Technology News
03/01/2000
Lithography's nemesis exposed
Crucial to the control of airborne molecular contamination in 248nm deep-UV lithography operations, particularly for chemically amplified resists, engineers have now identified an indeterminate number of "hidden" pollution sources that are neither obvious nor well documented. Well-documented sources include ambient air, polluted exhaust from other manufacturing areas re-entering the cleanroom through air handlers, certain cleanroom construction materials, etc. However, new hidden sources were identified by Carl Larson from IBM and Oleg Kishkovich of Extraction Systems in work reported in a poster session at the recent SPIE International Symposium on Microlithography.
|
Larson told attendees, "Some of these sources have existed for many years, but had little impact on earlier semiconductor manufacturing processes because the contamination levels were so low as to be tolerable. However, in the current era of DUV lithography, it has become apparent that even sub-ppb levels of these contaminants may adversely affect the manufacturing process without the fab staff realizing it." Larson's report at the SPIE conference focused on personnel, sealants and adhesives, and cleanroom gloves.
Even in a cleanroom with constant vertical laminar airflow, ammonia produced by a human body may produce a noticeable difference in local molecular base concentration. "The human body is surrounded with a cloud of ammonia at concentrations that may be significantly higher than ambient levels," said Larson. The airflow drags this ammonia towards the floor where some exposure tools intake air. "In addition, this means that human molecular base pollution may affect DUV lithographic performance especially when the track and exposure tools are not linked together and wafers are hand-carried between them."
Larson and Kishkovich found that some silicone caulks marketed as "odor-free, low off-gassing" in fact off-gas at a high rate and are, in reality, not at all suitable for use in DUV facilities (see table).
Sampling a variety of gloves routinely used in semiconductor manufacturing facilities, including the DUV area, Larson and Kishkovich found that some samples produced high levels of ammonia for more than two hours. "This time is reasonably close to the lifespan of a pair of gloves in a cleanroom. Using such gloves in a DUV area may lead to uncontrolled molecular base contamination, possible yield loss in production fabs, and confusing experimental results in research facilities," said Larson.
Larson and Kishkovich have concluded that performing a few simple tests on materials and supplies before bringing them into the cleanroom and DUV fab may substantially reduce chances of molecular-base contamination. "DUV practitioners involved in specifying cleanroom supplies and materials should be aware of the fact that even the most innocent supplies may aggravate the molecular base situation in the fab," said Larson. P.B.
Practical in situ FTIR for CVD yields revised TEOS-O3 model
Practical in situ sensors that provide real-time chemical reaction data during CVD processing are closer to a reality. Tom Whidden, principal scientist at Xylaur Enterprises (www.xylaur.com) recently described in situ spectroscopic characterizations and reaction modeling of the gas-phase reactions in TEOS-O3 SAP-CVD for SiO2 films. His report in Electrochemical and Solid State Letters (Vol. 2, p. 527, Oct. 1999) discussed in situ IR analyses of gases directly above a wafer surface during the deposition process.
|
In situ gas-phase spectrum of vapor in TEOS-O3 reaction during CVD process.
The work, being performed in conjunction with the University of New Brunswick faculties of physics, and electrical and chemical engineering, continues to focus on development of an FTIR-based sensor technology for practical implementation of real-time sensing and control in production CVD environments. Whidden tells Solid State Technology, "CVD system constraints, especially in single-wafer tool configurations, dictate that the bulky FTIR optics must be remote from the actual sensing port in the deposition tool. Fiber optics seems to be the most effective means to transmit the IR probe beam from the spectrometer to the sensing port, but implementation is complex."
Commercially available optical fibers that transmit in the mid-IR absorb light unevenly across the normal FTIR spectral bandwidth (2-20µm), causing the IR spectrum to exhibit detection sensitivities that vary depending on the wavelength being analyzed. "Spectral analysis of CVD reactions often requires simultaneous determination of the intensities of different absorptions in different regions of the spectrum and sensitivity variations of this type are unacceptable in quantitative measurements needed for control purposes," says Whidden.
