Current ramp tests are performed in a number of experiments to check their validity as an in-line monitoring tool for electromigration evaluation. We discuss our various findings from a series of experiments conducted to validate the usefulness of the current ramp test, using the standard NIST test structure. First, we check whether the current ramp test, using breakdown energy of metal (BEM test), shows the effect of a Ti/TiN underlayer on the reliability of A1–0.5wt%Cu unpassivated interconnects. The second experiment compares current ramp testing with fuse testing. Next we check if current ramp testing gives any indication of interconnect quality following different stages of our sputtering process, and in particular different degas temperatures. The last experiment tries to correlate the breakdown energy obtained from current ramp testing with the lifetime obtained from conventional long term electromigration testing of different combinations of line stack and line stack dimensions. From the results of the experiments, we conclude that current ramp testing using the breakdown energy of metals cannot be used quantitatively for electromigration evaluation using conventional NIST test structures.

In this paper we describe three phenomena that have been observed in via terminated test structures. The first is the effect of stress voiding on the electromigration current density exponent, n. We have tested stress-voided lines and the results show that the current density exponent is close to 1. This is consistent with the prediction that stress-voided lines have a lower lifetime. We propose that the current density exponent must be measured and not assumed for such lines. The second phenomenon is the reservoir effect in via-terminated lines. Void locations following electromigration stressing are shown for a via-terminated structure containing a reservoir, prepared by Focused Ion Beam (FIB), and a discussion of the reservoir effect is presented. Finally, we report on the correlation between void location and the bimodal distribution of failure mechanisms that have been observed.

The electrical reliability of Ta/NiMn/NiFe/Co/Cu/Co/NiFe/Ta top spin valve sensors was studied. Electrical current densities in the range of 20–50 MA/cm2 were applied at 70°C ambient temperature. The combined Joule and ambient heating correspond to peak sensor temperatures of 170–300°C. Resistance and giant magneto-resistive (GMR) amplitude of the sensors were measured with time up to 800 hours. Device reliability was predicted from the Arrhenius plots of time to failure (TTF). Microstructure of sensors after stress was studied by scanning transmission electron microscope (STEM) and energy dispersive X-ray spectroscopy (EDX). The device failure mechanism is discussed on the scope of inter-layer diffusion and electromigration.

Electromigration reliability of a GMR spin valve device has been investigated at elevated temperatures at various DC current densities. Significant degradation of the magnetoresistance was observed when the current density was higher than 20 MA/cm2. Either the electromigration induced damage or the rotation of the magnetization vector of the Mnlr anti-ferromagnetic layer are possible mechanisms of degradation. Furthermore, significant voiding caused by electromigration and thermal gradient induced atomic motion was observed at current densities higher than 35 MA/cm2. High-resolution temperature distribution measurements by scanning thermal probe microscopy revealed the existence of a huge temperature gradient of 106°C/cm, which is a strong driving force for atomic diffusion.

The interplay between stress and electromigration has been recognized since I. A. Blech et al. used x-ray topography in 1976 to demonstrate that stress gradients developed during electromigration. Availability of high brightness synchrotron x-ray sources, high stability energy dispersive detectors, high resolution area detectors, and pinholes, capillaries and other optical elements for forming x-ray microbeams, has made possible more quantitative, real time measurements of strains and composition changes which develop in polycrystalline metal conductor lines during electromigration. This paper describes advances made in this area, implications of results which have been obtained, and prospects for further progress.

We have used x-ray microbeam fluorescence and diffraction for in-situ measurements of Cu concentration and strain during 36 hours of electromigration with current density 5×105 A/cm2 at temperature 310°C for a passivated 10μm-wide, 200μm-long Al(0.5 at.% Cu) conductor line. These measurements indicate that Cu is quickly depleted at the cathode end of the line and is piled up at the anode end, where the Cu concentration first increases, then decreases, and finally remains constant. A void begins growing at the cathode end of the line about two hours after the start of current flow, with a nearly constant void growth rate of 1 μm/hr for the remainder of the test. The evolution of stresses within the line during electromigration involves changes by 100's of MPa and is more complex than seen previously for electromigration under similar conditions for pure Al conductor lines.

Synchrotron x-ray sources provide high-brilliance beams that can be focused to submicron sizes with Fresnel zone-plate and x-ray mirror optics. With these intense, tunable or broadbandpass x-ray microbeams, it is now possible to study texture and strain distributions in surfaces, and in buried or encapsulated thin films. The full strain tensor and local texture can be determined by measuring the unit cell parameters of strained material. With monochromatic or tunable radiation, at least three independent reflections are needed to determine the orientation and unit cell parameters of an unknown crystal. With broadbandpass or white radiation, at least four reflections and one measured energy are required to determine the orientation and the unit cell parameters of an unknown crystal. Routine measurement of local texture and strain is made possible by automatic indexing of the Laue reflections combined with precision calibration of the monochromator-focusing mirrors-CCD detector system. Methods used in implementing these techniques on the MHATT-CAT beam line at the Advanced Photon Source will be discussed.

