Search

This paper reports experimental observations showing that a current flow produces an effect of strengthening a solder joint against a shear load. This conclusion is found from a single-joint shear test conducted on Sn–1Ag–0.5Cu wt% (SAC 105) joints with in situ current stressing varied from 700 to 1400 A/cm2 at room temperature. To isolate the current effect from the Joule heat effect, the same type of tests were conducted at room temperature, 40, 50, and 80 °C, without current applied to the joint, thus mimicking the condition of steady-state temperature resulting from Joule heating. Comparative testing was also conducted after aging the samples at 150 °C for 200 h. These tests produced indications suggesting that the current flow causes the maximum shear load to increase, while the rise in temperature by the Joule heat effect results in the opposite effect. Experimentally, as much as a 9.5% increase in the maximum shear load was observed from the isothermally aged sample tested under a current density of 700 A/cm2, while an ~ 25% reduction was estimated to result from a temperature increase by Joule heating. The potential mechanism for these observations is discussed.

Wirebonding with bare Cu and palladium-coated copper (PCC) wires have reemerged as mainstream packaging technologies. This paper investigates the failure mechanisms in the PCC wirebond packages under electromigration (EM) tests. First, the directionality of failures at high current density and temperature conditions is discussed. Results illustrated that the intermetallic compound (IMC) growth rate is different when the EM flux direction is in the opposite direction of the chemical flux. Energy dispersive X-Ray spectroscopy (EDX) results paved the way for a fundamental understanding of wirebonding failure. Intermetallic compounds of CuAl 2 and Cu 9 Al 4 are observed after interfacial focused ion beam (FIB) cuts are made to obtain clear images of the failure site through scanning electron microscopy (SEM) and EDX. The Pd coating relocates from the Cu wire surface towards the ball area during the initial bonding process. Therefore, a small concentration of Pd alloys with the bulk Cu within the wire, which is small enough to go unnoticed during the manufacturing process. However, the Pd concentration makes a large enough difference to reduce the contact area between the ball bond and the Al pad. This is due to the solid-solution hardening that causes the ball to undergo a greater amount of work hardening during the bonding process. The smaller contact area yields a greater current density and resistance, enhancing the EM effects in PCC wirebonds. Throughout the EM test, Pd atoms migrate toward the Al pad, creating a higher concentration(∼6at%) of Pd near the interface between the ball bond and the Al pad. Interestingly, Pd has a minor effect on the failure mechanism and crack propagation owing to its concentration and resultant increased hardness at the interface. These results can lead to future studies that may lead to improvements in PCC wirebonding reliability.

This paper concerns the electromigration (EM) failure mechanisms in solder interconnects under low frequency pulsed direct current (DC) conditions. In our study, the accelerated EM tests of Wafer-level Chip Scale Package (WCSP or WLCSP) samples are conducted under 4 different pulsed DC conditions: 0.1 Hz pulsed DC with duty factors (DFs) of 33%, 50%, 75%, and 100% (DC). The result of our testing suggests that there are at least two competing factors affecting the failure rate in an opposite manner under pulsed DC EM conditions. Specifically, when compared with the cumulative damage model (estimates the damage only during the on period), the failure kinetics is found to be more accelerated at high DF and decelerated at a lower DF. This conclusion is drawn from the observation that the EM failure rate shows an extremely nonlinear relationship with the DF and also that the failure occurs faster under a high DF than under solely DC conditions, which is not possible without a mechanism assisting the EM failure. Furthermore, it is found that there is a 2- stage resistance change before EM failure, an indication of the change in the dominant failure mechanism. The results may indicate that the pulsed DC effect on the EM failure mechanism is far more complex than anticipated with the possible involvement of a damage mechanism other than EM such as thermal fatigue.

