We present a new approach to measuring mobile charge in SiO2 without MOS test structures. The measurement is done in non-contact mode and offers a whole-wafer mapping capability of the mobile charge in single minutes. Determination of the mobile charge is based on monitoring the ion drift by measuring corresponding changes in the oxide voltage. The drift is induced on the whole wafer by corona charge deposited on SiO2 followed by 200°C annealing. Surface contamination is distinguished from bulk contamination by selecting the appropriate sequence of the corona-temperature-stress, which injects ions from the surface into the SiO2.

Traces of iron contamination as minute as one part per trillion atom fraction can be detrimental to the gate oxide integrity of the very thin oxides, 10 nm or less, used in the most advanced ICs. Fe contamination monitoring discussed in this paper is done with the surface photovoltage (SPV) technique which measures the minority carrier diffusion length, L, before and after the recombination efficiency of iron is enhanced by optical splitting of the iron-boron pairs. Wafer-scale mapping of iron gives fingerprints of contaminating tools and processes. In this paper, we also present an extension of SPV to oxidized wafers. In the past, such measurements were rendered impossible due to optical interference in the SiO2.

The apparatus incorporates a whole wafer, optical Fe activation station and it provides whole wafer maps of Fe in a total time of 6 to 20 minutes per wafer, depending on probing density.

Of crucial importance to all areas of the microelectronics industry is the characterisation of silicon wafer quality. An important indicator of this is carrier lifetime, and a convenient non-destructive method for its measurement is Pholothermal Radiometry. This involves the photo-generation within a semiconductor sample of electron-hole pairs. Periodic generation of carriers leads to modulation of free-carrier absorption of mid-ir radiation provided by a black body source. The fr-radiation detected from the photo-excited region is inversely proportional to the optically induced carrier density. As the modulation frequency is increased, a point is reached at which the photo-generated carriers no longer have sufficient time to decay between pulses. This frequency is dependent on the carrier lifetime. We present a description of the Photothermal Radiometric lifetime scanning instrument built at UCL. This instrument offers an accurate method for producing detailed maps of carrier lifetime across whole or part wafers. The problem of surface-state effects has been addressed by employing a broad-band uv source to optically fill the surface states of the sample under investigation. The instrument is capable of producing maps of lifetime variation with 0.5 mm resolution. Alternatively, for selected points on a wafer, the instrument can generate detailed frequercy scans of free-carrier absorption. From these, it is possible to obtain information on surface recombination velocity and diffusion length.

The deformation of film/substrate systems due to a through-the-thickness distribution of mismatch strain is discussed for cases when the layers are thin, compliant and freestanding. The connection between substrate curvature and mismatch strain is reviewed within the framework of small strains and small deflections. These results are applied to determine the second-order correction to the Stoney formula and to note the connection between mismatch strain distribution and curvature history. Then, for small strains but large deflections, it is observed that a symmetric system with equi-biaxial mismatch strain can undergo a bifurcation in deformation mode. It is also shown that the bifurcation response is sensitive to system imperfections. Finally, the critical condition for introduction of an interface misfit dislocation in an epitaxial system is established, thereby extending the Matthews-Blakeslee criterion to the case of free-standing compliant systems.

Nucleation of voids due to vacancy condensation in passivated aluminum lines is analyzed within the context of classical nucleation theory. A discussion of sources of hydrostatic tensile stress in such lines provides a reasonable upper limit of 2 GPa. The void nucleation rate is then calculated at various sites within the line. Results suggest that nucleation rates are far too low to account for observed rates of voiding. Void nucleation at a flaw at the line/passivation interface is then considered as an alternative nucleation mechanism. Such flaws may be created by contaminants introduced during fabrication of the line. In this case, nucleation is feasible at greatly reduced stresses, well within the observed values. Furthermore, a simple model of void growth indicates that a fast atomic transport path, such as a grain boundary, must intersect the void for an appreciable growth rate. These results suggest that void nucleation in aluminum interconnect lines occurs at flaws at the sidewall of the line and that stress-induced and electromigration-induced voiding can be controlled by eliminating interfacial contamination.

We have modeled the effects of grain aspect ratio on strain energy density in (100)-oriented grains in a (111 )-textured Cu film on a Si substrate. Minimization of surface energy, interface energy, and strain energy density (SED) drives preferential growth of grains of certain crystallographic orientations in thin films. Under conditions in which the SED driving force exceeds the surface- and interface-energy driving forces, Cu films develop abnormally large (100) oriented grains during annealing. In the elastic regime the SED differences between the (100) grains and the film average arise from elastic anisotropy. Previous analyses indicate that several factors (e.g. elimination of grain boundaries during grain growth) may alter the magnitude of the SED driving force. We demonstrate, using finite element modeling of a single columnar (100) grain in a (111) film, that changes in grain aspect ratio can significantly affect the SED driving force. A minimum SED driving force is found for (100) Cu grains with diameters on the order of the film thickness. In the absence of other stagnation mechanisms, such behavior could cause small grains to grow abnormally and then stagnate while large grains continue to grow. This would lead to a bimodal grain size distribution in the (100) grains preferred by the SED minimization.

