Page 1
27
Chapter 3
Fabrication Process for Surface Micro machined Metallic Structures
This chapter covers major unit processes used for the fabrication of MEMS devices. The
main focus is on the fabrication of reliable RF MEMS devices using surface
micromachining. The unit processes that are of prime concern are polysilicon deposition,
diffusion, metal deposition and etching, photoresist coating and patterning, lift-off,
electroplating and release of photoresist/metal sacrificial layers. The major challenges
toward reliable switch fabrication/commercialization are structural deformation/ out of
plane buckling after sacrificial layer release. Reasons for device buckling are built-in
residual stress which is developed in the structure during metal deposition and difference
in thermal expansion coefficient of stacked layers. The issue of out-of plane
deformation/buckling is addressed in this chapter using optimised process of metal
deposition by electroplating and sacrificial layer release techniques.
3.1 Fabrication processes for RF MEMS devices
To understand the effect of fabrication process parameters on the structural reliability,
various fabrication iterations on RF MEMS switches (capacitive and ohmic), Test
structures, Beam array, Torsional Mirrors, SP4T(single pole four throw) and Inductors
are carried out. All the fabrication processes are implemented on 2” silicon wafer. The
fabrication process included the following steps:
a. Oxidation – thermal oxidation (field oxide) for initial one-micron thick layer.
b. LPCVD and PECVD – polysilicon electrode deposition, actuation electrode
insulation and dielectric layer for capacitive switches.
c. Diffusion - for low resistivity polysilicon ‘dc bias lines’, actuation pads and
resistor for ohmic contact switches.
d. Sputter deposition (dc and rf sputtering) and E-beam – Au, Cr and Ti for seed
layers and adhesion promoter for gold, Ti, TiN, Al-Si for underpass area (metal
stack) and Ni for contact pads.
e. Lithography – nine lithography steps involving the use of three type of photoresist
ranging in thickness from 1 – 10 microns.
f. Reactive ion etching for SiO2 and polysilicon etching.
g. Electroplating - for structural layer deposition, using lift–off and etching
techniques.
Page 2
28
h. Plasma Ashing – for thick hard baked photo resist sacrificial layer removal.
i. Critical point drying – for low temperature removal of sacrificial layer.
Among the above mentioned processes, the main processes those need optimisations are:
1. Polysilicon deposition and doping.
2. Metal (Cr, Au, Ti, TiN, Al) deposition, etching and lift off.
3. Photolithography optimisation for three different photoresist.
4. Electroplating process optimisation.
5. Sacrificial layer removal using plasma ashing and CPD.
Figure 3.1 shows the 9 level mask layout of the devices to be fabricated [1]. The
fabrication process is based on the IC fabrication techniques using surface
micromachining and gold electroplating. The general process compatibility issues,
material selection, thermal annealing during different processes, etc. decides the
overall performance of the fabricated device. Figure 3.2 shows the schematic of the
fabrication steps for RF MEMS Ohmic switch and details are explained in this
section.
Figure 3.1: Complete layout with marking of position on specific devices.
STS Ohmic
Switch
Meander
Ohmic
Switch
STS
Ohmic
Switch
Meander
Ohmic
Switch
Bridge
Test
Structure
Bridge
Test
Structure
Bridge
Test
Structure
Bridge
Test
Structure
Bridge
Test
Structure
SP4T
Switch
SP4T
Meander
Ohmic
Switch
STS Capactive
Switch
CPW
Torsionl mirror
Test
Structures
Test
Structures
SP4T
CPW
Self Actuated
Capactive Switch
CPW
S
P
D
T
Inductor
Torsion
mirror
AM(L) AM(R)
COMPLETE
LAYOUT
Page 3
29
A low-loss, 2", p-type, <100>, 5 KΩ silicon wafer is used as substrate for device
fabrication. The wafer is cleaned using four different cleaning steps in succession.
• Degreasing: Removes organic and oily contaminants from the surface.
• RCA 1, RCA2: Removes organic and metallic contaminants from the wafer surface.
• Piranha Cleaning: Removes heavy metal particles from the wafer surface.
• HF Dip: HF dip removes native oxide from the wafer surface formed during piranha
cleaning.
An insulation layer of 1000 nm of silicon dioxide is grown over silicon surface using
thermal oxidation. The thickness obtained in the case of dry oxidation is limited to few
nm only; a combination of dry-wet-dry oxidation is used to obtain larger oxide thickness.
The oxidation parameters for 1 µm silicon oxide are as follows:
• Furnace temperature in 3 zones: 846ºC, 1050ºC, 821ºC.
• N2 flow rate for purging: 85cc/min for 3 hours, saturate furnace temperature.
• O2 flow rate for dry oxidation: 35cc/min for 15 minutes
• O2 flow rate for wet oxidation: 20cc/min for 155 minutes
• O2 flow rate for dry oxidation: 35cc/min for 15 minutes
3.1.1 Polysilicon deposition and doping
The actuation pads, resistor for ohmic contact switches and the DC bias lines for the
switching structures are patterned in polysilicon. A polysilicon layer of 0.6 µm thickness
is obtained using LPCVD techniques and the process parameters are shown in table 3.1.
Since undoped polysilicon has low electrical conductivity, it can’t be used as metal lines
for DC biasing. The conductivity of polysilicon layer is enhanced by phosphorus
diffusion at high temperatures followed by annealing. The same polysilicon layer is also
used to make biasing resistors for ohmic contact switches which requires a precise value
of sheet resistance (240Ω/□). Table 3.2 gives the summary of experiments conducted to
obtain desired sheet resistance value for the polysilicon. Diffusion process is optimized
by multiple iterations until a repeated value for sheet resistance is obtained. Table 3.3
shows the results of optimized phosphorus doping recipe used for fabrication process.
Phosphorous doping is carried out for 7.5 minutes at 950°C and 15 minutes at 900°C, to
obtain a sheet resistivity of nearly 240Ω/□.
Page 4
30
Figure 3.2: Fabrication process flow for RF MEMS Ohmic switch.
Table 3.1: Polysilicon Deposition Process
S.No. Polysilicon Process Steps
1 Temp. 6250C 6300C 6200C
. 1st Zone 2nd Zone 3rd Zone
2 Gas Flow: Silane (SiH4)
3 Flow Rate: 50 cc/min (38.4%)
4 Process Time: 29 minutes
5 Pressure: 0.240 Torr
6 Thickness: 6000Å
Page 5
31
Table 3.2: Polysilicon sheet resistance optimization
Table 3.3: Polysilicon optimized doping recipes
First level lithography is used to define polysilicon actuation electrodes and interconnect
lines for the RF MEMS switch. The reactive ion etching (RIE) of polysilicon is followed
by the photoresist removal in acetone, piranha cleaning and dip in dilute HF. Figure 3.3
shows the first level mask layout, optical images of patterned actuation electrodes and
process cross section view. An over etching polysilicon is observed after the optical
inspection of defined patterns. Over etching of polysilicon can result in change in
geometry, leading to deviation of final switch performance from the designed results. For
example, over etching of actuation electrodes would result in small actuation area and
hence increase in actuation voltage. The next step is the deposition of 350 nm of silicon
dioxide layer using PECVD process. For 350 nm of SiO2 deposition, silane SiH4 (5.4cc),
Nitric oxide N2O (100%) and Nitrogen (31%) are reacted in a plasma chamber at 300ºC
for 14 minutes. This isolation layer prevents the electrical short between the movable
bridge and the actuation electrodes.
Second level lithography defines the contact holes in the SiO2 layer. The contact holes are
used for interconnection between polysilicon lines and metal lines. These metal lines are
connected to metal pads for biasing voltage. Reactive ion etching (RIE) technique is used
Step W#S1 W#S2 W#S3 W#S4 W#S5
Time 25 min 15 min 7.5 min 7 min 5 min
Temperature 9500C 9500C 9500C 9500C 9500C
Anneal Temp. 10000C 10000C 10000C 10000C 10000C
Anneal Time (min) 20+10 (O2 +N2) 20+15 (O2+N2) 10 (N2) 10 (N2) 10 (N2)
SHEET RESISTANCE MEASUREMENT
Before Annealing (i) 28.8Ω/□
(ii) 28.9 Ω/□
45.3 Ω/□
44.8 Ω/□
164Ω/□
165Ω/□
191.3Ω/□
200.0 Ω/□
301 Ω/□
350 Ω/□
After Annealing (i) 38.2 Ω/□
(ii) 37.2 Ω/□
(iii) 38.0 Ω/□
76 Ω/□
78 Ω/□
76 Ω/□
237 Ω/□
248 Ω/□
258 Ω/□
260 Ω/□
266 Ω/□
274 Ω/□
400 Ω/□
420 Ω/□
489 Ω/□
Step W#A1 W#S3
Time (min) 15 (with POCl3) + 10(without POCl3) 7.5(with POCl3)
Temperature 9000C 9500C
Annealing Temp 10000C 10000C
Annealing Time (min) 5 (O2)+ 10 (wet) + 5 (O2) + 5 (N2) 5 (N2)
Sheet resistance measurement
Before Annealing 135± 10 Ω/□ 160± 5 Ω/□
After Annealing 220 ± 10 Ω/□ 240 ± 10 Ω/□
Page 6
32
to open the contact holes in SiO2 layer followed by photoresist removal. Figure 3.4 shows
mask layout, optical micrographs of the contact holes openings in the oxide layer and
cross section view after second level of lithography.
(a) (b) (c) (d)
Figure 3.3: (a) Mask layout (b) Optical micrographs of patterned photo resist on
polysilicon electrodes and polysilicon connection pads. (c) Optical micrographs of
patterned electrodes after RIE and (d) cross section view of pattern layer.
3.2 Multilayer metal (underpass area) deposition, etching and Lift off
The part of the transmission line under the bridge is fabricated using a multi-layered stack
of Ti/TiN/Al/Ti/TiN. The stacked multilayer provides good adhesion, smooth contact
surface and low resistivity path for signal transmission. The important steps are:
Pre deposition metal clean dip in 5% HF solution for 2 - 3 seconds.
Sputter deposition of Ti / TiN / Al / Ti / TiN, for deposition time 6 min / 10
min / 14 min / 12 min / 16 min respectively, resulting in thickness for Ti ≈ 34
nm, TiN ≈ 80 nm, Al ≈ 400 nm.
Patterning of multilayer metal (lithography 3): Positive photoresist (S1813) is
coated at a speed of 4500rpm, RAMP 8000, for 30 seconds and prebaked for
25 minutes at 90⁰C. The PR is exposed for 5 seconds (intensity 8 - 9 mW/cm2)
SiO2
Polysilicon
Si
Page 7
33
followed by developing time of 60 seconds. The post bake step is at 120⁰C for
30 minutes.
Figure 3.4: (a) Mask layout of polysilicon electrodes covered with PECVD silicon
dioxide. (b) Optical micrographs zoomed view of contact holes, opened in SiO2 layer. (c)
Cross section view of pattern layer.
