A. Peker and W. L. Johnson, Appl. Phys. Lett. 63, 17, 2342-2344 (1993).
X. Lin and W. L. Johnson, swbm. Mater. Trans. JIM (1997).

Annual book of ASTM standards, Vol. 02.05, ASTM, Philadelphia (1991).

14.

15.

16.

17.

18.

52
W. L. Johnson, Materials Science Forum Proc. ISMANAM-95 (edited by
Robert Schultz), 225-227, 35-50, Transtec, Switzerland (1996).
A. Peker and W. L. Johnson, US Patent 5,288,344 (1994).

Handbook of Chemistry and Physics, 12-120, 72nd edition, CRC Press,
Boca Raton (1991).

R. Busch, A. Masuhr, E. Bakke and W. L. Johnson, MRS 1996 Fall Symp.
on Glasses and Glass Formers (1997).

Handbook of Chemistry and Physics, 6-169, 72nd edition, CRC Press,
Boca Raton (1991).

CHAPTER 3
MICROSTRUCTURE OF THE BULK-METALLIC-
GLASS MATRIX COMPOSITE

3.1 RESULTS OF MICROSTRUCTURAL ANALYSIS OF
COMPOSITES REINFORCED BY TUNGSTEN AND STEEL

3.1.1 X-Ray Diffraction

Samples of the composite were characterized by x-ray diffraction. Both
tungsten-wire and steel-wire uniaxially reinforced samples were tested with
this technique. Unless specified, all samples were made with the processing
parameters specified in Chapter 2 (15 min preheat at 925 °C, 30 min
infiltration at 800 °C). After a sample was made, a cross-sectional slice was
cut perpendicular to the axis of the wires. The edge of the slice was ground off
to remove the reaction layer between the metallic glass and the quartz tube.

Two different x-ray diffractometers were used for the results reported here.
One was manufactured by Inel, and uses a 120° position-sensitive detector.
This instrument is particularly useful because diffraction peaks over a broad
angular range can be simultaneously detected. Cobalt Ka radiation with
wavelength of 0.1790 nm was used with this apparatus. The other instrument
used was a Siemens D-500 diffractometer with a copper x-ray tube (for Ka
radiation, A= 0.1542 nm). This apparatus uses a 0-26 goniometer
configuration for the sample and the x-ray detector to mechanically scan
through a range of angles. This diffractometer provides better resolution
capability compared to the Inel machine. For calibration, peaks from a silicon
powder sample were measured and compared to tabulated values; this gave a

correction function for the diffractometer.

Samples of various fiber volume fractions were tested. The results are show
in Fig. 3.1. Composites with steel wire reinforcement and ones with tungsten
wire reinforcement are shown in this figure. The percentage refers to the
nominal volume fraction of reinforcement. In the middle, there is also a
diffraction pattern from an unreinforced metallic glass rod prepared in the
same manner as the composites. The boxed annotations denote the reflections
from the reinforcement material. These patterns were taken on the Inel
diffractometer with position-sensitive detector.

Both tungsten and iron have body-centered-cubic (BCC) structures. The
diffraction peaks with boxed labels from the tungsten reinforced composite
correlate well with tabulated interplanar spacings of tungsten. In the steel-
reinforced composites, considering the composition of the steel and the heat
treatment it receives during processing, we expect a mixture of bainite and
martensite to form. Bainite is a two-phase structure of ferrite (a-Fe) with
cementite (Fe3C). Martensite has a tetragonal structure; the c-axis lattice
parameter is slightly different than the other axis. This difference creates a
splitting in the diffraction peaks. The magnitude of the splitting is quite
sensitive to carbon content. We shall see later than the matrix-reinforcement
interface is carbon-rich, probably from diffusion of carbon from the wire; thus
there will be a lower carbon concentration in the steel than expected from the
starting composition.

The diffraction patterns from Fig. 3.1 show that the matrix is mostly
amorphous, with small crystalline peaks which do not correspond to the wire
reinforcement pattern. The pattern from the unreinforced samples display
only the broad bands characteristic of glassy structures. Experience has

shown that this generally indicates that the sample is at most about 2 vol%

rotirritarertsrrrlipr rtp tipi tis list titi lis ii l i

{(110)
(200) (211) (220)
steel 20% :
ee || Oe Pe AW

steel 40% \

steel 40%

tempered a A. _
3 | steel 60% _ A
fas)
—/
P> teel 80%
2 i A A. _
fa
2 metallic glass
q (unreinforced)

[(110)
(200) [ein] — [20] (220)

i Lt
aad

PEPEPPEPPPerperrep errr prevep errr pereepererprere rp erergae
40 60 80 100 120
Two Theta (degrees)

Fig. 3.1. X-ray diffraction patterns of metallic glass matrix composites.

Cobalt Ka radiation (A=0.1790 nm) was used.

crystalline; any smaller fraction is not detected by x-ray diffraction techniques
of this type. The small crystalline peaks only appear in samples with low
volume fractions of wires, for both types of reinforcements. In the samples
with higher volume fractions of reinforcement, the pattern from the wire
obscures the lower intensity patterns from the amorphous matrix and the
crystallized portions of the matrix.

Fig. 3.2 shows x-ray diffraction patterns of a 20 vol% steel reinforced
sample, a 20 vol% tungsten wire reinforced sample, and a pure metallic glass
sample for reference. The y-axis plots intensity as usual, but the x-axis is
scaled using Bragg's Law to plot the interplanar spacing. Thus, the
appearance of the patterns is slightly distorted compared to a plot versus 20.
Numbers next to each crystalline peak correspond to the interplanar spacing.
Numbers in boxes correspond to the diffraction peaks from the reinforcement
material. Each boxed number also has a set of coordinates referring to the
family of planes responsible for the given diffraction peak.

Fig. 3.3 shows two x-ray diffraction patterns from a sample of 20 vol% steel
wire composite. Like the pattern in Fig. 3.2, the peak intensities are plotted
versus interplanar spacing. Labels with both an interplanar spacing and a
specific family of crystallographic planes refer to the BCC peaks from the
reinforcement. Both patterns were taken by the Siemens D-500 machine.
Again, there are small crystalline peaks visible, but different sets of peaks are
visible in each of the patterns for the same sample. This is because the
distribution of these crystalline phases are inhomogeneously distributed
throughout the cross-sectional slice. All three patterns from steel reinforced
samples share the peaks located at 1.65 A, 2.34 A, and 2.70 A. These peaks
match the three highest intensity peaks of ZrC.! Later in this chapter, we will

metallic glass (Vitreloy™ 1)

Intensity (a.u.)

2.03
1.17 (110) 2.34:
(211) 2.54
1.43
1.01 (200) 2.71
steel 20%
1.29
(211)
2.24
1.00 (110)
(310) 1.58
(200)
1.12
W 20%

LL Jue. " 10 ere petit

re re ae | a a ae | re | |

1.0 1.5 2.0 2.5 3.0 3.5 4.0
interplanar spacing (A)

Fig. 3.2 X-ray diffraction patterns of unreinforced metallic glass and metallic-
glass-matrix composites with low fiber fractions. Measured on Inel
diffractometer with Co Ko radiation (A=1.790 A). X-axis converted to

interplanar spacing from 26 using Bragg's law.

Intensity (a.u.)

ee eee ee eee eee ee ee ee ee
2.03 A
(110) 2.34A
Metallic glass-20 vol% steel wire
1.43A
(200)
2.70A
165A
| 3.20A
2.54A
shee cote on.
rrrtlrriitlrrsritlrrretirrrrtirerdsriy
1.0 1.5 2.0 2.5 3.0 3.5 4.0

Interplanar spacing (A)

Fig. 3.3 X-ray diffraction pattern of 20 vol% steel music wire composite.

Siemens D-500 was used on the same sample for both patterns.

see that this is most likely from the interfacial layer between the matrix and

steel reinforcement.

3.1.2 Scanning Electron Microscopy

Scanning electron microscopy (SEM) is a useful tool for microstructural
analysis of the composites. The SEM gives high resolution images of details
from millimeters to microns in size. In general, backscatter images contain
contrast effects from both surface topography and the average atomic number
of the material. All of the samples analyzed in this section were sliced normal
to the wire orientation and polished down to 0.25 um diamond grit. A final
polish was also applied using a colloidal silica suspension. Polishing sections
removes topographical differences; in these samples, image contrast comes
only from compositional variations.

Fig. 3.4 shows an SEM micrograph of a nominally 80% tungsten wire
reinforced composite with a metallic glass matrix. Since tungsten has a higher
atomic number than the average atomic number of elements composing the
matrix, the wires appear lighter than the matrix. A few dark blemishes are
visible in the image, particularly along the bottom edge; these are
contaminants on the surface of the sample. The contamination is composed of
low-atomic-number material, so it appears dark. The array of wires is seen in
cross section, and is nearly close-packed. From measurement of the image,
the exact fiber fraction can be calculated from the number of wires in the
image, the diameter of the slice, and the diameter of the wires. Partial wires on
the edges can be counted by measuring their area with a digital image
processing program (NIH Image). This sample was measured to be 87 vol%

wire reinforcement. This is approaching the theoretical density limit of two-

Fig. 3.4. SEM micrograph of metallic glass matrix composite. Reinforced
with 80 vol% W wire. Light areas are wires. Image taken with backscatter
detector.

Fig. 3.5. SEM micrograph of metallic glass matrix composite. Reinforced
with 20 vol% steel music wire. Dark areas are wires. Image taken with
backscatter detector.

dimensional close-packing of 90.7 vol%. The nominal fiber fractions were
calculated from the number of wires put into the casting tube before
processing and the diameter of the tube. Apparently, many of the packing
defects occur near the wall of the tube, and are subsequently ground off,
leaving a higher volume fraction of reinforcement. Hereafter, all fiber fractions
will be nominal values unless otherwise stated.

Fig. 3.5 shows a SEM micrograph of a metallic glass matrix composite
reinforced by 20 vol% steel music wire. Again, a backscatter detector was
used to acquire the image. In this figure, the reinforcement wires appear
darker than the matrix. The slice appears non-circular because some carbon
paint, which is used for mounting the samples on stubs for viewing in the
microscope, was drawn up onto the edges of the sample. It is clear from this
micrograph that the fiber distribution is quite irregular. Many of the wires are
on one side of the sample. This is unfortunately typical of the samples with low
fiber fractions. Because of the slight bends introduced into the wires in low
fiber fraction samples, the fiber distribution does vary along the length of the
sample; this provides better spatial distribution than with straight wires alone.
However, there is a trade off: there is also more angular misalignment. The
lower fiber fraction samples are consequently less nearly uniaxial than the high
fiber fraction samples. Crystallized regions of the matrix appear as small
irregularly-shaped dark regions smaller than a wire diameter distributed
around the sample.

Fig. 3.6 is a backscatter SEM image of the same sample as in Fig. 3.5 at
higher magnification. The round cross section of the wires are clear, although
one of the wires has an oblong cross section. This may be due to dissolution of

the steel into the matrix during processing, damage to the wire, or a

Fig. 3.6. Metallic glass matrix composite. Reinforced with 20 vol% steel
music wire. Round dark areas are wires. Dark faceted regions are crystal-

lized portions of the matrix. Image taken with backscatter detector in SEM.

Fig. 3.7. Close up view of metallic glass matrix composite. Large round dark
area is single steel wire. Crystals are visible at some locations along the

wire-matrix interface. Image taken with backscatter detector in SEM.

manufacturing inhomogeneity prior to processing. Matrix crystals are visible
around the wires, and alone in the matrix in the bottom center of the
micrograph. Facets in these areas are clearly visible. These crystals appear
darker than the matrix, indicating that they have a lower average atomic
number.

