edfas.org 21 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 25 NO . 1 for process control; it is too slow.” And yet 30 years later, it is used for process control. Arguably, the advancement of semiconductors would have been slower after introduction of fin field-effect transistor (finFET) structures, if not for the invention of FIB lift-out. BEGINNINGS OF OMNIPROBE IN SITU LIFT-OUT In September 1995, I was a new hire into the Central Research Labs at Texas Instruments (TI) in Dallas. I joined colleague Rocky Kruger in reporting to Tom Moore, manager of the SEM and Acoustic Microscopy Labs, and I had arrived in the right place at the right time. An FEI 820 dualbeam FIB-SEM microscope had just been installed, and Rocky and I became its dedicated operators. We supported internal TI customers fromall over theworldacross a broad range of applications such as cross sections, TEM sample preparation, grain mapping, AFM tip fabrication, and circuit edit. In 1995, the H-bar method (also called the “ribbon” method at the time) was the standardway to prepare TEM samples by FIB.[4] It was time-consuming but tolerated, as it delivered unprecedented accuracy to prepare sitespecific specimens. It required producing a 3-mm long sample from a wafer that was then polished to a width of ideally < 25 µm with the target of interest contained in the center. This “Si ribbon” was glued to a slot TEM grid that had its upper portion removed to enable access by the gas injector nozzle and ion beam. Ion beam column performance at that time was much less than that available with today’s state-of-the-art FIB technologies. Due to large beam tails on our system, we were limited to 10 nA maximum current for bulk milling. As a result, it could take four hours or longer to remove the bulkmaterial froma 25-µmwideH-bar sample. The typical H-bar sample’s milled volume in our lab ranged from 25 x 5 x 5 µm3 to 50 x 10 x 5 µm3, with the largest number representing the H-bar sample’s cross-sectional polished thickness as prepared by the TEM lab. The bulk milling time to thin to a 100-nm thick sample ranged from 4 to 8 hours, and one could not walk away during the bulk FIB milling, due to the need to stay alert and frequently correct for sample movement or beam drift caused by charging. Tom Moore, as manager of the lab, had specified that only samples needing site-specific TEMpreparation could be submitted for FIB preparation. No bulk samples or thin blanket filmsamples were allowed. Nonetheless, the TEM lab submitted all kinds of samples, including the forbidden types, at the rate of 4-5 per day. In just one week, 20 samples accumulated in the queue, representing 80 to 160 hours of labor. Although the thin-film samples were returned to the TEM lab, the writing was on the wall. The volume of samples needing site-specific TEMpreparation would continue to grow for as far as one could see into the future, thanks to Moore’s Law, and the slow speed of FIB millingwas a huge bottleneck toward enablingproduction of high quantities of TEM samples that would be needed for characterization and failure analysis in the future. Tom, Rocky, and I stood around the FIB talking about the pain of slowH-bar TEM sample preparation times. Not enjoying babysitting the long, boring bulk mills to manually compensate for charge-induced beam drift, I said in frustration, “it would be so much faster if you could just FIB right around the sample and then take it out.” Immediately, Tom had an idea to leverage the pneumatic actuation technology used by the gas injectors to enable amanipulator that could be inserted and retracted on aGISport. Rocky suggestedusing ion-beam-inducedPt deposition for attaching to the sample to pull it out. But thenwhat? You had to put it onto something to get it into a TEM. Having just arrived at TI froma prior career as a TEM microscopist at UT Southwestern Medical School, I suggested taking one of the common cross-hatch TEM grids used by biologists and cutting it in half with a razor blade to create V-shaped landing sites for a FIB lift-out sample. Rocky got towork designing and building themanipulator while Tom led the project and worked on securing the needed parts. I called my former mentor Alasdair McDowall at UT Southwestern Medical Center and asked to borrow cross-hatch TEM grids. He readily complied, donating holey lacy grids as only thosewere readily available. This explains why you can see the remnants of the holey lacy film in the earliest lift-out images such as that shown in Fig. 2. You may not recognize Alasdair’s name. Hismentor was Jacques Dubochet, Nobel Prizewinner for cryo-EM. It was Alasdair who discovered, after two years of testing, that the secret to freezingwaterwithout formation of ice crystals, a process called vitrification, was to use ethane-cooled liquid-nitrogen.[5-7] Alasdair was recognized by Jacques at the Nobel ceremony,[8] and by loaning TEM grids to TI, became connected to the powerful invention of FIB lift-out, in addition to his groundbreaking discovery that enabled cryo-EM. Soon the first prototypemanipulator was built (Fig. 3). Establishing the lift-out process was not straightforward, and Rocky and I developed several strategies. The early lift-out effortswere not pretty, withonly coarse, open-loop motion leading to plenty of bent probe tips, but we were often successful. We were excited and well on the way toward a commercially ready solution to transform the
RkJQdWJsaXNoZXIy MTMyMzg5NA==