3.2.1 Etching of metallic layer
The multilayer metal etching is one of the most important steps in fabrication and is
performed in following steps:
TiN Etching (Type TiN 22-7): TiN etchant is used at temperature 20⁰C, for 25
seconds. The etchant composition is HF + Nitric Acid – a proprietary composition
of the vendor (Transene). Although the etchant should etch TiN only, it is
observed that TiN and Ti both layers are etched out for the time and temperature
combination shown above. The problem is solved by adopting the lift off
techniques mentioned in next section. Although this increases the process
complexity, but the reliability is better.
Ti Etching (Type TFTN): In order to remove any traces of Ti left from the
previous step ‘Transene Ti etchant’ is used at recommended temperature of 70⁰C,
(a) (b) (c)
Page 8
34
for 40 seconds.
Au Etching (Type TFA): Etch rate of Au etchant is 28Å/sec at 250C. The
temperature of Au etchant is kept at 25ºC and etching time is approximately 70 -
75 seconds, 8 - 10 seconds of ultrasonic vibrations in small steps (2 seconds each)
is required to remove Au seed layer from perforations.Wafer is rinsed in DI water
after etching.
Cr Etching (Type TFD): Etch rate of Cr is 25 Å/sec at 40oC, the solution consist
of sulfuric acid and nitric acid. Cr etchant is kept at 40ºC and etching time is
approximately 20 seconds with one second of ultrasonic vibrations after each five
5 seconds. Over-etching is desirable to ensure removal of Cr seed layer from
perforations. After Cr seed layer etching using Transene etchant (TFD), wafer
surface retains contaminants. To avoid this contamination, wafers is dipped in 2%
H2SO4 solution for 2 - 3 seconds immediately after Cr etching followed by 5 min
rinsing in DI water.
Al Etching (Type F): Etch rate of Al etchant is 100Å/sec @ 50oC. The etchant
constituted of phosphoric acid and acetic acid .Wafer is rinsed in DI water after
etching.
3.2.2. Lift off process
The lift off process is used as an alternate to metal etching and pattern metal at desired
area [2]. In this case lithography of desired pattern is followed by metallisation as shown
in schematic of lift off process (figure 3.5). The following steps are considered during lift
off:
Thickness of photoresist must be more than 2-3 times of metal thickness.
Metal deposition should be carried out using E-Beam.
Sputtering can be used for metal deposition with low power (400W) and
minimum distance between electrodes.
Sample should place in acetone for 30 minutes.
Sometimes manual cleaning of the substrate with clean room wipes helps
to remove photoresist residue after acetone dip.
Page 9
35
Figure 3.5: Fabrication process step for lift off.
The total thickness of the multilayer stack is matched with the thickness of actuation
electrodes (600nm). The height difference between the underpass and the actuation
electrodes would lead to a complex and non-uniform bending during actuation.
During the second run of fabrication only aluminium metal of thickness 600nm is used as
underpass layer, due to unavailability of multi-layer metal stack. The Al is deposit at two
different RF power 400 and 600W, at base pressure 5x 10-3
torr for 20 minutes. The high
RF power (600W) and distance between electrode results in photoresist (PR) burning on
the lift off sample as shown in SEM micrographs (figure 3.6). This PR burning is avoided
by using low RF power of 400W and E-beam for metal deposition. The third lithography
step defines the underpass area of the transmission line in aluminium and the contact pads
on the polysilicon connection lines, mask layout and respective optical micrograph after
Al pattern is shown in figure 3.7.
The deposition and patterning of underpass area is followed by the 100nm of PECVD
SiO2 deposition, which acts as a dielectric layer for the capacitive type switch. The down
state capacitance of the capacitive switch depends on thickness, surface roughness and
dielectric properties of this layer. The fourth lithography defines via holes through which
underpass area gets connected to the input and output transmission ports in thick gold.
Figure 3.8 shows mask layout and optical micrograph of fabricated structures with via
holes opening in oxide layer over central overlap area and zoomed view of polysilicon
line, connection pads and underpass bump area.
PR profile for Lift-off
Metal deposition from E-beam or sputtering
Metal pattern after lift off
Page 10
36
Figure 3.6: Photoresist burning after Al metallisation.
Figure 3.7: (a) Mask layout of polysilicon electrodes covered with PECVD silicon
dioxide, metal on underpass and contact pad. (b) Optical micrographs of fabricated
underpass metal contact area, poly resistor with contact pad and ground line.
In the next step an additional Cr (30 nm)/Au (150 nm) metal layer is sputter deposited
using lift off process over the underpass area oxide layer to form a floating capacitor. The
photoresist mould is defined as a result of fifth level of lithography for the lift off process.
This fabrication step is only useful or dedicated to capacitive switches. This floating layer
helps in improving the down state capacitance of the switch and hence the capacitance
ratio. During the switch actuation, the movable beam comes in contact with this floating
metal layer and connects the floating capacitor to ground, thus achieving the maximum
down state capacitance.
Al
Al+PR
Al
Al+PR
(a) (b)
Al (600nm)
Page 11
37
Figure 3.8: (a) Mask layout, (b) optical micrograph of fabricated structures with via
holes opening in oxide layer over the central overlap area and (c) zoomed view of via
holes with polysilicon line, connection pads and underpass bump area.
3.3 Photoresist Optimization
The thickness variations for defining the complete bridge ranges from 1 to10 microns and
requires three different kinds of photo resists. The normal pattern definition such as
polysilicon, contact holes (SiO2), metal and via holes is achieved by using high resolution
resist (S1813) where thickness ranges from 0.5 - 1.5 microns. The spacer layer HIPR
6517 (high resolution positive photoresist) is optimized to achieve the thickness ranges
from 2-4 microns. The spacer layer PR also needs to be stabilized by baking at high
temperatures and large number optimization experiments are conducted to characterize
the profile of baked layer. The formation of bridge using electroplating requires photo
resist thickness in the range of 3-5 microns and capable of withstanding the subsequent
electroplating environment for longer durations. For this purpose appropriate resist (AZ
9260, AZ P4620) and developer (AZ 400K) is used. Figure 3.9 shows the various resist
details with typical profile requirement specific to applications [2]. Table 3.4 shows the
methods and process required for photoresist to achieve the desire profile.
(a) (b) (c)
Page 12
38
(a) S 1813,1818 (1-2 m) (b) HiPR 6517 (2-4m) (c) AZ 4620,9260 (4-10m)
Figure 3.9: Positive photoresist typical profiles.
Table 3.4: Methods for positive photoresist to achieve desire profile
3.3.1 Photolithogrphy surface preparation
Photoresist adhesion to typical wafer surface such as SiO2 and metals is critical to
faithfully replicating the photoresist pattern in etching process. There are several factors
which can interfere with photoresist adhesion [3]:
If the base solvent has a higher degree of interaction with the wafer surface than
the photoresist base resin, the solvent can push the resin away from wafer surface.
Later during soft bake the solvent may evaporate, but the resin may be constrained
by the photoresist film from the resulting gaps. This issue is addressed by good
photoresist formulations [4].
The resin-resin interaction can be stronger than the resin-substrate interface
interaction, particularly if the base solvent is badly matched to the resin. The resin
will then form tight coils to minimize its energy, lowering its interaction with the
substrate. The intrinsic interaction of the resin with the wafer surface is however
the most important factor in adhesion and once again this addressed during PR
formulation [4].
Surface contamination particularly due to moisture. Water has a very high affinity
for SiO2. Moisture interaction with SiO2 is a particular problem because SiO2
readily becomes hydrated due to humidity. Surface moisture may take the form of
Required Process, Applications Photoresist Thickness Methods/ Process
Normal, general S 1813,1818 1-2 m Moderate dose
Moderate develop time
Spacer, anchor posts
HiPR 6517 2-4m Low dose
Developer dominant
Lift-off, e-plating
AZ 4620,9260 4-10m High dose
low dependence on developer
Si-Substrate
Si-Substrate
Si-Substrate
Page 13
39
adsorbed surface moisture or chemically bonded OH-Silanol group. Surface
moisture is reduced by bake at less than 100C, but silanol groups require a much
higher temperature bake immediately before photoresist coating.
In order to minimize moisture effects on the PR adhesion, a variety of surface pre-
treatment techniques are used. A liquid "adhesion promoter", such as
"hexamethyldisilazane" (HMDS), is applied to promote adhesion of the
photoresist to the wafer. The phrase "adhesion promoter" is a misnomer, as the
surface layer of silicon dioxide on the wafer reacts with the agent to form
Methylated Silicon-hydroxide, a highly water repellent layer. This water repellent
layer prevents the aqueous developer from penetrating between the photoresist
layer and the wafer's surface, thus preventing lifting of small photoresist structures
in the (developing) pattern.
Photoresist Coating
During spin coating, the centrifugal force of the substrate spinning with several 1000
rounds per minute (rpm) distributes few ml of resist over the substrate. The resist film
thickness attained by spin coating represents the equilibration between centrifugal force
and solvent evaporation, both increasing with spin speed. Figure 3.10 shows the
photoresist spin coat cycle and Headway PMW 32 coating unit used during PR
processing.
Influences on the resist film thickness
The resist film thickness given for a certain spin speed holds for sufficiently long spin
times, when the resist spin-off is completed. As shown in the figure 3.11, the spin times
required to attain the final film thickness increases with the resist viscosity. Low spin
speeds also increase the time for a constant resist film thickness. This time dependency
not only allows the film thickness to be adjusted for spin speed, but also the spin time.
However, this is only a reasonable procedure for attaining thin and thick films. For thin
films (approx. < 10 μm), the spin times is long for more reproducible results. The
photoresist spin speed curve is an essential tool for setting the spin speed to obtain the
desired resist thickness (figure 3.11). The final resist thickness varies as one over the
square root of the spin speed and is roughly proportional to the liquid photoresist
viscosity [2, 4].
Page 14
40
Figure 3.10: (a) Simple photoresist spin coat cycle and (b) Spin coater unit.
Figure 3.11: Effect of spin speed on PR film thickness [4]
Hot Plate vs. Oven baking
Hotplates have several advantages over convection type ovens [5]:
Decreased bake time
Increased reproducibility
Better film quality.
In conventional ovens formation of different temperature zones, is a problem and can
severely affect film quality and reproducibility. In an oven the heating rate of a substrate
depends not only on the heated air flow past a substrate but also on its proximity to other
cold substrates. Thus the heating rate for each substrate in a cassette of substrates is less
than if each substrate is baked alone. Substrates near the ends of a cassette heat faster than
the substrates in the middle, thus producing a non-uniform heating. Particle generation
also occurs within a standard oven. In a forced-air, convection oven, substrates are
Spin Time
Ramp
Spread
time
Time
Spin
spee
d
Resist
Dispensed
Page 15
41
commonly exposed to a flow of particle laden air for at least 30 minutes. The PR film is
very susceptible to these particles.