This paper investigates the effects of the under bump metallization (UBM) on Electromigration (EM) reliability of SAC solder joints and presents evidence suggestive of hidden mechanisms controlling the EM failure. This conclusion is based on the observation that the EM resistance does not show a monotonic increase with UBM thickness, but rather decreases above a certain critical limit. Since a thicker UBM would provide a greater supply of Cu, our observation contradicts the conventional view on the role of UBM on prolonging EM failure. We initially ascribed it to the increased of EM-prone microstructures or weak- links at the SAC/UBM interface active in thicker UBM layers. This is considered to be due to a more even distribution of EM flux across the joint by reducing current crowding at the corners of the joint. However, switching the test configuration, reducing current crowding, yielded results that disagreed with our proposition. It is found that the impact of the current configuration, and thereby the level of current crowding, on EM resistance does not vary much with the UBM thickness. This defies initial prediction that there would be a smaller impact in samples with thicker UBMs when the current configuration is switched. These results suggest that the EM failure mechanism is affected by UBM in a more complicated manner. Considering hidden factors such as the change in thermal stress with UBM thickness that act against the growth of EM void is necessary in order to better understand the mechanism.

This paper reports experimental observations evidencing that the intermetallic compound phase interfaced with Cu in the Al-Cu diffusion couple is most likely α2-Cu3Al phase, not γ-Cu9Al4 phase as previously assumed, and that its growth to a critical thickness may result in interface failure by stress-driven fracture. These conclusions are made based on an interdiffusion study of a diffusion couple made of a thick Cu plate coated with ∼ 2-μm-thick Al thin film. The interface microstructure and lattice parameter were characterized using scanning electron microscopy and x-ray diffraction analysis. Specimens aged at temperature between 623 K (350°C) and 723 K (450°C) for various hours produced consistent results supporting the main conclusions. It is found that disordered α2-Cu3Al phase grows in a similar manner to solid-state epitaxy, probably owing to its structural similarity to the Cu lattice. The increase in the interface strain that accompanies the α2-Cu3Al phase growth ultimately leads to interface fracture proceeding from crack initiation and growth along the interface. This mechanism provides the most consistent explanation for interface failures observed in other studies.

This paper describes the kinetic and microstructural mechanism of electromigration (EM) failure found in low-profile solder joints where EM and intermetallic phase formation compete for the same volume of Sn. The low-profile solder joint used in our study was made of 20-25um thick solder situated in between a Cu pillar and a Ni coated Cu lead frame (LF). The samples were EM tested in a temperature range of 140-170oC with the current densities varying between 35-45 KA/cm2 in an oil bath to induce failure without Joule Heat induced artifacts. Our studies on EM failure kinetics and microstructural mechanism have produced two key findings. The first finding suggests that the EM diffusivity (Z*D) of diffusing species (Sn, Ni, Cu) in the solder matrix can be uniquely ranked from microstructural analysis, and it is estimated to be (Z*D) Cu> (Z*D) Sn>(Z*D) Ni. This difference in EM diffusivity causes Cu-Sn and Ni-Sn intermetallic compounds (IMC) to develop in distinctively different manners under EM, leading to different EM failure mechanisms. The second finding is that EM in low-profile solder joints consists of multiple failure stages: a) with EM-related voiding in Sn dominating at lower temperatures; while b) thermally-induced IMC growth and invasion competes with EM-induced Sn voiding at high temperatures leading to the complete failure of each joint.

This paper reports the extreme sensitivity of the microstructure evolution in Sn-Ag-Cu (SAC) micro-solder joint, consisting of 15–20 µm thick solder alloy and electrolytic Ni plated Cu electrodes, to electromigration (EM) conditions resulting from a kinetic competition between the growth of Ni3Sn4 intermetallic compound (IMC) and EM of Sn in solder matrix. Microstructural analysis of samples tested at various conditions reveals that the voiding in the solder joint develops to the full extent only when the EM rate of Sn in the solder alloy is sufficiently greater than the growth rate of Ni3Sn4 IMC because the IMC phase is EM-inactive. The decisive evidence for such behavior is found from the samples tested at 160 and 170 °C under 35 kA/cm2. The samples tested at 170 °C show the classic solder joint microstructure that typically develops under the influence of EM, that is the exaggerated growth of Ni3Sn4 at the anode as a result of Ni-EM and voids at the cathode interface of the joint as a result of Sn-EM. Contrarily, such characteristic features are not found in samples tested at 160 °C, instead, the joint is fully converted to Ni3Sn4 without sign of its biased growth and voiding activities. The near absence of biased EM microstructure at 160 °C is believed to result from the stress-gradient developed by Sn EM counteracting the EM force on Ni. No such effect is at work at higher temperatures due to the significantly higher initial Sn EM rate which leads to voids and reduces the stress gradient.

