ABSTRACT
Title of thesis: A DEMONSTRATION OF GLASS BONDING
USING PATTERNED NANOCOMPOSITE
THERMITES DEPOSITED FROM FLUID
Juan Carlos Rodriguez, Master’s of Science, 2015
Thesis directed by: Professor Michael Zachariah
Department of Chemistry and Biochemistry
Ceramics and other nonmetals are widely used in industrial and research applica-
tions. Although these materials provide many advantages, they often pose unique
challenges during bonding. This work aims to expand on current processes, which
have much narrower applications, to find a more universal method for nonmetal
bonding. We utilize inks comprised of aluminum-based nanoenergetics (a heat
source) and tin (a bonding agent). Requirements for successful bonding are ex-
plored and four key criteria are established. Through statistical simulation and
thermochemical equilibrium calculations, we conclude that the presence of a diluent
in large percentages negatively impacts reaction kinetics. Conversely, we show small
percentages of added tin enhance gas generation and drive faster reaction rates. The
bulk bonding material, thermite plus tin, forms a continuous structure during reac-
tion, adhering well to the substrate surface. In some cases, these bonds failed above
1200 kPa.
A DEMONSTRATION OF GLASS BONDING USING
PATTERNED NANOCOMPOSITE THERMITES DEPOSITED
FROM FLUID
by
Juan Carlos Rodriguez
Thesis submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Master’s of Science
2015
Advisory Committee:
Professor Michael Zachariah, Chair/Advisor
Professor Bao Yang
Professor Christopher Cadou
c© Copyright by
Juan Rodriguez
2015
Acknowledgments
I would like to thank Dr. Zachariah for all his of guidance and suggesting this
stimulating research topic, as well as Alex Tappan for his input throughout this
process. I am also appreciative of the support of my family and co-workers who
made this experience enjoyable. I would like to express my deep gratitude to San-
dia National Laboratories and the National Physical Science Consoritium for their
financial support. I would also like to acknowledge the support of the Maryland
NanoCenter.
ii
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction To Thermites . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 A Brief History of Welding and Joining With Thermites . . . . . . . 5
1.2.1 Joining Ceramics and Non-Metals . . . . . . . . . . . . . . . . 5
1.2.2 Welding and Joining With Nanomaterials . . . . . . . . . . . 9
1.3 Bonding Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1 Pressure Cell Measurements and Thermochemical Modeling . . . . . 15
2.2 Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.1 Electrospray . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Direct Write . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Combustion Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Bond Strength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1 Pressure Cell Measurements and Thermochemical Modeling . . . . . 28
3.2 Combustion Velocity Results . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Bonding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Stress Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
iii
List of Tables
1.1 Overview of common thermite reactions, reproduced from [14] . . . . 2
2.1 Summary of reactant properties . . . . . . . . . . . . . . . . . . . . . 15
iv
List of Figures
1.1 Maximum enthalpies of combustion of selected energetic materials,
reproduced from [11] . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Diagram of proposed bonding system requirements . . . . . . . . . . 12
2.1 Diagram of constant volume experiment, reproduced from [40] . . . . 16
2.2 Number of layers vs. total weight, thickness, and width . . . . . . . . 23
2.3 Profile of ideal line . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4 Profile of rounded line . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Profile of line with channel and material pushed to side . . . . . . . . 24
2.6 Profile of line with channel and flat top . . . . . . . . . . . . . . . . . 24
2.7 Burn speed experimental setup as viewed by the camera . . . . . . . 25
2.8 Diagram of stress test . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.1 Pressure and optical signals for combustion of pure Al/WO3 . . . . . 29
3.2 Pressure and optical signals for combustion of Al/WO3 + 100% tin . 29
3.3 Peak pressure magnitude (experimental and calculated) and adiabatic
temperature (calculated) as a function of added tin . . . . . . . . . . 30
3.4 Relationship between peak pressure time and peak optical time . . . 32
3.5 Burning time as a function of added tin. . . . . . . . . . . . . . . . . 33
3.6 Combustion velocity as a function of thickness for 0%, 50%, and 100%
added tin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.7 Interface diagram for an ideal six particle system . . . . . . . . . . . 36
3.8 Interfaces, as a percentage of ideal, vs. added tin . . . . . . . . . . . 37
3.9 Average length of tin string as a function of added tin . . . . . . . . . 37
3.10 Figures of substrates post reaction . . . . . . . . . . . . . . . . . . . 38
3.11 Large hole in bonding material structure caused by excessive gas gen-
eration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.12 Thermal profile into substrate . . . . . . . . . . . . . . . . . . . . . . 41
3.13 Heat of reaction vs. added tin . . . . . . . . . . . . . . . . . . . . . . 42
3.14 Continuous bonding material layer is porous on a micron-scale . . . . 43
3.15 Tungsten silicon system phase diagram . . . . . . . . . . . . . . . . . 44
3.16 EDS line scan of substrate surface bonding . . . . . . . . . . . . . . . 45
3.17 EDS map showing tungsten silicon alloying away from substrate surface 45
3.18 Shear strength vs. thickness for 50% and 100% added nanotin . . . . 47
3.19 The bonding material remains attached to both sides of the substrate
during the first type of failure mechanism. . . . . . . . . . . . . . . . 48
v
3.20 Thermal fratures seen in substrates (1) . . . . . . . . . . . . . . . . . 49
3.21 Thermal fratures seen in substrates (2). . . . . . . . . . . . . . . . . . 50
3.22 Fragments of adjacent substrate remains bonded to bulk bonded region. 50
vi
Chapter 1: Introduction
1.1 Introduction To Thermites
The thermite reaction was first described by German Hans Goldschmidt in 1897 and
later patented as a means of producing metals and alloys. Goldschmidt noted the
reaction was “...very energetic...” to the point that material was “...thrown out of
the crucible...” [17]. A thermite reaction takes place between a metal (the fuel) and
a metal oxide (the oxidizer). The reaction forms a more stable oxide and releases
a significant amount of heat. This general reaction is depicted in Equation 1.1. In
this equation M is the fuel (metal or alloy), with corresponding oxide MO, and A
is the oxidizer (metal or alloy), with corresponding oxide AO. The heat released is
denoted by ∆H [44].
M + AO→ MO + A + ∆H (1.1)
The amount of heat released and the adiabatic reaction temperature are largely
dependent on the fuel and oxidizer combination and are summarized in Table 1.1
(reproduced from [14]) for several common combinations. These high-energy reac-
tions are used in a variety of military and civilian applications including bulk metal
welding [18], cutting torches [16], hand held incendiary devices [38], and large scale
1
incendiary bombs [20].
Tab. 1.1: Overview of common thermite reactions, reproduced from [14].
Constituents Adiabatic Reaction Temperature Heat of Reaction
T (K) −Q
(
cal
g
)
−Q
(
cal
cm3
)
2Al + 3CuO 2843 974.1 4976
2Al + Fe2O3 3135 945.4 3947
2Al + Bi2O3 3253 506.1 3638
2Al + WO3 3253 696.4 3801
4Al + 3(O2)g 4019 7857 1.424
Reactant particle size has a large effect on the reaction rate and reaction mecha-
nisms. Goldschmidt was the first to note this principle in his original patent. He
stated “the aluminum should be employed in a finely-pulverized state if the result is
to be a success” [17]. Further research in this area demonstrated that the reaction
rate between the metal and the oxide is largely dependent on the mass diffusion
rate and therefore the particle size of the reactant species. Traditional thermite re-
actions take place between micron-scale fuel and oxidizer particles. Research shows
that these large particles result in a slow reaction rate [35,43]. In addition, micron-
scale particles are insensitive to ignition and have a significant ignition delay.
More recently, effort has been put into decreasing particle size. These efforts strive
2
to bring the reaction rates and ignition characteristics between metal and oxide par-
ticles closer to those of monomolecular high-energy materials. Monomolecular ener-
getic materials contain fuel and oxidizer species within the same molecule, separated
by angstrom-scale chemical bonds. These short diffusion distances allow for very
fast reaction rates and relatively little ignition delay. However, these monomolecular
energetic materials have relatively low energy density, both in terms of mass and
volume, compared to metal fuels. This is shown in Figure 1.1 (reproduced from [11]).