Disadvantage of conventional ovens over hot plate
In normal oven baking results from baking substrates from the "outside in". Since heat is
applied to the outer surface of the film first, a skin forms on the surface of the film thus
trapping solvents. Upon vaporizing, these solvents form blisters or bubbles which results
in adhesion loss or even bulk film failure. This problem prevails in processes involving
thick film resins.
No skin effect occurs on a hotplate since hotplate baking heats the substrate from the
bottom up. This "inside out" approach offers advantages for thick films since solvents in
the film nearest the substrate are baked off before the film surface seals over. A well
designed hotplate insures uniform baking across the substrate. Since the substrate
intimately contacts a surface of a known constant temperature. Bake times is measured in
seconds, rather than minutes or hours, as in conventional ovens (figure 3.12). Reduced
vulnerability to particulate contamination is a major advantage of hotplate baking. Only
condition is ambient clean room air passes over the substrate. In general positive photo
resist baked for 30 minutes at 90°C in conventional ovens while 30 seconds at 115°C on
hot plate.
Edge Bead: Especially in the case of coating thick resist films (HiPR 6517, AZ9260 and
AZ P4620), edge bead is formed which may cause sticking to the mask as well as an
undesired proximity gap during exposure (with a reduced lateral resolution as a
consequence). In case no automatic edge bead removal is possible, initial stages solution
to avoid/lower the edge bead are:
An elevated spin-speed for a shorter time.
A ‘spin-off’ of the edge bead by abruptly increasing the spin speed at a
certain stage of spin coating.
A multiple coating with an elevated spin speed for each coating cycle.
For thick or solvent rich resist films, a delay between coating and soft
bake prevents the edge bead growing during soft bake due to the thermally
reduced viscosity. The delay time depends on the resist film thickness and
is shortened by applying moderate temperatures.
Page 16
42
Edged substrates: manual cleaning of the substrate from the edge bead
with clean room wipes. Exposure and development of the edge bead is also
possible using a ring mask.
Figure 3.12: Typical profile of various baking methods [5].
3.3.1.1 Recipe for resist S1813
Resist S1813 allows thicknesses from 1.2 to 2.0 µm. The spin speed is adjusted for each
recipe so that the desired thickness is obtained. The positive photoresist S1813 is used in
wide variety of process flow to perform wet etch, dry etch and even lift-off processes. In
present process S1813 is used for polysilicon electrode patterning, contact and via hole
opening, metal etching and lift off. The process parameters for S 1813 are shown in table
3.5. The effect of exposure time (3,6 and 9 seconds) is observed using Focus Ion Beam
(FIB) SEM cross section micrographs shown in figure 3.13 (a-c), the wall angles become
smaller as exposed areas grow with increased dose.
Table 3.5: Process parameter for S 1813
Coat 4500 rpm for 30 sec
Soft bake 90C for 25 min. conventional oven
Expose 6-7 sec
Develop 1 min.
Wall angle 75
Line width 2.5μm
Conventional Oven IR Hot plate
30
32
4
3
3
2
1
2
2
Pre bake time (minutes)
150
32
100
32
50
32
25
32
Tem
p. (0
C)
Page 17
43
Figure 3.13: S 1813 (a) exposure 3 sec. (b) exposure 6 sec. (c) exposure 9 sec.
3.3.2 Spacer layer photoresist optimisation
Sacrificial layer (spacer payer) for fabrication of suspended membrane is deposited and
patterned using positive photoresist. The use of material as a sacrificial layer depends on
the application, e.g. the temperature range in the following processing steps, the chemical
treatments followed after this step and surface profile after lithography. A variety of
metals, dielectrics and photoresists have been reported as sacrificial layers [6-9].
However, in the present case HiPR 6517 positive photoresist is used as it provides desired
contour by controlling using pre bake and post bake temperatures. After high temperature
baking at 180-200ºC PR turns into hard plastic layer and withstand in harsh environment.
The photoresist HiPR in general is used to obtain a thickness of 2.2 μm without prebaking
and without problem of edge beads. However, a thickness of 3.6 μm after post baking is
required for MEMS switches to achieve a gap height of 3 μm from the transmission line.
Thus, the photoresist coating process is optimized and includes the following steps (figure
3.14):
Step 1: Acceleration, filling and thickness decision,
Step 2: Edge bead removal,
Step 3: Deceleration.
Figure 3.14 shows the timeline graph for the photoresist coating to achieve the desired
thickness. Optimised recipe for the HiPR 6517 is shown in table 3.6, using the same
recipe optical micrograph of the fabricated structure with profile measurement is shown
in figure 3.15 (a, b). Surface profiler measurement of spacer layer before and after the
baking of photoresist is shown in figure 3.16. The overall thickness of 0.5 micron reduced
after baking at 1800C as depicts from figure 3.16. Cross section views of structure are
helpful to find the resist angel, next paragraph gives the details and requirement of resist
angel tuning.
(a) (b) (c)
Page 18
44
Figure 3.14: Spin coating programme for HiPR 6517HC photoresist
Table 3.6: Optimised Recipe for HiPR 6517
Coating Step Step 1 Step2 Step3 Exp.
time
(Sec)
Dev.
time
(Sec)
Thickness (µm)
Before
Bake
After
Bake
Speed (RPM)
Ramp (RPM/Sec)
Time(sec)
800
4000
25
2600
4000
1
0
500
0
20 60 4.1 3.5
Figure 3.15: (a) Optical micrographs after spacer layer and (b) surface profiler
measurement (Dectek 6M).
Figure 3.16: Surface profiler measurement of spacer layer thickness (a) after deposition
4.1 µm and (b) after baking 1800C for ½ hour 3.5 µm.
Sp
eed
(R
PM
) Time (sec)
Step 2 Step 3 Step 1
Direction of
profile scan
Spacer layer thickness 3.5-3.6µm
(a) (b)
(a) (b)
Page 19
45
Figure 3.17: Cross section profile of HiPR 6517 using Focused Ion Beam (FIB) SEM
(CeNSE IISc Bangalore).
Figure 3.18: Spacer layer profile control with exposure time.
Resist angle tuning
The mechanical reliability of MEMS switch depends on the quality of anchor posts, as
major failure in MEMS switches occurred at anchor points during switching. For reliable
operation the anchor thickness is comparable to the bridge thickness [6-9]. The resist
angle is fewer than 450 at the (steepest) position of the anchor. If we have a steeper angle,
more than 450, the metal thickness of the anchor is significantly thinner than the rest of
the bridge. Figure 3.17(a-c) shows cross section view of spacer layer near the anchor post
is measured using Focused Ion Beam (FIB) SEM. The resist slope angel of HiPR 6517 is
less than 450, the angel is tuned with exposure, developing and baking time [6]. The
figure 3.18 shows the structures (alignment mark, spacer layer pattern) (a) after 15
seconds and (b) 30 seconds of exposure time. The variation of the resist angle with time
and temperature is seen from the SEM cross section images shown in figure 3.19 (a, b).
(a) (b) (c)
Bridge
Anchor
Spacer layer profile
<450
(a) After15seconds of exposure (b) After 30seconds of exposure
Page 20
46
For comparison the SEM images for HIPR 6517 have been scaled to the same
magnification and compared. It can be easily seen that the resist angles decreased with
both baking temperature and baking time. The spacer layer profile is optimised and
following section covered the thick resist recipes optimisation used for electroplating.
Figure 3.19: The SEM images for HIPR 6517 resist (a) Baking time 25min. at120oC (b)
Baking time 25min. at1850C.
3.3.3 Thick resist optimisation
For the suspended bridge fabrication a photoresist mould is formed using AZ P4620
positive photoresist. For this step, the required resist thickness is 5 – 5.5µm. The AZ
P4620 photoresist can be used for coatings up to 20 µm in a single step [3]. For 5 µm of
thickness, the resist is coated at very high RPM (6000 RPM, 20000 RPM/sec) arising
problem of uniform filling of photoresist on wafer. Generally, there are two ways to attain
resist film thicknesses of 10 μm via spin coating: either by reducing the spin speed, or by
reducing the spin time. Low spin speeds, however, cause a pronounced edge bead and
sometimes even prevent the resist from tearing off the substrate with a non-reproducible
resist film thickness as a consequence. Thus, spin coater is programmed to coat the wafer
in two steps. In first step photoresist is spread on the wafer uniformly and second step
decides the thickness of the photoresist. Multi step spin coating programme as shown in
figure 3.20 is used for edge bead removal. Prebaking is also a crucial step in AZ P4620
lithography as oven baking causes cracks in photoresist film. The cracking of photoresist
during prebake in oven occurs due to early drying of photoresist surface layer. As
thickness is high, the solvent in bulk photoresist is trapped due to surface layer drying.
The release of this solvent from surface causes cracks in photoresist. Therefore, prebaking
in oven is done by providing temperature ramp. The cracking problem is solved by
prebaking the photoresist on hot plate as discussed above in section 3.3.1. On hot plate,
(a) (b)
Page 21
47
heating or drying of photoresist starts from the bottom surface and edges of the
photoresist, thus avoiding cracks. Also, AZ P4620 is highly sensitive to atmospheric
conditions, such as: humidity, temperature and time between consecutive steps during
lithography, e.g. a delay between prebaking and exposure can change the exposure dose
and developing time. In that case AZ 9260 is good choice for process repeatability, this
resist is pretty stable than AZ P4620. The process steps for AZ P4620 photoresist for 5-
5.5 µm are:
HMDS coating at a spin speed of 4500 RPM, 8000RPM/sec RAMP for 30
seconds and oven baking at 90ºC for 5 minutes.
Photoresist is coated using two step recipes shown in table 3.7.
After coating, prebaking of photoresist for 35 seconds at 60ºC on hot plate is
followed by lithography exposure time of 5 seconds and a developing time of
80 seconds.
In some cases, delay between prebaking and exposure changes the exposure
dose and developing time. This result in some PR traces, these PR traces is
resolved by using O2 plasma ashing.
Table 3.7: Recipe of AZ P4620 photoresist for 5µm
Coating Step Step 1 Step 2
Spin Speed (RPM) 400 6000
RAMP (RPM/min) 20000 20000
Time (sec) 5 15
Figure 3.20: Multi-step spin coating program for minimizing the edge bead.
t 1 t 2 t 3 t 4 t 5
Speed (RPM)
Time (sec)
Step 1 Step 2 Step 3
Sp
eed
(R
PM
)
Page 22
48
Table 3.8: Recipe of AZ P4620 photoresist for 10 µm
Coating Step Step 1 Step 2
Spin Speed (RPM) 500 2500
RAMP (RPM/min) 300 300
Time (sec) 20 10
The optimized recipe for AZ P4620 photoresist for 10-11 µm is:
Photoresist is coated using recipe shown in table 3.8.
After coating, prebaking of photoresist for 55 seconds at 60ºC on hot plate is
followed by lithography exposure time 16.5 seconds and a developing time of
130 seconds on SiO2 substrate.
Exposure time reduce to 14 seconds on gold substrate.