The mechanical stability of Sn-Ag-Cu interconnects with low and high silver content against mechanical shock at room and elevated temperatures was investigated. With a heating element-embedded printed circuit board design, a test temperature from room temperature to 80°C was established. High impact shock tests were applied to isothermally pre-conditioned ball-grid array interconnects. Under cyclic shock testing, degradation and improved shock performances were identified associated with test temperature variation and non-solder mask defined and solder-mask defined pad design configuration differences. Different crack propagation paths were observed, induced by the effect of the elevated temperature test conditions and isothermal aging pre-conditions.

This paper examines the mechanics of ball shear testing with the objective of understanding the mechanism by which the maximum shear force and the rate of crack growth is dependent on the solder bump size. For this, Pb-Sn solder bumps with diameters between 460 lm and 760 lm are soldered to 400 lm-diameter Cu pads and subjected to ball shear testing. In spite of the constant interface area, the bump size significantly impacts the measured shear fracture force and the crack growth rate. Both the fracture force and the crack growth rate increase with bump size, and in the case of the fracture force, the increase is almost linear. Our analysis finds that the linear increase in the fracture force is a result of the bump deformation force, which increases with bump size. A simple model that accounts for the deformation force component is developed and used to extract the true interface fracture force. The estimated true interface fracture force is found to vary little with bump size, tightly converging to the 40 MPa to 48 MPa range. On the other hand, the dependence of crack growth rate on bump size is found to result from the higher degree of rotational moment associated with larger bumps.

This paper is concerned with the mechanics of interfacial fracture that are active in two common testing configurations of solder joint reliability. Utilizing eutectic Pb-Sn/Cu as a reference system and assuming the presence of a predefined crack size in the intermetallic compound (IMC) layer, stress intensity factors (K I and K II) at the crack are numerically calculated for the two given configurations. The analysis of the tensile test configuration reveals that the fracture occurs by the crack-opening mode (K I mode), as anticipated, but that it is greatly assisted by the viscoplasticity of the solder. With nonuniform viscoplastic deformation across the joint, K I is found to increase much more rapidly than it would without the solder, decreasing the critical crack size to the micron scale. The same mechanism is also responsible for the development of a K II comparable to K I at the crack tip, that is, |K I /K II| ~ 1. It is also found that the predominant fracture mode in the bump shear configuration is crack opening, not crack shearing. This is an unexpected result, but numerical analyses as well as experimental observations provide consistent indications that fracture occurs by crack opening. During shear testing, bump rotation due to nonzero rotational moment in the test configuration is found to be responsible for the change in the fracture mode because the rotation makes K I become dominant over K II. With rotational moment being affected by the geometry of the bump, it is further found that the fracture behavior may vary with bump size or shape.

Expanding use of electronic systems in close proximity to biological systems, marine environments, and polluted air introduce elements that attack solder joints. Mechanisms that cause damage and their effects on the lifetime of electronic system are discussed.

On-off operation of electronic systems leads to thermal cycling, which introduces thermal and strain energy into solder joints, resulting in microstructural evolution mechanisms of recovery, recrystallization and damage nucleation. The thermal strain history is very sensitive to the package design. The effects of microalloying on microstructure evolution bring both a challenge and an opportunity to improve the reliability of solder joints.

