by Michael A. Fury, Techcet Group
April 17, 2009 – Among the Thursday papers at the MRS Spring meeting, including several on eCMP, an “elegant” application of CMP to fix SiGe strain channels, and still-stealth “digital CMP.” Other notable papers described work on CNT fabrication and electronic devices, inkjet printing to replace sputtering in displays, and optical screening for metal hydrides.
Bruce van Dover at Cornell has developed an combinatorial procedure with optical detection for screening Pt alloy catalysts that will be effective for fuel cell applications. A primary motivator is to allow the Pt with less expensive components without sacrificing performance. A graded composition catalyst library is sputtered onto a silicon wafer, annealed, and placed in a methanol bath with appropriate electrodes. The added quinine (sans gin) fluoresces in the presence of generated hydrogen, indicating catalytic activity. To date, 141 Pt-X-Y ternary combinations have been screened, covering a significant percentage of the periodic table. While no home runs have been hit so far, several cost-effective candidates with competitive performance have been identified for further study.
The CMP symposium had several papers devoted to eCMP, a technique that had been viewed as a threat to the CMP status-quo several years ago, but had since been declared dead because it could not adequately address copper barrier removal. Rumors of its death appear to have been greatly exaggerated. This technique appears to have grown a small but loyal following in several companies that have independently found a prospective niche for it, including Selete, Seagate, and SoloPower. Others may emerge from the shadows as semiconductor-like process technology continues to spread into tangential fields.
Jang-Sik Lee of Kookmin University in South Korea showed how inkjet printing of nickel nitrate solution on α-Si can replace the need for nickel sputtering and liftoff in metal-induced lateral crystallization (MILC) for large area TFT displays. The Ni salt is reduced to Ni0 in hydrogen; on further annealing, the array of Ni sites seeds the crystallization to polysilicon. Performance of substrates formed with this new method is equivalent to the original liftoff Ni, affording a significant reduction in processing cost.
John Rogers of the U. of IL continues to earn my recognition for the highest “wow” factor at this conference. I reported yesterday on his method of fabricating devices on thin Si layers and transferring them to flexible substrates, providing single-crystal silicon mobilities to printable electronics. In a talk today, he showed a method for growing SWNT on a single crystal quartz template, aligned and of equal lengths. Using his PDMS stamping method, a CNT cluster then can be transferred to a substrate and integrated with other device elements. In one demonstration, a functional transistor radio was fabricated with CNTs comprising every circuit element. In a demonstration of hybridization on flexible substrates, he fabricated a circuit of 30 separate components with a mixture of Si MOSFETs, SWNT TFTs, and GaAs HEMTs.
In a separate presentation, Rogers demonstrated an electronic eyeball camera. Individual photodetectors were printed onto a flex substrate and wired together with serpentine interconnects. The array was then transferred to the inside of a hemispherical case the size of a human eyeball, with a simple lens that does not require peripheral correction for a planar detector. In the same talk, he showed a conformal brain monitor consisting of an array of 288 detectors with seven transistors each, designed to detect brain activity by sitting directly on the outer brain tissue. (A call for demonstration volunteers went unanswered.) As if flexible were not impressive enough, Rogers also demonstrated how to induce systematic folding of a large-area thin silicon membrane on an elastic substrate, allowing it to stretch 10%-20% in any direction without rupture. With periodic surface treatment of the elastic substrate to differentiate adherent and delaminated zones with the silicon membrane, stretchability of 100%-150% can be achieved.
Jonathan Melman of Intematix, Inc. developed a high-throughput optical screening method for metal hydride systems that are critical to the storage and release of H2 for large-scale fuel cell applications. The catalyst alloy library is sputtered onto alumina or graphite fiber paper to encourage discrete catalyst sites rather than continuous thin films. The library is then treated with a metal hydride slurry and tested in a window cell that operates up to 350°C and 600 psi H2 (nervous people need not apply). Optical properties are monitored for transitions associated with hydrogen uptake and release. The procedure and apparatus were qualified using a known material system; thus new candidates for further study can be screened rapidly.
Kentaruo Sawano of Musashi Institute of Technology (the MIT of Tokyo) presented an elegant application of CMP to a fundamental problem with SiGe strain channels. The lattice mismatch between the Si substrate and the SiGe buffer can propagate to the layer surface, resulting in a roughness of ~20nm. This imperfect planarity degrades optimum electron mobility in the strained Si or strained Ge layer deposited on top of it. Sawano reduced the SiGe roughness from 23nm to 0.5nm by removing a 200nm layer of SiGe with a CMP process that is essentially a replica of the process used for silicon wafer final polish. He then replaces the lost SiGe with another deposition, restoring the lost material but with a fresh, planar surface. The resulting strain channel layer showed mobility as high as 520,000 cm2/Vsec.
The big tease of the meeting from my perspective came from Rajiv Singh of the University of Florida, in collaboration with startup Sinmat of Gainesville, FL, who spoke in broad terms about the concept of “digital CMP” which has been demonstrated on a modified Westech 372. In a DCMP environment, the removal rate can be held constant, independent of down force and slurry flow rate — behavior that is decidedly non-Prestonian. Non-uniformity across the wafer was <2% with minor edge effects. For more information, we will all have to wait about another six months.
Michael A. Fury, Ph.D, is senior technology analyst at Techcet Group, LLC, P.O. Box 29, Del Mar, CA 92014; email [email protected].