This paper presents the methodology employed in the determination of the stress tensor for thin crystalline films using x-ray rocking curves. Use of the same equipment for the determination of the average stress in poly- or non-crystalline thin films attached to a crystalline substrate is also discussed. In this case the lattice curvature of the substrate is determined by measurement of the shift In the Bragg peak with lateral position in the substrate.

Strains in single crystal layers may be measured using Bragg diffraction from the layers and from the substrate or a reference crystal, with the highest strain sensitivity of any known technique. The difference in Bragg angles for a strained and an unstrained crystal is related to the change in d spacing of the Bragg planes, and the elastic strain is related to'this angular difference. The separation of two peaks on an x-ray rocking curve is generally not equal to the difference in Bragg angles of two diffracting crystals, so diffractometer measurements must be carefully Interpreted in order to obtain x-ray strains in crystalline films (x-ray strains are strains relative to the reference crystal). The unstrained d spacings of the film and the d spacings of the reference crystal must be known to obtain the elastic strains in the film, from which the stress tensor is determined.

A new, differential method for determining the stiffness of a sub-micron indentation contact area is presented. This allows measurement of elastic modulus as well as plastic hardness, continuously during a single indentation, and without the need for discrete unloading cycles. Some of the new experiments that become possible with this technique, especially at the nanometre scale, are described. We show quantitatively that electropolished tungsten reproducibly exhibits the ideal theoretical lattice strength at small indentation loads.

Thin films of [111] oriented CdTe have been MOCVD grown onto [111] GaAs substrates. When thicknesses exceed 1000Å the epitaxy is quite good (backscattering minimum yield of approximately 15%) in spite of a 14% lattice mismatch. A narrowing of the Cd angular scan suggests a displacement of some of the Cd atoms in the lattice. A model based on a Te vacancy is presented to describe the data.

Residual stress has been measured as a function of layer thickness in thin films of an Al alloy on oxidized Si by the x-ray “d” versus sin2ψ technique. Samples with and without a passivation layer were examined. The results show an increase in residual stress with decreasing film thickness for the passivated samples and indicates that the interface between the metal film and SiO2 may be a region of high stress.

The perpendicular x-ray strain of epitaxial CoSi2 films grown on Si(111) substrates at ∼ 600°C by MBE was measured at various temperatures. Within experimental error margins, the strain decreases linearly with rising temperature at a rate of (1.3±0.1) × 10-5/°C from room temperature up to 600°C. Over that temperature range and the duration of a complete measurement (∼5h to ∼ 2h), these strain values remain reversible. At 593°C, the x-ray strain is -0.85%, which is about the strain that a stress-free CoSi2 film on Si(111) would have at that temperature. This results show that the stress in the epitaxial CoSi2 film is fully relaxed at the growth temperature. Strains below the growth temperature are induced in the film by the difference in the linear coefficient of thermal expansion of CoSi2 and Si. They were calculated by assuming that the density of misfit dislocations formed at the growth temperature remains constant. The slope of the strain-temperature dependence obtained that way agrees with the measured slope if the unknown Possion ratio of CoSi2 is assumed to be VcoSi2 = 0.35. A film stress of ∼ 0.8 GPa at room temperature was calculated using the above value for the Possion ratio, 130 GPa for the Young modulus, and the measured x-ray strain.

Measurement of the curvature induced in a wafer (or other flat plate) by the stress in a thin film has long been used as a convenient and accurate technique for the determination of the stress. Numerous improvements over the years have led to instruments that provide simple and rapid measurements of stress as a function of the time and temperature for any desired thermal history. A computer controlled instrument using laser scanning will be briefly described and its capabilities and limitations discussed.

Applications of the technique to a variety of thin film materials will be discussed. In addition to the effects of differences in thermal expansion, stresses associated with various deposition techniques, gain or loss of material, phase transformations and flow will be considered. In aluminum based systems, themal expansion, plastic flow and phase transformation play major roles. Refractory metals show, in addition, large stresses associated with the deposition process. In inorganic dielectric systems thermal expansion effects are usually relatively small; deposition effects and the gain or loss of material are the dominant effects. Silica based glasses formed by chemical vapor deposition, for example, show large stress changes due to gain or loss of water, and plasma deposited silicon nitride films show large effects associated with hydrogen. Overall, determination of the stress as a function of time and temperature is a valuable part of the evaluation of a thin film material for use in a VLSI device.