As mentioned above AZ P4260 is highly sensitive resist. Thus, new resist AZ 9260 is
procured and recipes are optimised for thickness ranges from 5 - 10 micron. The AZ 9260
is a thick photoresist with high viscosity and transparency. It is optimized to fabricate
mould for electroplating bridge and reinforcing in MEMS switch with stability and
repeatability. The photoresist (AZ 9260) is spin in three steps as shown in table 3.9. The
foregoing recipe (Table 3.9) is used to obtain films of varied thickness. The second step
mainly decides the thickness of the film. During this step, almost all of the solvent is
evaporated and a solid film is formed. The third step does not affect the thickness; it only
accelerates the solvent evaporation. The ramp plays a significant role in deciding the
uniformity of the deposited film.
Table 3.9: Coating recipe for AZ 9260 photoresist for thickness optimisation
Process Step 1 Step 2 Step 3
Spin Speed (RPM) 300 Speed varied* 0
RAMP(RPM/min) 5000 5000 1000
Time (sec) 5 15 0
* Speed varied from 1000-6000 with a step of 500
Two iterations are performed to analyse the effect of the ramp on uniformity as shown in
figure 3.21. First iteration is accomplished at lower ramp (300rpm/sec) and other at
higher ramp (5000rpm/sec). The 3D surface profiles of photoresist uniformity are shown
Page 23
49
in figure 3.21 for lower and higher ramp respectively. The standard deviation for lower
and higher ramp is 0.8500 and 0.0933 respectively. It is observed that ramp affects the
uniformity and for deposit a uniform layer higher ramp is required.
After coating prebake is given for 15 sec at 60ºC (hotplate). The exposure time is 20sec
and photoresist is developed using a potassium-based alkaline developer (Hoechst,
AZ400K), diluted by 1:4 in deionized (DI) water. The developing time is 130 sec. Effect
of spin speed on the photoresist thickness is illustrated in figure 3.22. Post baking is not
required for this photoresist. Figure 3.23 shows the SEM micrographs of fabricated AZ
mould and perforation. Photoresist profile is observed using tilted view of the
micrographs. After the optimisation of photoresist process next step is deposit metal as
structural layer. The electroplating process for metal deposition is optimised in next
section.
Figure 3.21: Effect of ramp on uniformity of photoresist AZ 9260.
0 1000 2000 3000 4000 5000 6000
4
6
8
10
12
Th
ick
ne
ss
(u
m)
Spin speed (rpm)
Ramp=5000 rpm/sec
Figure 3.22: Spin speed vs. thickness of AZ 9260.
Wafer 1, ramp=300 rpm/sec Wafer 5, ramp=5000 rpm/sec
Page 24
50
Figure 3.23: SEM micrographs of AZ mould and perforation.
3.4 Electrochemical Deposition
Electrochemical deposition involves the reduction of metal ions from aqueous,
organic, or fused-salt electrolytes. The reduction of metal ions Mz+
in aqueous solution
is represented by
Mz+
(metal ion in solution) + z e− (electrons) → M (metal deposit)
Two processes are used to provide the electrons for the reduction reaction: (1)
electroplating (or electroforming), where an external power supply provides the
electrons, (2) electroless deposition, where a reducing agent provides the electrons [10,
11]. In MEMS electrochemical deposition is commonly used to deposit surface
coatings, or in the case of electroforming, for producing an entire microstructure or
device. In electroforming, micro structured moulds of different materials (e.g.,
polymers/resist, silicon) are electrochemically filled with metals such as gold, nickel
and copper.
3.4.1 Electroplating
The material properties of electroplated metal is strongly governed by electrolyte
chemistry (type and concentration of ions, pH of solution and additives), physical
parameters ( temperature, current, fluid), and the property of the substrate (surface
quality, shape). The electroplating process is optimized depending on the shape of the
desired microstructures and metal to be deposited. The basics of electrochemical
deposition are found in several excellent books [10 - 14] and are summarized in this
section.
Page 25
51
3.4.1.1 Electrochemical Reactions
The general setup and operation of an electrochemical deposition cell are shown in
figure 3.24. Two electrodes are immersed into an electrolyte. By applying an electric
current, reduction (electron intake) occurs at cathode, and oxidation (electron liberation)
occurs at anode. Metal ions are released at anode via oxidation and adsorbed at cathode
via reduction. Substrate is used as cathode so that metal ion is reduced to form a solid
metallic layer. The anode is dissolved so that concentration of electrolyte is maintained
during electroplating. The two partial reactions are expressed by the following equations.
Figure 3.24: Schematic of a general electrochemical deposition cell.
Reduction (cathode): Mz+ + z e− → M Deposition of metal
Oxidation (anode): M → Mz+ + z e− Dissolution of metal (for a soluble anode)
The steady oxidation of the anode (a metal to be deposited) ensures a constant
replenishment of metal ions in the electrolyte. Sometimes inert anode such as platinized
mesh is also used. In that case, replenishment of electrolyte is made by manual addition
of metal salts into plating bath. The theoretical mass of metal deposited at cathode
( ) is governed using Faraday’s law as
⁄ (3.1)
Where, M = Molar mass the deposited metal;
I = Current,
t = Time
z = Valency
F = Faraday constant (96500 C).
Page 26
52
Decomposition of water leads to some additional reactions at anode and cathodes.
Oxidation of water leads to production of oxygen gas at anode and reduction of water
leads to release of hydrogen gas at the cathode. Total overall current is distributed among
these different reactions. The percentage of the total current associated with the reduction
of metal is defined as the cathodic current efficiency γ, and is calculated by the ratio of
the deposited effective mass meff to the theoretical mass mtheo
⁄ (3.2)
If production of hydrogen gas is not suppressed at cathode, it leads to reduction in
current efficiency.
3.4.1.2 Deposition Process
In the bulk electrolyte, cations are enclosed in a complex shell. This complex shell
consists of water molecules (hydration shell) or other complexing agents such as
sulfite or cyanide. Before applying a current, the ion concentration is homogeneous at the
electrode surface and in the bulk solution. When a current is applied, metal ion is
consumed at the electrode and a depletion region so created extends farther away into the
bulk as the deposition proceeds.
Movement of the complex metal ions in the electrolyte solution is governed by three
different mass transport mechanisms: migration, convection, and diffusion. In most of the
deposition processes, the conductivity of the electrolyte is relatively high and applied
potentials are moderate. Thus most of the electrical field drops across the electrical
double layer in front of the electrode and field -induced migration is minimal. Therefore
the predominant transport mechanism is convection (due to stirring or agitation of
electrolyte) and diffusion with former dominating in the bulk and later at electrode
surface. Figure 3.25 illustrated the overall deposition process.
In summary, the reduction of the metal ions at the cathode is divided into four parts: (1)
diffusion of the solvated or complexed metal ions from the bulk to the electrode
surface, (2) dehydration and transport of the cations through the electric double layer,
(3) cationic reaction at the solution–solid interface (ion uptake and electron transfer),
and (4) surface migration and incorporation of the adsorbed metal atoms into the lattice
[11].
Page 27
53
Figure 3.25: Schematic diagram of the electrochemical deposition process.
Figure 3.26: Concentration of metal ions as a function of distance from the cathode
(a) for direct current and (b) for pulse current plating [14].
In case of pulse-plating, the pulse current density is limited by the depletion of ions in
the pulsation layer, whereas the average current density is limited by the concentration
gradient in the outer stationary d i f f u s i o n layer [14]. Thus two diffusion layers are
defined: a pulsation layer in the immediate vicinity of the cathode and a stationary layer
up to the point where the mass transfer is controlled by convection (figure 3.26).
The final step in the formation of a crystalline metal deposit is the incorporation of the
adatoms into the lattice. The adatoms are preferentially incorporated at active lattice
sites such as grain boundaries, imperfections, or pre-existing built-up adatom clusters on
the surface. If the adsorption of an adatom ensued away from an energetically stable
Alignment H2O molecules Removal of H2O Hydrated metal ion
+ - +
(4)
(2) (3) (1)
+ +
+ cation
- electron
water molecules or
complexing agent
Bulk Solution Diffusion layer Electric
double layer Cathode
(a) (b)
Page 28
54
position, surface diffusion may transport the adatom to another active lattice site on the
surface. The process of either building new grains (nucleation) or contributing to the
growth of existing grains is define the formation of metal deposits in electroplating [10].
3.4.2 Electroplating Process Parameter
The parameters generally controlling the composition, structure and properties of the
deposit are briefly reviewed in following section.
3.4.2.1 Conditions affecting the structure and properties
The environment in the immediate vicinity of the cathode influences the characteristics
of electrodeposited metals. Electrodeposits are crystalline in nature and the form of
deposit mainly depends on two factors: (a) Rate of formation of the crystal nuclei at
cathode by ion discharge and (b) rate of nuclei growth into larger crystals. When
condition favours the rapid formation of fresh nuclei near cathode surface, the layer
consists of small, fine-grained crystals leading to a smooth and hard deposit. On the
other hand, if the circumstances favour rapid increase in nuclei size, the deposited layer
consists of large crystals which are rough in appearance. Many parameters do influence
the formation of crystal nuclei and its size. Some of them are considered in this study.
Current Density: At low current densities, discharge of metal ions occurs at a slow rate
allowing for sufficient crystal-nuclei growth time. Consequently the formation of fresh
nuclei is unnecessary. The deposits obtained at low current density exhibit a coarse
crystalline structure. At increased current density, the rate of discharge of the ions is
large, and fresh nuclei tend to form and the deposited layer consists of smaller crystals.
Thus increase in current within certain limits yield deposits that are fine grained.
However, there is a limit to this improvement because at very high current densities the
crystals tend to grow at cathode towards that regions where the solution is more
concentrated, thereby creating trees or nodules in the film [13].
Electrolyte composition: This includes the compounds supplying the metal ions (to be
deposited) and the supporting ions. The basic functions of the supporting ions or
compounds are to stabilize the electrolyte, improve solution conductivity, prevent
excessive polarization and passivation (especially anodic), and provide compatibility
with the desired plating conditions. Supporting ions or conducting salts reduce the
current shared by the metallic ions or complexes, making convection (agitation) a more
significant factor.
Page 29
55
Temperature: Increasing the temperature of electrolyte promotes the diffusion of ions
to the cathode preventing impoverishment which can lead to a rough deposit. It also
increases the rate of growth of the crystal nuclei leading to a coarser deposit. Maintaining
the bath at moderate temperatures, first effect dominates and the quality of deposit
improves. However quality of deposit deteriorates at higher bath temperatures.
Impurities: Electroplated films contain various types of inclusions/impurities. The
sources of these impurities are from one or more of the followings: added chemicals
(brighteners, levellers), particles (for composite films), cathodic products (complex metal
ions), hydroxides of metal to be deposited, and bubbles (hydrogen gas, etc.). Excessive
amount of additives in the electrolyte may cause a brittle deposit leading to break at
crystal interface.
pH: pH of the bath influences the discharge of hydrogen ions, causing the solution at
cathode layer to become alkaline and precipitate hydroxides or basic salts. The inclusion
of a significant amount of these compounds makes the resulting deposit exhibit a fine
grain structure.
