Technology news
09/01/2006
Helium ion microscopy promises better resolution than SEMs
A new helium ion-based scanning technology shown at SEMICON West by ALIS Corp. (since acquired by Carl Zeiss SMT) reportedly can generate higher-resolution images with more contrast than scanning electron microscopy (SEM).
The technology, incorporated into ALIS’s new LookingGlass LG-2 microscope, uses helium ions instead of electrons as imaging particles. The He ions can be focused into a high-current probe that, compared to an electron probe, excites a smaller interaction area, creates less scattering, and produces a greater ratio of secondary electrons/particle, enabling magnification at factors greater than 1,000,000×, according to the company. It also includes a technique to enhance the image contrast between different materials in a sample by assigning grayscale values according to the atomic number of the target atoms.
The LG-2 operates much like a focused ion beam (FIB) or SEM, using a source that produces a stream of charged particles; a column that accelerates, collimates, focuses, and scans the beam; a vacuum chamber that contains the sample; and detectors for the secondary particles used to generate images. However, compared to a FIB or SEM source, the He ion source is much smaller. The He ions are generated by an atomically precise, pyramid-shaped metal tip that can focus the beam to a probe size as small as 2.5Å. Given the smaller probe, the system is able to reveal details not normally visible with a SEM or FIB, the company claims.
A significant advantage of using He ions, according to ALIS, is that they have 300× the momentum of low-voltage SEM electrons. Correspondingly, the ions have a De Broglie wavelength that is 300× smaller, which means that they can pass through an aperture without behaving as much like a wave. Instead, the helium ions act more like particles and can propagate down the column toward the sample without significantly diverging.
Given how helium ions at this velocity interact with the atoms in the sample, they are able to penetrate deeper into the target material before dispersing. This creates an excitation volume with a much smaller surface area at the point of entry compared to a volume diffused more widely at the surface by an electron beam.
At the same time, the ion beam is said to improve secondary electron yield. Whereas in SEM instruments each incoming electron typically ejects only one secondary electron from the substrate, in the LG-2, each incoming He ion ejects 2-8 secondary electrons depending on the material, according to ALIS. Because a greater number of secondary electrons used in constructing images are ejected from a smaller surface area, a higher level of detail can be imaged.
Moreover, higher secondary electron yield enables greater differentiation between materials on the substrate. To that end, in addition to imaging with secondary electrons, the LG-2 uses a detector that collects so-called Rutherford backscattering ions (RBI), which scatter back from the nuclei of the more massive materials being targeted. The ions’ probability of being backscattered is proportional to the square of the atomic number of the targeted material-i.e., the heavier the material, the greater the backscattering of incident ions. This probability can be used to assign different grayscale values to each material-for example, tungsten could be displayed as nearly white, while copper would be light gray, and silicon a darker gray.
 Two different materials can be more readily distinguished using RBI (right) vs. secondary electron emission (left). |
In a comparison of a standard SEM image and an RBI image of a laser mark that has penetrated a substrate surface (see figure), with the higher contrast revealed in the RBI image, the two different materials can be more readily distinguished. -P.L.
Auto-fix for hot-spots in nanometer node designs
DFM start-up Takumi Technologies, building on its mask-data preparation (MDP) work for NEC and other customers, is now promoting its ability to automatically detect, classify, and repair yield-limiting design “hot spots”-areas of a design layout which, due to process or geometric conditions, fall outside of process windows, resulting in potential catastrophic or parametric failure.
Two new software tools work in coordination to optimize designs for manufacturability. Takumi’s Inspect software takes in the GDSII layout information after design-rule check (DRC) and then detects and rates hot spots against multiple yield-loss mechanisms. Defects are rated in terms of failure potential in parts-per-billion (ppb), allowing for a realistic estimation of manufacturing yield prior to tapeout. Yield loss mechanisms are evaluated concurrently, including those related to RET/OPC, lithography, random defects, systemic defects, and manufacturing tolerances, eliminating iteration time and minimizing the risk of missing interdependent phenomena. All ratings are performed using foundry-specific defect data.
Takumi’s Enhance software operates on GDSII layout data to detect, rate, and automatically repair hot spots based on critical area, single-contact hole or via, printability and edge-placement errors due to misalignment margins, and contrast issues. Enhance also uses inputs from third-party pre- or post-OPC verification tools to drive a 2D layout optimization, which can reduce the failure rate of a real chip from 96.3 ppb to 79.3 ppb. Users also can develop their own criticality rating functions.
Rules for fixes are fully programmable by the end-user, such as doubled contacts and vias on an “as needed” or a “where possible” basis. On average, each standard cell in a leading 65nm node library takes ~30 sec of Enhance software runtime to perform hot-spot correction-a large SoC with 500 standard cells can still be run overnight on a single Opteron 64-bit microprocessor with 32Gbit DRAM.
Working with Toshiba as an early customer, Takumi was given the tough challenge of finding ways to reduce mask costs by 20%. Analysis showed that tape-out labor was one of the greatest variable costs, adding up to ~$400k/month costs just for salary. S. Inoue, group manager, lithography process development, Toshiba Semiconductor Co.’s Process and Manufacturing Engineering Center, remarked that despite complexity and large data files associated with sub-65nm SoC technology, “we were able to reduce the number of hot spots from over 47,000 to just 40 in 12 hours.”
Takumi says that third-party design solutions like theirs are gaining favor as permanent augmentations to the design-flows of major IDMs. “We were working with 45nm data in 2004,” commented Tom Wong, Takumi Technologies’ VP of marketing. “We asked the customer when they would go into production-and they said 2008.” -E.K.
Promising progress in monitoring nanostructure fabrication
Researchers at the Georgia Institute of Technology have developed a new technique for growing nanotubes that may ultimately enable in situ monitoring of fabrication of nanoscale structures.
Instead of using a large furnace as part of the chemical vapor deposition process, bundles of nanotubes were grown on a microheater built into the tip of an atomic force microscope (AFM). By measuring the change in resonance frequency of the cantilever, researchers could accurately measure the nanotubes’ growth process.
In their work, a 10nm iron catalyst film was deposited onto an AFM cantilever using electron beam evaporation, and heated to form islands that provided catalytic sites for growing nanotubes. The cantilever was placed in a quartz tube and heated to ~800°C for 15 min; then a methane-hydrogen-acetylene mixture was flowed into the chamber, and an internal resistive heater in the cantilever ensured that chemical vapor deposition occurred only on the tip. Calibration of the cantilevers over a large temperature range using Raman spectroscopy was key to the research, the scientists noted.
 SEM image showing carbon nanotubes growing on the heated portion of an AFM cantilever. |
After removal from the tube, SEM inspection showed vertically aligned carbon nanotubes (5-10µm × 10-30nm), growing only on the heated region (see figure). Comparing vibrations of the cantilever before and after nanotube growth showed a resonance frequency decline from 119.10kHz to 118.23kHz, which then was used to calculate the nanotubes’ mass (4 picograms).
The next step of the research will be to combine the growth and measurement processes, creating a platform for in situ study of mass change during nanostructure growth. Arrays of cantilevers could enable parallel measurement of many different growth temperatures and conditions, accelerating the task of charting the growth kinetics to determine optimal settings. Because the cantilevers can be heated and cooled more rapidly than a traditional furnace, batches of nanostructures could be produced in 10 min, compared with ≥2 hrs for traditional processing, according to the scientists.
The process could apply to hundreds of electronic, magnetic, and optical materials grown using a similar thermally based technique, noted William King, assistant professor in Georgia Tech’s School of Mechanical Engineering. The production process could be scaled up once optimal conditions are determined. -J.J.M.