The mechanical stability of solder joints with SnAgCu alloy on various board thicknesses were investigated in a high G level shock environment. A test vehicle with 31mil, 62mil and 93mil board thickness, which has three different strain and shock level condition combination per board, was used to identify the joint stability and failure modes. The results revealed that joint stability is sensitive to board thickness and that the first failure location shift from the corner location near the stand off to the center with increased board thickness. Also the impact of isothermal aging and fine grain structure transformation on mechanical shock performance of solder joints were investigated. The results revealed that joint stability during shock loading is sensitive to the level of shock that can be absorbed during each shock cycle based on the capability of single to multi grain transformation. The localized fine grain structure distributions were analyzed to identify correlations between the microstructure evolution and shock performance.

This paper reports the fatigue properties of Sn-Ag-Cu (SAC) and Pb-Sn solder alloys determined from isothermal shear fatigue testing and analysis of resulting data using modified Coffin-Manson fatigue model. In our study, a series of cyclic shear fatigue testing was conducted on the solder joints in PBGA assembly with variation in testing temperature, strain range, and strain rate (or cycle frequency). The number of cycles to the first joint failure in the assembly was determined from the resistance of each solder joint and taken as the fatigue life of the assembly. Analysis of the resulting data reveals that isothermal fatigue behavior of both SAC (Sn-Ag-Cu) and Pb-Sn alloys follow the classic Coffin-Manson fatigue model and provide constants indicative of fatigue properties of the alloy, namely fatigue ductility coefficient and ductility exponent. These fatigue constants are consistent with what is generally expected from a metallic fatigue system except for their dependence on temperature. This deviation, attributable to the involvement of the thermal strain that is added to the applied mechanical shear strain, suggests that consideration of the thermal strain effect needs to be included in the fatigue analysis of solder joints even if the major fatigue mode appears to be pure mechanical shear. The fatigue constants gained from our study provide insights helpful in understanding the mechanism behind variation in fatigue reliability with solder alloy compositions and solder microstructure.

The work reported here concerns the effect of a brief exposure to a reversed current on the electromigration failure of narrow Al-Cu thin-film conducting lines. While the precise mechanism by which Cu retards electromigration in AlCu alloys is not fully understood, the consistent observation that electromigration failure is preceded by the sweeping of Cu from the failure site can be used to improve electromigration resistance by stabilizing the distribution of Cu. One way of doing this is to expose the Al-Cu line to a reverse current for some period of time. The present work shows that this method is particularly effective in thin lines with “quasi-bamboo” microstructures. It has the effect of building a reservoir of Cu at the upstream ends of the polygranular segments that are the preferred failure sites, and significantly increases both the mean time to failure, and the time to first failure of a distribution of lines. It can be inferred from these results that Al-Cu lines that conduct alternating current should be exceptionally resistant to electromigration failure.

This letter presents evidence of an oxidation-driven failure mechanism in Cu interconnects integrated with ultralow-κ materials. It is found that the open pore structure of ultralow-κ materials allows oxidants in the ambient to reach the interconnect structure and induce oxidation of Cu. In contrast to a normal oxidation process where Cu is in contact with the oxidant, oxidation is controlled by the outdiffusion of Cu through the barrier layers, Ta and SiCN, to form Cu oxide in the pores of the dielectric material. The loss of Cu by outdiffusion induces extensive voiding and subsequent failure in Cu interconnects.