3.4.2.2 Current Waveform
Direct current is the preferred mode of current supply for electroplating, a variety of
current modulations techniques also exists viz. triangular, saw-tooth and rectangular
shaped waveforms. Rectangular waveforms are further divided into two variants:
unipolar and bipolar. Table 3.11 illustrates the current waveforms of direct current,
pulse forward current, and pulse reverse current. Different current modulation scheme
affects the plating mechanism differently and thus the chemical and microstructural
properties of the deposited layer can be controlled by varying modulation parameters
[14]. The current waveform for pulsed electrodeposition (forward current) consists of
cathodic pulses (tc), separated by a current pause (tp). Pulse reverse electrodeposition
consists of a cathodic pulse (tc) followed by an anodic pulse (ta), where the current is
reversed for a short time. In addition, a pulse pause (tp) is also used. During the
cathodic pulse, metal ions are deposited on the cathode surface and loosely bound ions
are removed from plating sites during anodic cycle. In pulse plating, mean current density
(im) is defined by amplitude and duration of the various pulses. A pulse waveform having
the same mean current density as that of a direct current has larger peak amplitude.
The advantages of pulse plating over DC plating have been studied extensively. Various
Page 30
56
metal alloy compositions have been optimized for morphology, magnetic properties, or
mechanical properties [15-17]. Pulse reverse electroplating of copper is used in printed
circuit boards (PCBs) manufacturing to obtain uniform via and trench filling.
3.4.2.4 Theory and types of pulse supply
Current is toggled between ON and OFF state during pulsed plating. During ON period,
the concentration of metal ions next to the cathode surface is depleted and a layer rich in
water molecules is left. During OFF period, metal ions from bulk solution diffuses into
the layer next to the cathode and the cycle repeats itself. During OFF period, desorption
of gas bubbles and impurities occur at cathode.
Pulse plating is being used in real-to-reel selective plating, in automatic tab platters, on
barrel lines, in still plating, electroforming, anodizing, electro-cleaning, electro-polishing
and machining and, recently has been adapted by the semiconductor bump and wafer
plating technologies [18-20]. Pulse power supply have following advantages: increased
plating speed, improved distribution, lower deposit stress, finer grain structure, increased
ductility, improved adhesion, increased micro-throwing power, reduced hydrogen
embrittlement, and decreased need for additives [18].
3.4.4. Electroplating equipment
Various equipment are used for electroplating, ranging from simple to very complex. For
laboratory use, the setup is simple, as shown in figure 3.27. The setup consists of
power supply, hotplate, glass beaker and oscilloscope. Beaker contains the electrolyte
and rotating magnetic stirrer. The temperature of bath is maintained at a specified
constant temperature using the hotplate. A metal plate or platinized titanium mesh is
used as the anode. The power supply is equipped with a pulse module to enable pulse
plating when desired. An oscilloscope may also be used to monitor the applied pulses.
One such custom design electroplating station build in CEERI is shown in figures 3.28.
It can hold a large volume of electrolyte and has monitors for liquid (electrolyte) level
display, pH, and additives, as well as a continuous filtration of electrolyte and a
dummy plating cell for cleaning of the electrolyte. Filtration rids the electrolyte of
particles, which are interfering with the deposit. For large-scale manufacturing,
continuous filtering, salt replenishment, and pH maintenance are important issues.
Also, some baths may generate gaseous by products, so exhaust of these gases are also
considered.
Page 31
57
Table 3.10: Current waveforms for direct current, forward current, and pulsed reverse
plating
3.4.5 Electrodeposition of gold for MEMS
A gold film is a widely used material for many MEMS devices because of its excellent
stability, low electrical resistivity and high reflectivity for infrared rays [21, 22]. The
application field of the gold film includes low loss RF-MEMS, high performance
electromagnetic, and MEMS mirror with high driving current, high reflective Optical
MEMS device, biocompatible package, etc. In these devices, the gold film has been
deposited by vacuum evaporation, sputtering, electroless or electroplating using surface
activation process or seed layer [21]. Comparing to vacuum evaporation and sputtering,
Current waveform Mean current
im i
im ic c
c + p
im ic c − ia a c + a + p
Direct current
i m
Curr
ent
Den
sity
i
Time t
i c
tc tp i m
Curr
ent
Den
sity
i
Time t
Forward current i m
i c
ta
i a
tc tp
Curr
ent
Den
sity
i
Time t
Pulsed current
Page 32
58
the gold electroplating process was suitable for deposition on silicon three-dimensional
structure, because it can cover hidden planes conformally. The advantages of both
deposition technologies are in table 3.11.
Figure 3.27: Schematic of a laboratory scale electroplating unit.
3.4.5.1 Comparison of gold electrolyte
Comparison of gold electrolyte is shown in table 3.12. The mostly used electrolyte
electronics industry are cyanide based. But due to the toxicity of the free cyanide formed
during electrolysis, sulfite-based gold processes are the traditional alternative for wafer
applications. However, sulfite-based processes have suffered from issues with solution
stability and the necessity for annealing to achieve the desired deposit hardness.
The gold (I) cyanide complex Au(CN)2- has a stability constant of 1039
, which makes the
solution very stable, and shifts the standard reduction potential of gold (I) from 1.71V to
–0.61V (SHE) [23]. However, in recent years, the health and safety concerns regarding
the use of cyanide have overwhelmed the benefits it brings. In addition, cyanide
containing solutions are incompatible with positive photoresists used in the
microelectronics industry because they attack the interface between the resist film and the
(1) 50°C
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Cathode: Wafer
Temperature
Anode: e.g., Titanium
filled with metal
Inert gas inlet
Glass
Magnetic stir
Hotplate/stirring
Power
Pulse module
Oscilloscope
(10)
Page 33
59
substrate, lifting the resist and depositing extraneous gold under the resist. Therefore, the
demand for cyanide-free gold electroplating processes has increased significantly.
Figure 3.28: Photograph of a custom design electroplating system for 2-4” wafer,
electroplating unit for use in a cleanroom (a) Wafer holder j ig with platinized mesh
(b) plating facility for cleanroom (c) Dynatronix pulse power supply.
Table 3.11: Advantages of both deposition technologies
Vacuum Deposition Electroplating Aqueous Deposition
Close tolerances Lower costs
Wide choice of substrates Thicker coatings
Wide choice of coatings Coating complex shapes
Control and modification of deposit properties
Control of residual stress
The gold sulfite complex Au (SO3)2 3-
is the most common alternative to cyanide. This
complex has a stability constant of 1010
, much less than that of the gold cyanide complex.
(b) (a)
(c)
Page 34
60
Consequently, it is susceptible to disproportionation resulting in the formation of Au (III)
and metallic gold, especially at pH less than 7, where sulfite protonates to form bisulfite.
Therefore, stability is always an issue for sulfite-based gold electroplating solutions, and
stabilizing additives are usually necessary. Gold (I) thiosulfate complex Au (S2O3)2 3-
is
another alternative to cyanide. Its stability constant (1028
) is between cyanide and sulfite
complexes and it is stable in weakly acidic solutions. However, thiosulfate ions
themselves are not stable without the presence of free sulfite. They undergo
disproportionation and produce sulfite ions and free sulfur, which precipitates. Recently,
gold plating solutions containing both sulfite and thiosulfate have also been reported.
Table 3.12: Comparison of gold electrolyte
3.4.6 Gold electroplating experiments
Initially, experiments are carried out for characterization of the electroplating process to
achieve an optimum deposition rate and surface roughness. In the present study, a sulfite
based solution TSG-250TM (Transene, USA) is used for gold electroplating due to its
compatibility with a positive tone photoresist. Process flow decided for the gold film
evaluation is illustrated in figure 3.31.The starting substrate is a chromium/gold (30/200
nm) thin seed layer on top of a Si/SiO2 wafer. Subsequently, the photoresist is patterned
on the seed layer which is used as the cathode during the electroplating process while a
platinized mesh electrode is used as the anode.
Cyanide Based
Cyanide Au(CN)2-
Non-Cyanide Based
Sulfite Au (SO3)2 3-
thiosulfateAu(S2O3)2 3-
Toxic Non Toxic
Better for health and safety
concerns
Non Toxic
Better for health and safety
concerns
Incompatibility with
positive photoresists (attack
the interface between the
resist film and the
substrate)
Compatible With Photoresists Compatible With Photoresists
stability constant of 1039 stability constant of
1010(solution instability require
annealing for deposit hardness)
stability constant 1028
(instability of thiosulfate in
the absence of free sulfite)
SHE 1.71V to –0.61V
pH >7 pH less than 7 stable in weakly acidic
Sulfite protonates form bisulfite
cause instability
Sulfite ions and free sulfur
Recently both sulfite and thiosulfate Plating sol. used
Page 35
61
Figure 3.29: Process flow used for electroplated Au evaluation.
During the electroplating process, an anode/cathode surface area ratio of 1:1 is used with
a distance of 3 cm between the electrodes. Electroplating is carried out at different current
densities from 1mA cm−2
to 7mA cm−2
for a fixed duration of 10 min at a temperature of
60 0C. The deposition rate and mean surface roughness (Ra) of the electroplated gold are
measured using a Dektak6M profilometer and AFM. The final process parameter is fixed
to deposit gold thickness of ∼2μm at current density of 3.5 mA cm−2
and temperature 600
C for 10min. The measured thickness of electroplated gold on metallic structures of
various devices showed an average deviation of ±0.1 μm. The current density of 3.5
mAcm−2
is selected to obtain an electroplated layer with reduced stress and uniform
deposition [24-28]. Stress is characterized using surface profiler by curvature
measurement method and later using commercial stress measurement system (FSM).
Seed layer formation using
sputtering/evaporation
Cleaning of Wafer Degreasing,
RCA, Piranha and HF dip
Electroplating of Gold for MEMS
mechanical structures
Bath selection - Non Cyanide
DC Rectifier Pulse Rectifier
Constant Current source
Low stress deposit metallic structure with
minimum grain size and low roughness
Characterise: Deposit thickness, surface
roughness, grain size, resistivity using surface
profiler, SEM,AFM,XRD and four point probe.
Parameters Varied: current density, temperature
of electrolyte, plating Time (deposit thickness of
gold), Duty Cycle
Page 36
62
Finally, the Cr/Au seed layer used for the plating process is removed in their
corresponding etching solutions to get the desired metallic structure.
3.4.6.1 Gold film evaluation
Nucleation of plated gold layer is observed after plating of 10 seconds with other
parameters fixed as mentioned in section 3.4.6. Figures 3.30 depicts the SEM
micrographs during nucleation, the average grain size varies from 20-70nm. Figure 3.31
(a) shows the deposition rate of the electroplated gold for different current density values
using DC plating. The deposition rate is higher for higher current density, and hence the
thickness of the plated gold increases with current density for a fixed time (10min.).
