edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 25 NO . 1 6 the incident ionbeamandonto the lamella surface. During the thinning process, excess beam current can cause heating that results in precipitation of any volatile elements present (e.g., indium, gallium) to the surface, and excess beam energy can implant beam ions 10s of nanometers into the sample surface. Fortunately, techniques developed for atomic resolution imaging of FIB-prepared samples include steps to mitigate these effects,[3] though it does not seem likely that implantation can be avoided entirely. The most commonly used ion for FIB is Ga+, and surface implantationcanproducea layer of conductingGa, however plasma FIB (PFIB) systems, where the beamconsists of inert Xe+ ions, are becomingmore common. There are also a number of other elements (e.g., He, N, O, Ar) currently used in ion beams, each with varying contributions to implantation,[4] damage, and surface chemistry, making the ionparameter space apotentially fruitful place to explore in pursuit of higher quality sample preparation. For optimal sample thickness and uniformity there are a number of established procedures (e.g., tilting, rocking, post-FIB ionmilling) that may also improve the electronic quality of samples. While there is still a significant amount of development necessary to standardize processing for bias-enabled TEM samples, several successful samples have been reported recently.[5-10] IN SITU BIASING AND STEM EBIC In situ biasing allows for electronic and thermal manipulation of devices while imaging in the TEM, potentially revealing nanoscale device characteristics that are otherwise only studied via electrical testing or modeling. A sample can be electrically or thermally stressed while imaging, allowing live observation of the resulting nanoscale defect formation and real-time correlationwith electronic signatures of failure. The interplay between structure and function in contemporary components can be studied directly, providing rational directions for improvements in future iterations. The operatingmechanisms behind incompletely understood, next-generation components may be studied directly, expediting their development and mainstream implementation. In many cases in situ operation of devices may yield actionable information, but the ability to operate devices in situ does not guarantee that bias-induced changes will bedetectablewithstandardTEMimaging techniques. For a given device geometry under bias, somematerial systems may produce physical changes that generate obvious TEM contrast while others generate no TEM contrast despite electrical measurements confirming their operation (see, for example, references 11 and 12). This is largely due to the nature of TEM imaging itself. TEM’s scattering-based contrast only provides information related to the physical structure of a sample—the number and type of atoms and their arrangement. TEMmay be blind to electronic signals in samples, which are often the dynamics of interest for devices. The function and failure of devices often originate with changes to electronic structure that onlymanifest as a detectable physical changewhen such changes become pathological and produce defects. To take full advantage of clean, electrically connected TEM samples, TEM-based imaging techniquesmust be employed that directly probe the electronic structure of devices. One such technique is STEM EBIC imaging. EBIC is the measurement of current generated in a sample as it is irradiated by an electron beam, with EBIC images formed by measuring the current pixel-by-pixel at each beam position. The most commonly used mode of current generation, the separation of beam-induced electron-hole pairs by a local electric field, is referred to as the “standard” EBIC mode. Standard EBIC has been in use since the 1960s[13] and is routinely performed in an SEM, alongwith associated nanoprobing techniques such as electron beam absorbed current (EBAC) and electron beam-induced resistance change (EBIRCH).[14-16] Standard EBIC can also be measured in the TEM (Fig. 3), however it often yields images similar towhat can be obtained in the SEM; the large interaction volume associated with standard EBIC often negates the superior resolution of TEM. Recently,[1] an additional EBIC mode has been demonstrated in the TEM, called secondary electron emission EBIC (SEEBIC), which measures the holes left behind by emission of secondary electrons. SEEBIC is typically a much smaller signal than standard EBIC for a given beam condition and therefore requires much more sensitive current measurements for detection, often by several orders of magnitude. Owing to the higher beamenergy in TEMand the electron transparency of samples, the SEEBIC interaction volume is very small, enabling the measurement of SEEBIC with atomic resolution.[17] A unique property of SEEBIC is the ability to generate high resolution resistance contrast images, as the holes preferentially reach ground either via the EBIC amplifier or another path to ground depending on the relative resistance of the two paths.[12,18] SEEBIC resistance mapping can be performed while biasing, producing, for example, obvious contrast related to local conductivity changes in the early stages of dielectric breakdown.[12] The samplepreparationconsiderations for TEMbiasing samples are in alignment with those for obtaining highquality STEMEBIC: samplesmust be electron transparent,
RkJQdWJsaXNoZXIy MTMyMzg5NA==