AlO bonding: a method of joining oxide optical components to aluminum coated substrates
ABSTRACT With the increased use of photonic packages, there are needs for
reliable and low cost methods of attaching optical components. Packages
based on silicon optical bench (SiOB) technology include oxide coated
ball-lenses and silica fibers which are generally epoxied in
anisotropically etched features of silicon substrates. The reliable
attachment of these micro-optical components requires the application of
small (approximately 1 nanoliter) quantities of epoxy at precise
locations on the substrate. This is a time consuming process and
requires considerable operator training and skill. Dispensing too much
epoxy can deteriorate the optical performance of the device and
dispensing too little results in an insufficient holding power. AlO
bonding is an alternative attachment technique, under development, which
forms solid-state bonds directly between these oxide-components and
aluminum thin film coated silicon optical bench substrates and therefore
does not require the handling of additives, such as epoxy, at the bond
interface. This paper includes: methods of bonding ball-lenses and
fibers; an interfacial analysis and proposed bonding mechanism derived
from SEM/metallographic photomicrographs and thermodynamic data;
destructive test results as a function of bonding and material
parameters and; in situ loss measurements through AlO bonded components
(multi-, single mode fibers and ball-lenses) during the bonding
procedures and subsequent thermal cycling (-40 to 80°C and from
ambient to -195.8°C) tests
AIO Bonding: A Method Of Joining Oxide Optical Components
To Aluminum Coated Substrates
William R. Holland, Casmir R. Nijander and Robert E. Woods
Albert M. Benzoni*,Mindaugas F. Dautartas*,Ranjan Dutta,
AT&T Bell Laboratories
Princeton,N.J. and *Breiningsville,PA
With the increased use of photonic packages, there are needs
components.Packages based on silicon optical bench (SiOB)
technology include oxide coated ball-lenses and silica fibers
which are generally epoxied in anisotropically etched features
of silicon substrates.The reliable attachment of these micro-
locations on the substrate.This is a time consuming process
performance of the device and dispensing too little results in an
insufficient holding power.
of epoxy at precise
AIO bonding is an alternative attachment
development,which forms solid-state bonds directly between
these oxide-componentsand aluminum thin film coated silicon
optical bench substratesand therefore does not require the
handling of additives, such as epoxy, at the bond interface.
This paper includes: methods of bonding ball-lenses and fibers;
thermodynamicdata; destructive test results as a function of
measurements through AIO bonded components (multi-, single
mode fibers and ball-lenses) during the bonding procedures and
subsequent thermal cycling (-40 to 80°C and from ambient to
The transition from electronic to optoelectronic .and optical
packaging has introducedpassive transmission
with material properties that are not directly compatible with
  Packages, based on silicon optical bench
technology,include silica fibers and oxide ball-lenses which
are epoxy bonded in chemically micro-machined
and pyramidal cavities of silicon substrates.
these components in position at their permanent sites for the
purpose of guiding and collimating optical signals in an optical
switch.  At present, the reliable attachment of a 300 micron
diameter ball-Iense requires the application of small quantities
(approximately1 nanoliter) of liquid epoxy in the cavities,
positioning the ball in the cavity and holding it in place without
rotation while curing the epoxy.
process and requires considerable
since it has been found that dispensing too much epoxy can
deterioratethe optical performance
dispensing too little results in an insufficient holding power. A
similar procedure is used to epoxy fibers in V-grooves.
Figure 1 shows
This is a time consuming
operator training and skill
of the componentand
This paper describes an alternative bonding method which
forms solid-state bonds directly between an oxide component
and an aluminized' silicon substrate and therefore does not
requirethe handlingand curing of additives
interface.The objectiveof this investigation
bonding oxide coated ball-lenses
simplicity presently used to join metal conductors in electronic
packages. For brevity,the
technique described herein is referred to as AIO bonding. The
report includes an interfacial analysis and mechanical testing of
the AIO bonds, and in situ loss measurements through various
AIO bonded componentsduring the bonding sequence and
subsequent thermal cycling tests.
at the bond
is aimed at
and silica fibers with the
There were three forms of <100> oriented silicon substrates
usedin this investigation.
chemically micro-machined substrates with 128 micron wide
parallel V-grooves to accommodate fibers, and substrates used
in the Silicon-Based Moving Mirror Optical Switch which
contain both V-grooves(107 microns wide) and pyramidal
cavities.These features are anisotropically etched into <100>
oriented silicon wafers using a thermally grown silicon dioxide
mask. The oxide mask is removed after etching the features.
thicknesses of I and 2.5 microns.
the sputtered aluminummetal and silicon was achieved by
using an in situ RF Ar+ ion beam cleaning, followed by the
sputterdepositionof the aluminum
nominal diameter of 300 microns.
of silicon dioxide was deposited by chemica! vapor deposition,
in this study were AT&T multimode and single mode fibers
having a core/clad diameter of 62.5/125 and 9/125 microns
respectively.Relevant material properties of an epoxy used for
attaching optical fibers and a commercially pure aluminum is
shown in Table I.
