An investigation is made into the capability of one-particle models in modeling electromigration-induced resistance changes. Analytically obtained solutions to the model equations are used to gain more insight into the vacancy distribution as a result of the applied boundary conditions. These solutions are then used to determine the resistance change modeling capabilities of three types of one-particle models.

The effect of TiN Anti-Reflective Coating (ARC) layer capping of metal, in a multilevel interconnect system, on the reliability of the tungsten (W) plug has been investigated. EM study on Kelvin via showed that the thin TiN ARC Kelvin via lifetime exhibited a reduction of one order of magnitude compared to that of the thicker ARC thickness (1000Å) when they were stressed in a certain direction. A new testing technique, used to investigate the effect of ARC thickness on the W-plug electromigration (EM) performance, is discussed. The simulation results of current distribution confirmed the proposed failure model of EM performance of W-plug via with different ARC thickness.

Submicron high density devices require multilevel interconnections. In multilevel interconnections, issues relating to planarity, stress voiding of metals and stress cracking of the dielectric(s) play an dominant role.As the levels of interconnect go up so does the stress level. A high stress level in the intermediate dielectric can cause cracks which ultimately lead to device failure. In this paper we present a novel method to detect stress cracking using focussed ion beam. This method is correlated with wet chemical etch process and RIE.The phenomenum of cracking of dielectric(s) is studied and different technology aspects are considered. Using the focussed ion beam a robust dielectric is developed.

The electromigration lifetime of Al-Si (1 at.% Si) can be greatly increased by alloying with V and Pd (0.1 at.% V, 0.1 at.% Pd). This study reports on l/fα (α≈l) noise measurements in Al-Si, Al-Si-V and Al-Si-V-Pd alloy films. Samples were prepared by direct current (DC) sputtering, e-beam lithography and reactive-ion etching. The samples were annealed at 450 °C for 30 min and were not passivated. Test line dimensions were 800μm×l.2μm×0.5μm. 1/f noise was measured by a high-resolution alternating current (AC) bridge technique, which ensured that no electromigration occurred during the measurements. The sample-to-sample variation in normalized 1/f noise intensity for Al-Si, Al-Si-V and, Al-Si-V-Pd, measured at room temperature, was less than 10%. The smallest noise intensities were observed in the Al-Si-V-Pd films. The temperature dependence of the 1/f noise of Al-Si films shows a maximum at 338 K allowing the determination of the activation energy of the noise generating process (Ea = 0.75 eV). Above 400 K the resistance of the samples became unstable preventing reliable noise measurements. The instabilities were probably caused by the dissolution of very small Si precipitates (<10 nm). This is important information because the dissolution will also take place during electromigration lifetime experiments.

Statistics of fracture fatigue are derived based on essentially general models of crack propagation and initiation. In fact, the form of the distribution is invariant with respect to the crack propagation law, however, the parameters do depend on the details of the phenomena. When this dependence is known, i.e. extracted from experimental data, our results can be applied even in the case of changing environments. Thus, such problems as projections to use conditions and the efficiency of burn in procedures can be precisely formulated and resolved, often without the requirements of further testing. These advantages derive from our use of physical models in deriving the failure distribution. This was not previously the case where for the almost universally used log-normal distribution each new situation requires the establishment of new rules. In fact, the log-normal is usually pessimistic and requires ad-hoc rectification to account for curvature when plotted on log-normal paper!, and generally does not relate to the fatigue process.

The bulge test provides a convenient way to measure the mechanical properties, including fracture strength, of thin ceramic membranes. Thin (≈ 3000Å) films of non-stoichiometric SiNx, nominally defect-free and containing flaws introduced byfocused ion-beam machining, were examined. The use of such membranes in masks for projection electron-beam lithography systems makes it important to determine their ultimate strength and reliability. The mean fracture strength of the defect-free membranes was found to be 1.79±0.04GPa, and the Weibull modulus was estimated to be 49±15, giving a safe operating stress of 1.5GPa. The results of fracturing the samples containing flaws indicated that holes smaller than 15x0.5μm had no effect on the fracture strength of the membranes. By estimating the stress concentration factor for the slots, it was determined that the critical fracture stress was 16±3GPa.

Quantitative tests for interfacial adhesion between films and substrates are of criticalimportance in micro-electronic applications. However, many of the available tests are applicable to a limited array of materials systems, or are experimentally complex. In thispaper, a constant depth scratch test, which has been designed to circumvent the limitations associated with currently available tests, is presented along with theoretical developments for the quantification of interfacial shear strength. Because of a fixed scratchgeometry, the test is amenable to straightforward analytical formulations unlike other versions of scratch tests. It is unique in its experimental simplicity, and allows evaluation of interface strength as function of position on the sample. Sample outputs from the test based on Cr films on glass substrates are presented.

