edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 27 NO. 3 22 both depositions are conducted using the (methylcyclopentadienyl)trimethyl platinum precursor with a chamber pressure of 1.5E-5 mbar, almost an order of magnitude increase over the chamber base pressure of 2.0E-6 mbar. The standard precursor-limited deposition is conducted with a 30 kV, 90 pA beam giving an instantaneous current density of 286 nA/µm2 and a global value of 6 pA/µm2 for a total time of 120 s using a standard Pt deposition recipe (15 µm by 1 µm box, dwell time = 100 ns, overlap = -100%, spot size = 20 nm, pitch = 40 nm, refresh time = 1.2 ms). The defocused, beam-limited deposition is conducted with a 30 kV, 750 pA beam with total spot size of 1 µm giving an instantaneous current density of 0.95 nA/µm2 and a global value of 15 pA/µm2 for a total time of 10 s using the following recipe (15 µm line, dwell time = 100 ns, overlap = 50%, pitch = 500 nm, spot size = 1 µm, refresh time = 10 ms). The dimensions of both deposits are nominally equal (14.6 µm long, 1.3 µm wide, 300 nm high). Note that the order of magnitude increase in deposition rate cannot be explained by the 2.5× increase in global current density. The most obvious difference between the two depositions is that the deposition rate of the defocused line is approximately one order of magnitude greater than what the traditional rule of thumb is capable of. This is thanks to the beam-limited condition that is created due to the combination of the three orders of magnitude decrease in the instantaneous current density, the fast dwell times, and the relatively long refresh times. This style of line deposition is useful when depositing a protective layer onto a substrate, or attaching a TEM liftout that is embedded in a substrate to a micromanipulator in 10 s. However, it fails when trying to attach a TEM lamella onto a TEM grid because the grid geometry does not allow for enough precursor gas to accumulate and act as an endless reservoir around the deposition. The other minor drawback to the defocused deposition is that the corners and edges of the defocused line are rounded because the shape directly reflects the ion beam spot geometry. It should be noted that a similar recipe can be used to optimize gas-assisted milling as well, for example when using XeF2 to mill Si, because the basic gas-assist milling mechanism also works best in a beam-limited condition. One could imagine combining this gas-assist milling recipe with the following approach for curtain minimization of FIB-milled cross-sections. MINIMIZING CURTAINING Minimizing curtaining can be done by combining the defocused ion beam with a few degrees of sample overtilt into the ion beam, e.g., if normal tilt is 52°, then add up to 5° of stage tilt, in addition to specific milling parameters using a modified strategy including short dwell times, a large y-overlap and a medium x-overlap. It is well known that sputter efficiency (the number of target atoms sputtered due to one ion atom) is maximized at an incidence angle of approximately 75° to 85° (where 0° is a normal impact into the surface and 90° is parallel to the surface) and depends on several factors including ion type, ion energy, and target materials to be sputtered. Figure 3 shows sputter rate of materials Fig. 4 The FIB-milled cross-section of a mechanically thinned silicon substrate exhibits minimal curtaining using a defocused ion beam in addition to overtilt and a modified milling recipe (stage tilt of 55° with a beam spot diameter of approximately 5 µm, a dwell time of 0.2 µs, an x-overlap of 50%, and a y-overlap of 99%). While this type of result is also possible using a rockingmill, the stage movements that are required for rocking create much longer mill times. Fig. 3 Sputter yield estimation of materials for 30 kV Ga+ ions impacting as a function of incidence angle where 0° is normal and 90° is parallel to the surface using SRIM 2008.04 with 10,000 Ga+ ions.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==