Optimal adhesion between
film in a MRC-603
The anti-reflective coating
 to a thickness of 2240 ± 200 A. The fibers used
III. BONDING PROCEDURE, MECHANICAL TESTING
AND EXAMINATION OF BONDED BALLS
In practice, AR coated sapphire ball-lenses and silica fibers
are attached to multi-facetedwalls of pyramidal cavities and
V-grooves in silicon.This multi-point contact strengthens as
well as accurately fixes their optical position relative to other
testing and interfacial analysis of an AIO bond, the formation
of a relatively fragile single area bond will first be described
between a 300 micron diameter AR coated sapphire ball and a
:-pla.narsilicon substrate coated with 2.5 microns of aluminum.
The AIO bonding of balls in cavities will then be shown.
A 2.5 micron thick layer of aluminum coated on a planar
silicon substrate is positioned over a small vacuum hole of a
heated base. The vacuum ensures that the substrate reaches a
setting of the heated base.The ball is then placed on the
substrate (Figure 2). After pre-heating the workpieces,
faced metal bonding tool is loaded on the top surface of the ball
fora pre-determined time.
conductivityof the oxide ball and tangential contact that the
tool makes on the ball results in an efficient condition
heating the bond region without requiring to heat the tool. If it
is necessary, a heated tool could be an effective way to reduce
the required base or substrate temperature.
for a given temperature
The bonded workpieces were then transferred to the base of
a commercial Vertical Bond Tester.1 A loop-shaped adhesive
tape was attached to the jaws of the tester, and then
lowered onto the top portion of the ball and stopped before the
tape made contact with the surrounding
Destructivelytested balls conveniently remained tacked to the
tape for examination. A lower than maximum pull-strength was
usually obtained since most bonded balls were partially peeled
during testing. Thiswas
observatioris during testing which showed that the adhesive
'contact was usually not at precisely the top of the ball.
substrate (Figure 3).
Figures4, 5, and 6 show the interface
tested pair of workpieces
force of 8 ±2 gms.The bonding parameters were: pre-heat
time~1 min; bonding temperature=
and; bonding time= 3 sec. Figure 4 shows a top view of the
ball tacked to the test-tape and the bond area off-center relative
to the circumferenceof the ball.
majority of bonded balls were partially-peeled
which would result in a lower test value than if it was pulled at
a 90 degree angle.For example, balls tested nearer to 90-
degrees had recorded strengths of 18 ±2 gms. Balls, bonded on
a planar surface and at a single contact spot, have survived an
expedient thermal cycling test where the bonded workpieces
Nere immersed in a liquid nitrogen bath (-195.8°C) for about 3
conditions,the substrate was tapped on a table and the balls
destructivelythat failed at a pull-
325°C;z load= 600 gms.
were removed to ambient
The substrate workpiece in Figure 5 shows a bond region
circular-shapedaluminum film which was reduced in cross-
section to ::;;1micron in thickness.
an annular-shaped ring of the underlying
significant amount of aluminum was removed. Metallographic
The film is surrounded by
1. A product of Engineering Technical Products, Inc. Z. Nominal bonding
tempratures reported in this report were determined by monitoring a
thermocouple with its junction positioned between the tool and substrate.