To address this problem, the engineering group has developed new fiber-optic technology capable of uniformly high transmission of mid-IR radiation between 2 and 20µm. Then, they modified a conventional FTIR instrument for practical implementation of fiber-optic input of the IR-probe beam to the CVD chamber and used this configuration for in situ analyses during the TEOS-O3 reaction.
"From the spectra obtained, both absorption areas and peak heights showed simple, linear correlations with gas-phase concentrations of the precursors in the CVD chamber," explains Whidden. "Such accurate, real-time measurement of internal concentrations of precursors in CVD systems has obvious utility in maintaining process repeatability." Further, this group found that in situ analyses performed during actual deposition detected residual precursors and fully oxidized reaction products (CO2 and CO) as well as absorptions due to acetic and formic acids and intermediate silanol compounds that are present in the gas phase near the substrate surface (see figure).
These researchers have used spectral data from the deposition process to formulate more accurate chemical kinetic models of the CVD process. In their proposed model reaction initiation occurs via O2 or O3 reaction at the central Si atom to form a pentacoordinate transition state. The transition state rearranges to a silicon peroxy-alkoxide that then undergoes hydride elimination to generate a reactive silanol intermediate compound and acetaldehyde. Subsequent reactions of the silanol or its oligomers on the substrate surface yield the SiO2 film. Parasitic gas-phase reactions that produce unreactive siloxane oligomers and gas phase nucleation were also included in the model.
The modeling studies showed that the best fit of theory to experimental data could only be achieved if an additional mechanism over and above the reaction with TEOS was consuming O3 in the process. The researchers proposed a mechanism by which O3 is consumed through two possible reactions with the acetaldehyde produced in the reaction initiation step. These latter reaction channels were shown to yield acetic and formic acids as stable products.
"We feel this work provides an improved understanding of why acetic acid is detected as the primary organic byproduct of the reaction, rather than acetaldehyde as had been proposed in previous models based on the FTIR spectroscopy of static mixtures of TEOS and O3," says Whidden. "To model deposition rates of the dynamic process accurately, it is critical that a reaction channel other than the reaction with TEOS exists that consumes O3 in the deposition chamber." P.B.
Powerful simulation yields perovskite transistors
Engineers at Motorola's Physical Sciences Research Lab (PSRL, Tempe, AZ) have built thin functional transistors using perovskites (pronounced per-AHV-skites) a class of crystalline oxides with unique material properties. Specifically, they are using strontium titanate crystalline material on silicon substrates. While this technology is still at an early stage of development, it has already produced working devices that are electrically 3-4 times thinner than their actual physical thickness, without rapidly increasing leakage current.
Jim Prendergast, VP and GM of PSRL, says, "The key to this development was computer simulation of each individual atom at the interface using some of the most powerful simulation techniques available today. By understanding the behavior of the atoms in the structure, we were able to solve fundamental problems that have frustrated other attempts at using such new materials for a gate insulator."
This is the first time that such detailed computer simulations have been applied to the design of such an interface. The computer predictions were confirmed with carefully controlled experiments and advanced analytical capabilities in cooperation with the
University of Arizona in Tucson and the Stanford Synchrotron Radiation Laboratory (California).
Although perovskite materials have been explored for many years in the industry, Motorola Labs is the first to produce a working CMOS transistor to prove the concept, in its initial attempts. The resulting device, which demonstrated electrical properties superior to existing CMOS transistors, performed close to theoretical predictions.
Perovskites are a class of crystalline materials having a metal atom inside an oxygen octahedron structure. This structure gives them unusual properties such as high dielectric constant and even ferroelectricity depending on the specific atomic elements incorporated. Rare in nature, perovskites are found in Tanzania, Brazil, and Canada. In the lab, these structures are made artificially by building them up one atomic layer at a time, resulting in a pure and nearly perfect crystal.
To get the crystalline oxide material to match up with the silicon crystal, the interface between the two materials must be precisely controlled. Motorola used molecular beam epitaxy (MBE) to research the physics involved to precisely join the two dissimilar materials into a single crystal structure. The strontium titanate crystal needs to be rotated 45 degrees on the silicon surface and the number of defects at the interface must be less than one atom out of place for every ten thousand atoms.