Fig. 1.1: Maximum enthalpies of combustion of selected energetic materials,
reproduced from [11]. Reprinted from Progress in Energy and Combustion
Science, Volume 35 / Issue 2, Edward L. Dreizin, Metal-based reactive nano-
materials, Page 142., Copyright (2008), with permission from Elsevier.
In an ideal scenario, the short diffusion distances of monomolecular materials would
be combined with the high energy density of metals. Nanothermites were developed
to meet this aim [7]. Nanothrmites consist of nanoscale metal fuel and metal ox-
ide particles that have the same general reaction shown in Equation 1.1. Although
3
nanothermites are not a perfect blend of the desired energetic properties, as they
are not as intimately mixed as monomolecular materials, they exhibit many desir-
able traits compared to their micron-scale counterparts. Many of the improvements
resulting from the decreased particle size are due to an increase in specific surface
area [24, 26, 34]. Particles on the order of tens of nanometers have nearly as many
atoms on the surface of the particle as in the center of the particle, allowing them
to react and propagate quickly [33].
Nanoscale metal particles often exhibit unique physical characteristics that dif-
fer from similar micron-scale particles. These characteristics include unique mag-
netic and optoelectronic properties [33] and lower melting and ignition tempera-
tures [13, 30]. Nanothermites exhibit several desirable improvements over micron
reactants including lower ignition temperatures and faster propagation, up to many
orders of magnitude faster than similar micron reactants [1]. The benefit of decreas-
ing reactant particle size reaches a point of diminishing returns caused by an inert
oxide shell present on most metal fuels. For example, a typical aluminum particle
has a 3 nm thick oxide shell. On the micron-scale this oxide shell is not important,
as it accounts for less than 0.1% of the total mass of a 50 µm particle. However,
as the particle size decreases, this oxide shell becomes a significant portion of the
total mass. For a 50 nm particle, this oxide shell accounts for 32% of the total
mass. Ideally, this oxide shell could either be eliminated or replaced with an ener-
getic alternative. Recent work has explored these possibilities but a widely adopted
method has not yet been established [9, 29].
4
1.2 A Brief History of Welding and Joining With Thermites
Hans Goldschmidt also pioneered the use of thermites for welding. Goldschmidt re-
alized the potential of thermite reactions to generate molten metal and impart large
amounts of heat to the surrounding material. Furthermore, the thermite reaction
is self-contained, needing little to no outside energy. Goldschmidt recognized the
potential application in construction and filed an additional patent for the use of
thermites to weld streetcar tracks in 1903 [18]. This process proved to be so simple,
inexpensive, and reliable that it is still widely used in the rail industry today [23].
Although the mechanism to weld steel with aluminum and iron oxide is well es-
tablished, current work continues to improve on this mechanism. Recent research
indicates the bond can be strengthened by combining the energy release of the ther-
mite reaction with the alloying of aluminum and nickel [28].
1.2.1 Joining Ceramics and Non-Metals
Extensive work has been performed to investigate and understand glass-to-glass and
glass-to-metal bonding [3, 4, 31]. Research efforts have demonstrated the ability to
bond glass to iron, silver, gold, and copper, as well as many other metals and alloys
[10]. Glass has been bonded to the surface of metals to prevent wear caused by acid,
abrasion, and oxidation since ancient Egyptian times. Glass coatings are still widely
used in turbines and printed circuit boards [10]. One type of glass to metal bonding is
mechanical bonding. Two proposed mechanisms, dendrite and electrolytic, postulate
5
that the rough surface of the metal substrate allows the glass to flow and form an
interlocking structure. The dendrite mechanism suggests metallic dendrites form
in the glass during the reaction between the oxygen in the glass and the metallic
element [21]. This is shown in Equation 1.2 for a reaction between cobalt monoxide
and iron, reproduced from [10].
CoOglass + Fesubstrate → Codendrite + FeOglass (1.2)
The electrolytic mechanism postulates the rough surface is the result of a corrosion
mechanism [39]. This mechanism builds on the reaction proposed in the dendrite
mechanism, but adds two additional steps. The additional reaction steps are shown
in Equations 1.3, 1.4, and 1.5 , reproduced from [10].
2Coprecipitates + O2from atmosphere → 2Co
2+ + 2O2− (1.3)
2Co2+ + 2e− → Co (1.4)
Fe− 2e− → 2Fe2+ (1.5)
It is believed this mechanism promotes the dissolution between the glass and metal,
forming an interlocked key structure at the interface.
Another method of glass-to-metal bonding is chemical bonding. This method pro-
poses a transition region between the bulk glass and the bulk metal where the
mixing of the two species occurs [25, 42]. Unlike mechanical bonding, this region
is homogeneous and the strength comes from chemical bonds, rather than key or
dovetail type mechanical joints. These glass to metal bonds generally occur by ad-
hesion chemistry between the metal oxide and the glass surface [10]. In the most
6
simple case, a metal oxide forms on the metal surface by a redox reaction with
the glass. Then, the metal oxide diffuses into the glass surface, producing a region
of intermixed glass and metal oxide. A similar process involves first doping the
bulk glass with the corresponding metal oxide. When the interface is subjected to
elevated temperatures, the highly mobile atoms diffuse between the two adjacent
surfaces creating a region of equilibrium and resulting in a strong chemical bond [25].
When bonding dissimilar materials, matching the coefficient of thermal expansion
is of paramount concern. Dissimilar coefficients induce unnecessary stress on the
bond during heating. Glasses are often categorized in two categories: “soft” glasses,
with low softening temperatures and large coefficients of thermal expansion, and
“hard” glasses, with higher softening temperatures and lower coefficients of thermal
expansion. Glass to glass bonds can be accomplished by using a third glass with
a lower softening temperature. This “softer” glass is used much like a lead based
solder in the joining of metals [10].
Ceramics and other non-metals are used widely in industry. In particular, there is
increased interest in joining silicon wafers. The most common method of joining
wafers is direct bonding. In direct bonding, also called silicon fusion bonding, the
inherent tendency of clean, mirror polished surfaces to adhere to each other is en-
hanced when exposed to heat (in excess of 1000 K) and pressure [37]. A derivative
of direct bonding called anodic bonding, uses large voltages (500 V) to bond silicon
to sodium rich glass materials such as Pyrex [8]. Anodic bonding allows the sub-
7
strates to remain several hundred degrees cooler than traditional direct bonding.
Two wafers, one with a sputter deposited coating of Pyrex, are heated to a temper-
ature close to 700 K. A voltage of roughly 50 V is applied to join the wafers. One
possible limitation of anodic bonding is its inability to bond materials with a sili-
con dioxide layer. Recent work attempts to extend the usefulness of direct bonding
by investigating near room temperature joining using the gold/silicon eutectic [45]
and processing in ultrahigh vacuum [19]. All of the previously introduced methods
require a surface roughness of approximately 1 nm.
Some non-metals require special considerations when bonding, such as an external
bonding agent in addition to an energy source. Some progress has been made using
independent heat sources and bonding agents. In these bonds, the surface is typi-
cally media blasted. A layer of solder is then sputter-deposited and the surface is
machined flat. A multilayer nanofoil, capable of generating heat through an alloy-
ing reaction, is placed between the prepared surfaces as a heat source [5]. Although
this approach has been proven effective, it requires extensive pre-processing and its
use is limited to substrates sturdy enough to withstand the necessary preprocess-
ing. Additionally, nanofoils are typically manufactured in solid sheets and must be
individually cut to the desired shape.
8
1.2.2 Welding and Joining With Nanomaterials
Investigations into welding and joining with nanomaterials is very much a new fron-
tier. Several recent studies have investigated their use in specific applications, but
the fundamentals are still being established. One of the most promising methods
utilizes energy release through the alloying reaction of aluminum and another metal
as discussed previously in Section 1.2.1. This alloying reaction uses multilayer metal
foils to provide heat for welding, soldering, and brazing metals and select metallic
glasses [41].