An interesting fact is that the faceted crystals in the matrix seem to
preferentially form on the interface with the wires, although not exclusively so.
A cluster of crystals in Fig. 3.6 are a fiber diameter or more away from any
wires. Since the wires are approximately normal to the polished surface, we
can be reasonably sure that there is no reinforcement very close to this
crystalline cluster. In Fig. 3.5, there is another cluster of crystals on the very
right edge of the micrograph, which is millimeters from the nearest
reinforcement wire. These crystalline clusters do not appear in the
unreinforced Vitreloy™ 1 samples formed by the same processing techniques;
thus we conclude that they are caused by the addition of the wires.

Fig. 3.7 shows a high magnification micrograph of the interface between a
steel wire and the amorphous metal matrix. The sample is nominally 60 vol%
wire reinforcement. Preferential crystallization on the interface is quite
obvious. Small, very dark regions are beginning to be visible along the
interface; they seem to correlate with the presence of the larger faceted
crystals growing into the matrix on the interface.

Fig. 3.8 shows an interfacial region in a tungsten-wire-reinforced sample.
Three wires are visible, along with the matrix material between them.
Crystals in the matrix are visible both in the middle of the matrix region and
directly on the interface. Nevertheless, the crystals on the interface appear

larger and seem to have nucleated on the interface.

Fig 3.9 is a higher magnification SEM image of the interfacial region
between a tungsten wire and the amorphous matrix. This sample was
processed for a longer time; it was held at 800 °C for 160 min instead of only 30
min. Dark crystals about 5 um in diameter appear along the interface, but not
directly on the interface. This image was taken on a demonstration Philips
SEM with a field-emission gun by the Philips representative. The grain
structure of the tungsten wire is also visible in the micrograph. The matrix
seems to penetrate between the grains of the tungsten. A few grains right on
the interface appear to be completely surrounded by matrix material. In
addition, the grains near the interface are significantly smaller than those
deeper inside the wire. This is our first evidence of grain boundary attack on
the wire material by the melt.

Fig. 3.10 shows a high magnification image of an interfacial region between
a steel music wire and the amorphous metal matrix. Again, the dark faceted
crystals are visible near the interface. However, they do not appear to
nucleate directly on the wire-matrix interface. Smaller, extremely dark regions
are visible in between the larger crystals and the wire. In Fig. 3.11, we can see
these smaller crystals even more clearly, which are clustered between the wire

and the large matrix crystal.

Fig. 3.8. SEM micrograph of metallic glass matrix composite. Lighter areas
are tungsten wires. Dark, faceted crystals are visible at some locations along
the wire-matrix interface. Image taken with backscatter detector.

Fig. 3.9. SEM micrograph of metallic glass matrix composite reinforced by 40
vol% tungsten wire. Matrix allowed to infiltrate wires for 160 min. Light
area is tungsten wire. Light gray area is matrix, and dark gray area are
regions of crystalized matrix. Image taken in backscatter mode with Philips
field-emission electron gun by Philips representative.

Fig. 3.10. Close up view of metallic glass matrix composite. Darker area on
the top is single steel wire. Crystals are visible at some locations along the

wire-matrix interface. Image taken with backscatter detector in SEM.

Fig. 3.11. Close up view of metallic glass matrix composite. Darker area on
the left is single steel wire. Crystals are visible at some locations along the

wire-matrix interface. Image taken with backscatter detector in SEM.

3.1.3 Scanning Auger Microscopy

The scanning Auger microscope (SAM) is very similar to the SEM. Both
have electron guns and are used on bulk samples. However, the SAM has a
detector for Auger electrons which are emitted during relaxation of ionized
atoms in the sample. This is an alternate mode of relaxation to emission of x-
rays, which are detected by energy dispersive x-ray spectroscopy (EDS or
EDXS). However, the preferred relaxation mechanism depends upon atomic
number, and thus different elements are detected more easily by one technique
than the other. In particular, the Auger technique is much more sensitive to
light elements such as carbon and beryllium. However, since Auger electrons
can escape only from within a few angstroms of the sample surface without
loss of energy, the depth of material analyzed by Auger is much less than that
by EDXS; Auger is best used as surface technique. As a result, a much better
vacuum is required for the SAM; the instrument used in this work was held at
10° torr. There was also an argon ion gun used for cleaning the surface of the
sample in situ.

Fig. 3.12 is an image of the interfacial region of a tungsten reinforced sample
taken by the secondary electron detector. Even though the contrast effect
from compositional variation is not as large in secondary mode as in
backscatter, there are still sufficient differences to distinguish different phases.
At the top of the micrograph is the tungsten wire, which appears bright. Each
white dot near a number represents a point of elemental analysis by Auger
spectroscopy. The accelerating voltage of the electron gun was 10 kV. Points
1 through 8 are in the tungsten wire, progressively nearer the interface. Points
4 though 6 are in the amorphous matrix, progressively farther away from the

interface. Points 7 and 10 are in two different larger, dark faceted crystals.

Fig. 3.12. Backscatter electron image of tungsten/metallic glass interface in

composite. Each white dot and number corresponds to a point of elemental
analysis in Table 3.1 with Auger spectroscopy.

Area
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10

39.5
37.8
41.0
16.6
41.6
40.0
20.9

18.4
19.6
18.9
5.1

17.0
19.2
8.2

9.1
9.5
8.7
6.5
8.7
9.5
6.3

5.4
5.5
6.0
5.9
6.6
6.4
6.7

27.6
27.6
25.4
65.9
26.1
24.9
57.9

100
100
100

Table 3.1. Atomic concentrations at different points of interfacial region of
tungsten wire reinforced metallic glass matrix composite. Location of points
is given in Fig. 3.12. Oxygen content was excluded from analysis (see text).

Points 8 and 9 are also in the matrix, in areas with a slightly different,
smoother appearance than the area analyzed in points 4 through 6.

The initial analysis performed yielded about 10 atomic percent oxygen at
each of the points in the matrix. Analysis 15 minutes later yielded significantly
more oxygen at the same points. This indicates that despite the precautions of
sputter cleaning the surface and maintaining ultra-high vacuum (UHV)
conditions in the chamber, oxygen contamination was significant. When a
specific area was under analysis, the localized heating from electron
bombardment probably served to getter oxygen at that spot. This effect only
occurred in the matrix because of the easily oxidized constituents, particularly
beryllium, zirconium, and titanium. As a result, it is impractical to measure
oxygen content in this sample with this technique, and is thus excluded from
analyses in this section.

Areas 1, 2, and 3 in Fig. 3.12 and Table 3.1 are pure tungsten according to
the compositional analysis. At the elevated temperatures used in processing
the composite, atomic diffusion into the wire was a possibility. Due to its small
atomic radius, beryllium is a likely candidate as a diffusing species; fortunately,
Auger analysis is quite sensitive to beryllium. As close as 1 um away from the
matrix/wire interface, to the limits of detection of the Auger analysis, there was
no detectable diffusion into the tungsten. This is valuable to know, since a
diffused species can drastically change mechanical properties of a material.
The most salient example is hydrogen diffusion causing embrittlement in steel.

At areas 7 & 10, which are located in the large crystals, we notice a huge
increase in beryllium concentration, and decreases in the relative amounts of
zirconium and titanium. Points 4, 5, 6, 8 & 9, which were at various points in

the amorphous matrix, all have comparable elemental compositions. These

are within the nominal composition of the alloy by about 5 atomic percent,
which is within the experimental error for this technique.

Table 3.2 below shows the results of elemental analysis by Auger for a
composite with steel music wire reinforcement and a metallic glass matrix.
Points 1 and 2 in Fig. 3.13 were in the wire, about 12 um and 2 um from the
interface, respectively. As we expect, both show the wire is mainly iron with
some carbon. However, the measured carbon concentration is higher than we
might normally expect for this alloy; it is nominally 3.6 atomic percent carbon.
Again, the experimental error is large enough that we cannot make any
conclusions about changing carbon concentration within the wire itself.
However, we can conclude that there was no detectable amount of diffusion of
any matrix species any greater than 2 um into the wire. In addition, there was
no detectable difference in composition between these two different points
inside the wire.

There are a number of small dark regions appearing in Fig. 3.13 which did
not occur in the image of the other composite sample. Some, such as the one
analyzed in area 3 in Fig. 3.13 and Table 3.2, are directly on the interface
between the steel wire and the matrix; others, such as in Area 5, occur out in
the matrix, but in the vicinity of the interface. Area 3 is approximately BeC,
and Area 5 has composition approximately SiC. These measured compositions
are consistent with the contrast in the image. The steel wire would be the
likely carbon source for the BeC on the interface. There is often a small
amount of BeC present in commercially available beryllium metal; it is
unlikely, however, that this would then migrate to the interface. The silicon
carbide arises from some impurity, since neither the reinforcement nor the

matrix nominally contains a significant amount of silicon. It is possible during

Fig. 3.13. Backscatter electron image of steel/metallic glass interface. Each

white dot and number corresponds to a point of elemental analysis in Table

Area ZY
#1

#4 15.5
#5

#6 28.9
#7 44.8
#8 17.0

3.2 with Auger spectroscopy.

Ti Ni Cu

7.9 5.8 5.6

5.4
23.9 11.5 7.9
5.8 6.1 5.6

Be Fe
91.5
90.3

56.3

65.2

22?

65.5

58.9
8.2
12.0

8.5
9.7
43.7

41.1
57.5

Table 3.2. Atomic percent concentrations at different points of interfacial

region of steel music wire reinforced metallic glass matrix composite.

Location of points is given in Fig. 3.13.

some stage of polishing that silicon carbide particles used as an abrasive could
have become embedded, but there has been no evidence for this occurring in
the metallic glass previously. It is more likely that the silicon entered the
matrix from the silica tube used as a processing container. From the analysis
of Area 7 in Fig. 3.18, we see also that there is substantial silicon
contamination even in the non-crystallized matrix. The beryllium
concentration, however, was not quantified due to overlap of its main peak with
a silicon peak. The silicon, however, had another isolated peak which could be
used for analysis. Thus, the beryllium concentration for this area was omitted,
although there is beryllium present.

Areas 4 and 8 in Fig. 3.13 are the same phase that crystallizes in similar
locations in the tungsten reinforced composites. We see from Table 3.2 that
the composition is roughly the same as areas 7 and 10 in Fig. 3.12. Area 6 in
Fig. 3.13 is directly on a part of the interface which is free of large crystals.
The results in Table 3.2 show that it is most likely composed of a mixture of

zirconium carbide, titanium carbide, and silicon carbide.

3.1.4 Transmission Electron Microscopy and Scanning Transmission
Electron Microscopy

Transmission electron microscopy (TEM) studies of the composite
microstructure offer a number of advantages. For one, the magnification
capabilities are usually much greater than other forms of microscopy. Thus,
we can image the interfacial region between the reinforcement and matrix in
much finer detail. Also, since electrons are transmitted through the sample,
the structure of individual phases can be analyzed, rather than only the

composition. Further, the use of thinned samples in TEM allows much higher

resolution analysis by EDS. Scanning transmission electron microscopy
(STEM) units can provide a spot size of 100 A or less for compositional
analysis. Unfortunately, the sensitivity of EDS to light elements is quite low.
In fact, in standard EDS detectors, a beryllium window filters out all x-ray
signals from elements with lower atomic numbers than aluminum. The
concentrations of light elements can be measured in a detector with a specially
made thin window, but getting good signal to noise ratios even with this type of
equipment is difficult. Thus, all EDS data presented in this work will neglect
beryllium concentrations. All the TEM work presented here was performed on
a Philips EM430 electron microscope at 300 keV. It is also equipped with a
EDAX 9900 energy-dispersive x-ray analyzer and a single crystal lanthanum
hexaboride filament.