This paper reports experimental evidences suggesting that fracture resistance of solder joints in packaging assembly is not necessarily uniform but can vary significantly with their locations in the assembly. Specifically, it is found that the solder joints located at the chip corner is much more prone to the fracture under thermo-mechanical or mechanical loads such as vibration and shock. This, the variation in fracture resistance, is found from the isothermal cyclic shear fatigue testing of 60Pb-40Sn, SAC305 and SAC105 solder alloys assembled in PBGA configuration. Under pure shear fatigue situation, the location of the joint failure should ideally be random because all joints are subjected to an identical shear strain. However, it is found that the solder joint that fails first is always the ones located at the corner site of chip. The inner joint, close to the center of the chip, shows much higher resistance against failure. Further study that investigates the joint microstructure with aging, produces evidences indicating that the corner is the place of high residual thermal stress or stored strain energy (by plastic deformation during cooling or heating). In an extreme case, where stored energy by plastic deformation is not released well by competing process, the stored energy triggers recrystallization, leading to polygranular structure of the corner joint solder. These results strongly suggest that the joints in package assembly have different reliability and microstructural evolution path depending on their location in the package and thermal history.

This paper presents the results of isothermal shear fatigue tests conducted on Sn-Ag-Cu lead-free and Pb-Sn solder alloys in fully assembled PBGA configurations. Our study, aimed to better understand the fatigue behaviors of solder alloys in the package assembly, reveals that the shear fatigue test is indeed capable of measuring the fatigue properties of the solder itself and may lead to material parameters necessary for the prediction of the solder joints fatigue reliability in general. The cyclic shear fatigue test conducted under various fatigue conditions yields the result suggesting the close linkage between the fatigue in PBGA and the fatigue property of solder alloy. This is a result of the existence of consistency in crack propagation path in solder joint, that is fixed to the chip-side of solder matrix (not interface), irrespective of test conditions that includes temperature, strain range, and frequency. As a result, fatigue failure of Pb-free and Pb-Sn solder can be fitted to the Coffin-Manson fatigue model with a reasonable consistency. The resulting numerical constants, namely ductility coefficient and ductility exponent, are consistent with what is expected from general consideration of metallic fatigue system. However, their temperature dependence shows a significant deviation from expectation, probably due to an addition of thermal strain to the shear strain and its variation with temperature.

Sintered magnetoelectric composites generally exhibit reduced sensitivity due to: 1) breakage and aggregation of ferrite coating during compaction forming path of reduced resistivity, 2) coarsening of piezoelectric particle leading to interfacial porosity, and 3) inter-diffusion between piezoelectric and ferrite resulting in coherency loss. In order to overcome these problems, we are adopting following approach to synthesize core-shell nanocomposites: 1) create piezoelectric particles (sub-micron range) with crystallographic facets which will result in uniform stress.

This work is principally concerned with the microstructure of electromigration failure in narrow Al‐2Cu‐1Si conducting lines on Si. Samples were patterned from 0.5‐μm‐thick vapor‐deposited films with mean grain size of 2.4 μm, and had linewidths of 1.3 μm (W/G≊0.5), 2 μm (W/G≊0.8), and 6 μm (W/G≊2.5). The lines were tested to failure at T=226 °C and j=2.5×106 A/cm2. Other samples were tested over a range of substrate temperatures and current densities to test the effect of these variables, and 1.3 μm lines were tested after preaging at 226 °C for various times to change the Cu‐precipitate distribution prior to testing. Three failure modes were observed: The 6 μm specimens failed by separation along grain boundaries with an apparent activation energy of 0.65 eV; the 1.3 μm specimens that were preaged for 24 h failed after very long times by gradual thinning to rupture; all other narrow lines failed by the transgranular‐slit mechanism with an activation energy near 0.93 eV. Microstructural studies suggest that the transgranular‐slit failure mechanism is due to the accumulation of a supersaturation of vacancies in the bamboo grains that terminate polygranular segments in the line. Failure occurs after Cu has been swept from the grain that fails. Failure happens first at the end of the longest polygranular segment of the line, at a time that decreases exponentially with the polygranular segment length. Preaging the line to create a more stable distribution of Cu lengthens the time required to sweep Cu from the longest polygranular segment, and significantly increases the time to failure. In the optimal case the transgranular‐slit failure mechanism is suppressed, and the bamboo grain fails by diffuse thinning to rupture. Preaging is particularly effective in increasing the lifetimes of lines that contain very long polygranular segments, and has the consequence that the time to first failure in an array of lines is much longer than predicted by a log‐normal fit to the distribution of failure times.