In this study, we have measured changes in hardness and residual stress state in polycrystalline nickel samples which have undergone a variety of ion beam surface treatments including self-implantation, amorphization by ion beam mixing Ni-Al multilayers, and the formation of a two-phase microstructure by elevated temperature aluminum implantation. Hardness-depth profiles of the the near-surface regions were determined using an ultra-low load microindentation system, and residual stress states induced by the surface modifications were analyzed using optical interferometry. These property changes were correlated with the accumulation of surface fatigue damage.

Picosecond ultrasonics have been used to measure the sound velocity and index of refraction in a thin film of AlAs. The AlAs layer was grown by molecular-beam epitaxy on a buffer layer of GaAs which was deposited on a semi-insulating GaAs wafer. The AlAs film was capped by a thin layer of GaAs, and a very thin film of InSb. To generate a sound wave effectively a picosecond light pulse was absorbed in the InSb film. As the sound wave propagates through the microstructure it changes the optical properties in the various films, and this produces a change in the optical reflectivity, which is measured. From the round trip time of the acoustic wave we have determined the sound velocity in the AlAs film.

Amorphous LPCVD and PECVD Si3N4 films have been patterned into fibers of various shapes using integrated circuit methods. A tensile test apparatus was developed which allows a simultaneous testing of up to 20 fibers. The average tensile strength of LPCVD Si3N4 fibers with a 30 μm gauge section was 7.0 ± 0.4 GPa. The corresponding value for PECVD Si3N4 was 2.4 ± 0.2 GPa. The following properties of the materials were measured: atomic composition (RBS, FRES), density, refractive index, internal stress, elastic modulus and micro hardness. A Weibull analysis of the tensile strength values was performed.

Aluminum alloys have virtually replaced aluminum for interconnections in VLSI because of their improved reliability. Mechanical stress is a problem of growing importance in these interconnections. Stress as a function of temperature was measured for thin aluminum films and several aluminum alloys and layered films consisting of silicon, copper, titanium, tungsten, tantalum, vanadium, and TiSi2. Solid-state reactions of the aluminum with the additives and with the ambient during thermal cycling will occur and depending on what compounds have formed and at what temperature, this will determine the morphology and reliability of the metallization. The measurement technique, based on determination of wafer curvature with a laser scanning device, directly measures the total film stress and reflectivity in situ during thermal cycling. Changes in stress were detected when film composition and structure varied and were correlated using x-ray diffraction with the formation of aluminides. Other phenomena that contribute to stress changes including elastic behavior, recrystallization, grain growth, plastic behavior, yield strength, and film hardening from precipitates.

Metal films used as conductors in integrated circuits must be able to withstand stresses created during their deposition, caused by mismatches in thermal expansion coefficients and by bending which occurs during manufacture and installation. These stresses may be biaxial or uniaxial in nature. This paper discusses testing techniques and gives examples of results of mechanical testing of thin metallic films of aluminum, aluminum-copper(2%), copper, and tungsten in uniaxial and biaxial tension. The film specimens were deposited on and removed from production silicon wafers in order to simulate as much as possible actual device conditions. Basic considerations for the comparison of the two testing methods are also presented.

The present authors recently introduced the mechanical deflection of cantilever microbeams as an experimental technique for measuring the strength and stiffness of thin films.[1] The technique utilizes conventional integrated circuit (IC) fabrication to process the samples, a Nanoindenter to deflect the cantilever beams mechanically, and simple elastic beam theory to analyze the data. This paper will review the technique and describe some of its current applications.

A method for determining mechanical parameters and residual stresses for thin films is described. Multi-layer cantilever beams (LPCVD SiNx/thermal SiO2) are fabricated utilizing standard IC processing technologies and micromachining of silicon. The elastic response of the beams to imposed deflections is then measured using a Nanoindenter, a sub-micron hardness testing machine. The elastic constants of the nitride films are calculated from the force vs. deflection slope and known elastic constants of the thermal SiO2 and silicon. By measuring the curvature of the multi-layer cantilever beams with a scanning electron microscope after successive etching of the LPCVD nitride films, average and differential stresses in the films were calculated.

