A review of common stress measurement techniques as related to CVD diamond is presented. A correction to the commonly applied Stoney thin film equation for the substrate curvature technique is presented for coatings with a large stiffness mismatch with the substrate. The difficulties associated with the sin2ψ and peak width broadening X-ray diffraction techniques in highly oriented, X-ray transparent diamond are highlighted. Stress measurement by Raman spectroscopy is complicated by factors such as domain size, temperature, non-hydrostatic stress, degeneracy lifting and the presence of multiple peaks. Substrate curvature and Raman spectroscopy stress measurement results can be reconciled through consideration of these complicating factors and application of the appropriate corrections.

Films of a-Si:H were deposited by dual ion beam sputtering using a new configuration in which both the argon and hydrogen beam sources are directed at the silicon target. This geometry also permits independent control of the hydrogen and argon energy and particle flux. Infrared absorption measurents show that even for high hydrogen concentrations, the 2000 cm−1 Si-H stretching band is dominant. This result is in contrast with the more conventional configuration in which the H source is directed at the substrate, resulting in films with dominant 2100 cm−1 mode. This suggests that the precursors resulting in H-incorporation are different for the two configurations. In fact, IR reflectance and SIMS analysis of the silicon sputtering target reveal hydrogen is incorporated, peaking at about 30 Å below the target surface. A strong increase in the photo and dark dc conductivity occurs as the hydrogen ion enery is reduced below 30 eV, suggesting the importance of preventing high energy back-scattered H ion bombardment of the film. At a H ion energy of 8eV, the values are 2×10−5 (AM1) and 2×10−9 (ohm-cm−1), respectively. Spectroscopic ellipsometry measurements of films reveal a Si-Si bond packing greater than that of low H-content a-Si prepared by LPCVD even up to H contents as high as 24%. Above 25% a microstructural transition is observed, verified by SEM, resulting in an increase in the density of voids, (which appears to be responsible for a sudden drop in the hydrogen-induced compressive stress) and accompanied by a shift in the dominant stretching mode energy.

The crystallinity for silicon and germanium films deposited by ion beam sputtering (IBS) as a function of substrate temperatures was determined using Raman spectroscopy, spectroscopic ellipsometry, electrical conductivity and x-ray diffraction measurements. The results show that IBS silicon crystallizes between 300–350°C while germanium crystallizes between 20–200°C. Reasonably good agreement is obtained among the four distinctively different characterization techniques in identifying the onset of crystallinity. A direct relationship is observed between the substrate temperature required for crystallization and the log of the operating pressure for various deposition techniques. Energetic particle stimulation during film growth appears to reduce the crystallization temperature at a given operating pressure. Raman data show that the crystallization temperature depends on the deposition rate. A graded structure is observed in films deposited above 300°C, probably due to oxygen contamination.

Because of the unfavorable mechanical properties of diamond. the source and type of stresses present must be identified and controlled at every stage of the CVD diamond deposition process in order to develop a high-yield, large-wafer-scale fabrication process. In this paper, we report on the types of defect and stress encountered in free-standing CVD diamond films deposited by d.c. are jet deposition. Substrate materials and designs, temperature non-uniformity, and process conditions all contribute to the type and magnitude of stresses present in a film. Results of mechanical strength testing by biaxial flexure are presented. Strength sample sets consisted of 20 disks from each diamond wafer. Results from over 50 wafers are summarized. Weibull analysis of strength data is presented, with a description of the effects of thickness and sample orientation. It is shown that by implementing tight process control and through judicious selection of a substrate material, d.c. are jet technology is capable of producing 175 mm diameter wafers at high yield with uniform properties. (C) 1999 Elsevier Science S.A. All rights reserved.

Finite element modeling suggests that the thermal performance of plastic and ceramic packages could be significantly improved through the insertion of CVD diamond substrates. The model was formulated by considering the thermal properties, dimensions, and spatial locations of the materials comprising the dominant conductive thermal path. Optimized designs were selected, targeting the minimization of die junction temperature, package maximum temperature, and package temperature gradients through the reduction of the heat source to heat sink thermal resistance. Selected designs were fabricated and thermally evaluated using infrared thermometry. Diamond-enhanced package designs using leadframe-substrate "overlaps" for plastic SOIC packages and through-flange "inserts" for ceramic power packages realized junction temperature decreases of greater than 50%.