Recent research has shown promise in the joining of dissimilar materials through
the use of disks comprised of nanothermite compositions of Al-NiO-Cu mixed to a
stoichiometric ratio of 4 [2]. In his work, Zanjani utilized disks with a diameter of
3 mm and a thickness of 0.7 mm. The disks were used to join copper wires to glass
substrates. The system was heated to approximately 700 K to initiate a thermite
reaction. Although it is shown the bond was capable of supporting a glass substrate
of around 3-5 g, extensive strength testing was not conducted. Furthermore, this
study relied on an additional heat source (external to the thermite reaction) to ele-
vate the system temperature.
Nanomaterials have also been used to join copper wires. This method of joining
makes use of a paste composed of water and silver nanowires. One advantage of
this technique is the ability to keeps the bulk material at room temperature. This
9
method is able to exclude organic compounds during bonding, forming ultra-low
resistance bonds [32].
The majority of the methods discussed require extensive preprocessing in tightly
controlled environments and have limited utility. We aim to extend Goldschmidt’s
idea of a simple and universal thermite bonding mechanism to modern applications.
The development of a single “ink”, containing both a heat source and a bonding
agent, would be ideal in many of these applications. This ink could be easily and
accurately deposited on many types of surfaces and would ignite readily and react
quickly. This reaction would provide adequate heat to promote bonding and would
be accurate enough to bond only the desired areas. The ink would be composed of
the heat source (the thermite) and the bonding agent (the diluent). The addition
of diluents to thermites is a demonstrated concept. Thermite additives are widely
used in the pyrotechnics industry [27] and are also used to perturb the thermite
reaction to study the effects on propagation and reaction rate [15,27,36].
1.3 Bonding Criteria
Traditional metal bonds are formed by creating a continuation of the substrate ma-
terial. Here, we consider a different type of bond. This bond, a “sandwiched bond,”
consists of a bulk bonding material ‘sandwiched’ between the substrate material on
either side, as seen in Figure 1.2. Although this approach to bonding does not pro-
vide the advantage of a continuous material, it allows for the bonding of materials
10
that would otherwise prove very difficult. Furthermore, after the bulk bonding mech-
anism and ink formulations have been established, the bonding ink can be used in
many different applications with only minor modifications to optimize heat transfer
to the substrate material and promote bonding. With these properties understood
for several materials, dissimilar material bonding will be very straightforward. We
propose four criteria for the successful bonding of two substrates:
1. A reactant system must not be chosen from the systems known to produce
gas at the adiabatic flame temperature or in final products.
2. A reactant system must have tailored heat release rate and quantity to the
specific substrates being bonded.
3. The product species must have a continuous bonding agent network through-
out.
4. The system must have a mechanism to bond to the substrates being bonded.
The first proposed criterion is the production of little to no gaseous species. A large
percentage of gas negatively impacts the bonding mechanism in two ways. First, the
expansion of gas in a confined area imposes unnecessary stress on the substrates,
potentially compromising the material strength by either cracking or completely
shattering the substrate. More importantly, gas generated within the reacting ma-
terial expands quickly, often forcing the surrounding material outside of the desired
bonding area. Some of the gas may also condense, depositing metal outside of the
bonding area as discussed later in Section 3.3.
11
Fig. 1.2: Diagram of proposed bonding system requirements: 1) little to no
gas generation, 2) tailored heat transfer characteristics, 3) continuous bonding
structure, 4) substrate/bonding binding.
.
The second proposed criterion is the ability to tailor the heat transfer characteristics
to the specific substrates being bonded. Depending on the substrate material, dif-
ferent heat transfer characteristics may be required. In the case of soda-lime glass,
little to no heat transfer is desired as an excess of energy leads to thermal fractures
along the surface. In contrast, bulk metallic substrates are not nearly as brittle and
may tolerate, or even require, a larger amount of heat transfer. These heat transfer
characteristics must be able to be tailored with regards to the both the amount of
heat released and the rate at which it is released. These characteristics correspond
to total enthalpy and burning time. One established advantage of nanothermite and
nanofoil bonding is the ability to direct a large amount of energy to only the sur-
face of a substrate. These methods allow a thin region near the surface to become
thermally mobile enough to alloy and bond, while maintaining safe temperatures
12
throughout the remaining substrate.
The third proposed criterion is the formation of a continuously interlaced bond af-
ter the reaction has propagated. In an ideal bonding scenario, a completely dense
bonded region would form, providing maximum strength and conductivity across the
bond. To create this dense region, the bonding agent must be present in adequate
quantities to fill voids between the substrate materials. Additionally, the bonding
agent should have a low melting temperature to ensure adequate mobility and re-
duce the amount of heat lost to the bonding agent during the reaction. After the
reaction front has propagated, the bonding agent must form a continuous structure
between both substrates and along the length of the bond. Finally, the bonding
agent must have adequate strength once bonded to withstand stresses placed on the
material.
The fourth proposed criterion is an adhesion mechanism between the bonding mix-
ture and the substrate’s surface. Some substrates, such as copper and printed circuit
boards (PCBs), readily adhere to potential bonding agents (such as tin) and easily
provide a bonding force between the substrate and bulk bonding material. Other
substrates, such as soda-lime glass and bulk silicon, require an additional mech-
anism to promote adhesion between the bulk bonding material and the substrate
surface. This adhesion mechanism can take the form of an alloying reaction between
a component in the bonding mixture and the substrate’s surface.
13
Chapter 2: Methods
This work focuses on the bonding of soda-lime glass substrates. The primary moti-
vation for this was the easy availability of flat substrates at low cost, and the ability
to image through the substrate. Microscope slides, purchased from a local chem-
istry supply store, were used as the substrates in this study. The transparent nature
of soda-lime glass facilitated optical combustion velocity measurements through the
adjacent substrate. Additionally, it allowed for visual inspection of the bonded joint.
soda-lime glass substrates are naturally flat allowing for even deposition of the ink
and ensuring good contact with the substrate surface. Although several other sub-
strates, such as silicon, are of great interest to the scientific community, this work
aims to provide a framework for aluminum nanoenergetic based bonding inks that
can be modified for many applications.
Aluminum nanoparticles with an 80 nm diameter were purchased from Novacentrix
for use as the metal fuel. Aluminum is a popular choice for fuel given the stability
of alumina, the corresponding oxide, its low cost, and high combustion enthalpy.
Thermogravimetric analysis of the bulk aluminum indicated that it contained 64%
reactive aluminum and 36% inert alumina by mass. Tungsten (VI) oxide (tungsten
14
trioxide, WO3) with a 100 nm diameter was purchased from Sigma-Aldrich and
used as the the oxidizer. Tungsten trioxide produces relatively little gaseous species
while maintaining a high adiabatic temperature and reasonable enthalpy. Tin was
chosen as the bonding agent due to its low melting temperature and latent heat
of fusion. The low melting temperature allows the tin to easily flow and form
continuous morphologies while the low latent heat of fusion means the tin has a
minimal thermal load on the reacting system. Tin requires only 4% of the energy
per mass to melt compared to silicon. Tin with a diameter of 100 nm was purchased
from Sigma-Aldrich and used as the bonding agent in this study. These selections
are summarized in Table 2.1.
Tab. 2.1: Summary of reactant properties.
Item Purpose Size Manufacturer
Aluminum Fuel 80 nm Novacentrix
Tungsten Trioxide Oxidizer <100 nm Sigma-Aldrich
Tin Bonding Agent <100 nm Sigma-Aldrich
2.1 Pressure Cell Measurements and Thermochemical Modeling
The effect of the tin diluent on the nanothermite reaction is investigated through
constant volume combustion experiments. Pressure and optical signals are simulta-
neously recorded during the combustion of a known amount of sample in a constant
volume chamber. The experimental setup is identical to that used by Sullivan, the
15
diagram is reproduced in Figure 2.1 [40]. The experiments maintained a constant
25 mg of Al/WO3 (the nanothermite) in stoichiometric proportions. Tin particles,
the bonding agent, are added to the 25 mg of nanothermite to examine the effect
on combustion. The amount of the bonding agent is given as a percentage of added
weight with respect to the base mass of nanothermite.
Fig. 2.1: Diagram of constant volume experiment, reproduced from [40].
The total mass during these tests increased as the percentage of added tin increased.
