Solvent induced damage bands formed in residually strained polyimide thin films on different substrates have been studied. Microscopy studies have shown that these bands resemble crazes. A mechanics approach Is taken to understand the band formation phenomenon.

The critical strain for damage formation has been identified. This strain decreases with increase in exposure time, but always exhibits a threshold value. In contrast to the cracking of brittle films, the critical strain has only a weak dependence on the film thickness over a wide range. This behavior obtains because the crazing of the polyimide films is nucleation controlled. Strain-enhanced diffusion of solvent into the films is considered to be responsible for the property degradation that leads to damage formation.

Relationship between the degree of imide-ring-condensation and the absorbed quantity of sodium ion was studied in a polyimide film of 100nm thick immersed in an alcoholic solvent containing sodium ion in order to elucidate a mechanism of incorporation of ionic species. Absorption and depth-profile of sodium ion in a film were quantitatively determined by chemical analysis and secondary ion mass spectroscopy. Concentration of sodium ion in a thin film by two orders of magnitude was observed, and distribution of sodium ion in a thin film was found to be uniform, which cannot be accounted for by any kinds of previous models based on diffusion, permeation, or solvation. Instead, chemical ion-exchange model is proposed and is found to satisfactorily explain the experimental results.

This paper reviews some of the research efforts in microelectronic packaging to develop composites with dimensions of the fillers ranging down to nanometer dimensions. Some initial experiments have been conducted with nanocomposites, where the thermal and dielectric properties of polymer/fumed silica composites have been studied. The bulk samples were prepared in different ratios of polymer to silica using polydimethyl siloxane and Cab-O-Sil. Differential thermal analyses carried out on these samples showed no changes in the samples till 330°C. The dielectric constant of the 70:30 (Cab-O-Sil- Polymer by Wt.%) composite is 3.8, similar to that of the pure silica. For the same composite the value of the thermal expansion coefficient is 350 ppm, closer to the value of pure polymer (439 ppm). It is suggested that the key to model these nanocomposites with ceramic and polymer phases is to understand the interfaces and their chemical bonding.

There is considerable recent interest in the dielectric properties of materials measured over broad electrical bandwidths. This follows from advances made in the performance of electronic devices which now produce pulses with risetimes in the order of 5 picoseconds. It is essential to know the dielectric constants and loss properties of the materials used in fabricating these devices and their interconnection structures, up to frequencies of about 100 GHz. Moreover, such information will be invaluable for the effective use of microwaves in materials processing [1]. The curing of polymers and the sintering of ceramics are two examples. Here, we discuss the recently developed coherent microwave transient spectroscopy (COMITS) technique which measures the complex dielectric constants of materials, from 10 GHz to about 125 GHz, in a single experiment [2,3].

A new generation of substrate materials with low dielectric permittivity have been developed. These new ceramic-based nanocomposites consist of a quasi two-dimensional, multilayered structure realized by the intercalation of molecular species into the galleries of a host lattice. The materials have been characterized by x-ray diffraction, electronic and infrared spectroscopy. The relative permittivities vary between 2.5 and 4 at 1 MHz. The new hybrids can be processed at relatively low temperatures (below 350°C) and allow for step coverage of submicron structures.

The thermal diffusivities and specific heats of several polymer matrix composites containing ceramic Fiber FP [1] and AIN fibers were measured between 20 K and 300 K. The anisotropic thermal conductivities inferred are related to specific fiber geometries and the thermal conductivities of the composite components. The high in-plane conductivities at 125K suggest possible applications for high-Tc, superconductor electronic packaging.

A study was conducted on the preparation of Al2 O3 -coated metal substrates for high density electronic assembly. Al2 O3 films of 3∼15μm thickness were deposited on metal substrates, such as aluminum, stainless steels and the like, at a deposition rate of 0.2∼5μm/min. They were examined in terms of chemical composition, microstructure, density, hardness, breakdown voltage, thermal conductivity and so on. It was found that the dense and stoichiometric films could be achieved by adjusting the process parameters, in particular, substrate temperature, deposition rate and distance between target and substrate. The results obtained led to the conclusion that this new Al2O3-coated metal substrate could be effective for high density electronic assembly.

A new silver lead phosphovanadate glass die attach material is discussed. Its sintering characteristics are examined by dilatometry and the organic system by thermo gravimetric analysis. Sintering is found to occur at lower temperature than conventional silver glass systems. This is caused by reaction between silver and one of the glass components during heat treatment. The vehicle is found to evaporate faster allowing void free processing for large area dice.