Fig. 3.14 is an electron micrograph of a tungsten wire/amorphous metal
matrix region. The amorphous matrix lies to the upper left, and the tungsten
wire to the lower right. There is a partially crystalline reaction layer about 240
nm thick between the wire and the amorphous region. It lies between the
white marker lines in the bright field image (A). Diffraction patterns (B)
through (E) are shown from different regions. A selected area diffraction (SAD)
aperture with effective diameter of 1.6 um was used to obtain the patterns in
(B) and (E), while an aperture 0.5 um in effective diameter was used for (C) and
(D). The interfacial reaction layer was too narrow to obtain a diffraction
pattern from it alone.

The diffraction spots in (B), (C), (D), and (E) in Fig. 3.14 all can be indexed to
tungsten, with the exception of the smallest ring in pattern (E). This ring is
most likely an artifact produced from the SAD aperture; it is too sharp to be
diffraction from small crystals. See Appendix 1 for details of this analysis. The

Fig. 3.14. Transmission electron micrographs of interfacial region between

tungsten wire and metallic glass matrix. (A) BF image; diffraction patterns

from (B) matrix, (C) reaction layer plus matrix, (D) reaction layer plus tung-
sten, and (E) tungsten.

presence of the broad amorphous ring in (D) indicates that the reaction layer is
partially amorphous. Since the spots in patterns (C) and (D) all correlate, we
can conclude that the reaction layer is simply small crystals of tungsten about
30 nm in diameter in the amorphous matrix. The faint <110> tungsten
diffraction ring in pattern (B) shows that there are some tungsten crystals as
far as 3 um into the amorphous matrix away from the interface, the region
from where the pattern was taken.

Fig. 3.15 shows a graph of composition versus position across a wire/matrix
interface. The composite under analysis was 60 vol% tungsten wire in a
Vitreloy™ 1 matrix, processed as described in Chapter 2. The analysis was
performed in STEM mode, with a spot size corresponding to a 4 nm probe
diameter. Along the x-axis, 0 corresponds to the boundary between the
tungsten and the reaction layer, and 240 corresponds to the boundary between
the reaction layer and the amorphous matrix. One fact to notice is that there
is no detectable diffusion of zirconium, titanium, nickel, or copper 30 nm off the
interface into the tungsten wire. We find that the reaction interlayer is
composed of all the detectable components of the amorphous matrix, plus 10
to 20 atomic % tungsten, depending upon the proximity to the wire. Outside
the reaction layer in the amorphous matrix there is no detectable tungsten.
The actual atomic concentrations would be slightly less to account for the
presence of the beryllium. There are slight variations in the measured
concentrations of the four components of the glass within the reaction layer,
but these do not appear to be significant. These data are consistent with the
hypothesis that the reaction interlayer is composed of small tungsten crystals

in a metallic glass matrix.

100-
J —O— W (at%)
905 A Zr
i — Ti
80-
i —@-— Ni
70- —A— Cu

Composition (atomic %)

-50 0 50 100 150 200 250 300 350 400 #4450
Position (nm)

W i Reaction layer | Amorphous matrix

Fig. 3.15. Compositional profile across tungsten wire/matrix interface in
tungsten wire reinforced composite. Measured by EDS in STEM. Beryllium

concentration omitted from analysis (see text).

8 um off interface
into steel

Fig. 3.16. Transmission electron micrograph (bright field) of interface be -

tween steel music wire and amorphous Vitreloy™ 1 matrix. Amorphous

region on right, and steel on left. Diffraction patterns taken with 0.5 una

diameter selected area aperture, except for pattern from 8 um into steel,
which was taken with 1.5 um aperture.

Fig. 3.16 is a transmission electron micrograph of the corresponding
interfacial region for a steel music wire reinforced composite. Diffraction
patterns from selected areas along the interface are shown. There is a slight
texture change in the image about 1.4 um away from the interface into the
steel wire, but upon analysis of the diffraction patterns, there is no detectable
difference. The only diffraction spots are from the steel. There are some faint
diffraction spots from the diffraction pattern from the matrix, but these can be
indexed to BCC iron. Possibly some of the wire was included in the selected
area, which was chosen right next to the interface. The details of the analysis
are given in Appendix 2. This difference in appearance is possibly due to a
change in the microstructure of the steel near the interface.

Fig. 3.17 shows a bright field/dark field pair of electron micrographs of a steel
wire/metallic glass interface. The interface is clean and with no visible reaction
layer. Fig. 3.18 shows another bright field/dark field pair of electron
micrographs of a steel/metallic glass interface. This region was the one used in
taking the EDS compositional linescan shown in the next figure. Diffraction
patterns from each area are also shown and were taken with 0.5 um diameter
SAD apertures about 1 1m away from the interface on either side. The
pattern from the steel wire shows only spots from steel (Appendix 2). Note the
crystal which appears in the dark field image (Fig. 3.18 (B)) in the matrix
slightly off the interface. Nevertheless, in all these images taken using
selected area apertures, we must remember that exact correlation between
image and diffraction pattern is not guaranteed.

Fig. 3.19 is a graph of the compositional profile across this region taken by
EDS in STEM mode. Like beryllium, carbon is too light to be effectively

detected by this technique. There are also problems with carbon

(A)

(B)

Fig. 3.17. Transmission electron micrographs of interfacial region in steel-
wire-reinforced metallic glass. (A) bright field and (B) dark field. Lower
right area is amorphous matrix, and upper left is crystalline wire.

(A)

Fig. 3.18. Transmission electron micrographs in (A) BF and (B) DF of steel

wire/metallic glass interface. Region used for EDS analysis profile shown in
Fig. 3.19.

100- O
: Fe
90-
7 2Yr
80- Ti
1 Ni
705 Cu

Composition (atomic %)
on
rm

| —xK
40-
30-
20-
: Oo
10-
-100 0 100 200 300 400 500 1000

Position (nm)

Fig. 3.19. Compositional profile across steel wire/matrix interface in steel wire
reinforced composite. Measured by EDS in STEM. Beryllium and carbon

concentrations omitted. Location of wire/matrix interface at 0 on x-axis.

contamination at the location of the electron beam. Like in the tungsten wire
reinforced samples, there is no detectable diffusion of matrix elements into the
steel 30 ym from the interface. However, there is some diffusion of iron into
the amorphous matrix. The concentration drops off very quickly, to a few
percent within 60 nm of the interface. This concentration appears to remain
level up to 1 um away from the interface. There is a change in composition in
the matrix between 100 nm and 250 nm away from the interface. In this
layer, the concentration of copper increases and that of titanium drops. This is
probably due to a slightly different composition of the small crystal visible in
Fig. 3.18 (B) in the matrix. The crystal is of the right size and position from the

interface to correlate to this compositional change.

3.2 ANALYSIS OF RESULTS

The crystals in the matrix are most likely due to impurities on the surface of
the reinforcement. Although these crystals sometimes are not directly on the
interface but only in the vicinity could be simply a consequence of only viewing
a single plane of material. The crystals could have nucleated on the wire below
or above the surface which is being analyzed. Preliminary analysis shows that
dissolved reinforcement material is not a significant constituent of the
crystals; however, Figs. 3.6 and 3.8 clearly show more crystallization near the
matrix-reinforcement interface. The first crystallization event which occurs
upon heating in unreinforced Vitreloy™ 1 which can be detected by x-ray
diffraction is the formation of a Laves phase.” From the Auger analysis, we
find that the large crystals visible in SEM are primarily beryllium, with an
approximate composition of Be,2Zr3TiNiCu. Measurements of the relative

areas in SEM images show that these crystals make up 1 to 5 percent of the

composite samples by volume. The composition can also be expressed in the
form Be»X; X is Zr, Ti, Ni, or Cu in the ratio 3:1:1:1. Again, we suspect a Laves
phase structure from this composition. Beryllium forms the binary cubic C14
Laves phases BegCu and BegTi. This Laves phase is also known as the
MgCu,-type structure. A detailed description of this structure is given in ref. 3.
From the results of microscopy, this Be-rich phase appears the most
prevalent, and is probably a Laves phase, but further work must be performed
to determine the exact structure.

In both the tungsten reinforced and steel reinforced composites, none of the
elements of the matrix alloy were found within the reinforcement. It is still
possible, however, that amounts below the detection limit of SAM or EDS could
still affect the properties of the reinforcement.

Upon closer inspection of the x-ray diffraction pattern in the steel-reinforced
samples, there is splitting in the steel peaks. The position of the second peak
in the (110) reflection at 2.898 A allows us to estimate the c lattice parameter
in the martensite structure. Correlating this with tabulated results, the
diffraction pattern indicates a carbon content of 0.27 wt% carbon.* Since the
initial composition was 0.80 wt%, and additional carbon at the interface is
observed, this is a reasonable estimate of the final carbon content of the steel.
The cause for the larger amount of silicon in the steel reinforced sample is
unclear. Embedded abrasive particles would not explain the silicon distributed
in the amorphous portion of the matrix. The best explanation is that the region
analyzed by SAM was closer to the quartz container during processing, and
thus has a higher concentration of silicon from reducing SiOg. It is also

possible that the silicon diffused out from the steel wire reinforcement.

In considering the tungsten-reinforced composites, we can infer from
estimating diffusion constants that the interfacial reaction layer was solid
during processing. The following argument assumes the opposite and leads toa
non-rigorous proof by contradiction. If we assume that the matrix was liquid
and the tungsten atoms were dissolving and diffusing through a liquid medium,
and then precipitating out upon quenching, what would be the approximate
length scale of that process? For most metals near their melting point, the
diffusion can be estimated at about 10 cm?/s. However, we must recall that
the viscosity of Vitreloy™ 1 at its melting point is much higher than elemental
metals, as first indicated in Chapter 2. We can approximate the diffusion
constant of tungsten in the molten matrix by using the Stokes-Einstein

equation:
kT

= 3.1
6mur (3-1)

where D is the diffusion constant, k is Boltzmann's constant, T is the absolute
temperature, pl is the viscosity, and r is the radius of the diffusing species.
Equation 3.1 assumes that the diffusing particle is much larger compared to
the particles of the medium, which is not precisely true. However, even in the
case of the Sutherland-Einstein formula where the radii of the diffusing particle
and the particles in the medium are equal, the diffusion constant simply

increases by a factor of 3/2:°
kT

= 3.2
4nur (32)

If we plug in reasonable numbers (T=1073 K, p=4 Pa:s, r=2.02 A) to Egn. 3.2,
we find D~1.5*10°° cm?/s. Using a processing time of 30 min, and taking the

diffusion length scale to be given by:
x= Dt (3.3)

we find x=52 um, which is much larger than the observed diffusion of tungsten
in the reaction layer. Thus, we conclude that the reaction layer is solid during
processing of the composite.

If we then assume that the particles are removed from the wire as grains,
and then diffuse off from the interface into the matrix, we can again estimate
the characteristic length. If we try to model the situation in Fig. 3.9, we can
use a time of 160 min, and a particle (grain) size of 0.5 um. We then get a
length of about 2 um, which is still larger than what is observed in
micrographs. It is possible that the particles become completely dissolved in
the matrix by the time they have diffused on the order of 240 nm from the
interface. This would explain the absence of any particles farther than this
from the interface.