For the case of 50% added tin, the mixture is comprised of 5.6 mg aluminum (64%
active), 18.4 mg tungsten trioxide, and 12.5 mg tin for a total mass of 37.5 mg and
a specific volume of 0.00288 g/cm3 . The total mass of the sample in the constant
13 cm3 combustion chamber increases with the addition of tin while maintaining a
constant mass of nanothermite. Samples combined in glass vials with hexanes are
16
mixed in an ultrasonic bath for thirty minutes before air drying. Large agglomera-
tions are broken up with a spatula before using the dry powder. Constant volume
equilibrium calculations are made using the Cheetah thermochemical equilibrium
code with the JCZS library [22]. These calculations use the same specific volume as
the experimental values.
2.2 Deposition
This bonding technique depends greatly on a reliable deposition method. The ideal
deposition would be consistent, customizable, and yield thin lines. It is also desir-
able for deposition to occur as quickly as possible to maintain a homogeneous ink.
The deposition process needs to yield a homogeneous fuel/oxidizer bonding agent
regardless of thickness. Special considerations need to be made when dealing with
materials of varying densities or size, as denser particles are more likely to settle
out of the deposition ink, resulting in a heterogeneous deposition. Additionally,
consistent results are desired regardless of substrate. With these requirements in
mind, two deposition approaches are evaluated: electrospray and direct write.
2.2.1 Electrospray
Electrospray produces fine droplets from a liquid precursor by applying an electric
field. The liquid precursor in the needle is drawn to a point until the concentration
of charge forces the formation of a jet at the tip. The small particles produced
17
continue to shrink as the solvent evaporates. Eventually, the surface charge once
again forces the particle to fragment creating an expanding cone of dispersed parti-
cles. The major benefit of electrospray in this application is the inherent ability to
produce dry depositions. By depositing dry material, surface tension is eliminated
along with the drying effects which plague wet deposition techniques. Dry deposi-
tion also prevents particles from selectively settling out during the drying stage.
Two problems become apparent when considering electrospray as a potential deposi-
tion technique: the desire for accurate, thin lines and the need to deposit on a variety
of substrates - both conductive and insulative. One solution to these problems is
the use of a metal mask. The electric field can be generated between the needle and
mask allowing for a consistent spray, regardless of the conductivity of the substrate.
Additionally, the mask produces a consistent line width. Unfortunately, two issues
arise while using a conductive mask. First, the mask walls exhibit a shadowing
effect along the edges of the line (near the mask walls) where the charged particles
are attracted to the oppositely charged wall. This shadowing effect is exasperated
as the line increases in thickness or decreases in width.
Second, in order to produce a uniform layer of material across the width of the line,
the diameter of the spray must be 4-5 times the desired width. The need for a much
larger spray diameter greatly reduces the rate at which material can be deposited,
as the mass flow rate through the mask to the substrate is only a fraction of the
actual mass flow rate. This effect can be offset by increasing the mass loading of the
18
ink. However, a solids loading over 100 mg/ml is not practical, as the suspension of
particles becomes problematic. The additional mass of nanothermite on the mask
surface is also a safety concern. In general, the quantity of dry thermite being han-
dled should be reduced to the minimal necessary. This safety concern is worsened
by the presence of a high voltage creating the possibility of arcing to the mask and
igniting the material.
Another possibility to address the issues with electrospray is through the use of
an aperture ring. An aperture ring is a conductive ring placed between the needle
and the substrate. An electric field is applied between the needle and ring in order
to provide the necessary electric field and drive the electrospray process (as in a
conventional setup). The ring serves two purposes in this application. First, the
electric field is applied between the needle and the aperture ring eliminating the
need for the substrate to be conductive. Second, the aperture ring increases the
near-field region of the spray geometry allowing for a more accurate and confined
deposition. Generally, a narrow diameter spray requires a lower mass flow rate.
Although there is no longer a shadowing effect or an excess of material, as seen in the
masking approach, the low mass flow rates make the deposition time prohibitively
long. Duan et al. developed a process to produce narrow lines by operating in the
near field range of the spray [12]. They employed a volumetric flow rate of 0.2 ml/h
with a solids loading of 100 mg/ml. Using these parameters, it would take 60 minutes
to deposit a single line of ink on a microscope slide. This long deposition time is
particularly problematic when using unstable inks as the particles may fall out of
19
suspension during the deposition process.
2.2.2 Direct Write
Direct write is an additive manufacturing technique in which a stage and deposition
nozzle are moved in relationship to each other while the deposition nozzle discharges
material. Controlling the X-Y-Z location of deposition as well as the rate of mate-
rial deposition allows complex three-dimensional geometries to be fabricated. Two
types of direct write were explored: automated direct write and manual direct write.
The automated direct write system utilized a modified MakerBot Replicator fused
deposition modeling (FDM) platform. Each of the FDM nozzles was replaced with
a pneumatically driven 5 cm3 syringe. When material was not being deposited,
a negative pressure was applied to the syringes to produce bubbles and facilitate
constant mixing. Although the solutions needed to be stable for several minutes on
their own, the bubbling allowed for the use of a wider range of solutions. The direct
write process proved to be quick and reproducible. However, many problems arose
as a result of the wet deposition, including the presence of surface tension which
forced the lines into an nonuniform geometry during drying.
A simplified version of direct write is manual direct write deposition. To create
uniform line widths and straight edges, a mask is used. Blue masking-tape is ap-
plied to clean, dry slides and a relief is cut through the center. The ink is deposited
with a micropipette and allowed to flow and fill the relief. Once the solvent has
20
evaporated, the process is repeated to gradually build layers of material. Although
this method is more crude than the previously discussed methods, it still provides
consistent layer thicknesses and profiles.
This work used a manual direct write deposition method. This allowed more focus to
be placed on formulating inks and understanding the nanothermite/bonding agent
mechanism, rather than refining deposition techniques. Line reliefs were chosen to
be approximately 3 mm wide by 75 mm long. This geometry ensured the thickness
of the lines (between 10 µm and 150 µm) was an order of magnitude less than the
next smallest dimension. This allowed the problem to be considered one dimen-
sional. Inks were prepared at a concentration of 100 mg/ml. The total solids loading
changed with the addition of tin. However, the solids loading of aluminum tungsten
trioxide was held constant. In these experiments, it was of more importance that
the concentration of thermite be kept constant than the overall solids loading.
This approach utilized the experimental setup described in Section 2.1. Isopropal
alcohol was chosen as the solvent because of its polar nature, which allows powders
of varying densities to be held in suspension. The inks were mixed in an ultrasonic
bath for thirty minutes, breaking up agglomerations to achieve a uniform state. The
ink was extracted and deposited on the glass slides with a micropipette while the
vial remained in the ultrasonic bath. The water was changed regularly to maintain
near room temperature. The solvent was allowed to evaporate off the slide between
depositions. This process was repeated until the desired number of layers had been
21
deposited.
After the final layer was deposited, the slides were dried for several hours at room
temperature, to ensure all of the solvent had evaporated, before the masks were re-
moved. The slides were weighed before and after deposition to determine the mass
of material deposited. The line profiles were examined using a Tencor Alpha Step
200 profilometer to determine line geometry and thickness. Although line thickness
varied slightly along the length of the lines as well as between slides with the same
number of layers, the average line thickness and amount of material deposited in-
creased as the number of deposition layers increased. This can be seen in Figure 2.2.
By design, the line width remained constant regardless of thickness.
The ideal line profile has a flat top and straight walls with ninety degree corners,
as shown in Figure 2.3. This profile has the most contact area when sandwiched
with an adjacent line of the same profile. Less optimal is the line with a rounded
top, shown in Figure 2.4. The most common error seen in the line profile produced
by manual direct write deposition is a valley in the center of the line. This is the
result of the accidental dragging of the micropipette down the center of the line
during deposition. It is seen more frequently in lines with a higher number of de-
posited layers. The material from this channel is either pushed to one side (as seen
in Figure 2.5) or removed by the micropipette tip leaving a relatively flat top profile
(as shown in Figure 2.6). Despite a variation in profile, the combustion velocity
does not noticeably fluctuate. Even in the presence of these channels, the width of
22
Fig. 2.2: Number of layers vs. total weight, thickness, and width.
the line is roughly 10 times the thickness of the line, leaving the line thickness (the
smallest dimension) as the controlling dimension.