Figure 3.31 (b) shows the RMS surface roughness and grain size for different current
densities. It has been observed from figure 3.31 (b) that the roughness and grain size
value increases with current density and deposition time [24, 25]. This is due to depletion
of gold molecules in the electroplating solution with time. Thus, subsequent addition of
the electrolyte during the electroplating process is necessary to maintain the pH value of
the plating solution constant and also to reduce the surface roughness.
In order to apply the gold plating to MEMS application it is important to evaluate the film
characteristics. An internal stress, grain size and a root-mean-square roughness with AFM
images of surface morphology versus plating current density for DC and versus duty
cycle for pulse plating is shown in figures 3.32 and 3.33. By measuring a curvature of the
gold plated silicon specimen, the internal stress is calculated in the gold film using
Stoney’s equation. As shown in the figure 3.31, internal stress of the gold film is below
70 MPa for pulse plating and 117MPa for DC plating, and it slightly decreases as the
plating current increases [25]. Figure 3.31 (b) shows low tensile stress values (35-85
MPa) that are practically independent of film thickness in the 0.4-2µm range. This is due
to the dominant contribution of the intrinsic tensile coalescence component. Increasing
the density of current from 2 to 6 mA cm-2
or the temperature of the bath from 40 to
70°C, results in higher tensile stress. The grain size in the electrodeposits varies with
current density as shown in figure 3.33(b) (from 20 to 70nm at 60°C with increasing
current density from 2 to 5 mA cm-2
).The surface morphology of gold film is measured
by AFM. At low current density plating, the plated surface looks rather sparse than at
high current density. The AFM micrographs of DC and pulse plating gold of 2µm
thickness are shown in figure 3.34. Figure 3.35 (a) shows the SEM micrographs of 2µm
plated gold surface topography at Si substrate and 3.35 (b-d) plated gold test structures.
Page 37
63
Figure 3.30: SEM micrographs average grain size varies from 20-70nm.
Figure 3.31: (a) Thickness versus current density and (b) RMS surface roughness and
grain size versus current density of the DC gold electroplating process.
Figure 3.32: (a) Stress versus current density for DC Plating and (b) Stress versus duty
cycle for pulse plating of gold.
0 1 2 3 4 5 6 7 8
0
1
2
3
4
5
Th
ick
ne
ss
(u
m)
Current Density (mA/cm2)
Measured Thickness
Theoretical Thickness
2 3 4 5 6
20
40
60
80
Current Density (mA/cm2)
Gra
in S
ize
(n
m)
40
50
60
70
80
90
100
110
120
Ro
ug
hn
es
s (
nm
)
Page 38
64
Figure 3.33: Grain size and roughness with surface morphology vs. Pulse duty cycle.
Figure 3.34: AFM micrographs of (a, b) DC plating 2D and 3D topography and (c, d)
after Pulse plating.
(a) (b)
(d) (c)
Page 39
65
Figure 3.35: SEM Micrographs of electroplated gold structures.
X-Ray Diffraction (XRD) analysis
X-Ray diffraction is a principal method for determining structure of condensed matter on
the atomic scale, including the lattice parameter and microstructure of alloy films. The
wavelength range employed for X-Ray diffraction varies from 0.01 to 10 nm, which is of
the order of the distance between molecules and crystal lattice constants. By comparing x-
ray data with the well-established reference data sources, one can obtain a detailed
crystallography and compositional profile. A simple approach to x-ray diffraction from a
crystal is to treat it as interference of rays scattered by a multitude of equidistant lattice
planes. When the path difference between the scattered waves is an integer multiple times
the wavelength λ, there is a maximum in the intensity and constructive interference,
described by Bragg’s law:
nλ = 2d sinθ (3.3)
where λ is the incident ray wavelength, θ is the angle between incident ray and the
scattering lattice planes, d is the distance between two lattice planes, n is an integer, and
2dsinθ is the path difference between the incoming and the scattered ray. If the unit cell
structure of crystal is known, with Miller indices (h, k, l), the lattice parameter, is defined
by equation 3.4 from the distance between atomic planes d as below:
Cantilever
Bridge test structure
Plated Au surface topography at Si substrate
Guckel Ring
(a) (b)
(c) (d)
Page 40
66
√ + + √ (3.4)
From the theory discussed above, the individual peaks of reflected x-ray intensity provide
information about elemental species and micro-structures of the samples studied. By
comparing the measured data to those of well-calibrated data, the chemical composition
and structure of the films can be profiled. In addition to the chemical information
obtained from the X-Ray diffraction spectra, we can also estimate the grain particle
diameters from the Debye-Scherrer relation.
The XRD data is used for grain size measurement for plated layer. According to Scherer’s
equation grain size is measured. The governing equations are as follows:
Crystalline Size
(3.5)
Microstrain
(3.6)
Estimation of number of crystallites
(3.7)
Where λ: X-ray wavelength (1.54178Å), B is the full width at half maximum, θ Bragg’s
angle, t film thickness and K constant average value (0.9). Figure 3.36(a) shows XRD
peak for 2μm thick plated gold film after deposition and 3.36 (b) shows after annealing
at 1500C. XRD peak are used to find out grain size using equation 3.5 and crystalline
orientation of Au deposit layer. Grain sizes are verified using AFM data. Finer grain
size and reduced roughness deposit using pulse power supply also measured using AFM
and XRD as shown in figure 3.33. The extracted parameters of electroplated gold for
minimum residual stress, fine grain size and roughness are finalized. The resultant
optimized plating recipes are used for fabricating single clamped cantilever structures as
well as RF MEMS switches.
Figure 3.36: (a) XRD patterns of plated Au film before and (b) After annealing at a
temperature of 1500C
Page 41
67
Nano indentation of plated layer
The Nano indentation is used to measure Hardness (H), Young’s Modulus (E) and the
stiffness of plated and oxide layer on silicon substrate. The extracted Young’s modulus is
used in next chapter for extracting residual stress from test structures. Measured data of
indentation is depicted in table 3.13, two samples are used, one is silicon oxide of 1µm
and another is plated gold of 2µm thickness. Figure 3.37 shows the plated surface before
and after the indentation.
Table 3.13: Nano indentation measurement on silicon and gold surface
Wafer #AO3 (Silicon dioxide)
Er (GPa) Standard
Deviation
( GPa)
H (GPa) Standard
Deviation (
GPa)
Stiffness
(mN/nm)
Standard
Deviation
(mN/nm)
166.02 10.72 10.28 1.44 58.61 3.62
Wafer # FDCO1 (DC plating Au 2μm on Si)
120.58 13.6 1.17 0.15 125.26 15.45
Figure 3.37: Gold surface before and after Nano indentation.
Copper (Cu) plating experiments for via filling
The via filling of silicon cavity is required for packaging of RF MEMS devices. Cu
electroplating is used for via filling. Experiments are conducted to fix the process
parameter for Cu plating. Copper film is electroplated on gold seed layer for via filling
using different current densities varying from 10mA/cm2to 70mA/cm
2 at constant
temperature (40°C) [29]. The table 3.14 shows SEM micrographs of the Cu plated
samples with resistivity at different current densities (CD). Silicon cavity is created using
TMAH (Tetra Methyl Ammonium Hydroxide) etching. Figure 3.38 (a) shows SEM
(a) Before indentation (b) After indentation
Indentation points
Page 42
68
micrographs of the silicon cavities before Cu plating and (b) single silicon cavity during
Cu electroplating.
Table 3.14: SEM micrographs of the Cu plated samples with resistivity [29]
Figure 3.38: SEM micrographs of (a) silicon cavities and (b) silicon cavity during Cu
filling [29].
Page 43
69
3.4.7 Summary of electroplating experiments
1. DC plating optimum current density is 3.5mA/cm2 at temp. 60°C.
2. Roughness and grain size are increase with current density.
3. Residual stress of gold in DC plating at current densities (3.5 mA/cm2) was found
100-120MPa. Whereas, residual stress in pulse plating is reduced to 35-60MPa at
duty cycle 5-15%.
4. Grain size as well as surface roughness decreases 10 times (6 -10nm) using pulse
power supply.
5. Reduced grain size results bigger grain boundaries hence resistivity decrease.
6. Grain size measurements using XRD and AFM are closely matched.
7. Young’s modulus measured for 2µm plated gold using Nano indentation is
99.3GPa, whereas hardness is 1.17GPa.
8. Copper plating parameters for via filling applications are fixed.
3.5 Sacrificial Layer removal using Plasma ashing and CPD
In order to control the in-built stress in electroplated structures, it is desirable to realize
the metallic structure in two steps. In the first one bridge membrane and CPW are formed
with thickness up-to 1.5 μm. In the second step the suspended bridge can be reinforced
selectively to reduce the in-built stress. In this step, seventh lithography defines the mould
for electroplating using AZ 9260. At this stage the spacer (HiPR 6517) pedestal is
covered with the Cr/Au seed layer. Though the required bridge thickness is only 2 μm, a
resist thickness greater than 5 μm is required. After pattering mould in AZ 9260, the
electroplating of 2 μm thickness is deposit using above optimised process parameters.
After the wet removal of electroplating mould in acetone, Au/Cr seed layer is etched in
commercially available Au and Cr etchants. The etching of Au seed layer from undesired
areas also etches the electroplated Au surface from bridge and CPW. Accordingly the
electroplated Au thickness should be more than the desired bridge thickness. In present
case, 1.5 μm of bridge thickness is desired; whereas, electroplating has been done to
deposit 2 μm thick gold layer. After Au seed layer etching, remaining bridge thickness is
1.5 - 1.6 μm. This step is followed by Cr seed layer etching. The etchant used for this step
is highly selective and does not affect or etch other layers. Slight over-etching is desirable
to ensure removal of Cr seed layer from perforations.
Final ashing of HiPR hard baked positive photoresist, which is used as a sacrificial layer,
is carried out in oxygen plasma stripper (dry release) to prevent stiction of the suspended
Page 44
70
membrane and Critical Point Drying (CPD) wet release. The final release step is the most
critical step in a switch fabrication process, step by step details of each releasing methods
are covered in the following section.
3.5.1 Dry Release method (Plasma Ashing)
Ashing is the process of dry removal of photoresist by generating mono-atomic reactive
species at high RF power. Oxygen and fluorine are the most common reactive species.
The reactive species combine with the photoresist to form ash which is removed using a
vacuum pump. The basic ashing reaction of oxygen-based plasma with photoresist
polymers is [30]:
+ i + + (3.8)
The initial experiments for removal of hard baked PR (HiPR6517) were carried out in
simple ashing equipment (laboratory made) typically used for removal of minor PR traces
before metallization. After a prolonged ashing cycle in oxygen plasma (approximately 7
hours) at 80 watts of RF power, the PR under the bridges are found intact whereas other
directly exposed areas are clean and free from PR residue. The residue remaining under
the metallic structures is attributed to the slow lateral etch rate of oxygen plasma. The PR
removal in a ‘Drytek megastrip 6 HF’ plasma wafer asher is also observed to be slow and
non-uniform, as the oxygen plasma (at 900watts) is inefficient in removing the hard crust
of the PR baked at 1800C. For the crust removal, a combination of oxygen and CF4 is
supplied to the plasma chamber in a ratio of 35:1.