This paper reports the experimental and theoretical exploration of the fracture mechanism active in BGA lead-free solder assemblies under high speed shear fatigue test conditions. Our investigation finds that, contrary to common assumption, the crack growth in shear fatigue is not governed by shear stress but more by crack opening stress. Our theoretical analysis indicates that fracture by crack opening mode prevails because non-uniformity in the shear deformation of solder joint creates a body rotation which results in crack opening stress rather than shear. While the crack growth in shear fatigue is found to vary sensitively with variation in the mechanical constraints on the assembly, such as solder shape and elastic modulus of the chip mold, it is also sensitive to variation in solder microstructure. This, the sensitivity to the assembly constraints and solder microstructure, makes it ideal in investigating fatigue properties of solder joints as well as identifying the structural and microstructural features responsible for reliability failure.

This paper reports a study of the effect of Cu precipitation on electromigration failure in Al‐2Cu‐1Si thin‐film conducting lines. The films were 0.5 μm in thickness, and patterned to widths of 1.3 and 4 μm, providing width‐to‐grain‐size ratios (W/G) of approximately 0.5 and 2. The lines were aged for various times at 226 °C, and were then tested to failure at a current density of 2.5×106 A/cm2. Scanning and transmission electron microscopy were used to study the Cu precipitate distribution, its evolution during aging and electromigration, and the microstructural failure mechanism. Aging produces a dense distribution of intragranular θ’ (Al2Cu; coherent), with stable θ (Al2Cu; incoherent) in the grain boundaries. The θ’ is replaced by θ as aging proceeds. In the wide lines (W/G≊2), the mean time to failure (MTF) increases slowly and monotonically with prior aging time. The failure happens through the growth and coalescence of intergranular voids. In the narrow lines (W/G≊0.5), both the MTF and the time to...

This study concerns the failure mechanism of a lead-free SAC305 solder joint under the isothermal bending fatigue condition. For this, series of bending fatigue tests are carried out with the variations of bending displacement and frequency, and FEM analysis is conducted to examine the relationship between mechanical behavior of a solder joint and fatigue life cycle. The results of this investigation suggest that the isothermal fatigue of a solder joint is strongly affected by plastic deformation process of SAC solder. It is found that the failure life cycle ‘N f ’ fits well to the model suggested by Coffin-Manson, that is N f ∼ (Δe p )−2.3. This result indicates that fatigue failure is determined by accumulation of damage by plastic deformation of solder. The fractography of a solder joint also supports our conclusion because crack is found to propagate along the solder matrix near to IMC interface.

This paper presents findings evidencing that electromigration(EM) imparts significant influence on the kinetics of contact failure of Au wire bonded to an Al pad. Contact resistance between Au wire and Al pad at moderately accelerated test conditions (T=150–175°C; j=5×104A/cm2) revealed that the failure rate depends highly on the direction of electron flow across the contact: electron flow from Au to Al resulted in far faster failure rate than the opposite direction. Microscopic inspection of the contact interface indicated that the wirebond contact failure is related to the growth of Au-Al intermetallic compounds (IMC). EM was found to influence the failure kinetics because it accelerates or decelerates the growth rate of the IMC.

This paper introduces new electrochemical cells and testing methods that are ideal for characterizing corrosion risk assessment of the components used for liquid cooling system with high surface to liquid volume ratio. Two cell configurations are described in this paper and they use three electrode and two electrode cells. These cells have the identical structure except for the number of electrodes. These cells are made by sandwiching the working electrode plate (sample) and the counter electrode plate (graphite) with a spacer (gasket), and by filling the cavity with the liquid under interest. In case of the three-electrode cell, the reference electrode is inserted through the hole in the graphite. The three-electrode cell is ideal for the quantitative characterization of the corrosion rate by utilizing conventional electrochemical techniques such as a Tafel method. The use of the two-electrode cell is similar to the case of the galvanic corrosion characterization as it measures the cell current that flows between the dissimilar metals that are in contact with the liquid. When coupled with computer assisted data acquisition, the two-electrode cell configuration allows the characterization of long-term corrosion reliability of a component with a variation in a large number of test variables. It is particularly useful in finding corrosion inhibitors.Copyright © 2006 by ASME