The electron diffraction patterns and the EDS data show that the interfacial
reaction layer about 240 nm thick is composed of small tungsten crystals
about 30 nm in diameter in an amorphous matrix. Tungsten composes about
10 to 20 atomic % of this interlayer depending on the distance from the
interface. A possible mechanism for creation of this interface is the diffusion of
molten matrix material into the grain boundaries of the tungsten wire. The
grains could then break apart and become interspersed in the amorphous
matrix. Some dissolution of the tungsten into the matrix would also occur,
yielding the smaller grain size in the reaction layer than is found in the original
tungsten wire.

In the steel reinforced samples, the portions of the interface without large
crystals seem to be mixed carbide layers. Diffraction data and SAM analysis
indicate ZrC is the most prevalent form. Using a lattice parameter of 4.693 A

for ZrC and a cubic structure, we find a peak at 2.71 A is a (111) reflection, at

2.35 A is a (200) reflection, and the peak at 1.66 A is a (220) reflection. An
interesting point is that this carbide layer seems to be passive with respect to
further crystallization of the matrix. That is, it does not appear to act as a
heterogeneous nucleation site for the metallic glass. This is consistent with
other observations in the laboratory. Masuhr at Caltech has done extensive
processing of Vitreloy™ 1 above the liquidus point of the alloy in a graphite
crucible.® These experiments show that even after keeping the melt at
elevated temperatures for hours, it can still be cooled to a glass; carbon and the
resulting reaction layers do not seem to be detrimental to glass formation.
Nevertheless, there often are small low-atomic number particles on the
interface where larger crystals nucleate in steel-wire-reinforced composites.
We found one example in Fig. 3.13 where this particle was determined by Auger
analysis to be BeC. The role of carbides in crystallization of the matrix is
unclear.

Further, in the steel reinforced samples, there does appear to be substantial
dissolution and diffusion of iron into the matrix. From the data in Fig. 3.19, we
see that the characteristic length for such diffusion is certainly longer than 1
um; this is consistent with the estimation given above for the diffusion length
calculated by the Sutherland-Einstein formula. Also, Peker and Johnson found
that Vitreloy™ 1 remains a bulk-glass-former even with additions of up to
about 10 atomic % iron.” So, we can expect that the concentrations of iron
which were found to diffuse into the matrix would not cause the matrix to

crystallize.

3.3 STILICON-CARBIDE AND OTHER REINFORCEMENTS IN
METALLIC-GLASS-MATRIX COMPOSITES

In the preliminary stages of this work, considerable effort was put into
trying a wide range of different types of materials to determine what
reinforcements could be successfully incorporated into a metallic glass matrix.
Frequently, the matrix of the samples had crystallized. Usually only a single
sample was made of any given reinforcement type; since processing conditions
depend upon the materials involved, the procedure was not optimized for each
type of reinforcement. This initial screening process determined which
reinforcements could be incorporated into a composite most easily. However,
significantly different results were obtained using different processing
techniques.

A number of experiments were performed in an attempt to incorporate
silicon carbide fibers into the amorphous metal matrix. The first generation of
these composites were made by inductively heating ingots above a fiber bundle
in a sealed and evacuated quartz tube. The melt flowed down and infiltrated
the fibers by capillary action. The melt was estimated to be 1500 K or higher
with this technique, but there was no way to accurately control and stabilize
the temperature. In the time required to infiltrate, enough of the fiber material
had dissolved to destroy the glass-forming ability of the matrix.

In an effort to achieve better temperature control, the induction coil was
replaced by a resistive furnace with an active temperature controller.
Different arrangements were tried until one was found with a sufficiently
spatially large and homogeneous temperature zone. An inert gas line was
added to provide back pressure to assist infiltration. One sample was made by
infiltrating polycrystalline silicon carbide fibers with Vitreloy™ 1 for 3 h at

1098 K. Fig. 3.20 shows a map of carbon concentration for a cross section of

Fig. 3.20. Carbon compositional map for SiC/Vitreloy™ 1 composite. Image
taken with SAM. Sample infiltrated for 3 h.

Fig. 3.21. Silicon compositional map for cross section of SiC/Vitreloy™ 1

composite. Image taken with SAM. Sample infiltrated for 3 h.

this sample from scanning Auger microscopy (SAM). The Auger analysis of
the silicon carbide-reinforced composites was performed by the Carborundum
Company. Areas high in carbon appear lighter. The fibers are slightly off-
axis with the plane of the section, and thus appear as light-colored ovals. Fig.
3.21 shows a map of the same area for silicon. The silicon clearly plays a role
in the crystallization of the matrix; these silicon-rich regions are faceted and
clearly not amorphous. Even though there is an equal fraction of silicon and
carbon available, the carbon does not segregate in this fashion, and does not
appear to encourage crystallization.

Figs. 3.22 and 3.23 show the same type of compositional maps for a similar
silicon carbide composite sample; however, this one was processed at 1098 K
for only 0.5 h instead of 3h. The compositional map in Fig. 3.22 shows no
appreciable carbon in the matrix, similar to Fig. 3.20. But Fig. 3.23 also shows
that for this sample there was no appreciable silicon dissolved into the matrix,
either. Clearly, the processing time and the reinforcement composition play
important roles in determining matrix microstructure.

However, further analysis of the silicon-carbide-reinforced samples
processed for 0.5 h showed some small crystals in the matrix. These are
beryllium-rich regions visible in the beryllium map shown in Fig. 3.24. These
crystals are clearly not the same type as those visible in Fig. 3.21. These
smaller crystals are Be-rich, and arise from impurities in the melt. To avoid
this type of crystal in the matrix, the preheat stage (up to = 1200 K) is
required, as noted in Chapter 2.

Fig. 3.25 is an optical micrograph of a porous SiC preform infiltrated with
metallic glass. The light areas in the photograph are regions of metallic glass,

and the dark areas are the preform. The small, very dark speckles are small

Fig. 3.22. Carbon compositional map for cross section of SiC/Vitreloy™ 1

composite. Image taken with SAM. Sample infiltrated for 0.5 h.

Fig. 3.23. Silicon compositional map for cross section of SiC/Vitreloy™ 1

composite. Image taken with SAM. Sample infiltrated for 0.5 h.

Fig. 3.24. Beryllium compositional map for cross section of SiC/Vitreloy™ 1

composite. Image taken with SAM. Sample infiltrated for 0.5 h.

Fig. 3.25, Optical micrograph of SiC preform infiltrated with bulk metallic

glass. Dark regions are SiC, light regions are amorphous metal.

pores in the preform which are closed off and unavailable to the matrix
material. The preform is made of silicon carbide particles sintered to 60%
density. Inspection of the micrograph shows full infiltration of the preform by
the metallic glass and apparently a fully amorphous matrix. X-ray diffraction
of this sample showed crystalline peaks only from silicon carbide. The sample
was cast at 1098 K, and allowed to infiltrate for 5 minutes under a back
pressure of argon gas. Note the extremely good wetting which has taken place;
very fine details in the particulate preform have been filled by the metallic
glass matrix.

As mentioned in the previous chapter, preliminary investigations also
included making composite samples with a variety of metal wire
reinforcements, many which were judged to be too reactive for further study.
One interesting example is the composite sample reinforced with tantalum
wire. Tantalum exhibits the rare combination of a high melting point (3287 K)
and good ductility, both properties of interest in making metallic glass matrix
composites. A composite rod of Vitreloy™ 1 and tantalum wires was cast
isothermally at 1173 K; the wires were in contact with the melt for 10 min
prior to water quenching. Fig. 3.26 shows an SEM micrograph of the
wire/matrix interface in backscatter mode. The light area at the lower left of
the image is the tantalum wire. There is a layer about 50 um thick filled with
eroded tantalum particles. Some of the tantalum is visible farther away from
the interface, but the amount is small. The presence of tantalum in this
interlayer was confirmed by EDS. Despite tantalum's high melting point, there
was still significant reactivity with the alloy melt during processing. Clearly,
the chemical compatibility between the matrix and other materials plays an

important role in choosing a suitable reinforcement.

3.4 CONCLUSION

The amorphous-metal-matrix composites as processed had a mostly
amorphous matrix. Crystallization of the matrix occurs preferentially on the
interface of tungsten and steel wires. Some matrix crystallization occurs in
the matrix away from wires, although not as frequently; the largest matrix
crystals have the approximate composition Be;,Zr3TiNiCu. There is more
matrix crystallization in composite samples than in unreinforced samples
made with identical processing. In neither tungsten nor steel wire reinforced
samples was reinforcement material found in the largest matrix crystals.
Hence, the crystallization is most likely due to heterogeneous nucleation on the
surface of the wire or some contaminant on the surface. Carbides appear to be
associated with crystallization near the interface in some cases, but not in
others; the role of carbides is unclear. In silicon carbide reinforced samples,
crystallization can occur by dissolution of excessive silicon into the matrix,
changing the glass-forming abilities of the alloy.

In tungsten-reinforced composites, the electron diffraction patterns and the
EDS data indicate that the interfacial reaction layer is composed of small
tungsten crystals about 30 nm in diameter in an amorphous matrix. Tungsten
composes about 10 to 20 atomic % of this interlayer depending on the distance
from the interface. For processing conditions which held the wires and melt in
contact for 30 min at 800 °C, the reaction layer was about 240 nm thick. In
the composites with steel reinforcement wires, there was no detectable
reaction layer in TEM, although SAM analysis suggests the formation of a
carbide layer at the interface.

Other ductile reinforcement materials tested included tantalum and

molybdenum, which were expected to perform well due to their high melting

points. The melting point of tantalum is 3287 K, and the melting point of
molybdenum is 2890 K. Wire bundles of each of these materials were used as
reinforcement for a composite sample. The wires were in contact with the melt
for 10 minutes at 1173 K, and there was substantial reaction in both samples.
The tantalum sample had a reaction layer approximately 50 wm thick with
substantial erosion of the wire by the melt, and the molybdenum reinforced
sample had a reaction layer about 10 um thick, with large areas of crystallized
material distributed throughout the melt. Similarly processed tungsten and
steel reinforced composites did not show any reaction visible at this scale.
Clearly, the chemical compatibility between the matrix and other

materials plays an important role in choosing a suitable reinforcement.

An interesting comparison is between the interfacial layers formed in
tungsten-reinforced composites (Fig. 3.9 and Fig. 3.14) and that of tantalum-
reinforced composites (Fig. 3.26). We might initially expect a similar interfacial
reaction, since both metals are extremely refractory. Although only SEM
analysis was performed on the tantalum composites, it appears that crystals
of tantalum were released into the matrix, which remained amorphous. This is
basically what was found in the tungsten/metallic glass interfacial region. It is
also visible on a larger scale in the tungsten-reinforced composites in Fig. 3.9.
However, the tungsten reaction layer seen by TEM is 200 times thinner with
proportionally smaller crystals compared to the tantalum. This certainly may
be due to different chemical interactions between the reinforcement and the
melt, or perhaps due to etching at surface features due to the drawing process
of the wires. But it seems most likely that there is an attack on the grain
boundaries of the reinforcement material. Analysis on the wire properties has

not been performed, but the tantalum wire was larger diameter (500 um) and

Fig. 3.26. SEM micrograph of composite interface. Light area at lower left is
tantalum wire reinforcement, and dark area at upper right is metallic glass.