Fig. 2.3: Profile of ideal line Fig. 2.4: Profile of rounded line
23
Fig. 2.5: Profile of line with channel and
material pushed to side
Fig. 2.6: Profile of line with channel and
flat top
2.3 Combustion Velocity
Line combustion velocities were measured with high speed video using a Phantom
V12.1 camera. Combustion velocities were measured both with the line open to the
atmosphere on three sides and sandwiched between two soda-lime glass substrates.
Slides were prepared as discussed in Section 2.2. Two slides with similar line thick-
ness were sandwiched together with the lines facing each other. The slides were
positioned such that half of each slide was left exposed and half was in contact with
the adjacent slide.
Two spring loaded clamps (one on either side of the line) apply a total force of ap-
proximately 127 N to the substrates along the length of the line. It is important the
clamps apply a constant pressure throughout the duration of ignition and burning
to ensure the substrates remain in intimate contact with each other during the reac-
tion. The slides are positioned with the line face perpendicular to the camera lens.
24
The lines were ignited by a small piece of resistivly-heated wire with pure Al/WO3.
A shield around the ignitor was used to restrict ignition to a several millimeter
area of the desired line. Figure 2.7 shows the experimental setup as viewed by the
camera. For each test, three independent measurements were taken: the unconfined
speed of line A, the confined speed of line A plus line B, and the unconfined speed
of line B. Although the first line is observed by looking at the top of the deposited
line and the second line is observed through the glass slide at the bottom of the line,
a difference in combustion velocity was not observed. Measurements were made for
each of the three regions using a position vs. time calculation for the entire length
of the line.
Fig. 2.7: Burn speed experimental setup as viewed by the camera.
25
2.4 Bond Strength
The shear strength of the bonded slides is also of interest. To evaluate this, a
simple one-dimensional stress test was developed. This test loaded the bonded
slides parallel to the inked line. This loading scenario resulted in the determination
of the shear strength of the bond, expressed as τ = FA . Here F is the force and A is
the area. A diagram of the experimental setup is shown in Figure 2.8.
Fig. 2.8: Diagram of stress test.
The fixed substrate is secured to a stationary base while a loading sleeve is placed
over the top of the unfixed substrate. This loading sleeve is constrained to movement
along the length of the deposited lines. The sleeve is attached to an empty bucket
26
and water is added in a constant stream at approximately 1 liter per minute to
gradually load the system. The slide is loaded until a failure is observed (i.e., the
bond breaks) and the final weight of the bucket is measured with an American
Weigh hanging scale (110 × 0.05 lbs). The shear strength is calculated and the
testing apparatus is reset.
27
Chapter 3: Results
3.1 Pressure Cell Measurements and Thermochemical Modeling
The experimental pressure and optical signals are collected and the maximums are
compared with the calculated values. Examples of these recorded signals for 0%
and 100% added tin are shown in Figures 3.1 and 3.2. Among the values generated
by thermochemical equilibrium calculations, several trends should be noted. First,
the temperature decreases sharply with respect to the pure thermite value before
plateauing and remaining relatively constant as seen in Figure 3.3. This initial tem-
perature decrease corresponds to a loss of sensible energy as the tin vaporizes. As
the percentage of tin increases, the reaction no longer generates enough energy to
vaporize all the tin and the temperature remains at the boiling temperature of tin.
Secondly, peak pressure increases sharply and is maximized at 30% added tin. As
added tin increases beyond this point, peak pressure declines at roughly the same
rate. The peak pressure (at 30% added tin) corresponds with the leveling off of
temperature. This correlates with the transition from gaseous tin species to liquid
tin species. We conclude this inflection point corresponds with the fastest reaction
rate. Cheetah predicts that added tin beyond 100% (corresponding to equal parts
tin and nanothermite) will thermally quench the reaction and fails to converge at
28
an equilibrium solution.
Fig. 3.1: Pressure and optical sig-
nals for combustion of
pure Al/WO3
Fig. 3.2: Pressure and optical sig-
nals for combustion of
Al/WO3 + 100% tin
This relationship is not observed in the experimental pressure measurements. The
experimental and calculated peak pressures are normalized by the peak pressure
of pure Al/WO3 and shown in Figure 3.3. Since we are more concerned with the
overall trends and relationship between experimental and calculated, this scaling
more adequately captures the features of interest. When added tin is between 0%
and 30%, peak pressure is expected to rise rapidly. However, in the experimental
data it appears to remain roughly constant. Similarly, beyond 30% added tin, peak
pressure is expected to decrease, but is observed rising slightly and then remaining
constant. Additionally, the observed peak pressure remained elevated, relative to
the pure Al/WO3 system, for even large amounts of added tin. For small amounts
of added tin, the tin particles are in such low concentrations that they are likely
clustered in groups rather than homogeneously mixed as expected in the equilibrium
29
calculation. The minimal gas generation does very little for the reaction as a whole
because the tin is isolated in a limited number of locations. Although it is likely the
tin remains in similar sized clusters irrespective of the amount added, these local
tin pockets become more numerous as larger amounts of tin are added. Increasing
the number of these pockets has a greater impact on the global reaction kinetics.
Fig. 3.3: Peak pressure magnitude (experimental and calculated) and adia-
batic temperature (calculated) as a function of added tin.
Equilibrium calculations predict the reaction will become limited by the vaporization
of tin as larger percentages are added. This is realized on an experimental level as
well. Pure Al/WO3 has a peak optical emission time well before the peak pressure.
30
This behavior is examined by Sullivan et al. and is indicative of a reaction where
convective heat transfer is not the driving mechanism. Nanothermites with large
quantities of gaseous species (e.g., aluminum/copper oxide) exhibit peak pressure
either before or simultaneously as peak optical emission is achieved. To examine
the gas release timing of this nanothermite and bonding agent mixture, the rela-
tionship between peak pressure time (PPT) and peak optical emission time (POT)
is investigated with the ratio PPT/POT. Figure 3.4 shows this relationship for the
experimental studies. Ratios greater than 1 indicate that peak pressure occurs af-
ter the peak optical signal. Ratios less than 1 correspond to peak pressure being
achieved before peak optical emission. A ratio of 1 indicates the two are achieved
simultaneously. Pure Al/WO3 has a PPT/POT ratio of approximately 1.9. At 50%
added tin, the average PPT/POT ratio approaches one and remains approximately
one for increased amounts of added tin. As discussed previously, tin promotes gas
generation and the addition of this easily vaporized tin will slowly force a PPT/POT
ratio close to that of other gas-generating nanothermites. By examining Figures 3.1
and 3.2, it is evident the addition of tin reduces the PPT, while POT remains rel-
atively constant. Additionally, the ratio approaches one asymptotically indicating
that tin vaporization occurs with the bulk reaction rather than as a result of pre-
heating.
Total burning time was defined as the time between the optical signal initially reach-
ing 50% of the maximum intensity and when the signal eventually fell below 50%
intensity after the peak as shown in Figures 3.1 and 3.2. Burning time was deter-
31
Fig. 3.4: Relationship between peak pressure time and peak optical time.
mined during pressure cell testing and is summarized in Figure 3.5. The pressure cell
results show an initial increase in burning time for < 25% added tin. At this point,
there is a sharp decrease, reaching a minimum at or slightly after 25%. Burning
time subsequently increases as tin is added. The sharp initial increase in burning
time corresponds to the lack of a peak pressure increase seen for small percentages
of tin in Figure 3.3. As discussed previously, the tin is most likely agglomerated in a
few pockets, increasing the total thermal mass of the system and therefore burning
time. However, the presence of this tin does little to change global reaction kinetics.
The sharp decrease in total burning time aligns with the increase in peak pressure
32
observed around 25% added tin. For larger amounts of added tin, the burning time
increases steadily. The added tin is likely adding so much thermal mass, it is lim-
iting the reaction. Since it takes very little energy to fully melt tin, regardless of
the amount added, it is likely that all the tin is melted and the vaporization of tin
moderates the combustion velocity. This idea is supported with the plateau seen
among experimental peak pressure measurements.
Fig. 3.5: Burning time as a function of added tin.