The new glass combined with a more volatile organic system has produced a single pass die attach material which can be fired as low as 360°C.

The possibility of fast firing a relaxor dielectric to suppress grain growth, mitigate the lead loss and reduce the cost of sintering has been investigated. The relaxor Pb(Mn1/3 Nb2/3).02 (Mg1/2 W1/2 ).48 Zr.23 Ti.27 O3 [1] exhibits a highly desirable broad phase transistion range when the prbcessing procedures are carefully controlled and fired conventionally. When fast fired, this material's density, percentage of secondary phases, and dielectric constant were highly dependent on the sintering time. A sample fired for 6 minutes possessed a dielectric constant nearly equal to the same material conventionally fired for 16 hours.

The status of the publicly available knowledge base for aluminum nitride and its processing has been briefly reviewed in an attempt to clarify future research needs. Driving forces for the development of a scientific understanding of the property/process relationships have thus far been focused on thermal conductivity. Future developments may well be aimed at (1) retention of presently available properties while attempting to lower ceramic processing temperatures and (2) understanding metal-ALN and polymer-ALN interfaces. Progress has thus far been sufficient to define first generation ceramic powder requirements and sintering conditions to give adequate properties for packaged device evaluation. In addition, metal-ALN interfacial bonding, although not well developed, has shown sufficient promise to insure long-term interest in this material.

We have initiated a study of the ac electrical properties (dielectric constant and dielectric loss) and the dc resistivity of sintered AIN. For this study, we have used samples purchased commercially as well as samples made by the Carborundum Company, a member of the BP Group. Variations are observed in the room temperature dielectric constant for AIN substrates obtained from different sources. The source-to-source range is from 8.5 to 8.7. This range is lower than the 8.8 to 9.2 that is generally reported in the product literature. Similar source-tosource variations in the room temperature dielectric loss are observed. All of the samples show thermally activated behavior in the dc resistivity. The activation energy observed for all of the samples is similar, EA ∼0.47 eV; however, different pre-exponential factors are apparent. As a result, there are large variations in the measured room temperature resistivities for samples from different sources.

The thermal conductivity of ceramic materials used for IC substrates and packages has increased in importance as chip sizes have decreased and heat loads have risen. AIN which has a room temperature (RT) thermal conductivity (λ) greater than 200 W/m·K and BeO with λ(RT) ∼260 W/m·K are the major candidates for applications demanding high conductivity. Conflicting reports of the temperature dependences of λ for these materials over the range of interest for packaging use (≤200°C) have been published, with some reports suggesting a crossover in λ. These reported differences may be due to the reported problems in measuring λ in AIN using the flash diffusivity method. For the present experiments, we have used a new long wavelength laser flash diffusivity system which has been shown to determine thermal diffusivity to better than ± 3% for AIN with sample thicknesses ranging from 0.3 mm to 5 mm. No absorbinq coatings are required and no correction factors are needed to fit the data to theory. We report λ from room temperature to 400°C for AIN from a number of commercial sources, and for BeO and SiC. At room temperature, BeO has the highest thermal conductivity, but as the temperature is raised, the values for BeO and AIN approach one another, with crossover observed at about 350°C for the highest conductivity AIN sample studied. Recent steady state heat flow measurements agree with our thermal conductivity values rather than with previous literature values.

The materials studied in this research, having a wide range of thermal conductivities, permit a study of the mechanisms which affect their behavior. Data on the thermal conductivity of several substrate materials, over a wide temperature range, were analyzed using the Klemens model. Parameters in this model, which include crystallite size and impurity concentration, are determined through a nonlinear least squares fitting routine and related, where possible, to values obtained by other techniques. This analysis predicts an achievable upper limit of 290 W/m.K for very high purity sintered AlN with 5 micron crystallites.

The thermal conductivity of aluminum nitride is significantly affected by its purity. We used CaCO3 as an additive to decrease the oxygen content. Calcium-aluminate is formed by the reaction of the aluminum nitride powder oxide layers and CaO during firing. We investigated the influence of firing gas pressure on the microstructure and thermal conductivity of aluminum nitride. The thermal conductivity reached 270 W/m°K when 4 wt% CaO was added and firing gas pressure was 700 kPa. Under these conditions, the amount of remaining oxygen was at its minimum and the grain size of the aluminum nitride was at its maximum.

