edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 25 NO. 4 22 results leveraged the flexibility offered by the RPM-3D enabling sensing scans to be inserted at any interval into a material removal process, giving the user the ability to frequently monitor the removal rate in a form of “live monitor” as the operation was executed on the surface.[6] However, controlling the rate of material removal is understandably not a straightforward exercise, largely due to the large number of parameters involved in sliding nanocontacts.[7] This results in a substantial scan-to-scan variability observed in the material removal control during the high-pressure scans. While accelerating the entire process for tomographic data acquisition, the probe switching automation does not address the main factor connecting RR and load force. For example, although frequent monitor scans were available to the users, the latter could not request a fixed RR during the entire acquisition, thus requiring human intervention for force adjustments during the process. Moreover, the RR is related to the local mechanical properties of the layer under the probe. Therefore, for real case devices, generally constituted by various layers, this inevitably presents a variation of the RR during the removal. CONTROLLING TIP-INDUCED REMOVAL OF MATERIAL A possible strategy for understanding and eventually compensating for this variability is a feedback system in which frequent measurements of the removal rate are used to adjust the load force on the removal probe. This possibility is enabled by the system due to the presence of multiple dedicated probes, allowing a fast data acquisition “depth scan” to be inserted at regular intervals into the workflow between pressure-induced material removal and sensing scans. As frequent monitor scans can be collected, different algorithms to adjust the load force on the probe can be considered, thus enabling a controllable and consistent RR during data collection. Figure 2b shows the block diagram strategy that is implemented with the objective of allowing the user to obtain a consistent RR. Here, the user inputs the initial force, the desired RR, and the total depth of the tomogram. Therefore, after the first scalpel scan performed at the initial load force, the RPM-3D executes the steps shown in Fig. 2b to automatically determine the removal rate, by monitoring the height difference inside and outside of the machined region and evaluating how to proceed. Depending on the user-defined RR, the applied load force can then be modified to increase or decrease the RR for subsequent scans. This data is used to adjust the load force as a function of the needs during the removal, as shown in the block diagram (Fig. 2b). A Python script has been developed for the automated parsing of the step height and the comparison to the target removal on the readout scans. On subsequent steps the newly acquired RR is compared with the desired value and if it does not match within a +/-20% error, the force is recalculated, and the RR is reevaluated on the next removal scan. In this way, the system can take control over the amount of the removed material and convert it into actionable changes to the load force (voltage). Importantly, this automated step can be repeated with an arbitrary frequency, thus allowing for active RR control on each removal scan of the 3D scalpel Fig. 3 (a) Automated RR control operation for SiO2, where the user requests a RR of 4 nm/scan, and the system obtains control with good agreement by continuous adjustment of the load force (here reported for two probes with different spring constant). (b) Comparison of depth removal without RR control for a 3D NAND dummy structure, in the inset. Note the top inset shows the continuous adjustments of the load force during removal for the case of RR control ON. (a) (b)
RkJQdWJsaXNoZXIy MTMyMzg5NA==