_ 3. Measurements were made using a Nikon Metallograph and attached
the centrally located aluminum film which was directly below
the ball. This indicated the extent of aluminum thinning that
took place in this region.The outer annular-shaped
(shown out of focus) is composed of an extruded aluminum
pile.Figure6 shows an annular-shaped
stuck on the surface of the mating ball having only traces of
aluminum in the central region.
removed from the substrate in the mid-region where the silicon
substrate was exposed(Figure
probableposition of a 300
penetrated2.5 microns into an aluminum film. A calculated
chordal length which intersects the radius at 2.5 microns is
equal to 55 microns. This closely compares to measured3 outer
dimensionsof the annular-shaped
shown in Figures 5 and 6 respectively.
where the surface of the penetrating ball contacts the aluminum
as it extrudes outwardedly and forms the outer pile. Fissures
are known to develop in an extended native aluminum oxide
coatings, as its underlying aluminum is deformed.  During
AIO bonding, the ball is expected to form contacts with the
exposed aluminum at the fissures.
data in the forms:
at l000x shows that the silicon is in-focus with
Its shape shows that it was
silicon and aluminum ring
It is in this mid-region
/)'G= -RT In K
and/)'G= A + BT log T + T
where: -/)'G=Standard Free Energy
K= Equilibrium Constant
indicates the stable as well as protective nature of the native
aluminumoxide coating (equ.-l-)
atmosphere.A reaction is also possible between a silica ball in
contact with the underlyingaluminum,
= 2AI + 1.5 Oz
and that exposed areas of
(equ. -2-) as shown
/)'G25C= +378 kcal
/)'G400C= +350 kcal
-2-2AI + 1.5 SiOz = Alz 03 + 1.5 Si
/)'G25C= -82 kcal
The integrity of the bond may further be enhanced with the
potential diffusionof silicon in aluminum.
equilibrium solubilities of Si in Al have been reported: 400°C
(0.28 at. %); 350°C (0.16 at %); 300°C (0.10 at %) and; 250°C
(0.05 at %). 
Less interfacial shear would be expected to take place in the
central region as a result of acting frictional forces between the
ball and aluminum film. This condition is similar to the onset
surface which commences just below the surface. [ ] During the
early stagesof film deformation
interfacial flow and therefore sticking just below the ball would
be expectedas shown in the central bond region (Figures 5 and
6). A further increase in the bonding parameter was observed
to increase the extent of adhesion below the center of the ball
where the film deformation spreads to the surface.
flow by a hard indenter compressin?a metallic
a minimal amountof
Figure 8 shows a destructive bond interface on the substrate
which was formed at a lower temperature (220°C). Though the
bond was too lcw to pull-test, it revealed a less deformed or
indented aluminum film which closely replicated
texture of the ball.Examination
under these conditions showed significantly
aluminum. The deformation and sticking of the aluminum film
at the lower bondingtemperatures
increasing the bonding load or time.
D. Bonded Balls In Cavities
of balls that were bonded
less sticking of
would be increasedby
accurately fix their optical position as well as form multi-point
contacts for added strength. Figure 9 show balls that were
bonded in a cavity of a 2.5 micron thick aluminum film on a
silicon substrate. It was found that these bonded balls could not
be destructively tested,as previously
adhesive repeatedly failed leaving the ball intact. The repeated
stressing of AIO bonds as the adhesive failed· on the same
bonded ball was typically 16 ±2 gms. In order to examine the
bond interfaces, the ball was pushed out of the cavity with a
probe. Figures10 and 11 show the dislodged
cavity respectively.Figure 12 shows a higher magnification of
a bond region on the silicon wall. It shows a greater amount of
aluminum piled in the lower contact region which implies that
aluminum occurs during bonding.
shear that occurs against the angular wall of the cavity would
be expected to enhance the integrity of the bond compared to
forming it on a planar surface as previously described. Figures
11 and 12 also show that more aluminum remained where the
ball made its closestcontact with the silicon wall as the
aluminum throughout the bond area on the silicon wall with a
higher percentage localized in the off-center region (Figure 12).
Figure 12 and a highermagnification
aluminum stuck on the ball also indicated that the aluminum
was reduced to ~0.25 micron where the ball made its closest
contact with the silicon wall.
foundwhen the ballwas bonded
previously described. Figure 14 shows the aluminum deposit
on the surface of the ball. Its shape clearly shows that it was
removed from the silicon wall (Figure 12) as well as showing
where the spot of closest contact occurred.
have also been pushedoff with a probe tester positioned
resulting in typical values of 35 ±3 gms. These values were
also obtained after pre-stressing
tester as described above.
IV. MECHANICAL SHOCK AND VIBRATION TESTING
OF ALO BONDED BALL-LENSES
noted, balls bonded in pyramidalcaVIties
Dall and the
The additional interfacial
(Figure13) of the
This was similar to what was
AIO bonded balls
the bond with the vertical
AIO bonded balls were prepared for shock and vibration
cavities were coated with 2.5 microns of aluminum.
coated ball-lenses were bonded in each of six substrates.
aluminum alloy test carriers were used to secure the substrates.
