No abstract

Ion-attachment mass spectrometry is capable of accurate and in situ analysis of multicomponent gases that may even include unknown components. We developed a compact apparatus that uses this method and examined its performance for perfluoro compounds, a class of greenhouse gases, using typical standard gases, and subsequently applied the method to the in situ analysis of exhaust gases from a dry-etching machine, where c-C4F8/O2/Ar plasma is generated. Perfluoro compounds of low concentration can be detected as quasimolecules without any fragmentation. In the exhaust gases, many components that are not considered in the current measurement protocol [J. Meyers, D. Green, P. Maroulis, and W. Reagen, Equipment Environmental Characterization Guidelines, Rev. 3.0 (INTEL, 1999)] are measured, as well as components already reported.

The insideness problem is an aspect of image segmentation that consists of determining which pixels are inside and outside a region. Deep neural networks (DNNs) excel in segmentation benchmarks, but it is unclear if they have the ability to solve the insideness problem as it requires evaluating long-range spatial dependencies. In this letter, we analyze the insideness problem in isolation, without texture or semantic cues, such that other aspects of segmentation do not interfere in the analysis. We demonstrate that DNNs for segmentation with few units have sufficient complexity to solve the insideness for any curve. Yet such DNNs have severe problems with learning general solutions. Only recurrent networks trained with small images learn solutions that generalize well to almost any curve. Recurrent networks can decompose the evaluation of long-range dependencies into a sequence of local operations, and learning with small images alleviates the common difficulties of training recurrent networks with a large number of unrolling steps.

No abstract

No abstract