The local strain and texture in Al interconnect wires have been investigated using white and monochromatic x-ray microbeams on the MHATTCAT undulator beam line at the Advanced Photon Source. Intergrain and intragrain orientations were obtained with ∼0.01° sensitivity using white beam measurements on wide Al pads (∼100 μm) and thin (2 μm) Al wires. Orientation changes of up to 1°were found within individual grains of the (111) textured Al interconnects. Deviatoric strain measurements indicate small intragranular strain variations, but intergranular strain variations were found to be quite large.

A technique for observing microstructure and morphology changes during annealing or electromigration of 100 nm-scale and smaller interconnects is presented. The technique is based on dynamic in situ characterization using a UHV field-emission scanning transmission electron microscope (STEM) and can enable full microstructural characterization (images, diffraction patterns and compositions) during electromigration testing. Initial steps in the validation of the approach include in situ electron-beam OMCVD of Al-containing wires and observation of these wires. Although the deposition conditions were not optimized, Al-containing wires, with high edge acuity, thickness uniformity, and with minimum interconnect widths as small as 15 nm have successfully been deposited. In situ imaging of the annealing of 100 nm Al films on 25 nm Si3N4 substrates at 350°C demonstrates that an image resolution of 10 nm should readily be attainable in passivated nanoscale interconnects during electromigration.

The mechanical stress state of conventional Al and damascene Cu lines of a 0.18 pm logic technology flow have been determined using a novel X-Ray diffraction method that permits measurement of stress on an array of critical-dimension lines on the product die. The effect of high density plasma oxide deposition and the influence of low-K dielectrics on the stress state of the Al lines is described. The effect of materials properties and fabrication methodology on the stress state of damascene Cu lines is shown with measurement of mechanical stress and strain in passivated lines at room temperature and during annealing. The effect of underlayer on the damascene Cu stress state is also quantified.

X-Ray diffraction (XRD) provides an excellent tool for the measurement of both stress and texture (preferred orientation) on fabricated damascene interconnect structures. Since x-ray diffraction provides a direct measurement of lattice spacings, film strain can be measured directly. Also, since the intensity of diffracted x-rays is proportional to the density of lattice planes oriented in diffracting condition with respect to the incident beam, both the direction and extent of preferred orientation can be accurately measured. Special techniques and considerations are necessary when examining damascene interconnect structures with XRD which are not necessary with blanket films. These techniques are discussed and described in order to aid in obtaining meaningful XRD data and a correct interpretation of the results.

The thermomechanical behavior of a Cu thin film can be dramatically influenced (“oxygen effect”) by exposing the film to a small amount of air or oxygen prior to passivation. Film exposure is accomplished by breaking vacuum at some point prior to passivation. When the top surface is exposed immediately before passivation, the oxygen effect is produced when the passivation material is Si3N4, but no effect is seen with AIN passivations. However, the effect is seen in AIN passivated films when the Cu film surface is exposed to air or O2 and additional Cu is deposited before passivation. This suggests that conditions at the film/passivation interface are an important factor in this anomalous plasticity effect. We have investigated the effects of film thickness and oxygen concentration on these behaviors by preparing films with different thicknesses and oxygen contents for thermomechanical analysis. The thermomechanical behavior is correlated with secondary ion mass spectrometry (SIMS) data that shows oxygen concentration vs. film depth. The presence of the “oxygen effect” in thermomechanical behavior is found to be correlated with an increased oxygen concentration near the film/passivation and film/barrier interfaces.

The resistance of Al thin film metallization interconnects has been measured during temperature cycling where thermal expansion mismatch between the Al and the substrate and passivation is used to induce interconnect strain. Both passivated and unpassivated interconnects were thermally cycled between 80K and 373K. The coefficient of piezoresistivity, defined as dp/dE, where ρ= resistivity and εv = volumetric strain, is measured to be 2.0 10−5Ω-cm in tension, consistent with previous results determined in bulk Al under hydrostatic pressure. We describe a model which incorporates the thermal, geometrical, and piezoresistance effects on the measured interconnect resistance during temperature changes. Thermal expansion mismatch strains and the degree of interconnect constraint by the rigid surrounding materials, which vary with interconnect aspect ratio, are considered. The application of piezoresistance analysis to measurement of local stresses and stress relaxation in advanced interconnects will be described.

Stress and electromigration voids are believed to nucleate at the metal/passivation interface. In this study, modified interfaces were synthesized to experimentally elucidate the role of the interface on stress voiding for comparison with theoretical predictions. Interfaces were modified by depositing photoresist in a controlled manner on Ti/Algfi lines prior to passivation deposition. Stress voiding susceptibility was assessed in as-received lines as well as those subjected to storage anneals for 1, 5, and 25 days at 190°C to accelerate void formation. The global textures of all the lines were <111> fiber and were not significantly affected by the interface modification. Differences in stress voiding were attributed to the different interfaces. Higher ratios of voids in modified regions to voids in unmodified regions were observed both in as-received lines and those given a 1 day storage anneal due to the heterogeneous nucleation sites provided by the photoresist contamination in the modified regions. However, the void ratio appears to decrease exponentially with longer time anneals. Lowered driving force for void formation in the modified regions leads to less voiding with time, while higher driving force and a limited number of nucleation sites in the unmodified regions leads to more voiding with time. Local textures at a limited number of voided sites in modified and unmodified regions were identical to the global texture of the line. In our limited sampling, voids did not form in local regions of weaker <111> texture.