Three substrates were epoxied to the face of each carrier. The
test carriers, each containing 12 AIO bonded lenses, were then
subjected to shocks up to 2000G's
20G vibration levels from 10 Hz to 2 KHz on all axes. All 24
bonded lenses remained bonded under these test conditions
(0.5 ms) on all axes and
resulting in a 100% yield. The test levels used in this run met
or exceededmilitary specifications
Figures15 and 16 showone
completing the shock and vibration tests.
V. AIO BONDING SILICA FIBERS
Epoxy bonding fibers in V-grooves of silicon substrates
feasible. This is a result of the high temperature properties of
an AIO bond and the ability to localize or pattern generate the
deposited aluminum film. Since the bare fiber is composed of
silica, the next phase of this investigation explored a method to
AIO bond fibers and test their mechanical
function of the bonding and material parameters.
A. Fiber Preparation
mechanically stripped (about 0.5 inches) at one end. Next, the
ends were wiped several times with a methyl alcohol soaked
tissue (to remove the residual coating) and then wiped dry.
lengthsofpolymer coatedfibers were
B. Bonding Fibers
grooves and coated with 2.5 microns of aluminum was fixed by
vacuum to the heated base of the bonder. After pre-heating the
substrate, the stripped end of the fiber was positioned in a V-
groove and held until a flat faced bonding tool was loaded on
the fiber for a given time. The end of the fiber extended beyond
the tool to ensure reasonable consistency in the bonded length.
The bonded fiber wouldtherefore
aluminum film whichwould not necessarily
maximum strength for a given bond length since a slight peel
action rather than a complete shear test was expected to occur.
Deposited aluminum lands or pads of a given dimension would
of course negate this concern and define the bond length.
should be noted that the polymer coated section of the fiber
was prevented from reaching its depolymerization
substratewith 128 micronwide parallel V-
tail-inand out of the
result in the
Figure 17 and 18 show regions of two bonded fibers where
the tool made contact during
residual polymer which remained on the surface of the fiber.
Figures 18 and 19 show smooth surfaces of bonded fibers with
no traces of the residual polymer
cleaning step prior to bonding.
portion of the fiber (shown in Figure 19) could be secured and
relieved of mechanical stresses by cementing or securing it in
the V-grooved substrate where the polymer coating begins.
bonding. Figure17 shows
due to a thorough
C. Test Procedure
horizontal stage of a commercial pull tester (Figure 20). The
free end of the fiber was then carefully z-positioned on a hard
wax surface which was predeposited on a metal tab clamped to
the pull-rod of the testing machine. The wax was momentarily
heat-softened to wet and stick the fiber to the metal tab for
pull-testing. The fiber was then pulled approximately parallel
to its axial length until a destructive failure occurred.
pull-strength and the mode of failure was recorded.
witha bondedfiber was clampedon a
D. Test Results
Individual Pull Strengths vs. Temperature
:bonding conditions are shown in Figure 21.
..modes of the plotted data points showed that the fiber remained
intact and that aluminum was stuck to their surfaces.
strengths are shown to increase with the bonding temperature.
This would be expected since the flow properties of aluminum
decrease with temperature which would result in an increase in
the real contact or bonded area.
strengths near one pound (454 gms) can be obtained
bonding temperatureof 300°C.
parameters indicated in Figure 21. A sample (five) of bonded
fibers wereputina temperature/humidity
(120°C/85%)for 100 hours and pull tested.
fibers remained in tact with strengths similar to those shown in
along with the
All the failure
Figure 21 shows that pull-
All the tested
Figure 22 shows additional data points corresponding
shorter bonding time of 3 seconds.
an increase in the bonding time appears to have more of an
effect on increasingthe pull-strength
temperatures.This could be explained by the inherent decrease
in flowproperties(or increase
deformation)and strain hardening
At the lower temperatures,
than at "the higher
index of aluminumwith
Figure 23 shows the same plot as in Figure 16 with the
addition of data points corresponding to a lower bonding force
of 500 gms. The data shows a lowering in the pull strength for
a given temperature compared to the points corresponding
the higher bonding force of 1200 gms.
explained by the flow properties of aluminum.
would be expected to produce higher strengths until the fiber
bottoms-outnear the silicon surface or one of the workpieces
Again this could be
Figure 24 shows the addition of a datapoint corresponding
destructivelytested fiber remained in-tact.
with a longer flat-faced bonding tool and indicates the effect of
bond length on the final pull strengths.