Long term package reliability is becoming an increasing concern as IC clock speeds and power densities increase while simultaneously the package foot print decreases. Significant contributors to increasing the reliability and performance of plastic packages are reducing the package/junction temperature Tj and improving the integrity of the interfaces between the encapsulating polymer and the constituent package materials. CVD diamond, with its superior thermal conductivity, low density and high electrical resistivity offers an attractive alternative to traditional metal heat spreader materials to insuring a low junction temperature and therefore long time between failures. Considering these goals and the ubiquitous performance, reliability, and economic requirements, GaAs devices were encapsulated in a plastic package with CVD diamond substrates as the thermal management platform. Both GaAs flip-chip with polymer underfill and conventional face up chip-wire configurations were evaluated. This paper will highlight the effect of chemical termination of the CVD diamond surface on encapsulant polymer adhesion using x-ray photoelectron spectroscopy, surface energy measurements, and boiling water immersion/peeling tests. Additionally, the enhanced thermal and reliability (mechanical/environmental) performance of plastic packaged GaAs devices using “optimally terminated” CVD diamond will be discussed. [S1043-7398(00)00802-1]

This paper presents the progress made in developing a production-worthy 7 inch diameter CVD diamond wafer fabrication process by DC arc jet. Technical issues associated with fabricating diamond wafers are presented, along with performance and repeatability data from a pilot plant production run

The use of diamond as a thin film or a coating enables us to exploit the wear resistance properties of the thin film while taking advantage of the bulk mechanical properties (toughness) of the underlying substrate base material. However, in order to achieve this goal in practice, it is important to optimize both the physical properties of the diamond film and the properties of the film/substrate interface. Early efforts to produce diamond-coated WC Co products did not improve the adhesion strength to a level that provides a diamond-coated material of sufficient utility to the end user. In this paper we describe the essential characteristics of the diamond/substrate interface which, when acting together, have recently led to significant improvements in the adhesion strength of the diamond film to the underlying substrate. These characteristics fall into three general categories: a) chemical composition, b) phase composition, and c) microstructural composition. All of these characteristics are produced using essentially one process step that was developed to produce these characteristics in such a way that other important physical features are not compromised. Finally, we will review the results of efforts to optimize the physical characteristics of the diamond deposits for metal cutting applications as well as overview results of worldwide testing of these diamond-coated WC-Co based cutting tool inserts.

As the clock speed and chip power density continue to increase, heat management is becoming an ever-increasing challenge for package designers. In addition, the demand for a narrow profile in consumer products and compact footprint may disqualify traditional heat dissipation materials as viable solutions. Diamond is a premier candidate for this application because of its unmatched thermal conductivity and low density. However, the outstanding issue is cost and availability. This paper discusses de are jet deposition technology and the stress issues that must be addressed to develop a reliable fabrication process. By adopting tight process control and matching the substrate's thermal properties to the film, a crack-free 175-mm-diameter diamond wafer fabrication process has been developed with uniform properties and wafer yields that exceed 90% for thicknesses greater than 200 mu m.

We have made carbon K edge reflectivity and absorption measurements using synchrotron radiation on diamond crystals and chemical vapor deposited diamond films to determine their electronic structures. Our spectra of diamond films show that both sp2 and sp3 bonded carbon atoms are formed during initial nucleation and growth. Transmission spectra of a diamond film with 30 nm diameter cystallites show striking features below the carbon sp3 K edge due to internal interface states and/or defects. We compare these absorption features to x-ray absorption spectra of clean diamond (111) surface, graphite, and hydrocarbon gases to understand surface chemistry involved in the deposition process.