Image taken with backscatter detector.

probably required much less drawing than the tungsten. Thus, we would not
expect as finely refined a grain structure in the tantalum wire as in the
tungsten wire. This could explain the different length scales observed in similar
processes. In addition, Fig. 3.9 indicates that there is some dissolution of
individual tungsten grains once they have been eroded from the wire. It
appears that they reduce in size the longer they have been in contact with the
melt.

These different pieces of evidence bring up at important point about
comparing different types of microscopic analysis. It is nearly impossible to
directly compare images of similar samples if different preparations are
required. It certainly appears that different areas on the same reinforcement
wire have different interfaces. Some appear to have crystallized reaction
layers, while others do not. Analysis of many more samples seem to be needed
to really get a good typical characterization.

A valuable direction of further study would be a systematic investigation to
quantify the tolerance of the metallic glass to impurities. Use of an
electrostatic levitator (ESL) would be ideal for such work. This technique is
described in ref. 8. During composite processing by liquid infiltration, we would
normally expect a small amount of reinforcement material to dissolve into the
matrix. The quantity would depend on the matrix alloy, reinforcement
material, duration of exposure, temperature of processing, and convection
effects in the melt. Metallic glass alloys could be made with small known
additions of other elements for testing in an ESL. The amount of undercooling
could be measured for each impurity fraction, which would give a good
indication of the glass forming ability of that alloy. In addition, recent work on

the TEMPUS facility aboard MSL-1 has indicated that overheating the melt

plays an even larger role in achieving large undercooling than previously
thought. More experimental data in this field would allow other improvements
in processing to be made.

Another interesting area for future work would be the fabrication of bulk
samples of similar composition and microstructure to the interfacial layer
found in the tungsten-wire reinforced composites. Further investigations of the
microstructure are warranted to confirm if in fact there are nanometer sized
tungsten particles in an amorphous matrix. If this is the case, this material
could have tremendous yield strength. It could basically provide a means to
strongly bond together nanocrystalline tungsten without grain boundaries.
This material could be produced by infiltrating and reacting ball-milled

tungsten powder with metallic glass.

REFERENCES

1. NBS Monograph, 25, Standard x-ray diffraction powder patterns
(Washington D.C., U.S. Department of Commerce/National Bureau of
Standards, 1984).

2. S. Schneider, personal communication (1995).

3. C.S. Barrett and T. B. Massalski, Structure of Metals, Pergamon Press,
Oxford (1980).

4. L. Xiao, Z. Fan, Z. Jinxiu, Z. Mingxing, K. Mokuang, and G. Zhengqi, Physical
Review B 52, 14, 9970-9978 (1995).

5. T. lida and R. I. L. Guthrie, The Physical Properties of Liquid Metals,
Claredon Press, Oxford (1988).

6. A. Masuhr, personal communication (1997).
7. W.L. Johnson, personal communication (1997).

8. W.K. Rhim, S. K. Chung, D. Barber, K. F. Man, G. Gutt et al., Review of
Scientific Instruments 64, 10, 2961-2970 (1993).

CHAPTER 4
KINETIC ENERGY PENETRATORS: A NOVEL
APPLICATION OF BULK-METALLIC-GLASS:-
MATRIX COMPOSITES

4.1 TRADITIONAL KINETIC ENERGY PENETRATOR SYSTEMS

4.1.1 Background of Kinetic Energy Penetrators

Currently, a large portion of the defense industry dedicated to anti-armor
applications are focused on using kinetic energy penetrators. These
applications use only the kinetic energy of the projectile to defeat the target.
The two primary families of materials used for this application are tungsten-
based alloys, and depleted uranium (DU)-based alloys; other alloys have been
excluded for reasons of high cost, limited availability, or inferior properties. DU
is the U-238-rich by-product of the extraction of fissionable U-235 from
uranium ore. Although DU cannot be used for nuclear fission, it still emits low-
level radiation. Hence, "depleted" refers to the material's lack of the fissionable
isotope rather than its lack of radioactivity. The best DU-based alloys
outperform the best tungsten-based alloys by about 10% in penetrator
performance.! As a result, DU kinetic energy penetrators are preferred for use
by defense industries. Nevertheless, environmental concerns and related
political pressure to stop using DU continues to spur research to find a way to
improve the penetrator performance of tungsten-based alloys to match that of
DU.

Significant work has been done to extensively characterize both DU and
tungsten alloys in an effort to determine how to optimize penetrator
performance. A number of mechanical and physical properties are considered
essential for this application, such as high density and the optimum

combination of hardness, strength, stiffness, and fracture toughness. High

100

density is required in order to focus the maximum kinetic energy for a given
penetrator geometry. The design of penetrators is based on the idea of
concentrating the most kinetic energy possible over the smallest area of the
target. Thus, most penetrator designs are long-rod penetrators, which have a
length to diameter (L/D) ratio of 10 to 20. High density is also important in
reducing aerodynamic energy losses during flight and reducing flight time.?
Alloys of DU and tungsten used in this application range in density from 16.2
to 18.6 g/cem?.4

Unfortunately, it is unclear exactly how and which mechanical properties
affect penetrator performance. A number of studies suggest that
improvements in the quasi-static mechanical properties do not necessarily
coincide with improvements in penetrator performance.®° Penetration is an
extremely high strain-rate event during which both the target and the
penetrator undergo large amounts of deformation. There is an extremely large
hydrostatic stress present at the penetrator-target interface, as shown in the
diagram in Fig. 4.1. These data are from a computer simulation of a typical
penetration event. Penetration continues until the penetrator is decelerated to
a complete stop, the penetrator is completely eroded away, or the target is
perforated.

The most salient difference between penetration tests of tungsten alloys
and DU alloys is the behavior of the penetrator tip. Fig. 4.2 shows
schematically the three different types of behavior during a penetration event.
DU tends to slough off material as it penetrates due to its tendency to fail by
adiabatic shear; the tip remains about the same diameter as the rest of the rod
and stays sharp. The kinetic energy is thus focused over a smaller area of the

target and the penetration depth is enhanced. This behavior is known as “self-

101

xno CD coms
oo 2 ee,

Boos, \
Velocity ~ \
ie \ \
\ \ \
Penetrator \ \ {
5 4 2 11 GPa
A of |
/ /
Inverted/back extrudued > / ;
penetrator material ‘ / yA
a Leow? vA
., a“

Serene

Ts Pressure contours

Fig. 4.1. Computer simulation of the penetration process of a long-rod kinetic

penetrator into steel armor. Reproduced from refs. 1 & 6.

~.
Ss.

Pr
any

aor
ee,
me,
=,
: ss,
eorn
aor
wee
— —,
wane, ~
a,
: =.
ss ~,

(b) (c)
Fig. 4.2. Three types of penetration behavior for various materials that show
(a) self-sharpening, (b) limited self-sharpening and (c) no self-sharpening
(“mushrooming”). Reproduced from refs. 1 & 7.

102

sharpening.” On the other hand, tungsten-based penetrators tend to expand
and “mushroom” out, leaving a cavity of increasing diameter as deeper
penetration is achieved. This is widely accepted to be the primary reason that
DU exhibits superior penetrator performance. Some materials, such as
certain tungsten heavy alloys (WHAs), exhibit a hybrid-mode behavior called

limited self-sharpening.

4.1.2 Adiabatic Shear of Materials

The mechanism by which DU fails at high strain rates is known as
adiabatic shear, or flow localization. In this failure mode, a material undergoes
large shear strains confined to an extremely thin region. This thin region of
deformation is known as a “shear band.” Usually, this will occur on a
plane of highest resolved shear stress. For a homogeneous material in uniaxial
compression, this plane will be at 45° from the axis of compression. A large
portion of the mechanical work put into deforming a material in this fashion
goes to producing heat within the shear band. The speed of this
process allows it to be called “adiabatic,” since very little heat has been
conducted away by the time the process has stopped.

There have been a number of investigations to determine which properties
of a material influence its tendency to fail by propagation of shear bands. We
can assume that for a material deformed under pure shear, the flow stress 7 is
a function only of strain y, strain rate y, and absolute temperature T:

T= U%4T) (4.1)
We can also assume that the condition which implies catastrophic shear is

given by:

dt
—=0 4.2
dy (4.2)

103

Expanding the total differential, we find:
a (22) (2) a (2) a 9 (4.3)
dy \oy ry Oy ry dy \0oT),, dy
The three terms in the above equation represent the contribution from strain
hardening, strain rate hardening, and thermal softening, respectively. The last
term competes against the other two terms in trying to meet the condition
required for catastrophic shear (Eqn. 4.2). Then, if we assume that at this high
strain rate all of the energy of deformation goes into thermal energy, we can
write:

art (4.4)
dy C
where C is the volume specific heat in kJ-m™*-K". Further, assuming that the

strain rate is fairly constant, we find that the critical shear strain is given by:

_ on (4.5)

Ycritical =
| (2% Ty

when the relationship t= ey" holds for the relationship between the shear
stress and shear strain (n is the strain-hardening exponent). This relationship
was tested experimentally by Staker® on 4340 steel, and it was a good
predictor of the onset of adiabatic shear band formation.

Another model described by Shockey? requires two critical conditions to be

met for adiabatic shear band formation. They are:

1. nC, (4.6)

Y > Yeritical =~ OT
(Yr),

yoy ____20nk
critical R? ( Ot,
Q €,€

where 1, is the yield strength, n is the work hardening exponent, C, is the

(4.7)

specific heat, and K is the thermal conductivity. R is a characteristic

104

specimen dimension, and « is a constant approximately equal to 2. Notice that
Shockey’s first condition is almost identical to that of Staker, with the yield
strength replacing the flow stress.

Qualitatively, we see from Equations 4.5 through 4.7 that if we are to have
adiabatic shear propagation, we want a material with low strain hardening, low
strain rate hardening, and high thermal softening. In addition to these
theoretical models, Dowding? found empirically through ballistic testing that
adiabatic shear is encouraged in a matrix with elements with low melting
points, low density, low crystal symmetry, and low average atomic numbers.

Numerous investigators have found that tungsten alone does not tend to fail
by adiabatic shear during ballistic testing. In an attempt to take advantage of
the high density of tungsten while encouraging adiabatic shear, the most
common tungsten-based penetrator materials are metal/metal composites.
Usually, a primary phase composed of packed tungsten particles is liquid-
phase sintered together with a binder material. One of the most frequently
used binder materials is an alloy of nickel and iron in the ratio of 7:3 in weight
percent. Thus, a typical tungsten penetrator material is composed of 93 vol%
W particles, with the remainder of the volume filled by an alloy of NizgFeg9.1

Table 4.1 contains various physical and mechanical properties which have
been found to influence the tendency to fail by adiabatic shear. The second
column indicates the desired property to encourage adiabatic shear, as
discussed above. The pertinent comparison is really between Ni7gF egg, the
material commonly used as a matrix for tungsten composites, and Vitreloy™
1, a potential replacement. Some data are not readily available and soa
complete comparison is not possible. Nevertheless, Vitreloy™ 1 compares

very favorably with NizoFeg, in all four categories shown in the table. In

105

Property Better DU Ni7zoFe3o9 Metallic Glass-
Adiabatic Vitreloy 1
Shear

Melting Point | LOW 11321° 14411 72012

(Thiquidus» °C)

Density LOW 19.05 (a) ~8.6 6.1

(g/cm?) 18.89 (B)

Avg. Atomic | LOW 92 27.4 26.85

Number

Thermal LOW 281° ~87 3.5

Conductivity

(Wm? K?})

Table 4.1. Comparison of various physical properties of penetrator materials.

addition, the Ni7zgFeg 9 matrix is primarily a FCC phase; Vitreloy™ 1 is glassy,
thus satisfying the condition for low crystal symmetry. Bruck!* 14 found no
strain hardening (n=0), and no strain rate hardening (m=0) in the Vitreloy™ 1
system. For most crystalline metals, the work hardening exponent n is in the
range of 0.02 to 0.50, and the strain-rate sensitivity m is in the range of 0.0 to
0.1 at room temperature.