This was obtained
Figure 25 shows an aluminized silicon substrate where two
bonded fibers of different bond lengths were pull-tested.
remainingfibers were left intact.
322 gms. were obtained with values that were approximately
proportionalto their bond lengths.
correlationcould not be made since the same forces
therefore different interfacial pressures were actually applied.
The average nominal pull-strength in shear for these two fibers
was approximately11,700 psi which was determined from the
dimensionsof the indentation in the film. This value closely
compares to reported shear strengths of aluminum
temperature (Table I). A closer determination would require a
measure of the true contact area. Figure 26 shows one side of a
pulled fiber with two parallel bands of heavier aluminum stuck
on one side of a pulled fiber which was transferred from one
side of the angular wall (Figure 27). This condition was also
observedon the other side of the fiber and wall.
aluminum was found where the fiber made its closest contact
with the angular silicon wall.
bond regions examined below bonded balls on planar surfaces
and on angular walls of a cavities as previously described.
Strengths of 513 gms and
In this case a closer
This condition was similar to
sheared aluminum. This further indicated that the controlling
strengthmechanismof fibers pulled parallel
length is the shear strength of aluminum .
of the bond regions showed deformed and
to their axial
Figure 28 shows a plot of Pull-Strength
Thickness. It shows that there is an increas~ jn .the strength
with aluminum thickness.This could be explained
increase in the final bond area.
IV. OPTICAL LOSS MEASUREMENTS
We will now describe in situ loss measurements
Measurements will be made through the following
components: (1) a fiber bonded on a planar surface and in a V-
groove; (2) a set of spliced fibers bonded in a V-groove and:
(3) two fibers and two intervening ball-lenses bonded in-line on
substrates used in the 2x2 Silicon-Based
Optical switch package (Figure 1). In-situ loss measurements
ofbonded components during
procedures were also made.
A. Measurement Procedure
will be subjected
Themajority of measurementsthrough thebonded
components utilized three meter lengths of AT&T multi-mode
and single mode fibers connected
(1300 nm) and power meter. The accuracy of the reported
losses were ±o.OI dB, unless it was otherwise
fibers were wrappedfive times around a 1/2-inch diameter
teflon rod near the input end to achieve mode mixing and the
resultant stabilized output intensity signal was recorded.
losses (a) were calculated by measuring the output intensity
before (10) and after (If) forming the bonded components.
between an LED source
a= -lOLog (If / 10)
B. Thru Bonded Fibers
The center section (=10 cm) of a 3 meter length of multi-
mode fiber was stripped of its polymer coating and cleaned.
The ends were then connected to a 1300 nm LED source and
power meter.The stablized output signal was then recorded.
A 2.5 micron thick layer of aluminum
silicon substrate was fixed by vacuum to the heated base of the
bonder.After the substrate was preheated (~l min), the bare
portionof the fiber was positioned
bonding tool (35 mil square) was then loaded (1400 gms) on
the fiber for 13 sec. at a bonding temperature of 298°C.
bonded workpieces were then removed and cooled to ambient
temperatures.The intensityof the- output
constant(loss=0.00 dB) throughout
bonding at 298°C to ambient.
loss measurement experiment was made with the addition of an
expedient thermal cycling test. The bonded workpieces were
held by their extendingfibers "jump-rope-style"
immerse it in a liquid nitrogen bath (-195.8°C) for about 3 min.
The output signal indicated no change for each of the three
cycles. The same results were obtained by temperature cycling
(three times) a bonded fiber from a liquid water bath at O°C to
coated on a planar
on the substrate.The
A similar bonding and in situ
in order to
C. Spliced Fibers
A three meter length of a single-mode fiber was connected
to a 1300 nm LED source and power meter. After the output
signal was recorded, the fiber was cut in half. The polymer
coating of thefiber was
approximately 1 inch of each cut end.
cleaved with a hand held tool and cleaned again. An aluminum
thin film (2.5 micron thick) coated silicon substrate, with 128
micron wide V-grooves,was fixed by vacuum to the heated
base of the bonder. After pre-heating
minute, a fiber end was positioned in a V-groove and held until
the tool was loaded on the fiber at a temperature of 298°C. The
face of the tool was allowed to extend approximately
beyond the end of the fiber. The second fiber-end was butted
against the bonded fiber and compressed by the tool. The face
of the tool was again positioned so that it extended about 10
mils beyond the end of the previously bonded fiber. During the
second bonding cycle, the power meter registered an increase
in the output signal. There was no change in the intensity as
the bonded splice was removed
ambient.A set of five typical splices with an index matching
gel in place resulted in an average loss= 0.09 dB v.iith a range
from 0.05 to 0.20 dB.