Production Engineering R‘‘void-free’ metallization is proposed. It is shown that void formation is suppressed when the TiAl3 intermediate layer is formed at the interface between the Al alloy and reactively sputtered TiN barrier metal films. We have investigated the relationship between void formation and coverage of the intermediate layer. It is found that a coverage of more than 60 % TiAl3 perfectly suppresses void formation. The interfacial reaction is achieved by using ‘soft TiN', which arises from the short migration length of the sputtered particles impinging on the substrate surface when dc power is decreased. The soft TiN film includes many vacancies and crystallographically disordered regions which easily cause rearrangement of the TiN films by movement of Ti atoms during annealing. It is thought that these Ti atoms compensate vacancies in the Al-Si-Cu film and suppress the formation of Al voids.

Void and hillock formation during annealing was studied depending on the deposition conditions. Aluminum films were deposited onto oxidized silicon substrates by the self-ion assisted technique. The bias 0 or 6 kV was applied to the substrate during deposition. The films were then annealed in vacuum for 1 hour in the temperature range from 1500 to 550°C. The structure of the films was investigated by transmission electron microscopy.

It was found that recrystallization and void and hillock formation in the films depend on the bias during deposition. Normal grain growth occurred in the films deposited without bias. Abnormal grain growth was observed in the 6 kV-films. It was also found that the mechanism of stress relaxation during thermal cycling depends on the self-ion bombardment. In the films prepared without bias, stress relaxation proceeds by diffusion creep. In the films deposited at the 6 kV bias, stress relaxation proceeds by plastic deformation.

We have developed a technique for measuring thin film stress during growth by monitoring the wafer curvature. By measuring the deflection of multiple parallel laser beams with a CCD detector, the sensitivity to vibration is reduced and a radius of curvature limit of 4 km has been obtained in situ. This technique also enables us to obtain a 2-dimensional profile of the surface curvature from the simultaneous reflection of a rectangular array of beams. Results from the growth of SiGe alloy films are presented to demonstrate the unique information that can be obtained during growth.

The evolution of stresses due to the patterning and thermal loading of thin lines on Si wafers, and the consequent changes in the overall curvature of the wafer are studied theoretically and experimentally. The analysis involves finite element simulations within the context of generalized plane strain models. The analysis is capable of predicting the wafer curvature in directions parallel and perpendicular to the lines. These predictions compare reasonably well with experimental measurements of curvature made on model systems. The thickness, width and spacing of the patterned lines have been varied systematically, and the associated changes in the evolution of stresses and curvature have been determined. The non-uniform stress field within the fine lines is also analyzed.

Mechanical stress and its relaxation in aluminum metallization in integrated circuits (IC) are a major concern for the reliability of the material. It is known that adding Cu improves the reliability but complicates plasma etching and increases corrosion sensitivity. The mechanical behavior of AIVPd, AlCu and Al blanket films is investigated by wafer curvature measurements. During thermal cycling between 50°C and 400°C the highest tensile stress is found in AIVPd. In a subsequent experiment, the cooling was interrupted at several temperatures to investigate the stress behavior during an eight hour isothermal treatment. Isothermal stress relaxation has been observed in the three types of films and is discussed.

800 nm thick and 1 jim wide Al-lwt.%Si-0.5wt.% Cu parallel lines with I μm spacing were passivated with PECVD oxide, oxynitride or nitride. Substrate curvature measurements as a function of temperature and XRD-measurements at room temperature were used to characterize macroscopic samples of these parallel Al-Si-Cu-lines. By using both techniques the average inplane stresses for the Al-Si-Cu lines as well as for the covering passivation material can be determined as a function of temperature. The highest and lowest stresses in the Al-Si-Cu are observed for lines with nitride and oxide passivations, respectively. Also the number of voids in the lines after a storage test at 250 °C is clearly highest for a nitride passivation and lowest for an oxide passivation.

The stress in the passivation itself and its temperature dependence is found to be very different from the stress in a blanket passivation film.

Passivated metal lines, commonly used in integrated circuits, show thermally induced stresses due to the difference of the thermal expansion coefficients of the lines and their surroundings. These stresses cause voidage and plastic flow of the lines. Aim of the analysis was to derive equations connecting experimentally measured strains or stresses by the X-ray diffraction and wafer curvature techniques with the magnitude of voidage and plastic shear deformation of the lines.

