ICMCTF2014 Session TS2-2: Advanced Characterization of Coatings and Thin Films

Metal-polymer thin assemblies ( h <100 nm ) are now increasingly used in many devices such as micro - components in micro- and nano- electromagnetic (MEMS and NEMS) and on flexible printed circuit. To ensure the high reliability of such devices, it is essential to control the mechanical holding and heat dissipation of assemblies. We know that these properties are strongly dependent of the characteristics of the metal/polymer interface which depends itself on several parameters such as the nature of the metal and polymer, the roughness of the polymer and experimental conditions of gold deposit (temperature, speed…).

To study the structural properties and electrical resistivity performance of nanothick metal/polymer assemblies, we used a combination of X-ray reflectivity (XRR) and AFM analysis. X-ray (XRR) analysis was used to elucidate details of the structural modifications induced by the deposition temperature increase. Clearly, the structure of nanometric metallic layers deposited on polymer films is complex and the representation of polymers with coatings as two layered systems with a well pronounced interface depends of the deposition temperature. We have shown the progressive embedment of the gold particles into the polymer film when the deposition temperature increases above the T_g of the polymer film. The effect of the structural evolution of the assembly is then emphasized by measuring the increase of electrical resistivity of the assembly with a conductive-probe atomic force microscopy.

For the first time, we used a new system to make electrical resistance measurements, using AFM ((Resiscope®).

After applying a voltage between the tip and the surface we recorded AFM images of the surface resistance which allowed us to confirm the diffusion of the gold in the polymer when the temperature is higher than the T_g of the polymer.

AFM images can provide information on the internal structure of a thin film but it can also be used to study the characteristic dimensions of a mounded surface when they are combined with a new statistical analysis method. The interfacial differential function (IDF) is an efficient method to extract characteristic dimensions such as grain size, inter-grain lengths and periodicity.

We have demonstrated that for model mounded surfaces, the method is the most effective one to determine the different characteristic lengths. This method was thus applied to study the evolution of the surface morphology of ultrathin gold film deposited on silica. We have demonstrated that the roughness increase with deposition temperature is mainly due to a grain height increase and not to a grain coarsening phenomena as it claimed before.

Since the early 1990s, coatings have been routinely characterized using nanoindentation, scratch testing and tribology testing in order to optimize them for their applications.

Most of the current international standards were written during the same period and unfortunately most users of mechanical testers only compare values between hardness, elastic modulus, critical loads, friction and wear but these may be insufficient for a good understanding of a coating design.

Optimization of a coating remains a challenge due to the many influencing variables such as coating adhesion, yield and wear mechanism, thermal and mechanical stresses, etc.

By the use of measured data from a nanoindentation test, a physical calculation of spatial stress profiles is simulated considering realistic material properties.

This analysis allows the user to perform real dimensioning of sophisticated scratch and tribology tests so that they are focused at specific regions in the coating design architecture.

This procedure allows more specific investigations of critical interfaces, transition layers and substrate regions with adjusted depth resolution.

Due to roughness and surface structure such optimized scratch, groove or tribology tests sometimes even need to take into account the true topography of the sample surface before and after the test. Such 3D scratch or tribology tests can provide much more information about the material stability and reliability than do classical approaches.

Using this valuable approach to quantify the interfaces between layer and substrate in a coating design allows a faster optimization of the coating design thus providing significant cost reduction.

To illustrate the features of this optimization method, various examples of characterization on coated samples will be presented, showing the flexibility of the technique for different situations.

Residual stress strongly affects the performance of thin films, in terms of adhesion, hardness, wear and fatigue resistance. Thus, when assessing innovative coatings or new deposition technologies, it is compulsory to evaluate the distribution of the stress by means of a reliable technique.

X-ray diffraction (XRD) is one of the commonly used techniques, because it is non destructive, surface sensitive and phase selective. Unfortunately, XRD can determine the stress only in case of crystalline materials and it may be not reliable in the presence of texture or stress gradients, often occurring in thin films.

Recently, a new class of methods for residual stress evaluation has been proposed, based on incremental focused ion beam (FIB) milling, combined with high-resolution in situ Scanning Electron Microscopy (SEM) imaging and full field strain analysis by digital image correlation (DIC).

The aim of the paper is to understand the different meanings of the stress values obtained either by XRD or FIB and to discuss the weaknesses and strengths of the two techniques. On this purpose, a Chromium Nitride (CrN) highly textured thin film samples have been deposited by cathodic arc evaporation PVD and the residual stress has been evaluated by using the two methods.

Considering the differences of XRD and FIB methods, a good agreement between the obtained stress values was obtained, provided that the issues related to (a) probe-to-sample interaction volume, (b) film’s texture and (c) elastic anisotropy are carefully taken into account.

In order to understand the principle differences between rheological or simple stress tests like the uniaxial tensile test to contact mechanical tests and quasistatic contact experiments with oscillatory ones this study resorts to first principles respectively effective first principles. It will be shown how relatively simple models simulating bond interactions in solids using effective potentials like Lennard-Jones and Morse can be used to investigate the effect of time dependent stress-induced stiffening or enhancement in these solids.

The usefulness of the current study is the possibility of deriving relatively simple dependencies of the bulk-modulus B on time, shear and pressure P. In cases where it is possible to describe, or at least partially describe a material by Lennard-Jones potential approaches the above mentioned dependencies are even completely free of microscopic material parameters. Instead of bond energies and length, only specific integral parameters like Young's modulus and Poisson's ratio are required. However, in the case of time dependent (viscose) material behavior the parameters are no constants anymore. They themselves depend on time and the actual stress field, especially the shear field.