This letter reports experimental observations evidencing the viscoplasticity of porous organosilicate glass thin films under conditions pertinent to their application in advanced low-k/Cu interconnect technology. Specifically, it is found that porous SiCOH thin films exhibit a significant level of viscoplasticity with a rate sensitive to the porosity, the degree of plasma damage, and hydration reaction when tested using a ball indenter at 150–400 °C. The activation energy of the viscosity (1.25–1.45 eV) is measured to be far lower than the bulk glass (>4 eV), suggesting that the viscous flow is affected by the presence of defective bond-network such as Si–OH or Si–H bonds.

We report on a collective body of work wherein we have studied the mass transport phenomena which are likely to be operative during stress driven changes in microstructure arising from electromigration and stress voiding. Our goal is to understand such microstructural evolution leading to failure of the metal lines or interconnects associated with integrated electronic circuits or chips. This work, when complete, will lead to improved electronics performance and reliability and faster product development arising from accurate and predictive models of wearout phenomena. We report on the role of thermal induced strain leading to hole and hillock formation, the influence of grains structure on the reliability of Al- based interconnects, and the observation of counter-current electromigration of Ua in Al grain boundaries.

Thin-film Al-2Cu conducting lines with “quasi-bamboo” microstructures were investigated to understand the microstructural mechanism of electromigration failure. Both conventional test structures and electron-transparent lines fabricated on silicon nitride windows were utilized to identify the “weakest” polygranular segments. Even when the current density was reduced to 0.75 MA/cm2 and the segment length was on the order of a few microns, failure occurs at the upstream termination of the longest polygranular segment in the line, at a time that decreases exponentially with the segment length. There is no apparent “Blech length” in quasi-bamboo Al- Cu lines; the longest segments are the failure sites, and their lifetime decreases with segment length in a regular way, even when the longest segments are only a few microns in length. It follows (and is observed) that the time-to-failure distribution of a group of lines is fixed by the distribution of the longest polygranular segments within them. This distribution can be effectively controlled by post-pattern annealing, which can refine the quasi-bamboo structure so that the longest polygranular segments are short and the distribution of longest polygranular segment lengths is narrow. Consequently, post pattern annealing is a very effective method for improving reliability by increasing the time to first failure.

The work reported here concerns the effect of grain structure on electromigration failure in pure Al and Al-2wt.%Cu-lwt.%Si lines. The grain structures of fine lines were controlled by annealing after patterning to promote the formation of "bamboo" structures. Significant improvements in the median time to failure (MTF) and the deviation of the time to failure (DTF) were observed with the development of near-bamboo structures with polygranular-segment lengths shorter than ~ 5 μm. The most common failure sites are voids or slits across bamboo grains at the upstream ends of polygranular segments. The time-to-failure decreases with the polygranular segment length, and can be significantly enhanced by controlling the grain structure.

Microstructural studies of electromigration failure were performed on A1-2Cu-1Si interconnects with 1.3 μm width. The lines were tested to failure under controlled conditions after pre-aging for various times at three different temperatures. Examination of the microstructure of the failure sites suggests that the lines fail at the bamboo grains that terminate the longest polygranular segments in the line. Transgranular slit voids form after Cu has been swept from the grain that fails. Hence, pre-aging lines to create a more stable distribution of Cu lengthens the time required to sweep Cu and significantly increases the time to failure. The optimal microstructure has a maximum intragranular density of stable θ precipitates. In this case transgranular slit failure is suppressed, and the bamboo grain fails by diffuse thinning to rupture.