Using the concepts of linear elasticity the volume averaged stresses of an array of parallel interconnects embedded in a passivation layer on a flat substrate are analysed. Equations are derived connecting the volume averaged stresses in the metal and in the passivation with the “eigen-strains” of the metal which characterize the true (stress free) thermal strains and plastic deformation strains of the metal. The coefficients entering these equations are determined from (elastic) finite element method (FEM) calculations performed for various geometries and aspect ratios of the metal lines. Choosing the proper values of the coefficients allows the eigen- strains to be determined from the experimental data.

By comparison of the evaluated eigen-strains with the purely elastic eigen-strains ΔaΔT the extent of voidage and/or plastic shear deformation of passivated metal lines caused by thermally induced stresses can be determined model independently.

Al and Al- alloys are commonly used as interconnect materials in integrated electronic devices. Stress induced voiding and degradation of metal lines by electromigration are closely related to the stresses in the lines.

We have studied the strain and stress evolution during thermal cycling, isothermal relaxation and due to electromigration in passivated Al and AlSi(1%)Cu(0.5%) lines by Xray diffraction with variation of experimental parameters such as the aspect ratio and the electrical current density. Furthermore the extent of voiding and plastic shear deformation has been determined from the experimental metal strains with the help of finite element calculations.

Main results are: 1) During thermal cycling the voiding is less than 2.10-3. The extent of plastic shear deformation increases with increasing line width and with decreasing flowstress. 2) During isothermal relaxation void growth occurs but no significant change in the plastic shear deformation. 3) An electric current in the lines causes no measurable additional change of the volume averaged stresses up to line failure.

Residual strains in thin films of 1.0 μm, 0.5 μm and 0.25 μm Cu capped with 24 nm Ta and a 62 nm Ta diffusion barrier were measured using x-rays during continuous temperature ramping. Cycles up to 300°C and 400°C were performed. At ambient temperature, the flow stresses after heating to 400°C were found to increase wth decreasing film thickness. The stresses were sufficiently high below 100°C to cause plastic yielding. The diffracted intensities from the in-plane and out-of-plane (111) planes were recorded with temperature and the intensity changes were attributed to the necessary dislocation density and distribution variation to accomodate the mismatch strain due to thermal expansion differences.

As interconnect technology moves deeper into the submicron realm, the physics governing the mechanical behavior, hence the reliability of metal lines changes. Events at the level of individual grains become more important. One would like to be able to measure the strain and orientation of single grains, and the distribution of strain across a grain. Such data may help us understand such effects as sunken grains, the erratic occurrence of stress voiding and Blechlength effects in electromigration. To get this information, techniques involving X-ray diffraction with submicron beams are being developed. These experiments involve detecting Laue spots from selected areas and energy-analyzing them to find the strain tensor. Motivations for such measurements, the specifications for a useful instrument, some designs now being developed and the intrinsic limitations of these approaches will be discussed. Design issues in the choice of focussing optics and energy measurement will be highlighted with calculations of how the Laue spots would look for different conditions of crystal perfection.

Spatially resolved residual stress and lattice distortion measurements have been made or experimental RF NPN devices which failed electromigration testing. The experimental devices underwent a steady state long term life test for 38272 hours at a die temperature of 100 degrees Celsius and device emitter currents of 100 mA. Residual stress contour maps of the Au interconnect metallurgy of these patterned devices were constructed with spatial resolution approaching 13 microns. The stress contour maps clearly show areas of the Au interconnect emitter lines, approximately 17 microns wide, which buckled when the local stresses exceeded the adhesion limits of these films. A stress gradient was observed going along the Au line from the emitter to the wire bonded pad which went from slightly compressive near the emitter to tensile approaching the wire bonded lead, and consistent with electromigration failures. In addition, lattice distortions as determined by integral line breadths, which were least square fitted by a gaussian*orentzian convolution of the instrument function, show a large continuous increase along the Au interconnect line in going from the emitter to the wire bond pad. The data was acquired using a 4-circle x-ray diffractometer which was specifically constructed to study electromigration issues in-situ. The diffractometer has a very low sphere of confusion (< 5 microns) and is equipped with tapered glass capillary optics, a mechanical, electrical, and thermal stressing stage and various energy dispersive (EDX) or position sensitive (PSD) detectors.

Design of reliable thin film metallization is a complex problem that requires information about local parameters such as composition, microstructure, residual and loading stresses, etc. On the other hand, such information has been traditionally obtained using techniques that have sampling volumes that are appreciably larger than the BEOL dimensions currently in use. In this article, the x-ray residual stress analysis technique and its extensions to micro-domain measurements are reviewed. These concepts are then applied to strain data from in-situ x-ray measurements, obtained in real-time from Al metallization lines subjected to thermal stressing and electromigration.