The influence of the time dependent pressure-induced Young's modulus change is discussed especially with respect to mechanical contact experiments and their analysis in the case of viscose thin films and substrates.

One of the major effects, when it comes to indentation testing, is surface roughness. It can lead to a split-up of the assumed single contact area into a number of contact spots caused by many asperities underneath the indenter. However, even in cases where such a split-up does not matter roughness can still severely influence the indentation test results. The smaller the load the more important it is to take the actual surface curvature around the indented area into account. The supposedly normal load subjected to a supposedly flat surface could in fact become a complex mixture of the external normal load plus lateral, tilting and twisting moments caused by the surface curvature and inclination. Also the influence of the curvature on the resulting contact area can have a tremendous influence on the results. Especially hardness and elastic modulus can be dramatically overestimated if being measured on rough surfaces.

This work will cover how to account for such surface when analyzing contact problems like nanoindentation measurements by the means of the effective indenter concept being generalized to non-axial-symmetric loading and a curved surface solution applying paraboloid coordinates. It will also be shown how to evaluate the complete elastic fields and how to extend the approach to layered materials. In addition, it will be elaborated why the usual arithmetic averaging of many indentation measurements done on a rough surface does not give the correct hardness or Young’s modulus, but still rather overestimated values leading to false ultra-hardness results.

Blind hole (BH) filling by metallic deposition is widely usedin the high density interconnection (HDI) technology of three-dimensional (3D) packaging due to its high thermal and electrical conductivity and increased the efficiency of space utilization. The circuits in build-up layers of a HDI component are generally interconnected by electrolytic Cu fillings/platings. The morphology and crystallographic orientation/texture of the Cu fillings have been reported to pose significant influences on the electrical, thermal, and mechanical properties of interconnects, affecting the overall reliability of the HDI packages. Therefore, it is of great important to investigate the microstructural and crystallographic characteristics of the electrolytic Cu deposition in the BH structure.

In this study, the BHs were made by CO₂ laser drilling with a configuration of 60 μm (diameter) × 40 μm (depth) in a bismaleimide triazine (BT)-based substrate. The BH walls were deposited with a 1-μm-thick Cu film by electroless plating prior to the Cu electrodeposition. Subsequently, the substrate was subjected to a Cu electroplating process, where different current densities (j = 1, 2, 4, 6 A /dm²) in direct current (DC) mode at room temperature were utilized. After electroplating to the “super-filling” stage, the BHs were cross sectioned and then were subjected to a metallurgical grinding-polishing process, to reveal the interior microstructure of the Cu platings in the BH structure. Additionally, a vibratory polisher (vibroMet^TM 2) was employed to eliminate the artifacts resulting from the metallurgical treatment, to gain exact crystallographic information, e.g., grain size and boundary.

The microstructural and crystallographic characteristics of the electrolytic Cu BH filling in different current densities were examined by using optical microscopy ( OM ) and field-emission scanning electron microscope (FE-SEM) equipped with an electron backscatter diffraction (EBSD) analysis system. Correlations of grain size, grain boundaries (e.g., twin structure), crystallographic orientation, texture with j will be established in the present study. These findings offer better understanding of the microstructural/crystallographic characteristics in the electrolytic Cu BH filling process.

Copper-tin (Cu-Sn) is one of the most fundamental metallurgical systems and is now widely utilized in the microelectronic packages to join two metal parts of electronic devices. During soldering, the Cu would quickly diffuse/dissolve into the molten solder first and then a pronounced Cu/Sn chemical reaction at the interface follows. These two tightly connected processes not only cause a significant depletion in the Cu pads but produce a certain amount of brittle Cu-Sn intermetallic compound(s) (IMC) in the joining part, affecting the overall reliability of solder joints. Thus, understanding of the Cu diffusion/dissolution behavior in the liquid solder alloy is very important for advancing the quality of microelectronic packaging.

In this study, a mathematical model combined with experimental data was proposed to estimate the diffusion coefficient of Cu (D_Cu) in eutectic Sn-Ag (Sn-3.5Ag) alloy. The diffusion was conducted at temperatures ranging from 235 °C to 280 °C by using Cu/Sn3.5Ag solder bump joints, where the Cu is a solid but the Sn-3.5Ag is a liquid at the temperatures examined. To capture the overall scenario of the Cu diffusion in liquid Sn-Ag, the Cu concentration distribution across the entire Cu/Sn-3.5Ag joint was carefully analyzed using electron probe microanalysis (EPMA) line scans and elemental mappings. The mathematical model predicted that the average Cu concentration (C_avg) as a function of reaction time would approach saturation after 70 s at temperatures above 235 °C, which agrees well with that obtained through experiments. The saturation caused the slow down of the dissolution rate of existence Cu-Sn compound and consequently induced an evidently growth of Cu-Sn layer at the Cu/Sn-Ag interface. The combined results of calculation and EPMA data further showed that the average D_Cu are 1.2(±0.4) × 10^-5 cm²/s ( 235 °C), 2.4(±0.6) × 10^-5 cm²/s ( 250 °C), 2.6(±0.7) × 10^-5 cm²/s ( 265 °C), and 3.9(±1.6) × 10^-5 cm²/s ( 280 °C), respectively, which are of the same order with that obtained in pure Sn system. The insensitivity of D_Cu to the temperature reflects that the activation energy (Q) of the Cu diffusion in a liquid Sn-based alloy is not very large by contrasting with that of the solid diffusion. The Q was estimated to be approximately 13.3 kJ/mol at 235 °C– 280 °C on the basis of Arrhenius relation, which is slightly less than that acquired from the capillary-reservoir technique (17.6 kJ/mol) reported by Ma and Swalin in the literature.