Three types of cracks associated with interfacial failure of a tungsten fine line on silicon oxide (SiO2) have been observed in microwedge scratch testing (MWST) of fine lines. They were identified by carefully controlling scratch conditions in a continuous microindenter system1, in conjunction with SEM micrographs. The first type is a hinge crack, which is responsible for nucleating an interfacial crack. The driving force for the hinge crack is the difference in stress conditions in the middle on theline (plane strain) and the outside of the line (plane stress). The second type is the interfacial crack, which is driven by the movement of the microwedge. This crack is modelled by the finite element method (FEM) to obtain an interfacial work of adhesion value. The third type of crack is a spallation crack, and is driven by tensile strains associated with buckling of the thin film once the interfacial crack becomes sufficiently long. Tensile strains in the film at the maximum crack length are calculated by FEM in order to arrive at a value of thin film fracture strength. A brittle failure mechanism has beenconfirmed by a micrograph of a spalled piece exhibiting no perceptible plastic deformation.

During the plasma etching of Al-Si-Cu alloy used as a metal interconnection, it is generally reported that the metal pillar (or conical residue) affecting the degradation of device yield is formed by the micromasking effect of copper compound. However, it is stilldisputed with the formation mechanism and composition of the micromasking material. Moreover, the elimination method of the metal pillar is not well known.

According to previous reports, it is argued that the micromasking material consists of Cu agglomerates, A12Cu, or CuC1, and the formation mechanism of the micromasking is due to byproduct during plasma etching or reaction product during metal depositionor etching. However, using scanning electron microscopy (SEM), energy dispersive of x-ray (EDX), and high resolution Auger spectroscopy (HRAES), it is newly found that the micromasking consists of three layered structure, that is copper aluminum oxide, A12Cu, and Cu agglomerates. These results are quite different from previous reports. In addition, the removal methods of the metal pillar are suggested, which are high power dry etch process and multilayered metal deposition.

Interference holographic methods are applied to thin film structures to determine their state of stress and their thermal and mass diffusion coefficients. Special sample preparation techniques are used to produce tensioned membrane samples that preserve the state of stress in coated parts. Then utilizing real time vibrational holographic interferometry methods, the total state of stress can be determined, two normal stresses and a shearstress. Similar samples exposed to moisture have their stress state altered by swelling and this can be monitored on a real-time basis and the influence on stress state and thethrough-the-thickness moisture diffusion coefficient can be determined. In related experiments using similar methods, the through-the-thickness thermal diffusion coefficient can be determined using pulse heating methods. These methods coupled with routine test methods enables one to fully characterize all of the orthotropic elasticity coefficients and transport characteristics needed for numerial analyses of thin film structures.

The edge delamination test (EDT) is being developed to measure the critical energy required to cause a thin film, under biaxial tensile stress, to debond from a rigid substrate[1]. The test uses circular features etched through biaxially stressed films adhered to a rigid substrate. If the stress is large enough, a stable debond ring grows radially about the feature. We use a finite element analysis to model the test, solving for the applied strain energy release rate as a function of crack length, feature hole radius and other geometrical parameters. The model identifies both mode I and mode II components of the strain energy release rate, and agrees with previous analytical solutions for the total debond energy. However, the model predicts, with a very refined mesh at the crack tip, the fracture process is pure mode I. To explore this result, critical strain energy release rates from the EDT and the island blister test (IBT) are compared. This agreement supports the model prediction that the failure process in the EDT is modeI peeling.

We describe an experimental method useful for in-situ real-time evaluation of viscoelastic, thermal and adhesive properties of thin films and multi-layer structures. We demonstrate how the technique is used to quantify the elastic moduli, and in-plane thermal diffusivity. We also show how it can be used to “spot check” for dis-bonds and to generate dis-bond “maps”.

Neutron reflectivity (NR) was applied to measure the concentration of water at the buried interfaces between polymer and silicon single crystal wafers. In the polyimide samplesexcess water was discovered within 30 Å of the silicon/polymer interface,where the water concentration reached 17% (by volume) for the samples without a couplingagent and 12% for the ones with coupling agent. Beyond the interface, the water concentration was measured at 2-3%, which is typical of bulk polyimide. Excess water was alsofound at the silicon/epoxy interface; more than 30% water was found within 150Å of the interface in the samples with their interface precoated with a coupling agent.The equilibrium moisture content in bulk epoxy used in this study is about 6%. The aboveresults demonstrate conclusively the unique power of NR in determining water concentration near a buried interface, and provide the first quantitative evidence for a water concentration profile which peaks in the interface region.