The hardness and modulus of thin (<1 μm) ceramic films have been investigated using the differential mechanical properties microprobe (DMPM). The DMPM consists of a computer controlled ultra-low load indentation system with a special system that allows the elastic stiffness of the contact to be determined continuously during the indentation process. The films were deposited in a molecular beam epitaxy (MBE) system utilizing unique high temperature effusion cells. Two specimens have been studied. The first is a wedge shaped layer of amorphous Al2O3 on a silicon substrate. The thickness of the layer varies from 0 to 990 nm at the two extreme sides of a 3 in. silicon wafer. The effect of layer thickness on the hardness and modulus as measured with the DMPM has been quantified for this system. The second specimen is made up of 38 repetitions of alternating layers of crystalline TiO2 and amorphous Al2O3. The wavelength of the composition modulation and the thickness of the film varies uniformly from 0 to 38 nm and from 0 to 1460 nm respectively across the 3 in. diameter of the silicon substrate.

Closed-form elasticity solutions are introduced, that predict the average displacement beneath square and triangular, uniformly loaded areas at the surface of a bilayer. The solutions aid in the application of depth-sensing indentation techniques for measuring thin film elastic moduli. The elasticity solutions agree closely with experimental data of Al, Si, 1 μm Al on Si, and 2 μm Cr on Si. The case of poor adhesion between the film and substrate is briefly examined.

Ordered GaxIn1-xP epilayers grown on GaAs substrates have been studied in the electron microscope. Dark-field images and high-resolution micrographs have directly revealed the presence of an ordering of the Ga and In (111) planes. Several different planar defects have been identified in the ordered structure. There is a very high density of faults parallel to the (001) growth surface. Models for these faults have been constructed in terms of stacking faults and twin boundaries in the ordered structure. Other flat planar defects which occur approximately parallel to (111) and (111) planes can also be described in these terms.

Micro-indentation and micro-scratch techniques were used to characterize the hardness and the adhesion strength of 0.11 μm thick sputtered carbon f-ilms on Silicon substrates. Hardness depth profiles and critical loads were measured using a microindenter under indentation and scratch testing modes, respectively. The carbon film with 6 mtorr argon sputtering pressure shows better practical adhesion (or higher critical load) and slightly higher hardness. The indentation fracture phenomenon observed on the 30 mtorr film is closely related to its poor adhesion. The failure mechanism will also be discussed.

Hardness and modulus values of bulk and epilayer ZnTe and CdTe samples and of ZnTe-CdTe superlattices are reported. Both hardness and Young's modulus values increase with increasing ZnTe content in the ZnCdTe samples. Alloying effects and strains in the superlattice structure are proposed to explain the strengthening.

Gold-nickel multilayer films with periods of 1.2 - 4.6 nm were deposited on silicon substrates by magnetron sputtering, and plan and cross-sectional specimens were examined by transmission electron microscopy. The cross-sectional specimens revealed well-defined layering and columnar growth features that extended through the film thickness. Dark striations extending normal to the layers were attributed to diffraction contrast from defect strain fields. Diffraction patterns showed that the films were highly textured and that short-period films had a single fcc structure, while long-period films separated into three fcc structures. High-resolution images of the layer interfaces showed local regions of epitaxy partitioned by regions of disorder.

Indentation tests using a Nanoindenter, a depth-sensing indentation device, were performed to measure the elastic modulus and hardness of the films. No modulus enhancement was detected, and a small variation in hardness was measured.

Application of the elastic continuum theory of dislocations to modeling of phenomena associated with elastic strain relaxation in strained layer epitaxial heterostructures is discussed. The concept of critical thickness for onset of strain relaxation in a strained epitaxial layer is first reviewed, and some extensions to periodic arrays of dislocations and to multiple layers are described. Then, two issues are addressed that arise when the assumptions underlying the critical thickness concept are not met. One issue concerns the nucleation of dislocations at the growth surface of an epitaxial film, particularly the influence of surface irregularities on the activation energy for surface nucleation. A second issue concerns the kinetics of glide of a threading dislocation as it lays down an interface misfit dislocation when the layer thickness exceeds the critical thickness. A generalized driving force for the glide process is defined, and a relationship between this force and the glide speed is proposed.