Predicting and measuring scummed contact holes

BackgroundCD-SEM after develop inspection (ADI) and measurement of contact holes can detect missing or undersized holes but may have trouble detecting a hole with an acceptable critical dimension (CD) that is scummed (that is, where the resist is not completely cleared out of the bottom of the hole). For a modestly high aspect ratio (height over width of the hole), the very bottom of the hole is often not sufficiently visible in the scanning electron microscopy (SEM) image. In fact, a previous SEM simulation study showed that the aspect ratio, rather than the hole width and/or height individually, and the sidewall angle control the SEM signal emanating from the bottom of a contact hole. In an inevitably noisy SEM image, the signal coming from the bottom of the hole must be strong enough to be reliably detectable over the variations in that signal due to SEM noise.AimIt is important to understand what circumstances lead to the reliable detection of scummed holes using ADI SEM images.ApproachAlthough the previous study was theoretical, this work will address this question experimentally while also validating the results of that prior simulation study. Here, post-develop CD-SEM images of contact holes were collected under a variety of mask biases with varying doses to generate varying aspect ratios and likely different levels of scumming. To develop a best-known method for scum detection ADI, we evaluate the contrast detected in the CD-SEM images, apply a model to detect scumming of contacts after development, and compare results to atomic force microscopy measurements.ResultsWe demonstrate the relationship between contact hole aspect ratio and CD-SEM image grayscale using a physical model and apply it to samples to determine the rate of scumming. Furthermore, we demonstrate how to determine the amount of scumming in a contact hole and the sensitivity of the model in determining that value.ConclusionsAlthough scumming can be difficult to detect with other experimental methods, modeling image grayscale shows a viable pathway for determining the rate of scumming for a particular process, useful for process development investigations of stochastics-driven defects at ADI.