T. Jach and E. Landree

Objective: Standard methods of determining the thickness and composition of gate dielectric materials for CMOS devices on silicon run into difficulties when the total layer thickness is less than 100 nm. This problem becomes crucial in cases of the most advanced gate dielectrics which are grown as heterogeneous layers. A recent measurement by six techniques of silicon oxynitride samples circulated by SEMATECH pointed to variations between the total measured thickness of the order of 50%. There are at least four reasons for this: 1) physical limitations in the accuracy of the methods on this fine scale, 2) accuracy of the physical quantities which are the inputs for the methods, 3) diffusion in the layers, and 4) lack of detailed knowledge of the physical-chemical structure of the materials. A different method is required to provide both chemical analysis and thickness information in a non-destructive way.

Approach: The method of grazing incidence x-ray photoelectron spectroscopy (GIXPS), as developed at the NIST synchrotron radiation beamline X-24A at Brookhaven National Laboratory is highly suitable for analyzing the complex heterogeneous layers in the gate dielectric materials. The x-ray photoemission provides chemical analysis of the elements present in the layers, and the behavior of the x-ray fields at the grazing incidence provide thickness information. Seven samples from SEMATECH that had been measured by other techniques were measured and analyzed.

Results: Analysis of the data shows thicknesses which lie near the middle of the spread determined by the other methods. We also performed analyses to estimate the variation in results expected due to inaccuracies of the published physical parameters which go into fitting the data. GIXPS fits of certain samples which had resisted interpretation by methods such as ellipsometry, angle-resolved x-ray photoemission spectroscopy, and x-ray reflectivity indicate the formation of an unexpected surface nitride layer. Future Plans: We intend to test the accuracy of the results by varying some of the parameters (x-ray energy, composition) in a controlled manner. There is general correlation in the thickness measurements among different methods such that we should be able to understand the reasons for discrepancies and determine correction factors which will bring them into some level of agreement.

Web Contact micro@nist.gov