Table of Contents

• Background and Introduction
• Processing
• Calculations/Predicted Performance
• Experimental Results
• Conclusions

The goal of this project was to create a flexible strain gage array. Possible applications of the device include force sensing at discrete locations such as on robotic fingers, human/computer interface devices, or a flexible biomimetic robot limb. All work except the high temperature nitrogen purge cure of the polyimide layers was done in Ginzton.

The array consists of five strain gage rosettes of three strain gages mounted on a flexible substrate. The structures consist of patterned platinum strain gages and gold interconnects encapsulated in 20 microns of polyimide, 10 microns top and bottom. The arrays are built up on a 1 micron sacrifical aluminum layer on a silicon wafer with an oxide layer. The final array structue is patterned with another 1500 angstrom aluminum layer which acts as an etch mask for the polyimide. The aluminum is then etched away to free release the strain gage arrays.

The major challenges of this project included: designing the fabrication so that the array may be removed from the rigid wafer, achieving good electrical contact between interconnects and the strain gages, and making contact pads on the flexible substrate that will be suitable for wire bonding.

Processing

The overview of the fabrication steps shows graphically the layers and procedures followed. Four mask patterns were defined in Autocad, printed at Type and Design on the 3048 dpi laser printer, and defined by contact printing onto pre-sensitized chrome glass. The mask fabrication is detailed on the maskmaking page. Five wafers were taken through the processing steps detailed below.

Aluminum - sacrificial layer
Wafers #1 through #3 were patterned with a sacrificial aluminum layer 1 micron thick using light field Mask #1 to allow partial release of the array with a rigid backing remaining behind the bond pads. The procedures followed for spinning and developing photoresist are on the photolithography page. Wafers #4 and #5 had a uniform layer of aluminum 1 micron thick to allow complete release of some arrays. The aluminum on wafers #1-#3 was etched back using a 40C etch of equal parts nitric, phosphoric, and acetic acid.

Polyimide - bottom encapsulation layer
The polyimide substrate was created by spinning and curing polyamic acid (Hitachi PIQL200 and adhesion promoter) on the wafers in two steps to create a 10 micron layer of polymide film. Procedures for the polyimide are posted on the polyimide page.

Platinum - strain gages
The strain gages are patterned with light field mask Mask #2 on AZ5214 photoresist and AZ400 developer in a reversal process to leave a dark field pattern of photoresist for liftoff. The reversal process is the one used by the Kenny group. An adhesion layer of 100 angstroms chromium then 1000 angstroms of platinum are evaporated onto the photoresist. The excess metal is lifted off by soaking the wafers face down in an acetone bath for several hours to remove the underlying photoresist.

Gold - interconnects
The interconnects are processed similarly using the reversal process with light field Mask #3 . The interconnect metal deposited is 100 angstorms of chromium and 5000 angstroms of gold.

Polyimide - top encapsulation layer
The second layer of 10 microns of polyimide is spun and cured over the gages and interconnects with the same procedures as the previous layer.

Aluminum - plasma etch mask
A second layer, 1500 angstroms, of aluminum is evaporated over the polyimide layer and is patterned with light field mask Mask #4. This layer of aluminum acts as an etch mask for the polyimide in the oxygen plasma, Phlegmatron, etcher. The mask defines the edges of the array structures, opens holes which act to mechanically isolate the strain felt by the rosettes as well as allow the aluminum etch to attack the underlying sacrificial aluminum layer, and opens holes over the gold bond pads for wire bonding. The openings over the pads are half the size of the bond pads to provide some mechanical resistance to gold tear out during wire-bonding.

The calculations for approximate resistance of the gages and the interconects as well as the predicted gage factors are provided on the calculations page.

The actual dimensions of the interconnects and gages were measured using the profilometer. Images of the resulting structures were captured using the microscope and videocamera station. The resistance of several gages and interconnects were measured using the probe station and multimeter.

Aluminum - sacrificial layer
The first layer etchback resulted in roughly defined sacrificial pads on the silicon wafer. After developing and etching, many of the alignment marks had disappeared and the outlines of the pads were wavy. This may be the result of the etch solution (Nitric-Phosphoric-Acetic 1:1:1) being very corrosive and undercutting the resist or poor resist adhesion perhaps caused by expired photoresist which did not adhere or develop well. First aluminum layer before polyimide.

Polyimide - bottom encapsulation layer
The polyimide layer spun and cured as expected.

Platinum - strain gages
The lift-off of platinum to create the strain gages did not perform very well; many of the strain gages and parts of gages were washed away (wafer 4 - bad gage, wafer 5 - bad gage, wafer 5 - bad gage intersection), and some of the platinum between the lines of the gages did not wash away (wafer 5 - poor adhesion/debris). However, on some parts of the wafer the gages patterned very well (wafer 5 - good gage). A suspected cause is poor adhesion of the metal to polyimide despite the chromium underneath the platinum. The photoresist also may not have developed completely. Additionally, the reversal lift-off recipe used has not been tested on a polyimide substrate before, and adjustments to timing and temperatures may be necessary. Since some parts of the wafer developed very well while others did not suggests the reversal only worked on some parts of wafer. The recipe may have to be revised in future work. More attention must also be paid to removing films (e.g. developer) between layers; the literature suggests that adhesion is improved by including an oxygen plasma descum before applying polyimide.

Gold - interconnects
The patterning of the gold interconnects suffered many of the same problems as the platinum. In places, the gold between the interconnects did not lift off properly and tore at the gold that was supposed to remain (wafer 1 - good liftoff gone bad). On some parts of the same wafer, however, the gold patterned very well (wafer 1 - good interconnects). Adhesion at the junctions of the gold and platinum was also inconsistent. Some overlaps looked good wafer 5 - good junction with typical resistance values of 7-10 ohms. With poor adhesion of alignment marks, some wafers exhibit poor alignment of the interconnects and gages wafer 1 - good junction, poor overlap. Several interconnects showed good adhesion to polyimide, but lifted off the platinum at the junction wafer 1 - bad junction, wafer 1 - another bad junction

Polyimide - top encapsulation layer
The second layer of 10 microns of polyimide was spun and cured, and all wafers looked good except wafer 5. Presumably, a film (of developer?) trapped between the 2 polyimide layers vaporized and caused the polyimide to bubble up and blister. However, this allowed the top layer to be peeled back easily by breaking the wafer; the underlying structures of platinum and gold peeled off with it.

Aluminum - plasma etch mask
The liftoff of the aluminum was a complete failure and the plasma etch to free release the structures could not be done. Changes were made to the reversal process in an attempt to improve the liftoff process and obviously made the situation worse. The exposure was increased to 70 seconds, the 95C pre-bake was 110 seconds long, and the 105C reversal bake was 3.2 minutes. These changes should not have caused the complete peel off of the liftoff. No Chromium was deposited with this layer of aluminum since the aluminum was to be etched away; this adhesion layer may be crucial and an etch which removes chromium and aluminum without attacking polyimide, platinum, or gold may exist. Before attempting this last step again, more research into the cause of failure is necessary.