edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 25 NO . 1 24 OMNIPROBE’S FIRST CUSTOMER AND PATENTS The first purchase order in 1999went to a Silicon Valley customer, MonicaVargas of the IBMMaterials Laboratory in San Jose, Calif. The November installation coincidedwith ISTFA 1999, held in nearby Santa Clara, and Omniprobe founders later celebrated at Gordon Biersch Brewery near the conference venue (Fig. 6). We had a lot to celebrate because the installation almost never happened. One month prior to Monica’s installation, I discovered a problem. It was Saturday afternoon, and I called Tom in a panic: “Tom, guess what? Somebody patented our invention, and you’ll never guess who.... Hitachi!” It was 1999. The internetwas still immature. Google didn’t exist. U.S. patent applications were not published until after the patent issued. IBM ran a patent server that allowed a limited level of free public searches. Finally in 1999 and for the first time, the public could use Boolean logic to search patents for free. I searched on the terms “FIB” and “TEM” and received one hit.[19] It was the very first of what would later be many U.S. lift-out patents that eventually numbered into the hundreds on TEM sample preparation. The finding of this patent was a surprise, as we had not seen any publications related to FIB lift-out sinceHitachi's 1991 paper[1] eight years earlier proposing the FIB as a “cutter and attacher” machine. Therewas urgent necessity for a newworkflowprior to Monica’s installation, so over the next few days we brainstormed. We realized it could be even faster to first cleave a sample and then do a lift-out at the edge after a single ion-milled U-shaped cut, which creates a wedge-shaped specimen. A provisional patent application for this “single cut” method was filed in May 2000.[20] Soon thereafter, we realized that the wedge-shaped specimen createdby this approach could alsobe obtained by replacing the cleave with a second ion beam cut, creating a completely freed specimen that was safely held within the “pocket” created by the ion beams. The probe needle could then be attached to the totally released sample to lift it out. Apatent for this “total release”method issued,[21] and it became Omniprobe’s preferred method. It was extremely efficient, and one customer prepared a standard-sized sample ready for lift-out in only three minutes using total release and a dose-controlledmilling strategy. The total release method was fast because it required a minimum of milling to isolate the sample as a wedge that can be extracted. In addition, it enabled efficient in situ repositioning of lift-out samples for planar TEMpreparationor sideways thinning to reduce curtaining.[22] Total release also enabled large-area lift-outs, including across a 50µm-widebondpadexhibiting a liftedmetal failure after a ball bonding attempt as detailed in Fig. 7. In late 2001, the first volume order from an OEM was received. It was for 10 units and Tom realized Omniprobe could no longer be a part-time job. He had to consider leaving TI and his reliable paycheck at a time that coincided with the infamous internet stockmarket crash, which was the worst crash in the history of the semiconductor industry. Likemany other technology companies, TI stock prices plummetedbymore than 80% with no visibility into future recovery.[23] Our employee stock optionswere now too expensive and risky to cash out as required when leaving the company. It seemed a bad time to initiate a startup venture. Tomand I met with another local entrepreneur, whowas themother of our co-op student and owned a small Fig. 7 The Total Release lift-out method was applied to an entire bondpad that failed awirebonding attempt. 1) Total Release, through itsminimal material removal, enablesa>50µm-wide lift-out sample. 2) The liftedsample iswelded to a lift-out grid. 3) The entire width is thinned to capture the edges. 4) TEM imaging at the initiation site of the failure.
RkJQdWJsaXNoZXIy MTMyMzg5NA==