Journal of Vacuum Science and Technology A: Vacuum, Surfaces and Films

Thin ZnO and ZnO:Al films have been prepared by reactive dc magnetron sputtering, using both metallic and ceramic targets. The influence of oxygen flow and total pressure on film stress, structure, texture, and surface roughness has been examined. The properties of the films strongly depend on the deposition conditions, in particular the total pressure. Increasing the pressure from 0.5 to 2 Pa resulted in rougher, and less strained films. Increasing the oxygen flow at a constant total pressure led to a slight increase in stress, but had no visible effect on the surface roughness. Structural investigations by x-ray diffraction (XRD) showed that polycrystalline films with pronounced preferential orientation were formed. Most XRD-spectra showed a coexistence of strained and unstrained ZnO in wurtzite structure. The relative abundance of the relaxed phase is a function of the intrinsic stress only, and does not directly depend on deposition parameters. However, the deposition parameters determine the stress within the film, and thus influence the amount of the relaxed grains.

Titanium aluminum nitride films (Ti1−xAlxN) have been deposited by reactive magnetron cosputtering. Elemental compositions of these films have been determined by core level photoelectron spectroscopy. Scanning electron microscopy reveals a columnar film growth. This is also reflected by the topography of film surfaces as studied by atomic force microscopy. By x-ray diffraction a crystalline atomic structure is revealed. Single phase samples can be obtained, consisting of the substitutional solid solution (Ti, Al)N. Crystallites show preferential orientation. The optical properties of these films have been investigated by spectrophotometry in the UV-VIS-NIR wavelength range. Depending on the elemental composition, the optical constants vary from metallic to dielectric behavior. For film compositions with x<0.5 typical features are a tunable transmission maximum and reflection minimum in the visible spectral range, a high infrared reflection, and a low infrared absorption. Due to these optical properties, Ti1−xAlxN films are promising candidates for applications such as coatings for solar control windows and optical selective solar absorbers.

The surface composition of coevaporated Cu-In-S films for photovoltaic applications has been investigated by x-ray and ultraviolet photoelectron spectroscopy. A comparison of the front and back surface, i.e., the interface between the Cu–In–S film and the Mo back contact, is made for films with different stoichiometries. We find that In-rich films are covered at both surfaces by an In-rich second phase which has the cation ratio of CuIn3S5. The front surface of Cu-rich films shows a segregation of CuS that is not found at the back surface. The formation of binary and ternary second phases in nonstoichiometric CuInS2 films and their implications on the growth mechanism are discussed.

The introduction of copper interconnects into integrated circuits has increased the use of dual damascene dielectric etch applications because copper films are difficult to plasma etch. Fencing and faceting around the via hole during the trench etch of the via-first dual damascene integration scheme are particularly detrimental and can lead to problems during copper metallization and ultimately to device failure. Therefore, it is imperative that the evolution of these features be understood so that they can be avoided. In this article we will begin with an overview of the via-first dual damascene integration scheme. Experimental results will then be presented that indicate the evolution of these features is heavily dependent upon the existing via profile and whether bottom antireflection coating and/or photoresist is in the via hole prior to starting the trench etch. An empirical model for fence formation was then confirmed by a simple profile simulator written in Visual Basic. Finally, several options for avoiding the evolution of fencing and faceting during the trench etch will be proposed.

The scratch test has been used to measure the adhesion of CVD TiC coatings on various steels by comparing the critical load factors. The critical load (Lc) of a coating–substrate interface determined by the scratch test depends not only on the adhesion but also the substrate hardness and coating thickness. For identical deposition conditions, the coating thickness depends strongly on the carbide forming elements in the steel. In this work, a careful separation of these effects has allowed us to quantify the adhesion of a given system from the critical load. Of the nine steels tested, two had improved adhesion which was shown to be caused by diffusion of substrate elements into the coating. Adhesion measurements of Ti(C,N) coatings deposited by PVD are discussed and compared to the adhesion values of CVD TiC on the same steel substrates.

The interfacial fracture energy between a thin polyimide film and a rigid Si substrate was measured via the blister test. A dynamic debonding mechanism was used instead of a static debonding mechanism. A nonlinear analytical method was developed based on local energy balance for the dynamic debonding process of the blister test. The main advantages of this method are that (1) the dissipated energies which accompany the debonding processes can be measured and (2) the mechanical properties of thin films do not enter into the calculation of adhesion strength. A minimum value for adhesion energy can be obtained which compares favorably with estimated values of adhesion.

Using a scanning tunneling microscope (STM), we have measured the microscopic surface roughness of light emitting tunnel junctions (LETJ) which consist of thin films of Al, Al oxide, and Au. The power spectrum of roughness obtained from the STM data was used to calculate the light emission spectrum, using a theory that takes into account the effect of surface roughness in first order perturbation. The calculated spectrum due to the term first order in roughness fits the measured emission spectrum very well, but when the term due to an unperturbed smooth surface is included, the fit becomes unsatisfactory. We conclude that the theory predicts a wrong ratio between the contribution from surface roughness and the contribution from a smooth surface.