1. Detailed Description of Process Flow for Ziptronix Chip
Camilla Dagum ’13 and George Epstein ‘15
Cornell Nanoscale Science and Technology Facility (CNF)
Cornell University, Ithaca, NY 14853
We worked with 6”, 1-10 Ω-cm, 650-700µm thick silicon wafers. The end product was a
patterned semi-conductor micro-circuitry to be bonded (externally at Ziptronix) into a
detector device. The following article describes the process to achieve that goal.
I. Mask Printing
We used the Heidelberg
DWL2000 Laser Mask Writer to print
two 7” photomasks based on layers of
the provided gds file from Ziptronix.
Layer 116 was printed first, and includes
alignment markings and the circular via
features. Layer 125 was printed second,
and includes alignment markings and the
pad/wiring pattern. Both masks were
written in 4-5 hours. At the end of the
writing, the masks were developed and
chrome etched in the Hamatech-Steag
HMP900 Mask Developer. The resist
was stripped with a combination of
Acetone and Isopropanol.
II. First Photolithography
The wafers were primed with
hexamethyldisilazane (HMDS) vapor in
the YES LP-III Vapor Prime oven. This
automated process takes approximately
25 minutes. Following this, each wafer
was spun with SPR220-3.0 i-Line
photoresist at 3,000 revolutions per
minute with a 2,500rpm/sec ramp for 60
seconds (Note: following this step, the
wafers should not be exposed to white
light until the methods in Section III).
Each wafer was subsequently soft baked
for 90 seconds at 115°C and then cooled.
The wafers were allowed to sit for 30
minutes while the resist hardened. We
performed a soft contact exposure on the
Suss MA6 Contact Aligner for each
wafer, using the first (gds layer 116)
photomask for 7 seconds with a 5 second
exposure gap and a 35µm alignment gap.
The wafers sat for another hour before
development for one minute in 726MIF
developer (Tetramethyl Ammonium
Hydroxide 2.00%). Inspection with light
microscopy was performed to ensure
that the pattern had exposed properly.
III. Dry Etch
After removing the edge bead
from each wafer with Acetone, the via
pattern was etched into the substrate
with a dry silicon etch on the Unaxis 770
Deep Si Etcher. Etching was performed
in 10 loops at 300nm/min anisotropically
for ~3µm. A pure release of SF6 is used
for 35 seconds to isotropically flare out
the bottom of the vias so that the
aluminum plugs will not be ripped out
during chemical mechanical polishing.

2. The photoresist was stripped in a
two Hot Strip Bath system and the
wafers were rinsed and dried in a Spin
Rinse Dryer. They were then put through
a Metal Oxide Semiconductor clean for
20 minutes. A 1µm layer of oxide was
grown in a wet HCl oxide furnace at
1100°C.
IV. Aluminum Sputter
We followed two different
methods for achieving the necessary
aluminum sputter for this next step. The
first was the AJA Sputter Deposition
with a Chromium adhesion layer. The
second was the CVC Sputter Deposition
with a Titanium adhesion layer. The
AJA has the advantage of being much
more uniform in its deposition, with the
drawbacks of only being able to process
one substrate at a time at a slow rate
making the entire process last
approximately 5.5 hours per wafer. The
CVC is able to sputter batch quantities in
roughly the same amount of time with
the drawback of not being as uniform.
However, since the sputter step is
following by chemical mechanical
planarization, the CVC is recommended.
A. AJA Sputter
When using the AJA Sputter
Deposition tool for our thin film
deposition, we followed this
methodology. The substrate was
Argonne cleaned for 60 seconds at 30
Watts, and 16 mTorr. Chromium†
for the
adhesion layer is struck and a shutter
delay of 20 seconds is observed before
deposition begins. Deposition takes
place for 119 seconds at 150 Watts, and
3 mTorr to achieve a 0.42 Å/sec rate and
a layer approximately 5 nm in thickness.
To begin aluminum striking, we found it
prudent to have a shutter delay of 5
seconds at 20 mTorr before beginning
the 20 second shutter delay at 3 mTorr.
The higher pressure allowed for more
effective plasma striking over the
aluminum. Finally the aluminum was
deposited for 19,900 seconds
(5h31m40s) at 400 Watts, and 3 mTorr
to achieve a 2.01 Å/sec rate and a layer
approximately 4µm in thickness to
completely fill the vias and create a top
layer for Chemical Mechanical
Polishing.
B. CVC Sputter
Our alternative method of
sputtering the aluminum with the CVC
followed the following steps. We
performed an ion beam cleaning applied
at a base pressure of 1.6µTorr for 5.0
minutes with an Argon flow of 40sccm.
We followed this by a titanium adhesion
layer at a power of 2kW for 2.0 minutes.
The aluminum deposition was also
applied at 2.0kW and was done for 200.0
minutes to achieve the ~4µm layer.
Presputter times of 5.0 minutes are
advised. Having tried this method after
seeing the first results of the second
deposition (see section VII. B.), the
decision was made to lower the speed of
rotation to level five (~7.0 revolutions
per minute) to attempt to achieve less
streaking of the metal. Once again, since
planarization is performed directly after,