Motorola Labs also demonstrated that it can grow these materials on silicon wafers up to 200mm, reportedly the first time that MBE has been used successfully on large diameter wafers. The precision of this technique produced layers in which the thickness varied only one layer of atoms across an entire wafer.
Low noise microchannels developed
With its micromachining technology and grant from NASA Goddard, NanoSciences Corp. is working to develop a high-gain, low-noise silicon microchannel plate (e.g., a micromachined grid of 6µm holes on 8µm centers) that can significantly improve the performance of particle detectors, astronomical observation instruments, and night vision systems. The key lies in applying so-called high secondary yield materials to the front end of a microchannel plate to enhance the first strike (photon) statistics, thereby improving both device gain and increasing noise immunity.
NanoSciences' unique technology lies in a significant discovery of a new process for producing ultra-high aspect ratio features deep into the bulk of semiconductor materials. In addition to it use for microchannels, company technologists believe the technology will enable a new through-wafer via technology for 3D VLSI interconnects, thermal vias, and other z-axis features.
Sub-0.1-µm filtration for 248nm resists
Crucial for fabricating linewidth <180nm, a study has determined that as filter removal rating becomes finer, resist performance in terms of photospeed, process window, and thermal stability does not change. These results where reported by Barry Gotlinsky and Michael Mesawich of Pall Corp. and James Beach of International Sematech at the recent SPIE International Symposium on Microlithography.
Gotlinsky said, "The implementation of filtration below 0.1µm within existing dispense systems raises concern as the removal rating of the filter approaches the size of large molecular weight components of the photoresist."
|
180nm lines fabricated with DUV resist dispensed through 0.04µm Nylon 6,6 membrane filtration.
The work looked for possible effects on 248nm DUV-photoresist performance when using 0.1-0.03µm filtration with PTFE, Nylon 6,6, and HDPE membranes. Filters were tested in two latest-generation dispense pumps, one using nitrogen pressurization, the other a stepper motor and diaphragm. Coated 200mm wafers, using constant dispense rates and volumes, were exposed at increasing exposure dose and focus in a DUV scanner to produce 180nm features.
After patterning, the wafers were analyzed for changes in critical dimensions (CDs), depth of focus (DOF), thermal stability, and resist profiles. Analysis of the data revealed that there were no changes in DUV resist performance when switching among filter membrane types. Gotlinsky said that at a dosage level of 29mj, all filters yielded 170±5nm CDs with a 0.6-0.8µm DOF, and that examination of cross sections before and after thermal treatment did not reveal any significant differences in the profiles of the finely filtered resists (see figure).
"The conclusion from this work is that, using existing dispense systems, photoresists can be filtered as fine as 0.03µm without significant polymer shearing or unintentional removal of important materials from the resist," said Gotlinsky. "Further, appropriate protection in terms of particle removal is possible as linewidths necessitate the use of finer filters in resist dispense pumps." P.B.
TECH BRIEFS
Technology developed at Simon Fraser University of Canada uses light in combination with metallic compounds to directly deposit metal and metal oxides. Advantages of this photoimaging process include low temperature processing, simplified process flows, elimination of vacuum processing steps, and low capital costs. Potential applications range from printed wire boards to advanced ICs. EKC Technology (Hayward CA) has exclusively licensed the process. EKC has research efforts underway to develop additional compounds for using the technology, including further research at Simon Fraser University by Prof. Ross Hill, the principal inventor.
International Sematech has installed it first 157nm F2 laser lithography microstepper from Exitech Ltd. This tool was installed in the Resist Test Center of the consortium's Advanced Tool Development Facility in Austin, TX. With a 1.5x1.5mm image field size, the Exitech tool will be used for high-resolution exposures at 157nm in resist development programs. The tool incorporates a 6W 157nm Lambda Physik Novaline F630 F2 laser and a 0.6NA 157nm imaging objective manufactured by Tropel Corp. Gerhard Gross, International Sematech director of lithography, says, "We are fortunate to be among only a few places in the world to have this exposure capability." The Exitech 157nm microstepper is the world's first F2 VUV laser processing system commercially available.