A controlled experiment was performed to determine whether residual stresses in silicon oxide films affect the apparent modulus and hardness of the films measured using microindentation. A set of wafers, each with a SiO2 film of different residual stress, were produced by PECVD deposition using different RF power settings. Residual stresses, determined by curvature measurement, were shown to become more compressive with the increase of the RF power. Correlation between indentation modulus/hardness and residual stress was established because the modulus was also observed to be trending with the RF power. To determine whether the residual stress change is the cause for the change of indentation modulus/hardness, one of the wafers was cut to pieces and bonded to steel substrates of different thickness at elevated temperatures. Because of the thermal mismatch between silicon and steel, the silicon pieces were stressed by the steel substrates and therefore provided additional stresses to the SiO2 film on the top surface. The applied stresses were determined using Raman piezospectroscopy. Micro-indentation tests performed on these samples showed no appreciable change of modulus/hardness due to additional stresses. It was concluded that the material change, such as density, was behind the correlation between indentation modulus and residual stress.

An improved microbeam bending technique has been developed for the study of mechanical properties of thin films on substrates. This testing method utilizes a triangular beam geometry and improved micromachining techniques compared to previously used methods. The technique permits the stress-strain law for a metal film on a substrate to be determined. Single crystal Si beams and bi-layer Si/Al beams of lengths 25–100 pgm have been fabricated and tested. The beams are deflected with a nanoindenter, which accurately imposes a load on the beam and measures the corresponding displacement. For the bi-layer beams, a simple numerical model utilizing a Ramburg-Osgood constitutive law the film has been developed to determine the stress-strain behavior of the Al film.

The electrostatic test structure was presented to measure the micromechanical properties of micromaterials as thin films forming the microactuators. The test structure was fabricated by using the surface micromachining processes and driven by the electrostatic force. In order to measure the fracture toughness, the sharp notch in the test structure was introduced by using the etching process. The displacement and deflection of the test structure under the electrostatic force was on the basis of the beam bending theory. An analytical model was proposed to evaluate the fracture toughness by experimental determination of crack propagation from the variations of capacitance and voltage. Additionally, an acceleration test for evaluation of the long-term reliability of the microactuator device was performed to predict the lifetime of a gold micromirror plane used in DVD (Digital Video Disk) by analyzing the resonant frequency and capacitance.

A methodology to study the stress distribution of a patterned thin film residing on a silicon wafer was developed. Si underlying the pattern was thinned down through etching so that the deformation caused by residual stress in the microstructure could be detected by a Twyman- Green interferometer. A procedure called "numerical etching" was implemented to simulate the etching process, which linked the stress state of the microstructure on a regular wafer to that on a Si diaphragm. An initial stress field on the pattern was assumed, and its effect on the deformation of the Si diaphragm beneath was calculated and compared with experimental results. The discrepancy between them was used to modify the initially assumed stress field and repeated until a satisfactory match was achieved. The stress field from numerical analysis accurately predicts the actual stress distribution in and around the patterned structure under investigation. The stress distribution in a Ti pad on a Si3N4/ SiO2/Si composite diaphragm is used as an example.

Thin film structures may fail by progressive or time-dependent debonding at stresses far below those required for catastrophic failure. Previous work has shown that progressive debonding in a typical interconnect structure occurs either along the TiN/SiO2 interface or parallel to this interface in the SiO2 Such subcritical debonding was found to span several orders of magnitude of debond growth rates and occur at significantly reduced driving forces. The presence of SiO2 at the failure location indicates that the mechanisms which give rise to stress corrosion cracking in bulk glasses may also play a role in the subcritical debonding behavior of multilayer interconnect structures. Accordingly, this work focuses on the effects of temperature and humidity on subcritical debonding and rationalizes them in terms of the relevant chemical reactions taking place at the debond tip.

The SiO2/TiN interface is prone to sub-critical decohesion by crack extension along the interface when exposed to moisture. As with conventional sub-critical crack growth in bulk silica, the rate of decohesion is dependent on both the relative humidity and the strain energy release rate. One of the unusual observations we have noted is that the decohesion velocity of narrow strips is highly sensitive to the exposure time in the humid environment, suggesting that the moisture diffuses into the interface ahead of the propagating decohesion crack, causing a form of hydrolytic weakening of the interface. In seeking to establish the origin of this behavior, we report measurements of moisture diffusion along the SiO2/TiN interface using Imaging Secondary Ion Mass Spectroscopy (SIMS) after immersion in water containing radioactive tracers, 2D and 18O. Analysis of images formed using 2D and 18O revealed that both diffuse along the interface with a diffusivity of (6 ±2)×10−13 cm2/sec, a value about 104 times faster than the bulk diffusion in the SiO2 dielectric.