The ends were then
the substrate for ~1
from the heatedbase to
Three of the above bonded
in a liquid nitrogen
described.The average additional room temperature change in
loss after the combined nine cycles was 0.01 dB. Also, there
were no detectablelosses monitored
nitrogen temperatures in each of the nine cycles. After the
cycling tests were complete, the extending fibers of one splice,
used in these experimentswere pull-tested
strengths of 336 gm. (bond length= 0.037-inch) and 241 gm.
(bond length= 0.027-inch).Both pull-tested fibers remained
intact with aluminum deposited on their surfaces.
from room to liquid
An earlier Ala bonded splice was mounted by taping the
extending fibers to a glass slide. The mounted splice was then
held "jump-rope style" by the extending fibers for immersion in
a liquid nitrogen bath. During the immersion cycle the glass
slide shattered while the attached splice remained intact and
continued to transmit the intensity signal with no detectable
loss. The quenching cycle~lls
holding the extending fibers for the immersion step. No loss or
change in signal was detected at room temperature as a result
of the three quenching cycles. Earlier thermal cycling (-80 and
40°C) of fibers spliced in a V-groove resulted in no change in
loss at room temperature after 33 cycles at which time the
coolant was exhausted.Similar results were obtained with
splices cycled in boiling water (33 min) and in ice water (33
min).In these early experiments
connected betweenan 850 nm LED source and a battery-
operated power meter (0.1 dBm resolution).
an SEM photomicrographof a spliced fiber bonded in a V-
groove. Figures 30 and 31 show a fiber bonded at its end by
overlapping the tool and its penetration into a 2.5 micron thick
repeated two more times by
multi-mode fibers were
Figure 29 shows
D. In-Line Bonded Fibers and Ball-Lenses
Two 300 micron diameter balls were placed into two in-line
cavities of an aluminizedsilicon
(Figure 1). A three meter length of a multi-mode
connected to a 1300 nm LED source and power meter. Two
the V-grooves behind each bonded ball (Figure 32 and 33.).
Duringbondingof the second fiber, an output signal was
detected.As in the case of splices, previously described, the
output signal did not change while it was transferred from the
heated base to room temperature.
bondedsubstrateswith an index
between each fiber and ball resulted in an average straight thru
loss= 0.49 dB with a range between 0.40 and 0.58 dB. The
measurementsof three samples were made by holding their
set-up as previously described and bonded into
preparedand connectedto the loss-
A sample of four in-line
matching gel deposited
loss for a straight
mtrogen and room temperature resulting in no change in loss at
R.T. There were also no losses shown from room temperature
to liquid nitrogen temperatures in each cycle while the bonded
sample was immersed in liquid nitrogen.
This paper has described and characterized a method which
permanently bonds oxide coated ball lenses and silica fibers to
aluminized silicon substrates with the use of heat and pressure.
The method offormingthe aluminum to oxide bond is referred
to as Ala bonding . Its objective is aimed at solid state
bonding optical components in packages with the simplicity of
joiningmetal conductors in electronic packages.
data was obtained after subjecting the components to handling
subsequent bonding which is expected to be representative
manufacturing conditions.The following is a summary of the
duringtheir preparation and
1. The aluminum
interfacial shear and contact between the silica component and
aluminum where it is exposed at fissures of its extended native
oxide protective coating. Thermodynamic
to show that the reaction was possible.
aluminum was stuck tothe
to oxide bond occurs as a result of
data was presented
It was also shown that
2. A means of pull testing bonded balls and fibers were
developed.Typical values of 35±3 gms resulted when bonded
balls are pushed out of aluminized (2.5 micron thick) micro-
machinedcavities with a probe testing machine positioned
parallel to the silicon substrate.
3. Ala bonded balls were also shock and vibration tested
at levelsthat met or exceeded
mechanical integrity and resulted in 100% yield.
5.Fibers bonded to aluminized
parallel to their axial length.
increasewith an increase
thickness of the aluminum
obtained when using bonding temperatures of 300°C.
investigation resulted in the tested fiber remaining intact. This
included a sample of bonded fibers that were pulled tested after
aging for 100 hours in a temperature!humidity
were shown to
in the bonding
film and longer
6. In situ loss measurements
(with an accuracy of ±0.01
formationofdB) duringthe Alabonded
and spliced in a V-groove and; two fibers and two intervening
ball-lenses bonded in-line into etched V-grooves and cavities.