33
3.2 Combustion Velocity Results
The combustion velocity data, shown in Figure 3.6, relates line thickness to propa-
gation velocity. This shows that the minimum thickness to propagate a reaction is
nearly constant and not a function of the percentage of nanothermite in the total
mass. Lines propagated selectively at a thickness of 15 µm and propagated with
a near 100% success rate by 25 µm. The propagation rate of compostions of pure
thermite (the blue line) and 50% added tin (the red line) increase initially with an
increase in thickness. The combustion velocity for compositions containing 100%
added tin (the yellow line) is relatively constant with regard to thickness. Composi-
tions containing 50% or 100% added tin both exhibit an increase in gas generation
compared to pure Al/WO3.
The data also indicate a thickness region where the combustion velocity of a com-
position with 50% added tin propagates faster than pure thermite. This region
corresponds to an increased amount of gas generation that promotes convective
heat transfer and results in faster combustion velocities. This thickness region also
corresponds to the strongest bond as discussed later in Section 3.4. We note that for
a constant thickness, the unconfined and confined burn speeds are identical. Adding
an additional substrate (e.g., sandwiching the line between another glass slide) does
not noticeably increase the minimum thickness required to achieve propagation. It
is possible a small increase in minimum thickness is needed, but not observed due
to the limited resolution available in these experiments.
34
Fig. 3.6: Combustion velocity as a function of thickness for 0%, 50%, and
100% added tin.
Presumably the large percentage of tin in the 100% added tin samples hinders the
kinetics of the reaction by decreasing the proximity of the fuel and oxidizer. It is
well established nanothermite reactions rely heavily on homogeneous mixing and a
large interface area between the fuel and oxidizer. When considering the pure ther-
mite case, with a constant particle size between all species, a stoichometric mixture
corresponds to 43% aluminum and 57% tungsten trioxide particles. These propor-
tions change to 30% aluminum, 39% tungsten, and 31% tin for a 50% added tin
sample. In the case of 100% added tin, there is 23% aluminum, 30% tungsten, and
47% tin. To achieve a more complete understanding of the possible kinetic impacts
of a diluent particle, a simple simulation study was conducted. We consider the
traditional scenario of sampling two different colors of balls from an urn without
35
replacement. Each ball represents a single species (either fuel or oxide) in the nan-
othermite system, where the initial probability of selection is proportional to the
species representation in the Al/WO3 system. The balls are exhaustively sampled
and the number of interfaces (i.e., any instance where a fuel is sampled immediately
before or after an oxide) is counted as shown in Figure 3.7.
Fig. 3.7: Interface diagram for an ideal six particle system.
Next, a percentage of diluent is added to the system and the simulation is repeated.
The results of this simulation are shown in Figure 3.8, which relates the average
percentage of interfaces to percentage of added tin. The percentage of interfaces is
given as a proportion of an ideally mixed sample, like the one shown in Figure 3.7.
An ideally mixed sample of a two species system would see one less interface than
number of particles. In Figure 3.7 there are 5 interfaces for 6 particles. The number
of interfaces between fuel and oxide particles decreases to only 25% of the ideal sce-
nario and nearly half of its initial value for a sample with 100% added tin. Based on
these results, it is expected a reaction which relies heavily on intimate mixing and a
large interface area between fuel and oxidizer species will be negatively affected by
36
large percentages of diluent particles. Additionally, we examine the average string
length of the diluent particle. The average length of the diluent particle is repre-
sentative of the ability of the tin to flow and coalesce with adjacent particles. As
the percentage of added tin increases, so does the the average length of the diluent
particle string as seen in Figure 3.9. This corresponds with larger clusters of tin in
the bonding agent mixture.
Fig. 3.8: Interfaces, as a percentage of
ideal, vs. added tin.
Fig. 3.9: Average length of tin string as
a function of added tin.
3.3 Bonding
The first criterion for successful bonding is the reaction should produce little to
no gaseous species. The justification for this criterion is given in Section 1.3. The
experiments showed increased gas production as the amount of added tin and/or
thickness increased. Both of these scenarios correspond to more tin in the system
that is available to be vaporized. Figures 3.10 shows images of several slides after
37
combustion velocity testing. The ideal, well-bonded case corresponds to a small total
mass of tin. The addition of tin, either by increasing the total amount of material or
percentage added, drives gas production. Figure 3.10 A, B, and C show depositions
of 50% added tin with thicknesses of 50 µm, 75 µm, and 100 µm respectively. For
comparison, pure Al/WO3 with the same thicknesses of 50 µm, 75 µm, and 100 µm
respectively is shown in Figures D, E, and F. It is clear gas generated by tin in large
amounts perturbs surrounding material and may create a gap in the continuous
bonding agent structure.
Fig. 3.10: Figures of substrates post reaction. Figures A, B, and C are
Al/WO3+50%Sn with thicknesses of approximately 50 µm, 75 µm, and 100
µm respectively. Figures D, E, and F are pure Al/WO3 with thicknesses of
approximately 50 µm, 75 µm, and 100 µm respectively.
Figure 3.11 depicts a scenario when excessive gas generation perturbed the sur-
rounding material, leaving a several hundred micron hole in the bonding material
structure.
38
Fig. 3.11: Large hole in bonding material structure caused by excessive gas generation.
The second criterion is the ability to tailor energy transfer for a specific substrate.
The main properties of interest are the depth at which a temperature increase will
be seen in the substrate and the slope of the thermal profile in the substrate. Both
of these properties can be investigated with basic thermodynamics. We will consider
the case of steady state heat flux into the substrate using the maximum value of
heat flux to establish an upper bound. This value is calculated by taking the total
amount of energy released and dividing by the total experimental burning time.
This assumption corresponds to a square wave of heat flux representing the reaction
zone. Consequently, the ramp at the leading and trailing edges of the reaction zone
will be lost. We will assume these portions of the reaction wave do not greatly effect
39
the average heat flux. Fourier’s law, given in equation 3.2, indicates how the energy
release rate effects thermal penetration. Here k is thermal conductivity, q˙ is heat
flux, and T is temperature. The X direction is defined as into the substrate.
q
t
=
k(Thot − Tcold)
∆X
(3.1)
∆X =
k∆T
q˙
(3.2)
Equation 3.2 shows the relationship between temperature change and distance into
the substrate. Since k can be considered relatively constant regardless of the per-
centage of added tin and the reaction temperature changes very little after 30%
added tin, the only parameter changing is q˙. The temperature change as a function
of depth into the substrate for q˙ = 1.0E6 Wm2 and q˙ = 2.5E5
W
m2 , comparable to re-
actions containing 50% and 100% added tin respectively, are shown in Figure 3.12.
Although no quantitative estimation of thermal penetration can be gathered from
these equations because many assumptions and simplifications were made, a quali-
tative relationship between heat flux magnitudes and thermal profiles can be estab-
lished. It is clear as the heat flux increases, the maximum distance into the substrate
that a temperature increase is experienced will decrease.
Thermochemical equilibrium calculations show the amount of heat release can be
modified by increasing the amount of tin in the reaction. Although this seems coun-
terintuitive since a diluent is being added to the system, the tin acts as a secondary
fuel and oxidizes with available oxygen. Total energy release increases to a maximum
at 30% added tin and then decreases, as shown in Figure 3.13 for thermochemical
40
Fig. 3.12: Thermal profile into substrate.
equilibrium calculations modeling the constant volume combustion chamber. Com-
bining the trends of burning time and heat release allows for the creation of ink
formulations with the same amount of total energy release, but with differing rates
of release.
The third criterion for successful bonding is a continuously interlaced bonding agent.
Tin, the bonding agent used in the previously described experiments, fulfilled this
criterion and provided a continuous structure on the macro-scale. We confirmed
the presence of a continuous structure by measuring the electrical resistance across
the entire length of the line. Despite a continuous path being present, we note
that images show the line is still porous on the micron-scale (see Figure 3.14). The
41
Fig. 3.13: Heat of reaction vs. added tin.
nanoscale tin particles agglomerate to form a bed of micron-scale particles sintered
together.
As discussed in Section 1.3, a mechanism for binding the substrate to the bulk
bonding material is required. In these experiments, we exploit the alloying reaction
between elemental tungsten and elemental silicon. Elemental tungsten is produced
after losing its oxygen to aluminum. Additionally, the aluminum scavenges oxy-
gen from silicon dioxide present on the surface of the soda-lime glass substrate,
yielding exposed elemental silicon. The feasibility of this reaction is tested with
thermochemical equilibrium calculations by reacting a composition of 80 wt. per-
cent stoichiometric aluminum tungsten trioxide and 20 wt. percent silicon dioxide.