A charged defect model for the formation of band-tail states, first proposed by Redfield et al. to understand the optical properties of Si doped GaAs [1], is applied to oxygen doped aluminum nitride (AIN) ceramics. In this case, oxygen substituting on a nitrogen site and (subsequently formed) aluminum vacancies comprise the charged centers. The results of optical absorption edge measurements as well as steady state and time-resolved luminescence between 10 and 300K are used to support this model. Finally, it is noted that the mixed covalent and ionic nature of bonding in AIN make it a model system for this charge defect picture, and thus explains the extremely large extent of the band-tail states in this material.

Aluminum nitride has a great potential as a semiconductor substrate material. However, the defects which occur in this material are very detrimental to the material properties. ‘Domelike’ or ‘D’ defects, consisting of a flat basal fault joined to a curved planar fault, are examined by TEM. They are shown to be inversion domains, using a newly developed technique for polarity determination in the TEM. Structural models are proposed for the defects which incorporate a layer of alumina one unit cell thick into the basal fault of the defect.

The microstructure and microchemistry of planar and curved defects in Aluminum Nitride (AIN) has been investigated using Conventional Transmission Electron Microscopy (CTEM), Convergent Beam Electron Diffraction (CBED), and Analytical Electron Microscopy (AEM) techniques. Both defect morphologies were identified as Inversion Domain Boundaries (IDB). Microchemical analysis revealed oxygen segregation to the planar faults; when present on the curved defects, oxygen was at a lower concentration than in the planar defect case. Annealing experiments on defect containing AIN support our microchemical analysis of oxygen segregation. A proposed model for the formation of these two types of boundaries is presented.

Aluminum nitride is known to have a large affinity for oxygen as an impurity. At high levels (>∼4 wt/o) the oxygen is incorporated in the form of planar stacking faults where “pure” 2H AIN is regularly interspersed with a layer of oxygen at the faults. At oxygen levels lower than ∼ 4 wt/o the structure shows an expanded c-axis. The present authors have not observed this effect, rather a random distribution of stacking faults is observed along with another, more prevalent, extended defect identified as an inversion domain boundary (IDB). The IDBs are significantly aplanar (indicating a low interface energy), and often have precipitates and other, faceted defects associated with them. The role of these defects in oxygen accommodation in AIN has been investigated both structurally and chemically by electron optical methods (SEM, TEM, STEM, HREM, CBED, EDS, EELS, and CL-TEM). The structural nature of the boundaries, in the absence of oxygen, requires Al-Al or N-N bonding to occur with some frequency across the boundary. Such bonding is unlikely due to the excess energy required. Chemical analysis (EELS) and luminescence studies (CL-TEM) reveal that oxygen is often associated with the boundaries and may mediate the bonding at the boundary. A model is proposed for the IDB which includes structural aspects combined with considerations of stoichiometry in an effort to understand the origin and energetics of this defect.

Aluminum nitride (AlN) films have been deposited by reactive DC planar magnetron sputtering at a low substrate temperature of 80°C. An Al disk and either N2 or NH3 were used as a target and an ambient gas, respectively. In films deposited in N2, the c-axis of AlN crystallites was perpendicular to the film plane. On the other hand, in NH3, X-ray diffraction peaks from (100), (110) and (200) were observed. The surface and the cross-sectional microstructures of AlN films deposited in N2 were significantly different from that deposited in NH3. The surface of films deposited in NH3 was smoother than that deposited in N2. The transmittivity of films deposited in NH3 was superior to that deposited in N2 in wavelength of 275∼450 nm. Resistivities of films deposited in N2 and in NH3 were 1014sim;1015 Ω · cm. Dielectric constants of films deposited in N2, and of that deposited in NH3, were 9.7, and 9, respectively, in a frequency range of 20 kHz to 20 MHz. The structure and the properties of AlN films deposited by this method depended on kind of ambient gas.

A technique for brazing aluminum nitride (AIN) using conventional (non-active) brazing alloys has been investigated. The process involves the in-situ decomposition of a metal hydride. This process alters the surface chemistry of the substrate and improves the wettability of the molten braze. The development of high strength bonding between braze and ceramic results. The ceramic-braze interface was studied using scanning electron microscopy (SEM). The nature of the interfacial reactions and the reaction products have been identified using x-ray diffraction (XRD). The progress of the reaction has been followed using differential thermal analysis (DTA).

The experimental results have been correlated with thermodynamic predictions of the reaction process. In addition to joining ceramic to ceramic, braze joints of AIN to a low expansion iron-nickel lead frame alloy were made.