No loss was observed through bonded fibers. These results
were shown through each sequence from bonding at 298°C to
ambient. Similar results were observed during thermal cycling
from liquid nitrogen to room temperature and, ice water to
average loss= 0.09 dB. Splices that were immersed in liquid
nitrogen and removed to ambient resulted in an average change
in loss of O.OldB at ambient. The average straight through loss
measurements between two bonded fibers and two intervening
bonded ball-lenses were equal to 0.49dB with a range between
0.40 and 0.58 dB. This was well within an allowable value of
1.1 dB for a reported optical switch which used the same
type of chemically micro-machined substrate.
through: a bonded fiber; a pair of fibers bonded
splice losses resultedin an
The authors would like to thank their colleagues for their
assistance during this investigation:
R. E. Fanucci, R. E. Frazee, J. P. Honore, S. A..Gahr, B. H.
Johnson D. M. Ors, R. F. Roberts, C. M. Schroeder, D. R.
Smithgall, L. S. Watkins, K. L. Komarek and M.A.C.
R. Borutta, J. G. Cavalli,
A Cure Schedule
Lap Shear Strength
Glass Transition Temp.Coef. of Thermal Exp.«130
Coef. of Thermal Exp.(>130 0c)
Outgas to 300 °C
Coef. of Thermal Exp. (20 to 300 0c)
1. Conti, R.I., "Thermocompression
Quarterly, vol. 6, pp. 29-35, 1966.
"Ultrasonic Welding in Electronic Devices," IRE International
Convention Record, vol. 10, pp. 3-12, 1962.
Peterson, J.M.,H.L. McKaig,andC.F. DePrisco,
Conference, pp. 539-556, 1970.
Coucoulas,A., "Hot WorkUltrasonic
4. Dautartas, M.F., A.M. Benzoni, Y.c. Chen, G.E. Blonder,
B.H. Johnson,C.R. Paola, E. Rice, and Y.-H. Wong,
Lightwave Technology, vol. 10, no. 8, pp. 1078-1085, 1992.
Of OpticalSwitch," Journal
5. Smolinsky, G. and R.E. Dean, "LPCVD Oxide Films In The
7), pp 256-260 (1986).
Materials Letters, vol. 4(5-
6. Metals Handbook, ASM, 1948 ed.
Pergamon Press, 1958.
Evan, E. LL.,"Metallurgical
8. Hansen, M., "Constitution of Binary Alloys," McGraw-Hill
Book Co, Inc., 1958.
9, Tabor, D., "The Hardness of Metals", Oxford Univ. Press,
1. Assembly view of a chemicallymicro-machined
TABLE 1- Relevant Propenies of a High TemperalUre Epoxy 1
And Commerci.a!ly Pure Aluminum2
Number of Components
150 °C for 10 min.
One Year At Room Temp.
10,000 psi13,000 psi7,500 psi
23.7x1O-6 in./in./"C25.6x1O'6 in./in./"C
Tensile Strength (25 0c)
Tensile Strength (400 0c)
Melting Point Coef. of Thermal Exp. (-60 to 20 0c)
1. Epoxy Technology Inc.
2. ASM Metals Handbook, 1948.(\'!o
Figure 2. Schematic cross-section of an oxide coated ball-lens
and flat aluminized silicon substrate prior to (a) and during (b)
Figure 3. Schematic views of the destructive test procedure.
looped-taped is first lowered to the top surface of a bonded ball
for tacking and subsequent pull testing.
~t..-.(''\ :~: f
~.,:- -.•~~' /
Figure 4. Top view of a destructively tested ball attached to an
adhesive tape shown in the surrounding background.
circumference of the ball. 50x (reduced 25% for printing),, .
. " .
Figure 5. Top view of the aluminum thin film coated silicon
substate showing the following three distinct regions: an inner
circular region of thinned aluminum: a partial annular-shaped
ring of underlying silicon and; an outer ring of an aluminum
pile shown out offocus.1000x (reduced 25% for printing).
nugget stuck on the surface of the tested ball. 400x (reduced
25% for printing).
6. Top view of a partial annular-shapedaluminum
Figure 7. Schematic of a 300 micron diame~~r ball pen~trati~g
a 2.5 micronsaluminumthin film on SIlIcon showmg
chordal length (55 microns) in-line with the surface of the film.