Microwave plasma chemical vapor deposition was used to deposit diamond films on polished silicon. The substrate was back-etched to form taut free-standing membranes up to 75 mm in diameter. Raman spectroscopy and X-ray diffraction verify that the films are diamond. Transmission electron microscopy and X-ray diffraction reveal that the films are fine-grained polycrystalline diamond with an average grain size between 15 and 110 nm, depending on methane fraction. Surface roughness measurement by atomic force microscopy and a stylus-type profilometer shows that the film roughness scales inversely with the grain size, which is a function of the methane fraction in the feed gas. The optical transmission at the HeNe laser line is 58% for a 1 μm membrane, uncorrected for reflection losses from both surfaces, absorption and scatter. The transmission in the blue is sensitive to the methane fraction in the gas stream. The hydrogen content in the films was measured by nuclear reaction analysis and elastic recoil detection spectroscopy. No correlation was observed between hydrogen content in the film and deposition temperature, however, a direct correlation was observed with the methane fraction. Tensile and compressive total (thermal and intrinsic) stress were observed, depending on the deposition conditions. The thermal stress is compressive and relatively constant (0.215–0.275 GPa) over the temperature range investigated. The intrinsic stress is tensile and its origin is interpreted in terms of the grain boundary relaxation model. The biaxial Young's modulus and the ultimate tensile strength of free-standing membranes were measured. Bulk values were observed at low methane fraction. Both properties decrease with increasing methane fraction due to sp2 carbon and hydrogen incorporation.

A review of the literature reveals that the intrinsic stress in sputtered thin films can be tensile or compressive depending on the flux and energy of particles striking the film. Stress data indicates that the normalized momentum P*n=γ√ME (where γ is the energetic particle/atom flux ratio, M the mass, and E the energy) may be the appropriate stress scaling factor. The forward sputtering model predicts a √E dependence. An idealized stress–momentum curve is constructed consisting of three regimes:(1) a region of increasing tensile stress at low P*n, due to a porous microstructure, followed by (2) a sharp transition from tensile to compressive stress at intermediate momentum, accompanied by a conversion to zone T-type microstructure and (3) a saturation region at high momentum, due to plastic flow. For any deposition process the sign and magnitude of the stress depends on P*n, which is a function of several deposition parameters. Calculations indicate that stress reversal in sputtered and ion-assisted evaporated films occurs at normalized momentum P*n of about 5–15×10−23 kg m/s/atom (Pn=0.3–1 √eV/atom).

Real time spectroscopic ellipsometry from 1.5 to 4.0 eV has been applied to monitor surfaces of optical quality diamond thin films during exposure to low energy Ar and H ion beams, and atomic H. Monolayer-level sensitivity to the conversion of diamond to optically absorbing sp2 bonded carbon upon ion impact is demonstrated. For Ar ions, a beam voltage of 50 V is found to be sufficient to convert the surface to the equivalent of a single monolayer of sp2 bonded carbon, whereas the topmost 12–16 Å is converted at the threshold for sputter etching, between 100 and 150 V. Atomic H and low energy H ion treatments (≤50 V) are found to eliminate optically absorbing defects within 15 Å of the surface. The overall results provide optimism that real time spectroscopic ellipsometry will be a highly sensitive real time probe of diamond surface and bulk microstructure in the reactive environments of enhanced chemical vapor deposition.

Quantitative characterizational methods are required to optimize vapor-deposited diamond thin films for optical applications. In this work, spectroellipsometry has been applied to deduce two important characteristics of diamond films: the volume fraction of sp2-bonded defects in the bulk and the thickness of the roughness layer on the surface. We have determined these characteristics versus substrate temperature and CH4:H2 flow ratio for optical quality films prepared to 1000–4000 Å by microwave plasma-assisted chemical vapor deposition. Under optimum conditions, uniform films with ∼100 Å roughness and 3 vol.% bulk sp2C are obtained.

The stress in diamond films prepared by microwave plasma CVD was investigated as a function of methane concentration (0.2%–3.0%) and deposition temperature (600–900 °C). Tensile and compressive total (thermal and intrinsic) stress were observed, depending on the deposition conditions. The thermal stress is compressive and relatively constant (0.215–0.275 GPa) over the temperature range investigated. The intrinsic stress is tensile and its origin is interpreted in terms of the grain boundary relaxation model. Calculations indicate a value of 0.84 GPa, using the grain boundary model, which is in fair agreement with the measured value. For the methane series, the tensile intrinsic stress decreases with increasing the methane fraction. The increasing compressive stress is ascribed to increased impurity (hydrogen and nondiamond phase carbon) incorporation with increasing methane fraction. 15N nuclear reaction analysis shows a linear correlation between hydrogen in the film and methane in the supply gas while spectroscopic ellipsometry shows a direct correlation between optically absorbing nondiamond (sp2) carbon incorporation and methane. For the temperature series, the intrinsic tensile stress increases with deposition temperature. The increase is ascribed to decreasing sp2 C incorporation with temperature, as confirmed by spectroscopic ellipsometry measurements.

