scispace - formally typeset
Search or ask a question

Showing papers in "Ibm Journal of Research and Development in 1969"


Journal ArticleDOI
TL;DR: A new, computer-generated, optical element called a kinoform that operates only on the phase of an incident wave and forms a single image by wavefront reconstruction without the unwanted diffraction orders characteristic of holograms is described.
Abstract: A new, computer-generated, optical element called a kinoform is described. This device operates only on the phase of an incident wave and forms a single image by wavefront reconstruction without the unwanted diffraction orders characteristic of holograms. The kinoform is also more efficient in the use of spatial frequency potential and reconstruction energy and can be synthesized in considerably less computer time than the digital hologram.

636 citations


Journal ArticleDOI
TL;DR: A new sequential decoding algorithm is introduced that uses stack storage at the receiver that is much simpler to describe and analyze than the Fano algorithm, and is about six times faster than the latter at transmission rates equal to Rcomp.
Abstract: In this paper a new sequential decoding algorithm is introduced that uses stack storage at the receiver It is much simpler to describe and analyze than the Fano algorithm, and is about six times faster than the latter at transmission rates equal to Rcomp the rate below which the average number of decoding steps is bounded by a constant Practical problems connected with implementing the stack algorithm are discussed and a scheme is described that facilitates satisfactory performance even with limited stack storage capacity Preliminary simulation results estimating the decoding effort and the needed stack siazree presented

635 citations


Journal ArticleDOI
K. C. Norris1, A. H. Landzberg1
TL;DR: In this paper, an analysis of the mechanics of the system, a model to establish the relationship among different fatigue testing conditions, and experimental verification of the model were presented. And the chip failure rate of the interconnection as used in present designs was predicted to be better than 10-7%/1000 hours for the mechanism studied.
Abstract: The use of solder pads to join multi-pad integrated circuit chips to modules provides a highly reliable, rugged interconnection technology. This paper reports some important aspects of the reliability evaluation that was carried out on the "controlled chip collapse" interconnection system developed by IBM. Included are an analysis of the mechanics of the system, a model to establish the relationship among differentt hermal fatigue testing conditions, and experimental verification of the model. In the course of this work, the chip failure rate of the interconnection as used in present designs was predicted to be better than 10-7%/1000 hours for the mechanism studied.

433 citations


Journal ArticleDOI
TL;DR: Computer holograms differ from a normal hologram in that the transmittance is binary, yet they are able to construct general wavefronts and images efficiently and have several practical advantages over holograms with a continuous range ofTransmittance.
Abstract: Holograms synthesized by computer are used for constructing optical wavefronts from numerically specified objects. Elimination of the need for a physical object has made new applications possible, for example, three-dimensional computer output displays, synthetic prototypes for interferometric testing, and filters for various optical data processing operations. Our computer holograms differ from a normal hologram in that the transmittance is binary, yet they are able to construct general wavefronts and images efficiently and have several practical advantages over holograms with a continuous range of transmittance. Recent improvements that simplify the production of binary holograms and improve their performance are described and experimental work showing reconstruction of two- and three-dimensional images is presented.

382 citations


Journal ArticleDOI
L. F. Miller1
TL;DR: In this paper, the authors describe a technique that prevents these solder pads from collapsing and permits large scale production by limiting the solderable area of the substrate lands and chip contact terminals so that surface tension in the molten pad and land solder supports the device until the joint solidifies.
Abstract: Solder reflow connection of semiconductor devices to substrates has been shown to be a reliable, effective, and readily automated technique. Rigid copper spheres, which remain rigid during solder reflow, have been used successfully for some timae as major element of the contact joint. However, to expand the capability of such joints to larger devices such as multiple-transistor chips in hybrid components, ductile metallic joining pads can be used on the devices instead of the copper spheres to reduce mechanical strains andp ermit multi-pad devices to make proper contactt o module lands during reflow. This paper describes a technique that prevents these solder pads from collapsing and permits large scale production. Termed "controlled collapse," the method is based on limiting the solderable area of the substrate lands and chip contact terminals so that surface tension in the molten pad and land solder supports the device until the joint solidifies. The result is a sturdy, testable connection of high reliability (bond strength30-50 gm, pilot-production yields exceeding 97%, predicted failure rate-based on laboratory tests-considerably lower than that of copper ball contacts). The process is economically adaptable to automation and offers considerable latitude in fabrication and control tolerances.

357 citations


Journal ArticleDOI
P. A. Totta1, R. P. Sopher1
TL;DR: The glass-passivated, face-down semiconductor chip joining technology employeidn IBM's SLT (Solid Logic Technology) has become not only a fundamental element in the hybrid circuitry of System/360 but also the basis for later metallurgical designs as mentioned in this paper.
Abstract: The glass-passivated, face-down semiconductor chip joining technology employeidn IBM's SLT (Solid Logic Technology), has become not only a fundamental element in the hybrid circuitry of System/360 but also the basis for later metallurgical designs. The "flip-chip," copper ball terminal, solder reflow technique is comprehensively reviewed and a discussion is given of its extension, through the use of ductile, all-solder terminals, to monolithic applications.