It is not quite clear which temperature regime we should consider when

comparing the thermal softening effect between the metallic glass and

Ni7z9Fe39; however, investigators have shown that the viscosity of Vitreloy™ 1

follows a Vogel-Fulcher-Tamman relation:'®
DT,
= 4.8
1 meen P| (4.8)

where 7 is the viscosity, T is the absolute temperature, D and Ty are fitting

parameters, and

106

AN,

V (4.9)

No =
where h is Planck’s constant, N, is Avogadro's number, and V is the molar

volume. We also know from elementary fluid dynamics that the shear stress 7,

viscosity n, and velocity gradient dv/dy can be related by:
T= N— (4.10)

Thus, we might expect the flow stress to go linearly with the viscosity if we
maintain a constant velocity gradient. Near the glass transition temperature
of Vitreloy™ 1, the viscosity changes by 33% over a span of only 2 °C. For
comparison, a commercially-used nickel alloy undergoes about a 20% decrease
in its yield point upon heating from room temperature to 370 °C. We thus

expect the thermal softening of Vitreloy™ 1 to also encourage adiabatic shear.

4.2 ADIABATIC SHEAR BANDING IN METALLIC GLASS AND
METALLIC GLASS COMPOSITES

4.2.1 Failure Mode of Unreinforced Metallic Glass
Experimentally, from quasi-static mechanical testing, we know that

Vitreloy™ 1, like most other metallic glasses, fails by adiabatic shear band
formation. Fig. 4.3 shows a cylindrical sample of aspect ratio approximately
2:1 which was stressed in uniaxial compression along the axis of the cylinder
until failure. The photograph shows that there is no perceivable deformation in
the bulk of the sample. The sample failed along planes of maximum resolved
shear stress at 45° to the applied compressive stress.

Fig. 4.4 shows an SEM micrograph of the fracture surface. The surface is

covered with the characteristic “venous” pattern associated with metallic

107

Fig. 4.3. Photograph of metallic glass (Vitreloy™ 1) cylinder after

compressive failure. Applied stress was along axis of cylinder.

Fig. 4.4. SEM micrograph of fracture surface of metallic glass in compressive
failure.

108

glass failures. The side of the cylinder is visible in the lower right corner of the
micrograph. Figure 4.5 shows the stress-strain curve for this metallic glass
sample. It is elastic up to 2% strain, and then fails with no appreciable global
plastic strain. This type of stress-strain curve is typical of a material which
fails in adiabatic shear. Bruck and coworkers“ performed dynamic
compression testing of Vitreloy™ 1 using the split Hopkinson bar apparatus.
The results showed that at strain rates between 10? and 104 st, the failure
mechanism was also by adiabatic shear, virtually identical to results from

quasi-static tests.

4.2.2 Quasi-Static and Low Strain-Rate Testing of Metallic Glass
Matrix Composites

Quasi-static mechanical tests were performed on composites with a
metallic glass matrix. The samples were reinforced by uniaxially aligned
tungsten wires. The wires were 0.254 mm in diameter. Tests of the 20 vol%
wire composites yielded fracture surfaces similar to unreinforced samples and
with virtually no global plastic deformation. In the composite sample with 40
vol% wire reinforcement, the failure mode changed. The cylinder began to
buckle and exhibited significant plastic deformation. It appears that shear
bands were still propagating through the matrix, but global failure was being
prevented by the presence of the reinforcement. Adiabatic shear was no longer
the only mechanism by which the sample was deforming. For this strain rate,
there appears to be a critical volume fraction of reinforcement between 20 and
40 vol% required for failure along a single adiabatic shear band.

Further compression testing was performed at higher strain rates using a

Hopkinson bar apparatus. Here, strain rates achieved were on the order of

Stress (GPa)

2.2

2.0

1.8

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

109

— Vitreloy™ 1 unreinforced
Young's Modulus: 96 GPa

retoritirrtisitirsitirrstiirirtispirtirirtiiirtiy

0.0 02 04 06 08 10 12 14 16 18 20 2.2

Strain (%)

Fig. 4.5. Stress-strain curve of unreinforced metallic glass cylinder

in quasi-static compression.

110

107-10° s?. Again, samples with lower percentages of wire reinforcement
tended to fail along a plane at about 45° to the applied stress by adiabatic
shear, while samples with higher volume fractions of reinforcement tended to
fail by axial splitting. However, the volume fraction for which this transition
occurs is different than for the quasi-static case. In this series of tests, the 60
vol% reinforcement samples showed evidence of failing in shear, while the 80
vol% sample did not. The failure surface for the 60 vol% tungsten wire sample
is shown in Fig. 4.6. Again, the micrograph shows the same characteristic
venous pattern on the surface, in addition to wires which have been sheared
off. Note, however, that the metallic glass matrix has been smeared over the
surface of the sheared reinforcement. Clearly, this smearing occurred after the
wires had sheared through. The liquid-like appearance of this surface indicates
that there could be a local drop in the viscosity of the metallic glass matrix,
perhaps precipitated by an extreme temperature increase from the local
deformation.

These mechanical tests indicate that catastrophic shear failures become
the preferred mechanism at higher strain rates, even in samples with a large

volume fraction of reinforcement.

4.3 PROCEDURE FOR BALLISTIC TESTING

Evidence from quasi-static and limited dynamic testing of metallic glass and
metallic glass matrix composites was very promising. There is evidence to
indicate that as strain rates increase, adiabatic shear becomes more
pronounced and could possibly be the preferred mechanism of failure in

composites with only small volume fractions of metallic glass. To further

111

Fig. 4.6. SEM micrograph of failure surface of 20 volume percent tungsten-
wire-reinforced Vitreloy™ 1 composite. Sample failed on single shear band

under dynamic axial compression on a Hopkinson bar apparatus.

112

investigate the high strain-rate behavior of the composites, ballistic tests were
performed; the experimental technique is described in this section.

Two types of testing were performed: reverse ballistic and forward ballistic.
Both were intended to test the penetrating capabilities of tungsten-wire-
reinforced metallic-glass-matrix composites. Both used an enclosed cannon at
the California Institute of Technology. A schematic of the cannon is shown in
Fig. 4.7. The cannon has a smooth-bore barrel which is 3.05 m in length and 35
mm in inner diameter. Acceleration of the projectile is provided by a hot-wire
ignited gunpowder charge. The velocity of each shot is primarily controlled by
the mass of the projectile, as well as by the mass and type of gunpowder used.
Each type of gunpowder burns at a different rate; generally more finely divided
powder burns more quickly and provides a higher final velocity. At the exit end
of the barrel, there are two fiber-optic light sources separated at a known
distance. Each is directly across from two other fiber-optic cables attached to
optical sensors. The time between when each light path is broken is measured,
and the velocity is calculated.

The projectiles were carried down the barrel in a nylon sabot, which is
machined to a slip fit inside the barrel. The sabot, made of a relatively soft
material, protects the barrel from damage during a shot, keeps the majority of
rapidly expanding gasses behind the projectile during flight down the barrel, and

stabilizes the projectile against tumbling.

4.3.1 Procedure for Reverse Ballistics
Initial testing was performed using reverse ballistics; that is, the penetrator
was held stationary at the exit end of the barrel, and a flyer plate was attached

to a sabot with epoxy and used as a projectile. This test was performed

113

light
source |}

fiber-optic
cables

powder

charge barrel

target
chamber

(- ~

optical sensors &
power . timing electronics
trigger

\ y,

Fig. 4.7. Schematic of cannon used for ballistic testing.

initially because many of the gun parameters were already calculated for this
configuration. The flyer plate was a circular disk of 0.983 cm thick C200
maraging steel. The plate was 3.152 cm in diameter, with a mass of 60.13 g.
The steel was aged at 490 °C for 6 h; the reported mechanical properties for
this alloy and heat treatment is yield strength of 1400 MPa, tensile strength of
1500 MPa, 10% elongation, and a fracture toughness in the range 155-200
MPa-m!?.!” The mass of the sabot was 24.07 g, so the total mass of the flyer
plate plus the sabot was 84.20 g. The composite rod was a circular cylinder
with a flat end; it was glued with epoxy into a steel mounting block with a
countersunk hole. The rod was 0.635 cm in diameter, 25.38 g, 4.71 cm long,
with 3.53 cm of exposed length. A schematic of this test configuration is shown
in Fig. 4.8.

114

flyer mounting
plate block

barrel composite rod /
wall momentum

trap

Fig. 4.8. Reverse ballistic schematic.

Only one successful test was performed in this configuration. Sufficient
gunpowder was used so that the predicted velocity for this flyer mass was
1000 m/s; however, there was an error in the velocity measurement circuit
during this test, so the exact speed is not certain. The composite was 80 vol%
uniaxially aligned tungsten wire in a Vitreloy™ 1 matrix. The wires were 254
um in diameter, and the composite was fabricated as detailed in Chapter 2.
However, the Vitreloy™ 1 used for composites in the ballistic tests was
provided by Amorphous Technologies International (Laguna Niguel, CA). The
rod used in this test was a right circular cylinder; however, since the rod was
solidly mounted, only a portion of the rod was exposed and available for
penetration. Behind the steel mounting block was the momentum trap, an
arrangement to absorb the shock of the impact. This test was also done with
the barrel under rough vacuum to maintain the same testing parameters from

previous experiments, although this is not necessary for the test.

115

4.3.2 Procedure for Forward Ballistics

In order to better simulate what an actual penetrator would experience
during impact, we also tested the composite in a forward ballistic arrangement.
For these tests, the penetrator was mounted in a sabot and fired at a target.
This testing configuration is shown in Fig. 4.9.

Two different target materials were used. The first was an aluminum (6061
T651) cylinder aligned along the axis of the barrel. The aluminum targets were
cut from a 15.2 cm (6 inch) diameter billet, and 15.2 cm thick. The second type
of target was a 15.2 cm (6 inch) diameter 4130 steel. The steel targets were
10.2 cm (4 inches) thick, and heat treated to a hardness of 24 to 37 on the
Rockwell C scale (HRC). Unfortunately, there was some inconsistency in the
heat treatment of the steel targets, and the hardness varied from the target
impact surface to the back surface over values differing by 10 or more on the
Rockwell C scale. A number of tests were already performed when this was
discovered, so the hardness is not consistent through the different targets.
Nevertheless, our experimental data suggest that this does not create a large
error.

The target was chained down to a V-block, and braced from behind by a
heavy spring to act as a momentum trap. Special care was taken in aligning
the block so that the composite rod impacted normal to the face of the block;
the yaw for all tests reported here is 0°. Aluminum was chosen for greater
penetration depth and deeper embedding of the penetrator; the penetrator
would remain intact after the impact, and the behavior of the composite during
the event could be studied more accurately. Steel is a better approximation for

armor commonly found in real-world applications.

116

3495
14.5
rare
\x

59
3359
349
3339
349

composite rod

3399
65
339
ote
ose

DISD
{$395
3399453945
3399253965
{$9535

powder a

charge

3 OOD
\ OOD
canot SOO

%)
X33
ses

target block

Fig. 4.9. Forward ballistic schematic.