The micro-wedge scratch technique was applied to characterize interfacial properties of copper fine lines on PMDA-ODA polyimide. A 20 pm wide, 0.1 μm radius wedge-shapeddiamond tip was used to mechanically debond Cu fine lines from polyimide substrate. Copper fine lines varies from 0.2 to 0.4 μm thick, and from 1.μm to 100 μm wide. Both cohesive and adhesive delaminations were obtained. Loaddisplacement curves were recorded and the scratch morphologies were observed by SEM. From a first order calculation, the fracture toughness of the polyimide and interfacial strength of Cu-PI were obtained.

In MCM-D applications, interlayer dielectrics separate and insulate metal conductors to form a three-dimensional interconnection structure. Due to the three-dimensional nature of these structures, the thermal, electrical and mechanical properties of the dielectricmaterials must be known for all orientations in order to correctly design and simulate devices. The most commonly used polymer in microelectronics, polyimide, exists in formulations which have been shown to have a high degree of orientation and exhibit anisotropicproperties.

This work reviews the dependence of reliability testing results on material processing and properties for polymeric materials used in high density interconnection technologies.Specific data are presented for CyclotenesTM 3022 resins containing divinylsiloxane bisbenzocyclobutene, (CAS 117732-87-3). Silane coupling agents discussed in this study are 3-aminopropyltriethoxysilane (CAS 01760-24-3)(APTES) and 3-methacryloxypropyl trimethoxysilane (CAS 02530-85-0) (MOP-TMS).

Moisture-assisted crack growth in polymer/glass interfaces was measured as a function of the applied energy release using a four-point flexure test coupled with an inverted microscope. The specimens consisted of two glass plates bonded together with a thin layer of epoxy adhesive. Roughening the surface of one of the interfaces increased its fracture resistance sufficiently so that crack growth on this interface was inhibited. Thus, crack growth would occur only on the “smooth” interface (one with the least fracture resistance) or in the case where both interfaces were smooth, dual cracking occurred. Finite element analysis was used to explain the observed crack growth results.

Using Tapping Mode Atomic Force Microscopy (TMAFM), the surface structure was determinedfor polymer sensors which incorporated polyaniline (PAn) films that were deposited electrochemically across narrow insulating gaps between interdigitated gold electrodes. The sensitivity and response time for such sensors, which can be used for the detection of low levels of gases and low concentrations of impurities in liquid media critically dependon the quality and structure of the polymer film in the gap region between the gold electrodes. TMAFM images of the PAn films ranging in thickness between ≈1.5μtmand ≈5μm reveal that films approaching 5μm (typically used in such sensors) develop deep cracks at the edges and along the length of the Au electrodes. Thecracks, which appear to be a consequence of stress build-up in thick films, can lead to reliability problems and inferior sensor performance. Simple modeling analysis of cracked films indicates that they can lower sensitivity and increase response time. TMAFM images of sensors with thinner ≈2.5μm PAn films show no cracks as well as continuous PAn bridges across the gap between Au electrodes. Analyses of TMAFM images are presented and compared for thin and thick film PAn sensors.

We have developed a technique to fabricate embedded variable conductivity paths in phenolformaldehyde, PF-resin, with resistivity as high as 109 Ω-cm. The resin used for this work consists of a monomer called Phenyl Methanol (Resol) which upon setting at 150ºC will form short chains of polymers with high resistivity. The conductive paths are produced at depths of 4 to 35 micrometers by changing the energy of thebombarding ions. The resistivity of each path is controlled between 109 Ω-cm to 10 Ω-cm by adjusting the ion dose. The ions used in this work are nitrogen, alphas and protons with energies between 0.5 and 6.5 MeV to form conductive paths as narrow as 0.5 mm and as wide as 4 mm.

To be able to implement light-sensitive properties of ionimplanted PMMA in optical devices, it is first necessary to investigate the optical stability of the implanted layers. For this purpose, PMMA samples were implanted with nitrogen using doses from 7×1013/cm 2 to 5×1015/cm2 as well as energies from 100 keV to 380 keV. Subsequently these samples were exposed to dry media such as H2, 02, CO2, N2, Ar as well as to dry and humid atmosphere. In a further investigation, the optical stability was examined under the influence of different wavelength ranges. The changes resulting in these layers were ascertained over various periods (maximum of 5 years) by ellipsometric measurement of the refractive indices and, in parallel, the chemical modification of the surface was examined by means of XPS. If any changes were caused by the media, they consisted basically in a decrease of the index values and of the original concentrations on CH— and COH—groups generated by the bombardment— i.e. in a recombination of molecular chains. In correlation to the chemical modification of the surface, the refractive index partly changed drastically.