Microwave plasma chemical vapor deposition was used to deposit diamond films on polished silicon and then the substrate was back-etched to form taut free-standing membranes up to 75 mm diam. Raman spectroscopy and x-ray diffraction verify that the films are diamond. Transmission electron microscopy and x-ray diffraction reveal that the films are fine-grained polycrystalline diamond with an average grain size between 15 and 110 nm. The films contain microcrystalline graphite and nondiamond carbon as indicated by the broad Raman bands at 1355 and 1560 cm−1, respectively. The measured x-ray transmission of a 3.25-μm-thick membrane at the Cu Kα line is 99.2% while the optical transmission at the He-Ne laser line is 58% for a 1-μm membrane, uncorrected for reflection losses from both surfaces, absorption, and scatter. The transmission in the ir is at the theoretical limit, 71%. The surface roughness of a typical 3-μm-thick membrane is about 30 nm on the side away from the silicon substrate and 20 nm on the side adjacent to the substrate. The surface roughness scales inversely with methane fraction in the supply gas due to decreasing grain size. The typical membrane stress is 0.03–0.15 GPa. The films contain between 0.8% and 7.0% hydrogen, as determined by 15N nuclear reaction analysis. No correlation was observed between hydrogen content in the film and deposition temperature; however, a direct correlation was observed with methane fraction. The average in-plane distortion is 70 nm for a 1.5-μm-thick diamond mask with 0.2-μm-thick tungsten absorber. No measurable out-of-plane distortion was observed for evaporated gold absorber metallization while tungsten produced 0.8 μm distortion.

Distortion of x-ray masks remains an important issue in the development of x-ray lithography. In x-ray masks employing the embedded absorber structure, the absorber material is deposited selectively in the trenches at an elevated temperature. Thermal stresses develop upon cooling the mask due to the mismatch in the coefficients of linear expansion between the absorber and the membrane resulting in distortions. Such x-ray masks were fabricated with 0.45 μm thick tungsten absorber and 2.5 μm thick silicon membrane; the tungsten was deposited by chemical vapor deposition (CVD) at 300 °C. The resulting in-plane-distortions (IPD) were experimentally measured by using a special purpose mask consisting of equal lines and spaces and a large array of fine crosses; the IPD was found by measuring the locations of these crosses at the beginning and the completion of the mask processing. It was found that the maximum IPD is 0.27 μm over a 7.5×7.5 mm2 square die for a typical value of fractional absorber coverage (FAC) of 50%. The results are in agreement with the calculations obtained from a previously developed analytical model suggesting that the distortions are entirely due to thermal stresses, predictable and correctable. Large IPD can be greatly reduced by using a membrane whose Young’s modulus is much larger than that of the absorber. A conventional mask was fabricated by patterning 0.2 μm thick evaporated tungsten on a 1.5 μm thick diamond membrane; it was found that the distortions were reduced to less than 0.07 μm over a 7.5×7.5 mm2 square die in spite of the large intrinsic stress (500 MPa) in the evaporated absorber.

The intrinsic stress in ion beam sputtered Fe, Si, and AlN films was investigated as a function of deposition temperature. At low deposition temperature (300 K) the intrinsic stress is high and compressive for all materials investigated. With increasing deposition temperature the stress initially decreases gradually followed by a rapid drop for Td >Tm /3. The initial slope of the stress variation with temperature appears to be material dependent, being steepest for iron and shallowest for AlN with silicon falling in between. Unlike AlN and Si, Fe converts from compressive to tensile stress at Td /Tm =0.25, where Tm is the absolute melting temperature. The inverse temperature dependence of the intrinsic stress is explained in terms of the grain growth and annealing of a strained microstructure models. Applying the forward-sputtering model [which relates the intrinsic stress to the elastic energy/mole, EM/(1−ν)D, where E is Young’s modulus of elasticity, D the mass density, M the atomic mass, and ν Poisson’s ratio] it is shown that the temperature dependence of the material properties do not contribute significantly to the intrinsic stress variation with temperature.