Thesis Title: Preliminary Results of a Micro Induction Motor <- download full pdf
Author: Jonathan L Klein
Location: University of Wisconsin - Madison
Date: 1998
(C) Copyright by Jonathan L. Klein, 1998
Downloadable is my thesis from 1998 as a searchable PDF. It contains unique information on Deep X-Ray Lithography (DXRL) for High Aspect Ratio Micromachined Structures (HARMS). I am publishing it on this site due to availability issues and errors in the original binding of the work (poor images and missing pages).
Some new technologies are:
- Measuring motor torque with a bending beam technique
- The real way to maximize magnetic field from a coil (not just maximize the number of turns)
- The real limits to deep x-ray exposure (high energy x-ray exposure requires a different look)
- Higher energy x-rays allow for deeper exposure for the same exposure time (speed of exposure is limited by thermal and rate dependent damage)
- Stacking photoresist samples to expose them all at the same time (increasing throughput and reducing cost)
- MKS version of the Green and Schwinger synchrotron equations (found a funny serious paper proposing centimeter-gram-second for MEMS since they are small)
- Computer exposure simulation software for source/mask/filters/photoresist and photoresist heating
- Based on a 1974 paper by H.D. Chai, a cool way to model permeance for the hand calculation of magnetic reluctance forces
Since original publication, most sources of Deep X-Ray Lithography for High Aspect Ratio Micro Machining have errors; namely:
- X-ray Lithography masks are expensive (we made hundreds of 1um critical dimension masks a year with 1um of plated gold on 1um of stress controlled, ultra pure, Silicon Nitride)
- X-ray exposure is expensive (compared to the $100-200M sub 10nm tools today, a $20M x-ray synchrotron is cheap, plus the area of exposure measured 10m away from the source in a solid angle is huge, and based on our patent (US5679502, JP11-502616, EP815565A1, AU1996057107, WO1996028829, Method and apparatus for micromachining using hard x-rays) multiple samples can be exposed in parallel by stacking
- X-ray exposure limitations can not be scaled to hard x-rays with a maximum top-to-bottom absorption ratio or total dose
Errors in the bound versions of this thesis include:
- At least 23 pages are missing (the last 20 pages from Appendix B, and all of Appendix C)
- The Appendix section was out of order (some pages missing, others duplicated, order wrong)
- Most photographs were reproduced poorly (this copy has the original SEM images in them)
Thesis Title: Fabrication and Comparison of All X-ray Lithography and All Optical Lithography Devices and Integrated Circuits <- download full pdf
Author: Jonathan L Klein
Location: University of Wisconsin - Madison
Date: 1992
(C) Copyright by Jonathan L. Klein, 1992
Downloadable is my Masters thesis from 1992 as a searchable PDF. It contains the first (and maybe only) MOS integrated circuit using x-ray lithography for all layers. I am publishing it on this site due to availability issues. The page numbers are a bit off from the original, I had to find a Mac Word 5.1 converter...
A little background: Integrated circuits (ICs) are getting smaller and smaller. One key technology in creating the transistors and connecting them together to form integrated circuits is photo lithography. Among other things, the wavelength of the light used is related to the smallest dimension transferred. Back in the 80's and 90's, SEMATECH sponsored research around the USA on x-ray lithography, expecting it to be needed in 10 years. IBM Fishkill got it working for critical integrated circuit fabrication layers in 1990 but did not release any information on it due to IP. Because of this thesis work, IBM started to release its information and by 2000, Motorola and a bit later ASML got it working in the laboratory. Everyone said that x-ray lithography is 10 years in the future, but it may always be 10 years in the future.
The photon source peak wavelength used by this project was around 4000eV or 0.3nm after filtering through 250um of Beryllium (EUV is currently near 10nm). The high energy x-ray source used my my PhD Thesis had a peak of 17000eV or 0.073nm after filtration (1mm Be + 400um Si) but could expose through many wafers stacked on top each other. For comparison, the diameter of a silicon atom is about 0.2nm.
What is the problem:
- At first everyone thought the penetrating photons of x-ray lithography would create a type of radiation damage in the transistor gate and substrate material that would make them inoperable. This thesis work showed that even with 100x more radiation, a mild post fabrication anneal (even at 480 degrees C for 30 minutes in 10% hydrogen 90 nitrogen) would completely remove all damage. IBM Fishkill data agreed.
- The second problem is the cost of the source of x-ray photons. I used a $20 million x-ray synchrotron. This is less of a problem today since an ASML photolithography tool is $150-400 million.
- The third problem is the difficulty in using lenses between the mask and photoresist on the silicon substrate due to the energy of x-ray photons. X-ray photons do not bend that easily. A mask drawn/created 1:1 with a very columnated x-ray source is one solution.
- The last problem is the mask itself. High Z (atomic number) absorber on a low Z substrate.
Many of these problems were explored in my PhD thesis on the x-ray lithography made induction motor. Very high energy x-rays allow the mask transparent material to be much thicker but the absorber material x-y dimensions still need to be 1:1. Parallel exposure through substrate stacking is a possible new advantage. Calculations showed a greater focusing with increasing depth which can prevent linewidth loss. So maybe x-ray lithography shouldn't be always 10 years in the future...
I also have a SEMATECH presentation of this research which I will post once I find a copy in my attic.
From "Preliminary Results of a Micro Induction Motor", Appendix A - Detailed Derivation of the Synchrotron Equations; a complete list of equations are presented with specified MKS units. From what I could find at the time, this is the first time these equations were presented in MKS (Julian Schwinger and Green used the esu cgs system). The purpose was to create a computer program (see Appendix B) to simulate the synchrotron and thick photoresist exposure to understand and explore optimization and limitations.
Synchrotron Parameters:
The radius of the electron orbit (synchrotron bending magnet radius) is:
where E the energy of the electron, R the radius of the electron path defined by the bending magnetic, B the magnetic field of the bending magnet, and γ the relativistic factor. All four parameters are derivable from two: either E and R, or E and B.
The relativistic factor (ratio of the electron's relativistic energy to the electron's rest energy )is:
The critical frequency expressed as a wavelength in Angstroms is:
The critical energy in units of eV is:
The energy ratidated by one electron per revolution is:
The total power radiated in Watts by a synchrotron is:
where B is the magnetic field, E is the energy in GeV, and I is the beam current in mA.
How was this all done at the University of Wisconsin Madison? We had a complete semiconductor/integrated circuit fabrication facility on site plus an x-ray synchrotron 20 miles away. The fabrication facility (WCAM) had all the equipment to design and fabricate masks (0.7um CD 0.1um delta CD), perform all depositions (12 diffusion/oxidation/anneal tubes, LPCVD poly/nitride), two ion implanters, photoligthography, metallization, dicing, packaging, and test. We were the only university with all this on-site. Fabrication classes would design a MOS process, run test devices, and then use the test results to design fabricate and test full integrated circuits; all in one semester. Most other universities used outside standard fabrication facilities and may have done the last fabrication step in-house with outside generated masks. For LIGA-MEMS we had a 24 hour process from design to working devices. ~8 hours to create the mask from CAD (custom CAD, pattern generator, step and repeat onto chrome plate, transfer to x-ray mask and plate in gold) and create the LIGA substrate (100-200um PMMA on thin metal film, on silicon substrate), 4 hours to x-ray expose, 1 hour to develop, a few hours to electroplate, a few hours to planarize and strip photoresist. Many iterations of designs could be done every year while most universities gave graduate students one chance to do one project. Pretty amazing. I was very lucky.