The sabot used was longer than the one for reverse ballistics. Initially, a
short sabot was used, but there were problems with stability of the rod. A
number of shots using a slightly different configuration resulted in no
penetration due to tumbling of the penetrator prior to impact. For tests into
aluminum targets, the composite rods were 3.81 cm (1.5 inch) long, and 0.635
cm (0.250 inch) in diameter; thus the length-to-diameter ratio (L/D) was 6. For
the tests against steel targets, the rods were 5.08 cm (2 inches) long and 0.635
cm (0.250 inch) in diameter; for these tests, L/D=8. The density of the
composites was 17.3 +0.2 g-cm™. The variation in density was due to the
variation in the exact fiber fraction for each piece.

The sabot remained with the penetrator until impact; it was not stripped
away during flight. The sabot encased 3.05 cm of the penetrator, so that a
significant portion of the rod was left exposed beyond the length of the sabot in

order to allow penetration to begin before the sabot impacted the target. As in

117

the previous tests, the composite rod was a right circular cylinder with flat
ends.

A C200 steel backing plate was added to the nylon portion of the sabot for
forward ballistic testing. Later, a 17-4 hardened stainless steel was used. The
particular alloy was not essential, as long as it was strong enough to support
the penetrator rod during acceleration. A schematic of the sabot is shown in
Fig. 4.10. The nylon portion of the sabot was split into two parts. The rear
part was bored out to allow the backing plate to sit flush inside it. Then, the
front nylon portion of the sabot was drilled out to a slip fit with the rod and
inserted into the rear portion. It was determined that without the backing
plate, during the acceleration of a shot, the rod would punch through the nylon
behind it. The backing plate was made thick enough so that the composite rod
would not deform it significantly during acceleration. Even if the rod only
dented the backing plate, it was enough to skew the penetrator off axis and

create tumbling before impact.

steel
backing
plate

nylon | | [xp foo Oe eee

Fig. 4.10. Sabot for forward ballistics.

118

Another change in the forward ballistic testing configuration was the free
flight of the projectile before striking of the target. There was approximately a
12 cm gap between the end of the barrel and the face of the target. At these
high velocities, it did not present a stability problem; the distance of free flight
did not allow sufficient time for tumbling to occur. As in the reverse ballistic
tests, the forward ballistic tests were performed under rough vacuum.

The WHA penetrators used with steel targets were the same dimensions as
the composite penetrators; that is, 5.08 cm (2 inches) long and 0.635 cm
(0.250 inch) in diameter with L/D=8. This removes any effects from different
penetrator sizes or aspects ratios in the results for these tests. The WHA
used is designated X-27C and is produced by Teledyne Firth Stirling. Selected
properties of this alloy are shown below in Table 4.2. The samples used were
provided by R. J. Dowding of the Army Research Laboratory at Aberdeen, MD,

and are used as a standard material for ballistic testing.

Property X-27C
Composition (wt %): Tungsten 90.73
Nickel 4.55
Tron 1.97
Cobalt 2.75
Density (g/cc) 17.45
Ultimate Tensile Strength (ksi) 171
Elongation (%) 11.9

Table 4.2. Selected properties of the tungsten heavy alloy X-27C.18

119

4.4 RESULTS
4.4.1 Results of Reverse Ballistic Testing

After the shot, the steel mounting block, penetrator, and flyer plate were
fused together. To analyze the penetration behavior, they were cut along their
cylindrical axis, and the photograph in Fig. 4.11 was taken. In Fig. 4.11, the
penetrator is upright in the center of the photo, and the flyer plate is on top.
The sabot melted and disintegrated upon impact. The large horizontal crevice
separates the flyer plate from the steel mounting block. Extensive
deformation from the impact is visible at the outer edges of the flyer plate.
Also, a hole in the middle of the flyer plate from the penetrator is visible.
Following cutting, enough stresses were relieved that the pieces could be
separated. A top view of half of the flyer plate alone is shown in Fig. 4.12. In
this photo, the hole is more obvious, although some debris still obscures it
slightly.

The most important part of Fig. 4.11 to notice is the shape of the penetrator
tip. The tip appears to come to a point, and is indicative of self-sharpening
behavior. In addition, the hole in the flyer plate appears to be approximately
the same diameter as the original penetrator diameter; this indicates that
there was no mushrooming of the penetrator head during impact. These
results imply that the composite undergoes self-sharpening behavior, but they

are not conclusive.

4.4.2 Results of Forward Ballistics into Aluminum Targets
Four successful shots into aluminum targets were made using forward
ballistics, all at different velocities. Following the shot, the aluminum block

was cut axially around the penetrator, and then split open to avoid cutting into

120

penetrator tip

flyer plate

Fig. 4.11. Photograph from reverse ballistic testing of tungsten wire rein-
forced metallic glass composite: cut-away view of penetrator and flyer plate

following reverse ballistic test.

Fig. 4.12. Photograph from reverse ballistic testing of tungsten wire
reinforced metallic glass composite: top view of half of flyer plate following

reverse ballistic test. Hole from penetrator is visible at top middle of plate.

121

the penetrator or the penetration path in the target.

Fig. 4.13 shows the embedded penetrator in the aluminum target for a shot
at 605 m/s. In all penetrator photographs, the initial impact between the rod
and target occurred on the left side, and as the penetration occurs, the rod
traveled to the right through the target material. Note the tip of the
penetrator is chiseled down to a sharp point, despite starting with a flat head.
The flat rear end of the penetrator rod is visible about halfway down the tunnel.
Also note the constant diameter of the hole bored out by the penetrator. There
is a slight indentation on the entry surface of the target from the impact of the
sabot and the backing plate. The rectangular light areas to the right of the
penetrator tip are artifacts from the cutting open of the block. The portion of
the target immediately around the penetration tunnel appears lighter and
rougher because it was split by force rather than cut.

Fig. 4.14 shows the embedded penetrator in the aluminum target for a shot
at 749 m/s. Also, this photograph is only of one quarter of the target block; two
cuts had to be made to actually find the tip of the penetrator. Thus, the
penetration tunnel appears smaller than it actually is. Again, the tip of the
penetrator is chiseled down to a sharp point. The penetration channel is not
as smooth as the previous case, but there is still no evidence of mushrooming
of the head. Significantly more of the penetrator has been eroded away, as we
expect for the higher velocity. In addition, near the end of the penetration, the
rod began to travel off axis. This probably reduced the total penetration depth
in this sample, and is not surprising that the stability against yaw is reduced
when the penetrator has been eroded down to a short rod. Penetration studies
using a tungsten-heavy-alloy rod showed similar results in a number of cases;

these tests will be further discussed later in this section. The penetrator is less

122

Fig. 4.13 Photograph of composite penetrator embedded in aluminum target

block. Shot in forward ballistic configuration at 605 m/s. L/D=6. P/L=1.29.

Fig. 4.14 Photograph of composite penetrator embedded in aluminum target

block. Shot in forward ballistic configuration at 749 m/s. L/D=6. P/L=1.21.

123

intact than for the previous case, probably because of the higher velocity and
the greater amount of yaw during penetration.

Fig. 4.15 is a photograph of penetration of a composite rod at 1032 m/s into
an aluminum target. Again, the penetrator tip is sharp, and some tilting off-
center has occurred, but it appears only at the end of the penetration. Note
that the rod is almost completely eroded away; the back of the rod can be seen
about 7/8 of the way down the penetration tunnel. Small bits of the penetrator
can be seen embedded in the side of the tunnel.

Fig. 4.16 shows a rod fired at 1257 m/s. Virtually all of the penetrator has
been eroded away; a small bit can be seen at the very bottom, and it has tilted
at the very end of penetration so that it sits almost 90° from its entry angle. A
substantial crater is visible from the sabot, as well as an even deeper one from
the backing plate. The penetration channel is not as clean and smooth as for
the lower velocity shots. In fact, about halfway down the channel, there are
four large gouges approximately evenly spaced around the axis. In each of the
gouges is a significant amount of composite material. It appears that outer
portions of the penetrator split off at this point, and the inner core continued to
burrow deeper. Judging from the amount of penetrator left from this sample,
this is probably very near the maximum penetration depth achievable with
this initial rod length; increasing the velocity will not be effective with no
penetrator left intact.

As a comparison, Fig. 4.17 shows the penetration of a rod of the tungsten
heavy alloy (WHA) W-Fes,-Ni;,4 into an aluminum target. This sample was
one of a number of tests performed by Sunil Yadav at Caltech. The behavior of
the tip is clearly different than that of the composite penetrators. The tip of

the embedded penetrator is much larger than it was initially, and extremely

124

Fig. 4.15. Photograph of composite penetrator embedded in aluminum target

block. Shot in forward ballistic configuration at 1032 m/s. L/D=6. P/L=2.17.

3.81 cm=Lo

Fig. 4.16. Photograph of composite penetrator embedded in aluminum target

block. Shot in forward ballistic configuration at 1257 m/s. L/D=6, P/L=2.33.

125

Fig. 4.17. Photograph of WHA (W-Ni;z ¢-Fe, 4) penetrator embedded in
aluminum target block. Shot in forward ballistic configuration at 694 m/s.

L/D=6. P/L=1.06. Test was performed by S. Yadav.

126

rounded. The test parameters were nearly identical, although the WHA rods
were 2.0 inches long, and 0.33 inches in diameter. They, like the composite
rods tested, have a L/D of 6, and were shot at a number of comparable
velocities, so they can still provide a basis of comparison. For rods of the same
L/D, we can calculate a normalized penetration depth, which is the penetration
depth divided by the initial length. Fig. 4.18 shows a graph of the normalized
penetration depth of the composite rods and the WHA rods. Linear fits have
been made for each, and are also shown on the graph. The fit for the composite

is significantly higher than that of the WHA.

2.5 ;
O Composite ea

A WHA — 1 27

ea -
aan A
1.5 se
LL Lo?
P/L 0 O-7
1-- = -A
f LID =6
0.5 Al target
500 700 900 1100 1300
Velocity (m/s)

Fig. 4.18 Penetration efficiency versus impact velocity. Target was aluminum
(6061 T651). Penetrators used were flat-ended rods with L/D=6. Composite
was 80 vol% uniaxial tungsten wire in a metallic glass matrix. WHA was the
liquid-sintered tungsten heavy alloy with composition in weight percent of

Wo3Nis gFe;.4. WHA experiments by S. Yadav.

127

4.4.3 Results of Forward Ballistics into Steel Targets

Three shots of metallic glass composite penetrators and three shots of
WHA penetrators were successfully performed into steel targets. The targets
with the embedded penetrators were cut open by a combination of electrical
discharge machining (EDM) and band sawing. EDM was used on the
penetrator itself to minimize damage during cutting.

Fig. 4.19 is a photograph of a steel target after impact with a metallic glass
composite rod. The penetrator was shot at 760 m/s. The penetration was
quite shallow, as we would expect for such a low velocity and hard target
material. The penetrator did not even remain embedded in the target. Fig.
4.20 shows the results of a similar test with an impact velocity of 956 m/s.
The head of the penetrator has not mushroomed extensively, although it is not
as clearly self-sharpening as in the case with the aluminum targets; the tip
has expanded to only about 1.5 times its initial diameter. Near the tip, a line
along the penetrator at about 45° from its axis is visible where the wires in the
rod begin to bend over to be sloughed off. Fig. 4.21 shows a test performed at
1253 m/s. Again, the tip is only about 1.5 times the original diameter. The
striated appearance of the tunnel is due to the lining of tungsten wires. It
appears that the wires in the outer layer of the penetrator get bent around
180° and are left nearly intact on the wall of the tunnel.