42
Fig. 3.14: Continuous bonding material layer is porous on a micron-scale.
A significant portion of silicon dioxide is seen to reduce to elemental silicon while
maintaining an adiabatic temperature around 2500 K. Therefore, we posit this alloy-
ing reaction is forming a bond between the bulk bonding material and the soda-lime
glass substrate as seen in the phase diagram in Figure 3.15, reproduced from [6].
An SEM image of the cross section of a bonded slide provides more concrete evidence
of the proposed adhesion mechanism. The EDS line scan in Figure 3.16 indicates
tungsten silicon concentrations following a constant 6:4 ratio across the bonded
region until the edge of the substrate, where the concentration of tungsten begins to
decrease. This 6:4 ratio of tungsten to silicon corresponds with a eutectic point at
43
Fig. 3.15: Tungsten silicon system phase diagram [6].
2433 K. Spacial EDS mapping in Figure 3.17 shows micron-scale particles of tungsten
and silicon that have formed away from the substrate surface. The approximate
location of this micron-scale particle is outlined in blue in Figure 3.16. EDS mapping
corresponds to elemental intensity but not necessarily to atomic percentages. We
are unable to identify the specific alloy with certainty. However, the consistency of
the ratio is indicative of a single alloy, rather than a mixture of alloys.
44
Fig. 3.16: EDS line scan of substrate surface bonding.
Fig. 3.17: EDS map showing tungsten silicon alloying away from substrate surface.
3.4 Stress Testing
Bonded slide pairs are observed to exhibit two possible failure mechanisms corre-
sponding to the two regions of the bond. The first bond region is between the
45
bonding agent particles (e.g., between tin particles). The second region of the
bond is between the bonding material and the substrate surface (e.g., between the
Al/WO3/tin mixture and the soda-lime glass surface). The weaker of the two bond-
ing regions will fail first during loading.
The first failure mechanism depends solely on the strength of the bonding agent
and quality of the bond between bonding agent clusters. Figure 3.14 indicates the
reaction reaches the melting temperature of the tin, which allows it to locally melt,
flow freely, and coalesce. This creates a macro-scale bonded network of sintered
micron particles. The bonds between these micron particles failed under loading.
Improving the ability of local tin to flow and combine with tin in nearby regions
increases the strength in this region of the bond. The easiest solution is to increase
the percentage of added tin. This would decrease the distance liquid tin must flow
to join a neighboring melted mass. However, increasing the percentage of added tin
also increases the burning time and heat transfer to the substrate.
The second failure mechanism is dependent on the interface between the bulk bond-
ing material and the substrate surface. During the previously described experiments,
the bond created between the bulk bonding material and the substrate through the
alloying of tungsten and silicon proved to be very strong. We were unable to load
the sample until this bond failed due to other weaker links in the system. The
bonding reaction proved to physically compromise the substrate surface by creating
thermal fractures. The heat transfer to the substrate must be better controlled to
46
avoid this mechanism of failure in the future.
Experimental data, plotted in Figure 3.18, show the strongest bonds corresponds
to 50% added tin with a bond thickness between 50-60 µm. This thickness aligns
closely with the minimum thickness of material required to propagate a reaction.
Fig. 3.18: Shear strength vs. thickness for 50% and 100% added nanotin.
The strongest bonded pairs failed in the bulk bonded material, i.e., bonds between
tin particles pulled apart. Figure 3.19 shows a pair of substrates after a failure of the
bulk bonded material. After failure, the tin remained continuous on the surface of
47
each substrate. An electrical resistance can be measured along the length of this line.
Fig. 3.19: The bonding material remains attached to both sides of the sub-
strate during the first type of failure mechanism.
Higher percentages of tin and thicker bonding layers failed at the bond/substrate
interface. In general, the soda-lime glass experienced varying degrees of thermal
fracture as shown in the plan view Figure 3.20. Regions with thermal fractures
proved to be very weak and failed under loading. During failure, the thermally-
fractured regions would remain attached to the bulk bonded region but pull away
from the substrate.
After failure, the substrate surface has regions of exposed bulk bonding material
and exposed substrate. Figure 3.21 shows a plan view of the substrate surface after
such a failure. A region of exposed thermally-fractured substrate is visible on the
left and outlined in red. A second region still bonded to the now-fractured adjacent
substrate is visible on the right outlined in blue. These two regions are separated
by only 50 to a 100 microns of exposed bonding material.
48
Fig. 3.20: Thermal fratures seen in substrates (1).
Interestingly, the bond between the substrate and bulk bonding material proved to
be very strong and did not typically fail in either case. Figure 3.22 shows a fracture
cross section view of a substrate with bulk bonding material and broken sections of
the adjacent substrate remaining attached to the bulk material after failure.
49
Fig. 3.21: Thermal fratures seen in substrates (2).
Fig. 3.22: Fragments of adjacent substrate remains bonded to bulk bonded region.
50
Chapter 4: Conclusion
This work investigated the requirements to join nonmentals using aluminum-based
nanoenergetics. Requirements for successful bonding were explored and four key
criteria were established. First, a reactant system must not be chosen from the sys-
tems known to produce gas at the adiabatic flame temperature or in final products.
Second, a reactant system must have tailored heat release rate and quantity to the
specific substrates being bonded. Third, the product species must have a continuous
bonding agent network throughout. Lastly, the system must have a mechanism to
bond to the substrates being bonded. Based on these criteria, tungsten trioxide was
selected as the oxidizer species due to the small amount of gaseous species gener-
ated during its combustion with aluminum. Tin was chosen as the bonding agent
for its low melting temperature and low latent heat of fusion. The effect of a diluent
on the nanothermite system was explored through simulation and thermochemical
equilibrium calculations. The simulation indicated the number of interfaces between
fuel and oxidizer particles decreases to only 25% of the ideal scenario and nearly
half of its initial value for a sample with 100% added tin. It is clear a reaction that
relies heavily on a large interface area will be negatively affected by the addition
of large amounts of diluent. Furthermore, thermochemical equilibrium calculations
51
show small percentages of added tin promote gas generation and aid in reaction
kinetics. However, higher percentages of tin will hinder the reaction, decreasing the
adiabatic temperature until the reaction is eventually quenched. These theoretical
results were compared with constant volume combustion experiments. Experimen-
tal results indicate a small or even negative effect on the reaction for percentages
of tin less than 25%. An increase in gas generation and decrease in burning time
is shown for added tin between 25% and 100%. Formulations with greater than
100% added tin experience substantially longer burning times and reduced reactiv-
ity. Measurements of combustion velocity indicated little change between lines open
to the atmosphere and those confined with an adjacent substrate. Strength testing
of bonded slides showed 50 µm thick lines containing 50% added tin produced the
strongest bonds. In several cases, these bonds failed above 1200 kPa. The bonds
were seen to fail by two mechanisms. The first mechanism was within the bulk
bonding material, and failure occurred between large tin clusters. A second pos-
sible failure mechanism was between the bulk bonding material and the substrate
surface. Tungsten and silicon alloyed, producing a very strong bond which did not
fail during our tests. However, the substrates often experienced thermal fractures
which compromised the strength of the surface and lead to failure. This work acts
as a framework which can be built upon to establish a library of “ink” formulations
that can be used with a range of materials in a variety of applications. The ink
formulation technique presented here has also been used to bond bulk copper to a
soda-lime glass substrate. Although an in depth analysis of this application has yet
to be completed, this method holds promise for providing an effective way to join
52
nonmetal surfaces.
53
Bibliography
[1] B. Bockmon, M. Pantoya, S. Son, B. Asay, and J. Mang. Combustion velocities
and propagation mechanisms of metastable interstitial composites. Journal of
Applied Physics, 98(6):064903–064903, 2005.
[2] G. Bohlouli Zanjani. Synthesis, characterization, and application of nanother-
mites for joining. 2013.
[3] M. P. Borom and J. A. Pask. Role of adherence oxides in the development of
chemical bonding at glass-metal interfaces. Journal of the American Ceramic
Society, 49(1):1–6, 1966.