3. these details are not detrimental to the
process flow, but it may help improve
results.
V. Chemical Mechanical Polishing
(CMP)
We used the A7100 Slurry recipe
provided by CNF on the Strasbaugh 6EC
Chemical Mechanical Polishing tool to
polish the top layer of aluminum away
leaving only the plugs in the vias. The
recipe includes a 5:5:1 mix of A7100A :
A7100B : 30%H2O2. The polishing pad
was spun at 60rpm, while the wafer was
spun at 35rpm on a rotating chuck. Back
pressure on the substrate was set to 15psi
as was down pressure on the chuck. The
slurry was dispensed onto the polishing
pad at 250ml/min. The process to
remove the entire top layer of aluminum
took approximately 26-30 minutes to
complete and leaves only small
variations in the thickness of the surface
oxide layer. Following polishing, the
substrate underwent a piranha clean in
the Hamatech Hot Piranha processor.
After the process was completed,
microscope observation revealed that the
aluminum plugs were still in the via
holes as expected.
VI. Second Photolithography
The wafers were again primed
with hexamethyldisilazane (HMDS)
vapor in the YES LP-III Vapor Prime
oven. We used SPR220-3.0 positive i-
line photoresist with an exposure and
development procedure ‡
designed to
give the positive resist good lift-off
capabilities. Resist was applied at 2,000
revolutions per minute with a
1,000rpm/sec ramp for 60 seconds to
achieve a coating of roughly 4µm (Note:
following this step, the wafers should
not be exposed to white light until the
methods in Section VII). The resist was
soft baked at 115ºC for 90 seconds and
held for one hour to allow rehydration to
take place. The wafers were then
exposed by soft contact on the Suss
MA6 Contact Aligner to the second
photomask (gds layer 125) for 7.0
seconds with a 5 second exposure gap
and a 35µm alignment gap. Following
exposure, the substrate sat for another
hour to again allow rehydration and was
put through a post exposure bake at
115ºC for 90 seconds. We then
developed the wafer using the novel
technique asserted by Zhu et. al. The
process involves a 30 second
development in MIF321 developer
followed by a spin rinse dry (SRD) cycle
followed by another 30 second
development in MIF321. The SRD step
between developments was intended to
allow for oxidation of the resist and was
the key step in creating lift-off
capabilities.
VII. Aluminum Deposition
Two methods exist for the
deposition of aluminum and we discuss
both here. The methods are evaporation
with the CVC SC4500 Odd-Hour
Evaporator, or sputtering with the CVC
601 Sputter Deposition System. Both

4. methods have been used and both
yielded favorable results on the final
outcome despite producing markedly
different intermediate profiles.
A. Evaporation
We applied a thermal Chromium
layer of 10nm to start. DC Voltage was
set to 10V and power was raised up to
54.2% before beginning deposition.
Following this, an Aluminum layer of
1µm was deposited via Electron Beam
Evaporation. Power was raised to a mere
6.4%, but the deposition rate was still
able to peak at a remarkable 11Å/sec. A
microscope evaluation of the wafer
surface after deposition showed a
smoothly uniform layer of aluminum
with no major flaws.
B. Sputtering
Our alternative method of
sputtering the aluminum with the CVC
followed the following steps. We
performed an ion beam cleaning applied
at a base pressure of 1.6µTorr for 5.0
minutes with an Argon flow of 40sccm.
We followed this by a titanium adhesion
layer at a power of 2kW for 2.0 minutes.
The aluminum deposition was also
applied at 2.0kW and was done for 50.0
minutes to achieve the same 1µm layer.
The entire time, the samples were
rotated at the maximum rotation setting
provided by the CVC (20 per minute).
The end result of this deposition was an
acceptable layer of aluminum deposited
that did not interfere with the final
product, though there were many
repeated surface defects in the aluminum
layer: Ripples formed between the pad
patterns and ridges were pulled up
around the wiring patterns. Since the
artifacts did not seem to directly affect
the pattern itself and were only present
on areas that were going to be lifted off,
the sputtering process proved to be just
as viable as the evaporation process.
VIII. Lift-Off
We submerged the substrates in
Microposit Remover 1165 and placed
the beaker into a sonicator. With 10-15
minutes in the sonicator, most of the
aluminum is lifted off. Upon removal
from the 1165, a rinse with DI Water is
recommended in order to remove the
larger strips of aluminum that are still
adhered. Following this, we found that
small traces of aluminum continued to
adhere in between the wiring especially
where the pattern was most dense. To
combat this, we performed a number of
rinses, swabs with acetone, and
sonication in acetone. Alternating these
steps and repeating them several times is
enough to remove the excess aluminum.
When swabbing, the areas where
aluminum still adheres are easily
recognized with the naked eye; they
appear brighter due to the higher
reflectivity of the continuum of metal.
It is worth noting that the
adhesion of aluminum to the bare silicon
was very strong and was very much
unaffected by the excessive amounts of
sonication and spraying. The removal
techniques following sonication in 1165
thus left the pad pattern wholly
unharmed.

5. As a procedural note, it is
important to begin sonication as soon as
possible after submersion in 1165.
Otherwise, enough resist is pulled out
into the solution that the aluminum to be
lifted off is more reluctant to do so.
†
For the second deposition layer (Section
VII) we attempted a Titanium deposition
rather than a Chromium one. The results
were comparable and until further notice,
either method may be used.
‡
Zhu et. al. 2005
http://www.pdx.edu/pnna/sites/www.pdx.edu.pn
na/files/spr220_liftoff_trick.pdf