For comparison, the next three figures show the results of the shots made
with penetrator rods of X-27C WHA. Figs. 4.22, 4.23, and 4.24 show
photographs of WHA penetrators fired at various velocities. The tips are
larger and mushroom out more noticeably than in the composites, creating
wider tunnels through the target. Correspondingly, the normalized depth is less
than for the composite penetrators. In Fig. 4.24, the head of the penetrator

128

Fig. 4.19. Photograph of metallic glass matrix composite penetrator in 4130
steel target. V=760 m/s, L/D=8, and P/L = 0.312.

Fig. 4.20. Photograph of metallic glass matrix composite penetrator in 4130
steel target. V=956 m/s, L/D=8, and P/L = 0.469.

129

Fig. 4.21. Photograph of metallic glass matrix composite penetrator in 4130
steel target. V=1253 m/s, L/D=8, and P/L = 0.812.

130

Fig. 4.22. Photograph of WHA (X27C) penetrator in 4130 steel target.
V=860 m/s, L/D=8, and P/L = 0.328.

Fig. 4.23. Photograph of WHA (X27C) penetrator in 4130 steel target.
V=932 m/s, L/D=8, and P/L = 0.375.

131

Fig. 4.24. Photograph of WHA (X27C) penetrator in 4130 steel target.
V=1269 m/s, L/D=8, and P/L = 0.734.

132

has expanded to about twice the original diameter of the penetrator. A graph
comparing the P/L ratios for the two different materials into steel targets is
shown in Fig. 4.25. Like for the aluminum targets, the penetration efficiency is

greater for the metallic glass composite penetrators.

O Composite -
A WHA

0.8

0.6 —| -°

P/L -7
0.4 — --

0.2 ~— 4130 steel target
L/D =8

700 900 1100 1300
Velocity (m/s)

Fig. 4.25. Penetration efficiency versus impact velocity. Target was hardened

4130 steel. Penetrators used were flat-ended rods with L/D=8. Composite was

80 vol% uniaxial tungsten wire in a metallic glass matrix. WHA was the X-27C
liquid-sintered tungsten heavy alloy with composition in weight percent of

Wo0.73Nig.55Fe1.97C 09.75.

133

4.5 CONCLUSIONS

Tungsten-wire-reinforced metallic-glass-matrix composites show promise as
a new kinetic energy penetrator material. They can be made with high
densities comparable to those of currently used alloys and exhibit more self-
sharpening effects in penetration tests than one of the best tungsten heavy
alloys. It has been reported in the literature that the WHA typically exhibits
limited self-sharpening behavior, while pure tungsten exhibits no self-
sharpening.! The metallic glass matrix composites would probably fall into the
category of self-sharpening, or at least a high degree of limited self-sharpening.

For velocities of about 1 km/s and for penetrator aspect ratio L/D of six to
eight, the data in Figs. 4.18 and 4.25 show that the metallic glass composite
has approximately a 20 percent improvement in penetrator efficiency over the
best WHAs. This improvement is comparable to the improvement in
penetrator efficiency provided by depleted uranium alloys. Figs. 4.26 and 4.27
plot the penetration data from this chapter against other published results
using similar test parameters. On these plots, the data obtained at Caltech is
overall somewhat higher than that from the references, but this may simply
be an effect of a lower L/D ratio, which is known to result in higher P/L
values.!? In addition, the different target materials used in the data shown in
Fig. 4.27 created a significant difference in the P/L values. To correct for this,
the Caltech data has been scaled: the P/L values were reduced so that the
WHA values in each data set matched. This provides a more accurate
comparison for the metallic glass composite and DU penetrator efficiencies.

There are a number of differences between the series of forward ballistic
tests presented in this work and traditional penetrator testing. One is the

presence of the sabot and the flyer plate behind the penetrator on impact. It is

134

| | Al target oOg Oo
Solid symbols: L/D = 6
Open symbols: L/D = 10 Ou
3-
eogy
P/L 2-
A a
& enh @ Metallic glass composite
Ij A WHA (Caltech)
O O WHA (Hohler & Stilp)
0 l
0 1000 2000 3000 4000
Velocity (m/s)

Fig. 4.26. Penetrator efficiency of metallic glass and WHA penetrators versus
aluminum targets. L/D values shown for the penetrator. Hohler and Stilp data

from ref. 19.

135

1.2
Metallic glass
@® composite-
L SCALED SP
WHA (Caltech)- O oO
SCALED
oO
0.84] O DU (Magness)
O WHA (Magness) Oo
Se)
P/L 0.6- ea
9 O a
0.44 0
e Solid symbols: 4130 target
L/D=8
0.2- | Open symbols: RHA target
e & LID = 20
0 l r l !
600 800 1000 1200 1400 1600 1800

Velocity (m/s)

Fig. 4.27. Scaled penetrator efficiency of metallic glass composite, WHA, and
depleted uranium penetrators versus steel targets. Caltech data scaled down
so that WHA data from both sources match. WHA tests performed at
Caltech with X-27C alloy. WHA used by Magness was liquid-sintered 93%
tungsten. DU alloy is U-0.75 Ti. Magness data from ref. 4.

136

conceivable that extra kinetic energy was transferred to the penetrator on
impact, thereby increasing the penetration depth. However, in both the tests
with aluminum and steel targets, the WHA penetrators gave a suitable
reference point for evaluation of the composite penetrator performance. In the
aluminum target tests, a backing plate was used, but it was a slightly different
size than the one used to test the composites. However, in the tests with steel
targets, the entire configuration of the sabot was identical for the composite
and the WHA. In addition, if the backing plate truly drove the rod into the steel
target, the backing plate itself should have significant deformation. This was
not observed.

Further work could include trying to fabricate full scale penetrators and
perhaps full scale ballistic tests. In addition, making the composite from
different metallic glass alloys, particularly from those without beryllium, would
provide advantages in lower cost, simpler processing, and easier material
handling. Increases in density would be possible by ensuring close-packing
with the wires and using a hafnium-based version of Vitreloy™ 1. It is known
that hafnium can be used to replace zirconium in the alloy with no detrimental
effects to the glass-forming ability.?? Through these techniques alone, the
density of the composite could be increased up to a theoretical limit of 18.6
gem”; 18.4 g-cm™“would be well within reach. Increasing the density would

certainly increase the penetration efficiency as well.

10.

11.

12.

137

REFERENCES

W. D. Cai, Y. Li, R. J. Dowding, F. A. Mohamed and E. J. Lavernia,
Reviews in Particulate Materials 3, 71-132 (1995).

T. W. Penrice, Progress in Powder Metallurgy (edited by M. S. Nayar, S. M.
Kaufman and K. E. Meiners), 39, 507, MPIF, Princeton, NJ (1984).

R. J. Dowding, P/M in Aerospace, Defense, and Demanding Applications,
(edited by F. H. Froes), 25, MPIF, Princeton, NJ (1993).

L. S. Magness and T. G. Farrand, Proceedings of the 1990 Army Science
Conference, 465 (1990).

K. J. Tauer, R. J. Dowding and P. Woolsey, Tungsten and Tugnsten Alloys-
1992 (edited by A. Bose and R. J. Dowding), 525, MPIF, Princeton, NJ
(1992).

J. Mescall, in Computational Aspects of Penetration Mechanics (edited by
J. Chandra and J. E. Flaherty), 47, Springer-Verlag, New York (1982).

L. S. Magness, by Tungsten and Tungsten Alloys-1992 (edited A. Bose and
R. J. Dowding), 15, MPIF, Princeton, NJ (1992).

M. R. Staker, Acta Metall. 29, 683 (1981).

D. A. Shockey, in Metallurgical Applications of Shock-Wave and High-
Strain-Rate Phenomena (edited by L. E. Murr, K. P. Staudhammer and
M. A. Meyers), 633, Marcel Dekker, New York (1986).

Smithells Metals Reference Book, 14.1, Butterworths, London (1983).

Binary Alloy Phase Diagrams, ASM International, Materials Park, OH
(1990).

A. Peker and W. L. Johnson, Appl. Phys. Lett. 63, 17, 2342-2344 (1993).

13.

14.

15.

16.

17.

18.

19.

20.

138

H. A. Bruck, T. Christman, A. J. Rosakis and W. L. Johnson, Scripta
Metall. Mater. 30, 4, 429-434 (1994).

H. A. Bruck, A. J. Rosakis and W. L. Johnson, J. Mater. Res. 11, 2, 503-
511 (1996).

T. H. Courtney, Mechanical Behavior of Materials, McGraw Hill, New York
(1990).

R. Busch, A. Masuhr, E. Bakke and W. L. Johnson, MRS 1996 Fall Symp.
on Glasses and Glass Formers (1997).

Metals Handbook, American Society for Metals, Metals Park, OH (1985).

R. J. Dowding, K. J. Tauer, P. Woolsey and F. S. Hodi, The Metallurgical
and Ballistic Characterization of Quarter-Scale Tungsten Alloy
Penetrators, MTL TR 90-31 (Watertown, MA, U.S. Army Materials

Technology Laboratory, 1990).

Hohler and Stilp, A penetration mechanics database, SwRI 3593/001 (San
Antonio, TX, Southwest Research Institute, 1992).

W. L. Johnson, personal communication (1997).

139

APPENDIX 1: W interface diffraction
measurement

0.74 in dia.

0.74 in dia.
1.05 in dia.
1.27 in dia.
1.47 in dia.
1.65 in dia.
1.95 in dia.

#4738-amorph.

9.4 mm r (2.21 A)<110>

13.3 mmr (1.56 A)<200>
16.1 mmr (1.29 A)<211>
18.7 mm r (1.11 A)<220>
21.0 mm r (0.99 A)<310>
24.8 mmr (0.84 A)<321>

#4743 rxn. + W

0.315, 0.73, 1.04, 1.26, 1.635, 1.95 in dia.

0.74, 1.05, 1.276, 1.47 in dia

#4742 amorph + rxn.

#4737 W only

0.315 in dia.->5.19 A (110 for Be ,2W->100 rel. intensity)
but no other lines appear (like 61 rel. ints. <101> line @ 3.65 A)
MOST LIKELY shadow of SAD aperture.

140

.600 in. dia-> APPENDIX 2
.870 in. dia.->

1.06 in dia.->

1.23 in dia.->

1.357 in. dia.->

1.635 in. dia->

.61 in dia.
0.86 in dia

8 um off interface 1.065 in dia.
into steel

-605 in. dia.->

-858 in. dia.->
1.062 m. dia.-> steel/metallic glass interface
1.24 in. dia->
138 in. dj 2.021.A
| 1.38 in. dia-> 6/5/96

| 1.487 in. dia->

_ 1.625 in. dia.->

dia.(in)->r(mm)->d spacing(A)

0.605 in->10.24 mm->2.03 A <110>
~ 0.87 in->14.73 mm->1.41 A <200>

1.5 um off interface into stl 1.065 in.->18.03 mm->1.15 A <211>

1.24 in->21.00 mm->0.99 A <220>
61 in dia. 1.37 in.->23.20 mm->0.89 A <310>
| 88 india 1.487in->25.18 mm->0.82 A <222>
ta 1.635 in->27.69 mm->0.75 A <321>
| 1.07 in dia
1.23 in dia
1.385 m dia all images at 75% scale
1.635 in dia 1200 mm nominal camera length
1054 mm actual camera length
.5 um off interface into stl
.603 in dia.
.88 in dia
1.065 in dia
1.248 in dia
1.38 in dia
1.63 in dia

ig. 3.18: 0.5 um SAD-
#5093 EDX sample

1.487 in dia (light)

i ght off interface