FABRICATION AND THERMAL ANALYSIS OF POLYIMIDE-BASED X-RAY MASKS AT THE SYNCHROTRON LABORATORY FOR MICRO AND NANO DEVICES, CANADIAN LIGHT SOURCE
Shen, Chen 1988-
In deep X-ray lithography (DXRL), synchrotron radiation (SR) is applied to transfer absorber patterns on an X-ray mask into the photoresist to fabricate high-aspect-ratio micro and nano scale structures (HARMNST). The Synchrotron Laboratory for Micro and Nano Devices (SyLMAND) at the Canadian Light Source (CLS) is a DXRL laboratory with continuous tuning capabilities for spectrum and power of the synchrotron beam. X-ray mask fabrication is one of the most demanding sequences associated with the entire processing chain and has always constituted a bottleneck in the DXRL technology. In this thesis, X-ray masks based on polyimide membranes are studied, including the development of a fabrication sequence as well as theoretical and experimental analyses of limitations of pattern accuracy. A 30 μm polyimide membrane was obtained by spin-coating photo-sensitive Fujifilm Durimide 7520® polyimide on a stainless steel substrate. Subsequently, a layer of TiO1.9 was sputtered onto the membrane as the plating base for the absorbers. On the plating base, 100 μm of UV-sensitive negative-tone resist Futurrex NR26-25000P® were spin-coated and patterned by UV-lithography. The patterned photoresist served as a template, filling the voids with 80 μm nickel by electrodeposition. These metal structures served as the mask absorbers in the test masks. Two test masks were fabricated, one with complete coverage and one with a center block absorber layout. In the next processing step, the sacrificial steel substrate was locally opened by etching with ferric chloride solution to create an X-ray transparent exposure window. The mask was finally bonded to a mask frame for support and rigidity. Polymer-based mask membranes are often avoided in DXRL because of large thermal distortions expected during X-ray exposure as a result of the low thermal conductivity. The power tuning capabilities at SyLMAND, however, allow the beam power to be adjusted and consequently limit thermal distortions. The heat load of the polyimide masks was analyzed by numerical analysis for the thermal and thermoelastic behavior of the test masks under synchrotron beam exposure. The beam power was calculated by the software LEX-D. ANSYS FLUENT® was used for the thermal analysis by computational fluid dynamics, and ANSYS Mechanical® for thermoelastic analysis using the finite element method. The thermal simulation results indicate that the main heat dissipation mechanism is from the mask absorbers by conduction across the rarefied helium gas in the proximity gap between the mask and its cooled surroundings. In DXRL, masks are vertically scanned through the synchrotron beam. Under the given conditions, this scanning speed of 50 mm/s was faster than the heat dissipation speed, such that a steep temperature gradient is observed between the exposed and unexposed mask areas as the beam scans across the test mask. The low thermal conductivity of the polyimide membrane can cause accumulation of heat in absorber structures such as the isolated center block absorber. At SyLMAND, an intensity chopper can effectively tune the incident beam power, thereby reducing the heat load in the mask during exposure. The temperature rises during exposure scale almost linearly with the incident beam power. Final temperatures of close to 39° C were obtained for both test masks at an incident beam power of about 14.5 W. To verify the numerical analysis, the actual temperature rises in the test masks during exposure were experimentally measured. Five thermocouples were bonded to the surface of the absorber to measure the local temperature. By comparing the recorded temperatures at different beam power settings, the temperature rises in the test masks were found to be proportional to the beam power, which verifies the numerical findings. Furthermore, the shape and size of the absorbers have a significant impact on the physics of the thermoelastic behavior. However, the increased power absorption associated with larger absorbers almost completely compensate the increased heat transfer capability along the higher conductivity mask absorbers in the examined cases. The experiments verified the simulations to a large extent. Deviations typically amount to 5° C at an overall temperature of approximately 40° C, which is mainly attributed to the size of the proximity gap varying in reality, and therefore differing from the model assumption of a constant gap. Finally, the thermal and thermoelastic behavior of the test masks was evaluated by an extended numerical analysis model for different typical exposure scenarios used for the DXRL exposure of 250 μm and 500 μm PMMA resist coated onto a silicon wafer substrate. In these simulations, the resist and substrate were also modeled. For a 25% duty cycle chopper setting, and spectral tuning as required for exposure, about 14 W to 20 W of incident beam power gets absorbed. A maximum mask temperature of 25.4° C is observed for 250 μm PMMA, and 31.0° C for 500 μm PMMA. While the resist deforms on the nanometer level, mask deformations in the lateral plane amount to approximately 2.3 μm for 250 μm PMMA, and to approximately 4 μm for 500 μm PMMA. These are worst-case values without further beam power reduction, and integrated over 6 cm large absorbers. Local deformations would be significantly lower. Such deformations are therefore deemed acceptable. The results prove that polyimide masks can be applied with acceptable thermal deformations under the conditions found at SyLMAND.
DegreeMaster of Science (M.Sc.)
DepartmentElectrical and Computer Engineering
CommitteeDinh, Anh van; Achenbach, Sven; Johanson, Robert; Sumner, David
Copyright DateAugust 2016