Figure 8. Top view of an aluminum thin film coated silicon
surface of the ball. 1000x (reduced 25% for printing).
Figure 9. AlO bonded ball-lenses. Bonding parameters were:
temp= 298°C: pre-heat~1 min.; load= 600 gms. and; bond
time= 13 sec.
sections stuck to its surface.
10. Dislodgedball-lense with four aluminumfilm
aluminum was removed during push-testing of the bonded ball
shown in Figure 10.
11. Pyramidalcavity showingbond regionswhere
concentration of aluminum remaining in the off-center region.
12. Bond region on the cavity wall showing a high
Figure 13. Aluminum deposit on the destructively
showing the thinned topographical
where the ball made the closest contact with the silicon wall
in the vicinity features
Figure 14. Aluminum deposit stuck on the destructively tested
ball which was pushed out of the cavity shown in Figure 12.
I II I
showinir positio~ing of the fiber end to the waxed pull-rod of
the testmg machme; the wax sealing of the fiber end and' the
destructive pull tested fiber remaining in-tact.
20. Schematicof thefiber pull-testingprocedure
Figure 19. Top view of a bonded fiber in a V-groove (right)
and the extending free end (left).
Figure 18. Top view of a fiber-end in the region where the tool
made contact during the bonding cycle.
shows no evidence of the residual polymer.
Its cleaned surface
Figure 15. Aluminum test carrier with three attached substrates
with all 12 bonded balls remaining
and vibration testing ..
in the cavities after shock
Figure 17. Top view of a bonded fiber in an aluminized V-
groove showing islands of the remaining polymer coating.
Figure 16. One of the three substrates attached to the aluminum
test carrier with all 4 bonded balls remaining after shock and
bolld fon:e.12OO±1OO JIllS.
bolld 1eflIth. 0.031·inch
substrals: 2.5 micronsAlISi v-JtOOve
bcftd time • 13 tee.
bolld 1eflIth. 0.031·inch
.•.boa4 time. 13 tea.
• bolld fon:e.12OO±IOOJIIIS.
Figure 21. Individual Pull Strengths vs. Temperature for a set
of bonding parameters.
subllrut: 2.5 micronsAlISi v_p'OOve
Figure 23. Individual Pull-Strengths
sets of bonding parameters having different bonding forces.
vs. Temperaturefor two
.•.bald. time. 13 teeS.
• bond time • 3 JlllCS •
subcue:2..$microns AlISi v'JI:'OOYe
bolld 1ilIle. 13•••.
• bolld IelIc1h • O.03I·inch
• bolld IelIc1h - 0.0.7·inch
Figure 22. Individual Pull-Strengths
sets of bonding parameters having different bonding times.
vs. Temperaturefor two
bonding parameters having different bond lengths.
24. IndividualPull-Strengthsvs. Temperaturefor
bond time • 13 sec.
bond (octo.I200± 100 ps
bond kqlIl• 0.03 I-inch
_ •••. """p • 2li3"C
Figure 28. Individual Pull-Strengths vs Aluminum Thickness
Figure 29. Side view of spliced fibers without index matching
gel in place.Top portion of the splice shows the relative z-
position of the fibers.
Figure 30. Fibers and a balllense
base. Lower bonded fiber was bonded at the required position.
The upper fiber was bonded out of position to view its end.
bonded to an optical silicon
Figure 25. Top view of a substrate with parallel V-grooves
showing one side of two bond regions of· different lengths
where the fibers was a pull-tested
strengths.of322 gms (upper bond region) and 513 gms (lower
bond regIOn). Both destructively
with aluminum stuck to their surfaces.
in shear with resultant
tested fiber remained in-tact
Figure 26. The bond region of a pull-tested fiber which was
in-tact after testing.It shows two parallel bands of aluminum
which was removed from one side of the V-aroove.
Figure 27. Bond region from one side of the V-groove where a
bondedfiber was pull tested.
observed on the opposite angular wall.
A similarbond regionwas
fiber into the aluminum film on the angular wall.. ,
31. Bondedfiber (which
of Figure 30) shows the penetration
is also shownin upper
Figure 32. Fibers and balls bonded in-line on a silicon optical
measu~ements during bonding and thermal cycling.
Figure 33. Higher magnification of AlO bonded fibers and balls
on a silicon optical substrate shown in Figure 32,