Figure 3.39: (a) Ohmic switch before release (b) After 50 minute exposure to O2 + CF4
plasma [31].
(a)
Metallic bridge Sacrificial layer
HIPR 6517 Metallic bridge etching
(b)
Page 45
71
Fluorine ions are reported [30] to react with the top surface to form CF3- which acts as a
catalyst for faster removal of hardened PR. However, longer exposure (40-50 minutes) to
fluorine plasma damages the Au structural layers as shown in figure 3.39. Figure 3.39 (a)
and (b) show the bridge of ohmic switch before ashing and after 50 minutes of ashing in
O2 + CF4 plasma. The crust removal process is optimized using combination of O2+ CF4
plasma exposure for 3 minutes followed by oxygen plasma for 90 minutes with 10
minutes on-off cycle. The on-off cycle is required to maintain the processing temperature
below a threshold level so that the effects like thin film delamination due to thermal
expansion coefficient mismatch, and unwanted dopant diffusion can be avoided. State of
art ashing system maintains low substrate temperature by circulating chilled water or
maintaining liquid nitrogen ambient around the plasma chamber. Figure 3.40 shows the
optical micrographs of PR residue (Au beam removed manually) during and after the
ashing.
Process parameters of dry release: In this section, the effects of reactive ion etching (RIE)
process parameters on the lateral etching rate of the sacrificial photoresist during the dry
releasing process are presented. The vertical etch rate of the photoresist varies between
230 and 350nm/min and is faster than the lateral etching rate 38-77nm/min. Therefore,
the total release time is mainly determined by the lateral etching of the sacrificial
photoresist layer. In the ashing process the chamber pressure is maintained at 400 mTorr
with 900watts of RF power as shown in table 3.16.
Figure 3.40: Optical micrographs after (a) 15 minutes (b) 45 minutes and (c) 90 minutes
of ashing [31].
(a)
Au layer
Clean surface inside
perforation
Remaining
Photoresist
under the
lifted bridge
Perforated Au metallic
bridge layer
(b)
Remaining
Spacer PR
beneath the
lifted bridge
during ashing
Perforated metallic
bridge layer
(c) Clean surface
after ashing of
spacer layer
Metallic bridge
structural layer
Page 46
72
The vertical etch rate affects, after 15 minutes of ashing are visible in figure 3.40(a), the
PR directly exposed to plasma is removed whereas, spacer layer under the bridge remains
intact. The lateral etch rate affects, after 45 minutes of ashing are shown in figure 3.40(b).
Due to lower lateral etch rate the surface is fully clean only after 90 minutes as shown in
figure 3.40(c). SEM micrograph of successfully released RF MEMS ohmic switch using
optimized plasma ashing is shown in figure 3.41.
3.5.2 Wet release processes
3.5.2.1 Air Dry
Etching of photoresist (sacrificial layer) has been done in mild piranha solution
(H2SO4:H2O2::10:1) kept at 50-70°C. It is followed by DI water rinse for 3-4 times. The
wafer is then immersed in acetone, methanol, and isopropanol (IPA) for 5minutes each
respectively. IPA immersed wafer is kept at a small inclination in a petridish. Afterwards,
the wafer is dried in an oven at temperature of 120°C. Most of the structures are found
stuck to the wafer (typically the central part of the structure) and curled up at corners, as
shown in Figure 3.42 (a-b). The wet release method (drying in hot air) though cost
effective and simpler, can reproduce high-k (stiffness) structures (> 20Nm) only. The
method is found unsuitable for releasing the low stiffness metallic structures (0.05-
15N/m) without stiction and deformation [32, 33].
Figure 3.41: RF MEMS ohmic switch after plasma ashing.
Anchor
Anchor
RF
input
RF
output
Suspension
spring
Actuation
electrode
Suspended
Perforated bridge
Page 47
73
Figure 3.42: (a-b) MEMS switches stucked to substrate and curled up at corners.
3.5.2.2 Critical Point Drying (CPD) release
As mentioned above the sacrificial layer is hard baked (1800C) positive photoresist (PR)
having a plastic like structure. The sacrificial PR layer removal is carried out using mild
piranha H2SO4 (10): H2O2 (1) solution while maintaining its temperature between 50-
700C. Wafers are held vertically in a beaker as shown in figure 3.43 and dipped for 2- 3
minutes in mild piranha to release the photoresist. After spacer layer etching samples are
rinsed with DI water. This rinsing step is crucial, because micro machined structures are
in hanging stage at this point of time. Samples are subsequently transferred into CPD
chamber filled with isopropanol (IPA). The next section gives the detail of working
principle and process of CPD [34].
Figure 3.43: Vertical held wafer in beaker with SEM of lifted structures.
(a) (b)
Au beam stick to
Si-surface Lifted beam
Lifted beam
Lifted Structure
Wafer dipping direction
Page 48
74
Working principle of CPD
Various mechanisms (see figure 3.44) have been used to alleviate sticking in MEMS by
reducing the surface tension of the final rinse solution. For example, IPA, which has a
lower surface tension, is applied after the final water rinse to displace the water. A unique
way to reduce surface tension is critical point drying. This technique takes advantage of
the vanishing surface tension of a supercritical liquid. The method is explained in detail
below. This study indicated that by CPD technology gave superior and consistent results
compared to air dry (evaporation) or sublimation drying. The results from other methods
are mediocre and the by-products in sublimation drying were toxic and environmentally
hazardous. No such by-products were present in supercritical CO2 drying, but the setup
was rather complicated. The methods investigated by these researchers are also
summarized in a pressure temperature phase diagram in figure 3.45, which outlines the
various paths that are followed to minimize stiction [34-38]. In CPD process, the wafers
immersed in IPA are brought into a pressure vessel at room temperature. Liquid CO2 is
then used to replace the IPA during a rinse cycle. IPA is highly miscible with liquid CO2.
Next, the temperature of liquid CO2 is raised above its critical point. The pressure vessel
is then vented at constant temperature at T > Tcr and the CO2 escapes as a gas. A liquid to
solid interface is never formed during the process, hence surface tension is completely
suppressed and structure released successfully [37].
Figure 3.44: Pressure temperature phase diagram.
Page 49
75
Experimental description of CPD equipment
Critical point drying mechanism is one of the most efficient and economical methods for
removal the sacrificial layers used in conjunction with wet etching methods. The basic
mechanism involves the use of wet chemistry to dissolve the sacrificial (photoresist,
metal or dielectric) layer and transfer of the process wafer to CPD equipment for drying
the liquids without any change in the volume. The extraction of IPA from micro
machined samples are performed in a CPD, rated at 2000psi maximum allowable working
pressure, which is depicted schematically in figure 3.45 along with chamber details. The
CO2 source gas is supplied by cylinders of technical grade CO2 which is delivered
through a regulator, to a pneumatic compressor. The compressor allows for pressurization
above the critical point for CO2. The gas flows through a 0.5 micron sintered stainless
steel filter, into a temperature-controlled extraction vessel. Temperature control is
provided by a constant temperature circulating bath which flows heat transfer fluid
through a water jacket surrounding the outer wall of the vessel. After leaving the
extraction vessel, the CO2 flows through a high pressure metering valve through which
the pressure is reduced. The metering valve is warmed with heat tape to prevent clogging
the valve with dry ice that might otherwise be formed due to Joule-Thomson cooling
upon expansion of the gas. The pressure is reduced into a temperature controlled 500 ml
vessel held at 0oC. Isopropanol (IPA) is not as soluble in gaseous CO2 as in CPD, so the
solvent condenses and can be trapped in the separator vessel. Finally, the gas is vented
out of the laboratory through a suction duct. System pressure is monitored and controlled
with a modular pressure controller with analog read-out which controls the pneumatic
compressor to maintain the desired system pressure. Over-pressure protection is provided
by a 2000 psi burst disc. All tubing is 1/4 inch O.D. (0.065” wall thickness) 316 stainless
steel tubing.
The extraction vessel is made of Teflon material. The vessel is supported horizontally. To
minimize excess vessel volume and to support the silicon wafer pieces being extracted;
one hard Teflon half-cylinders are machined to effectively fill the entire vessel volume.
At the intersection of the half-cylinders, a channel is machined to allow for CO2 flow over
the samples. On the bottom half-cylinder on which the wafer pieces rest, a deep trough is
milled to serve as a reservoir for the IPA. With the Teflon cylinder vessel inserts
installed, the large vessel is reduced to an effective volume of about 200 ml. This
arrangement forces CO2 to flow in more direct contact with the isopropanol being
extracted from the micro machined samples.
Page 50
76
Figure 3.45: Critical point dryer (CPD) System.
Process flow in CPD: There are five mode of operation on CPD are as follows, the
schematic of the system is shown in figure 3.46. Cool: The process mode in which
coolant material is used to reduce the initial chamber temperature. Fill: During this mode
LCO2 is filled inside the chamber through fine nozzle in Teflon vessel. Purge: The flow
rate with which material IPA is replaced by LCO2. Always maintain fill rate greater than
purge rate. If purge rate greater than fill rate, the liquid level in the process chamber will
gradually drop below the level of sample. Therefore it is important to maintain grater fill
flow rate in the chamber than purge flow rate leaving the chamber. This will insure that
sample is always covering with liquid during purge. Heat: During this mode temperature
of the chamber increase along with pressure. This stage is continue until it reaches critical
point in temperature (310C) as well as pressure (1072psi) scale on measuring gauge.
Table 3.15 illustrate the change in temperature and pressure with time during critical
point. Bleed: a stage at which LCO2 and remaining IPA residual drain out after attaining
critical point. Vent: The stage at which pressure reduces to atmosphere and temperature
reduce to room temperature, sample is clean and dry.
Critical Point Dryer- CPD
Process
Chamber
Pressure
gauge
Temperature
gauge
CO2 Cylinder
Exhaust duct Filter
Page 51
77
Figure 3.46: CPD schematic for process flow.
Table 3.15: Change in pressure and temperature during critical point
Critical Point passage Heat Pressure Reduction
Time
(Minute)
Pressure
(PSI)
Temperature
(ºC)
Pressure
(PSI)
Temperature
(ºC)
0 800 6 1460 34
2 1300 20 1180 36
4 1460 28 1140 37
6 1480 32 1060 39
8 1450 33 900 39
10 1460 34 710 38
12 500 38
14 380 39
Figure 3.47 (a)-(d) shows SEM micrographs of cantilever array, beams, switch and test
structure released using CPD. The abnormally high vertical deformation of structures is
attributed to the placement strategy (in vertical direction) of wafers while dissolving the
PR and rinsing. The vigorous chemical reaction and turbulence caused during rinsing lifts
the structures upwards if held vertically (figure 3.43).
Figure 3.47: (a-d) Micrographs of CPD released structures when wafer held vertically.
