Development of ICP-based methods for quantification of sulfur in electro-deposited copper samples

Abstract

Over the past 20 years, copper has become more and more important in the semiconductor industry. Its outstanding thermal conductivity, allowing miniaturization of chips, has led to a wide use in the microchip manufacturing process. During this process, by using different materials and components, variable layer structures can be manufactured. On silicon wafers, a copper layer is deposited on top to allow a better high temperature consistency, making use of its great thermal conductivity and low thermal expansion. Those wafers are used for microchips in power devices as used in e.g. cars or trains. Deposition of copper layers on wafers can be derived through a sputtering process. Since this is no longer cost effective for layers thicker than 5 -m, a newer process for manufacturing of copper layer is via electroplating. For copper layers derived with this process, the addition of different additives to the liquid copper solution is needed to guarantee a homogeneous deposition. These additives contain sulfur. If sulfur is build-in into the copper layer during the electroplating process, it is a contamination, leading to different mechanical properties of the copper layer. Therefore analytical determination of the distribution and the total sulfur amount, in wafers manufactured through electroplating, are important. Goal of this work was to determinate the sulfur content of several copper wafers. Analysis of sulfur using ICP (Inductively Coupled Plasma)-based methods is not very common. Different problems occur when analyzing sulfur using ICP-MS (Mass Spectrometry) or ICP-OES (Optical Emission Spectrometry). One disadvantage of using ICP-MS is the occurrence of spectral interferences. For the analysis of sulfur on masses 32S and 34S spectral interferences occur, e.g. 16O16O. Another problem is low ionization potential of sulfur. Measuring with ICP-OES, spectral interferences seldom occur. However, sensitivity is much lower compared to ICP-MS, resulting in higher detection limits. Existing measurement methods, especially for measuring sulfur in a metal matrix, are not widely spread or commonly used, resulting in non-satisfactory results. A thorough method development is required. In this work the development and comparison of different ICP-techniques should be elaborated. Methods were developed and optimized using a stock solution for liquid measurements and self-manufactured pellets for solid-sample measurements. The copper wafers were analyzed as solid sample and after wet chemical dissolution measured on both ICP-MS and ICP-OES. To overcome interferences deriving from solvents, liquid ICP-MS measurements were conducted using a reactive cell, the reactive gas was O2 and 16O32S was formed and detected on m/z = 48. Results show that analysis of sulfur for liquid analysis is not possible without reaction cell technology for the ICP-MS. When using the reaction cell, limits of detection (LOD) for liquid sample analysis are almost the same using ICP-MS and ICP-OES. LOD around low ng/g were obtained. With this approach, sulfur contents for the copper wafers could only be determined using ICP-OES, since the WTi-layer of the wafers was also digested during sample preparation, leading to the detection of titanium on m/z = 48. Using ICP-OES, it was shown, that the different used additives lead to different sulfur contents, some of them under LOD. It was also shown, that the tempering process has influence on the sulfur content. Within the investigated wafer area, a homogeneous distribution was derived. For solid sample measurements much lower LOD was gained coupling the LA with ICP-MS than ICP-OES, deriving from the lower sensitivity of ICP-OES. Compared to liquid analysis, LA analysis is less sensitive. The big advantage of LA is that sample preparation is less work-intensive and no liquids are added. It also has better spatial resolution, allowing analysis of smaller sample-areas. The gained sulfur concentrations from liquid measurements were almost the same with LA-ICP-MS for the non-tempered wafer.