[4] J. J. Brennan and J. A. PASK. Effect of composition on glass-metal interface
reactions and adherence. Journal of the American Ceramic Society, 56(2):58–
62, 1973.
[5] G. Caswell. Nanobond R© assembly–a rapid, room temperature soldering pro-
cess. In Microelectronics and Packaging Conference, 2009. EMPC 2009. Euro-
pean, pages 1–6. IEEE, 2009.
[6] T. Chart. Thermodynamic properties of the tungsten–silicon and chromium–
silicon systems. Metal Science, 9(1):504–509, 1975.
54
[7] W. Danen and J. Martin. Energetic composites, Nov. 30 1993. US Patent
5,266,132.
[8] M. Despont, H. Gross, F. Arrouy, C. Stebler, and U. Staufer. Fabrication of
a silicon-pyrex-silicon stack by ac anodic bonding. Sensors and Actuators A:
Physical, 55(2):219–224, 1996.
[9] D. Dlott. Thinking big (and small) about energetic materials. Materials science
and technology, 22(4):463–473, 2006.
[10] I. Donald. Preparation, properties and chemistry of glass-and glass-ceramic-
to-metal seals and coatings. Journal of materials science, 28(11):2841–2886,
1993.
[11] E. L. Dreizin. Metal-based reactive nanomaterials. Progress in Energy and
Combustion Science, 35(2):141–167, 2009.
[12] H. Duan, C. Li, W. Yang, B. Lojewski, L. An, and W. Deng. Near-field elec-
trospray microprinting of polymer-derived ceramics. Microelectromechanical
Systems, Journal of, 22(1):1–3, 2013.
[13] J. Eckert, J. Holzer, C. Ahn, Z. Fu, and W. Johnson. Melting behavior of
nanocrystalline aluminum powders. Nanostructured materials, 2(4):407–413,
1993.
[14] S. H. Fischer and M. Grubelich. Theoretical energy release of thermites, in-
termetallics, and combustible metals. Technical report, Sandia National Labs.,
Albuquerque, NM (US), 1998.
55
[15] T. J. Foley, S. Son, V. E. Sanders, and B. Asay. The effect of added alumina
on the propagation behavior of the nanoaluminum and molybdenum (vi) oxide
system. In Proceedings of the Thirteenth International Detonation Symposium,
Norfolk, VA, pages 23–28, 2006.
[16] R. Givens, J. Klein, T. Burky, and J. Reuther. Thermite charge, July 7 2009.
US Patent 7,555,986.
[17] H. Goldschmidt. Method of producing metals and alloys., Mar. 16 1897. US
Patent 578,868.
[18] H. Goldschmidt. Process of metal-welding., Aug. 4 1903. US Patent 735,244.
[19] U. Go¨sele, H. Stenzel, T. Martini, J. Steinkirchner, D. Conrad, and K. Scheer-
schmidt. Self-propagating room-temperature silicon wafer bonding in ultrahigh
vacuum. Applied physics letters, 67(24):3614–3616, 1995.
[20] C. H and H. B. Incendiary bomb mixture, July 20 1948. US Patent 2,445,312.
[21] J. Healy and A. Andrews. The cobalt-reduction theory for the adherence of
sheet-iron ground coats. Journal of the American Ceramic Society, 34(7):207–
214, 1951.
[22] M. L. Hobbs, M. R. Baer, and B. C. McGee. Jczs: An intermolecular poten-
tial database for performing accurate detonation and expansion calculations.
Propellants Explosives Pyrotechnics, 24(5):269–279, 1999.
56
[23] D. Housed. Welding of railroad railsa literature and industry survey’. Rail
Steels: Developments, processing, and use, 644:118, 1978.
[24] A. Il’in, A. Gromov, V. Vereshchagin, E. Popenko, V. Surgin, and H. Lehn.
Combustion of ultrafine aluminum in air. Combustion, Explosion and Shock
Waves, 37(6):664–668, 2001.
[25] B. W. King, H. Tripp, and W. H. Duckworth. Nature of adherence of porcelain
enamels to metals. Journal of the American Ceramic Society, 42(11):504–525,
1959.
[26] Y.-S. Kwon, A. A. Gromov, A. P. Ilyin, E. M. Popenko, and G.-H. Rim. The
mechanism of combustion of superfine aluminum powders. Combustion and
flame, 133(4):385–391, 2003.
[27] J. MCLAIN. Pyrotechnics: From the viewpoint of solid state chemistry(book).
Philadelphia, Pa., Franklin Institute Press, 1980. 252 p, 1980.
[28] E. B. Motlagh, J. V. Khaki, and M. H. Sabzevar. Welding of aluminum alloys
through thermite like reactions in al–cuo–ni system. Materials Chemistry and
Physics, 133(2):757–763, 2012.
[29] R. G. Nuzzo, B. R. Zegarski, and L. H. Dubois. Fundamental studies of the
chemisorption of organosulfur compounds on gold (111). implications for molec-
ular self-assembly on gold surfaces. Journal of the American Chemical Society,
109(3):733–740, 1987.
57
[30] M. L. Pantoya and J. J. Granier. Combustion behavior of highly energetic ther-
mites: Nano versus micron composites. Propellants, Explosives, Pyrotechnics,
30(1):53–62, 2005.
[31] J. A. Pask and R. M. Fulrath. Fundamentals of glass-to-metal bonding: Viii,
nature of wetting and adherence. Journal of the American Ceramic Society,
45(12):592–596, 1962.
[32] P. Peng, A. Hu, A. P. Gerlich, Y. Liu, and Y. N. Zhou. Self-generated local heat-
ing induced nanojoining for room temperature pressureless flexible electronic
packaging. Scientific reports, 5, 2015.
[33] A. Pivkina, P. Ulyanova, Y. Frolov, S. Zavyalov, and J. Schoonman. Nano-
materials for heterogeneous combustion. Propellants, Explosives, Pyrotechnics,
29(1):39–48, 2004.
[34] P. Pranda, K. Prandova, and V. Hlavacek. Particle size and reactivity of alu-
minum powders. Combustion science and technology, 156(1):81–96, 2000.
[35] V. Rozenband and N. Vaganova. A strength model of heterogeneous ignition
of metal particles. Combustion and Flame, 88(1):113–118, 1992.
[36] J. Shen, Z. Qiao, K. Zhang, J. Wang, R. Li, H. Xu, G. Yang, and F. Nie. Ef-
fects of nano-ag on the combustion process of al–cuo metastable intermolecular
composite. Applied Thermal Engineering, 62(2):732–737, 2014.
[37] M. Shimbo, K. Furukawa, K. Fukuda, and K. Tanzawa. Silicon-to-silicon direct
bonding method. Journal of Applied Physics, 60(8):2987–2989, 1986.
58
[38] E. Song. Thermite destructive device, Dec. 16 1997. US Patent 5,698,812.
[39] H. F. STALEY. Electrolytic reactions in vitreous enamels and their relation to
the adherence of enamels to steel*. Journal of the American Ceramic Society,
17(1-12):163–167, 1934.
[40] K. Sullivan. Ignition, combustion and tuning of nanocomposite thermites. Pro-
Quest Dissertations and Theses, page 255, 2010.
[41] A. Swiston, T. Hufnagel, and T. Weihs. Joining bulk metallic glass using
reactive multilayer foils. Scripta Materialia, 48(12):1575–1580, 2003.
[42] A. P. Tomsia and J. A. Pask. Bonding of dental glass to nickel-chromium alloys.
Journal of the American Ceramic Society, 69(10):C–239, 1986.
[43] M. A. Trunov, M. Schoenitz, X. Zhu, and E. L. Dreizin. Effect of polymor-
phic phase transformations in al 2 o 3 film on oxidation kinetics of aluminum
powders. Combustion and Flame, 140(4):310–318, 2005.
[44] L. Wang, Z. Munir, and Y. M. Maximov. Thermite reactions: their utilization
in the synthesis and processing of materials. Journal of Materials Science,
28(14):3693–3708, 1993.
[45] R. Wolffenbuttel and K. Wise. Low-temperature silicon wafer-to-wafer bond-
ing using gold at eutectic temperature. Sensors and Actuators A: Physical,
43(1):223–229, 1994.
59