Cool Fill Purge Heat Bleed Vent
(c) (d)
(a) (b)
Cantilever array Beam array
MEMS switch curl bridge
MEMS Test
structure
curling due to vertical release
Page 52
78
Improved CPD release method
The improvement in wet etching process is carried out by placing wafer in horizontal boat
during rinsing and etching. Placing the wafers in horizontal direction and gentle stirring
shows a drastic improvement. The method is found suitable for all types of micro
machined structures. The microstructures are an assortment of RF MEMS switches,
single pole double throw (SPDT), single pole four throw (SP4T), MEMS Test structure,
cantilever and bridge test structures [39]. Figure 3.48 (a,b) shows successfully released
RF MEMS ohmic switch and Au cantilever array extracted using CPD method which are
free standing with minimal deformation in case of longer cantilever.
3.5.3 Summary of releasing methods
CPD is demonstrated to release low stiffness metallic microstructures without stiction and
deformation. In contrast, approximately 90% of the air dried samples are stuck as shown
in figure 3.42. The air drying method resulted in surface stiction in the MEMS switches.
For some supercritically extracted samples, fabrication problems observed during etching
of vertically held wafers are resolved. Metallic structures are also released using modified
plasma ashing (CF4+O2) without any stiction problem. Some of the structures have
deformation due to rise in temperature of plasma asher. However intermittent plasma
ashing using a 10 minutes on-off cycle, to maintain temperature inside the asher
approximately at 800C, improves the planarity of structures. The structures released with
wet etching using CPD, are free from stiction and buckling as shown in figure 3.48. Table
3.16 summarizes the procedure and process parameters for all the releasing methods with
a comparison of their merits and demerits.
Table 3.16: Comparative summary of release methods
Release Methods Plasma Ashing Supercritical CPD drying Air Drying
Procedure
Parameters
O2
[100%]
CF4 [20%] + O2[80%]
(2-3min), O2[100%]
Mild piranha + DI water +
IPA
Mild piranha +
DI water + IPA
RF Power (watt) 80 900 N/A N/A
Chamber pressure (torr) 0.112 0.4 N/A N/A
Maximum temp. (0C) 60 80 0 N/A
Melting Temp. (0C) N/A N/A -56.4 [2] -97.5 [2]
Boiling Temp. (0C) N/A N/A -78.3 [2] 64.7 [2]
Surface tension (mN/m) N/A N/A 1.16 [2] 22.65
Process time (min.) 410 3+87 30 60
Advantage Compatible with IC processing Stiction free, No photoresist
residue, less process time
Simple
Disadvantage Complex recipes, residual stress,
time consuming
Wet chemistry control Mediocre results
Page 53
79
Dry release using plasma ashing typically results in damage to metallic structures or
stress related deformation due to rise in temperature (>800C). Plasma ashing with10
minutes on-off cycle is found appropriate to release structures without deformation. CPD
being a low temperature (31.10C) process is more suitable for compliant structures
without any deformation due to residual stress.
Figure 3.48: (a, b) RF MEMS Ohmic switch and cantilever array released using CPD.
3.6 Conclusions
Fabrication process for suspended metallic structures without any deformation or
buckling is presented. The out of plan deformation in microstructures are attributed to
residual stress of metallic layer. The metallic layer is deposited using electroplating
process. Electroplating process is improved using pulse power supply and fine grained (5-
8 nm), reduced stress (30-70 MPa) and uniform thickness layer is deposited. Photoresists
(S 1813, HiPR 6517, AZ 4620 and 9260) profiles and recipes for various thicknesses (1-
10µm) are optimised. The suspended microstructures are released using plasma ashing
(dry release) and CPD (wet release) techniques. Both releasing process are optimised for
minimum structural deformation. The improved processes (electroplating and releasing
using CPD) results in minimum deformation of 4-5µm in the micro machined structures.
Free standing
moveable beam
without deformation
Contact bumps
(a) (b)
Suspension
spring
Page 54
80
References
[1] K. J. Rangra, “Electrostatic low actuation voltage RF MEMS switches for
Telecommunication,” Ph. D. Thesis, International Doctorate school in information and
communication technologies, DIT, University of Trento, Italy, February 2005.
[2] Marc Madou, “Fundamentals of micro fabrications,” CRC Press, 2000.
[3] John N. Helbert, “Handbook of VLSI Microlithography principles, tools, technology,
and applications,” 2nd Edition, Noyes Publications USA, 2001.
[4] www.microchamicals.com
[5] www.brewerscience.com
[6] Shimul Chandra Saha, Hakon Sagberg , Erik Poppe , Geir Uri Jensen , Tor A. Fjeldly,
Trond Sther, “Tuning of resist slope with hard-baking parameters and release methods of
extra hard photoresist for RF MEMS switches,” Sensors and Actuators A,Physical 143(2),
452-461, 2008.
[7] H. Puchner, R. Castagnetti, W. Pyke, “Minimizing thick resist sidewall slope
dependence on design geometry by optimizing bake condition,” Microelectronics
Engineering, 53, 429 – 432, 2000.
[8] C.J. Hamel, “A low cost robust method to vertical sidewalls for thick resist
processing,” Proceedings of 2004 IEEE/SEMI Int’l Electronics Manufacturing
Technology Symposium, San Jose, CA, USA, 247–250, 2004.
[9] S. Audran, B. Faure, B. Maortini, J. Regolini, G. Schlatter, G. Hadziioannou, “Study
of mechanisms involved in photoresist microlens formation,” Microelectronic
Engineering, 83, 1087 - 1090, Feb. 2006.
[10] M. Schlesinger, M. Paunovic, “Modern Electroplating,” Wiley, New York, 2000.
[11] M. Paunovic, M. Schlesinger, “Fundamentals of electrochemical deposition,” Wiley,
New York, 1998.
[12] D.R. Crow, “Principles and Applications of Electrochemistry,” Stanley Thornes
Publishers Ltd., Cheltenham, 1998.
[13] David P.Arnold, “Additive process for metals,” MEMS reference Shell, 2011.
[14] J.C. Puippe, “Theory and Practice of Pulse Plating,” American Electroplaters and
Surface Finishers Society, Orlando, FL, 1986.
[15] D. Landolt, “Electrochemical and materials science aspects of alloy deposition,”
Electrochim. Acta 39,1075–1090, 1994.
[16] M.V. Rastei, S. Colis, J.P. Bucher, “Growth control of homogeneous pulsed
electrodeposited Co thin films on n-doped Si (111) substrates,” Chem. Phys. Lett. 417,
Page 55
81
217–221, 2005.
[17] F. Lallemand, L. Ricq, E. Deschaseau, L. De Vettor, P. Bercot, “Electrodeposition
of cobaltiron alloys in pulsed current from electrolytes containing organic additives,”
Surf. Coat. Technol. 197, 10–17, 2005.
[18] www.dynatronix.com
[19] Chandrasekar, M. and M. Pushpavanam, “Pulse and pulse reverse plating conceptual,
advantages and applications,” Electrochimica Acta 53(8), 3313-3322, 2008.
[20] Yeh, Y.-M., C.-S. Chen, et al., “Effect of Pulse-Reverse Current on Microstructure
and Properties of Electroformed Nickel-Iron Mold Insert,” Japanese Journal of Applied
Physics 44(2), 1086-1090, 2005.
[21] G Oskam, J. G. L., A Natarajan and P C Searson, “Review article electrochemical
deposition of metals onto silicon,” J. Phys. D: Appl. Phys. 31, 23, 1998.
[22] Ravi Sankar, A., S. K. Lahiri, et al., “Performance enhancement of a silicon MEMS
piezoresistive single axis accelerometer with electroplated gold on a proof mass,” Journal
of Micromechanics and Microengineering 19(2), 025008, 2009.
[23] Kai Wang, Rozalia Beica and Neil Brown, “Soft gold electroplating from a non-
cyanide bath for electronic applications,” IEEE/SEMI Int’l Electronics Manufacturing
Technology Symposium, IEMT-2004.
[24] B Margesin, A. B., I Guamieri, F.Giacomoui, A.Faes, “Stress characterization of
electroplated gold layers,” Dtip 2003: Design, Test, Integration and Packaging of
Mems/Moems 2003.
[25] Kal, S., A. Bagolini, et al, “Stress and resistivity analysis of electrodeposited gold
films for MEMS application,” Microelectronics Journal, 37(11), 1329-1334, 2006.
[26] Fujita, T., S. Nakamichi, et al., “Seedless layer - gold electroplating on silicon
surface for MEMS applications,” Sensors and Actuators A Physical 135(1), 50-57, 2007.
[27] Martegoutte, J., C. Seguineau, et al., “Influence of microstructure and internal stress
on the mechanical behaviour of electroplated gold freestanding thin films,” EPJ Web of
Conferences 6: 26009, 2010.
[28] He, S., J. S. Chang, et al., "Characterization of Young's modulus and residual stress
gradient of Metal Mumps electroplated nickel film,” Sensors and Actuators A, Physical
154(1), 149-156, 2009.
[29] Akshdeep Sharma, Deepak Bansal, Dinesh Kumar, and K. J. Rangra, “Experimental
study on sheet resistivity and thickness measurement in Copper Electroplating,” NSTI-
Nanotech 2013, 2,166-168, 2013.
Page 56
82
[30] H. Abe et al., “Developments of plasma etching technology for fabricating
semiconductor devices,” Japanese journal of applied physics, 47(3), 1435–1455, 2008.
[31] Akshdeep Sharma, Prachi Jhanwar, Deepak Bansal, Amit Kumar, Maninder Kaur,
Shilpi Pandey, Prem Kumar, Dinesh Kumar, and Kamaljit Rangra, “Comparative Study
of Various Release Methods for Gold Surface Micromachining,” J. Micro/Nanolith.
MEMS MOEMS 13(1), 013005, Jan–Mar 2014.
[32] J. Sharma et al., “Low-temperature wet-release process for low stiffness structures,”
J. Micro/Nanolith, MEMS MOEMS 7(4), 043007, 2008.
[33] Jaibir Sharma and Amitava Das Gupta, “Fabrication of reliable RF MEMS switches
in CPW configuration,” Advanced materials research, 74, 259-263, June 2009.
[34] I. Jafri et al., “Critical point drying and cleaning for MEMS technology,” Proc. SPIE
3880, 51–58, 1999.
[35] C. J. Kim et al., “Comparative evaluation of drying techniques for surface
micromachining,” Sensors and Actuators A 64, 17–26, 1998.
[36] C. W. Dyck et al., “Supercritical carbon dioxide solvent extraction from surface
micro machined micromechanical structures,” Proc. SPIE 2879, 225–235, 1996.
[37] R. Maboudian, “Surface processes in MEMS technology,” Surface Sci. Rep. 30,
207–269, 1998.
[38] http://nanolab.berkeley.edu/labmanual/chap2/2.13cpd2.
[39] D. Bansal et al., “Design of vertical packaging technology for RF MEMS switch,”
Proc. SPIE 8549, 854911-15, 2012.