233 citations


Journal ArticleDOI
Dennis Gabor1
TL;DR: In this article, a more general analogue is discussed and two two-step transformations that imitate the recording-reconstruction sequence in holography are presented, where the first transformation models the recall of an entire sequence from a fragment while the second is more like human memory in that it provides recall of only the part of the sequence that follows the keying fragment.
Abstract: Recently Longuet-Higgins modeled a temporal analogue of the property of holograms that allows a complete image to be constructed from only a portion of the hologram. In the present paper a more general analogue is discussed and two two-step transformations that imitate the recording-reconstruction sequence in holography are presented. The first transformation models the recall of an entire sequence from a fragment while the second is more like human memory in that it provides recall of only the part of the sequence that follows the keying fragment. Both models require only the three operations: shift, multiplication and addition.

186 citations


Journal ArticleDOI
Lewis S. Goldmann1
TL;DR: In this paper, the authors deal with the mechanical reliability of controlled collapse joints in modules subjected to the thermal fatigue conditions of machine usage, focusing on design variability and how the shape and dimensions of the joint and chip affect reliability.
Abstract: This paper deals with the mechanical reliability of controlled collapse joints in modules subjected to the thermal fatigue conditions of machine usage. Particular emphasis is placed on design variability and how the shape and dimensions of the joint and chip affect reliability. A systematic technique is presented to optimize pad dimensions. A new experimental method to characterize chip-to-substrate interconnections-the torque test-is described and analyzed. Its applicability to design evaluation is discussed and representative data are analyzed. The relationship between torque test measurements and fatigue is discussed.

182 citations


Journal ArticleDOI
D. P. Kennedy1, R. R. O'Brien1
TL;DR: In this paper, it was shown that the profile inferred from differential capacitance measurement of semiconductor junctions is not that of the impurity atom distribution but, instead, the majority carrier distribution.
Abstract: In a recent paper, 1 it was shown that the profile inferred from differential capacitance measurement 2–5 of semiconductor junctions is not that of the impurity atom distribution but, instead, that of the majority carrier distribution. For this reason, conventional differential capacitance measurements can be used to evaluate the impurity atom distribution only in charge neutral semiconductor material (where the majority carrier density equals the density of ionized impurity atoms). This requirement of charge neutrality limits the applicability of this measurement technique to semiconductor material containing a minimum impurity atom density of about 10 16 atoms/cm 3 .

100 citations


Journal ArticleDOI
TL;DR: In this article, the high-field electron distribution function for GaAs, its small-signal frequency response and its behavior in large sinusoidal electric fields were analyzed and the response speed is limited by the low scattering rate within the (000) valley.
Abstract: Numerical calculations have been made of the high-field electron distribution function for GaAs, its small-signal frequency response and its behavior in large sinusoidal electric fields-The response speed is limited by the low scattering rate within the (000) valley. With increasing frequency the threshold field for negative conductivity rises and the negative mobility and oscillator efficiency fall. The free-electron dielectric constant is positive at high fields, with a peak near the threshold field.

100 citations


Journal ArticleDOI
Jitendra V. Dave1
TL;DR: In this paper, two subroutines with which one can compute the various characteristics of the electromagnetic radiation scattered by an absorbing, homogeneous sphere of any reasonable size are presented.
Abstract: Details are provided for two subroutines with which one can compute the various characteristics of the electromagnetic radiation scattered by an absorbing, homogeneous sphere of any reasonable size. The necessary expressions for this purpose were first derived byM ie. The method of computations used is the so-called method of logarithmic derivative of one of the complex functions, introduced by Infeld. The main difference between the two subroutines is in the procedure used in computations of one of the functions. This function is computed by an upward recurrence procedure in one subroutine and by a downward recurrence procedure in the other. Sufficient results for demonstrating the reliability of these programs are presented and discussed for a sphere of 10µm radius illuminated by an unpolarized radiation of 0.4µm wavelength.

Journal ArticleDOI
TL;DR: In this article, a fourfold classification of current-controlled instabilities in amorphous semiconductors is proposed, and the experimenta evidence supporting a simple band model for the Amorphous covalent alloys is given.
Abstract: A four-fold classification of the current-controlled instabilities in amorphous semiconductors is proposed. The experimenta evidence supporting a simple band model for the amorphous covalent alloys is given. The present understanding of the reversible switching effects and of the switching with memory is discussed.

Journal ArticleDOI
William B. Pennebaker1
TL;DR: In this article, two dielectric loss peaks are believed to be caused in part by an oxygen deficient region near one electrode, and the variation in the dielectrics constant K with electric field is similar to that observed in bulk material.
Abstract: Two deposition parameters are important in rf sputtering of SrTiO3 films: the oxygen-argon content ratio, and the substrate temperature. More than 1% oxygen is needed to produce insulating films; the exact percentage required depends on system cleanliness. Both dielectric ccnstant and crystallite size increase with increasing substrate temperature. Films of 2400 A deposited at 500°C on gold have a dielectric constant of 200. The dc conductivity closely follows the Poole-Frenkel model. Two dielectric loss peaks are believed to be caused in part by an oxygen deficient region near one electrode. The variation in the dielectric constant K with electric field is similar to that observed in bulk material.

Journal ArticleDOI
P. Chaudhari1
TL;DR: The authors of as discussed by the authors showed that the stress required to operate dislocation sources within a grain, at a grain boundary, and at surfaces is larger than the intrinsics tresses observed in polycrystalline films.
Abstract: The stress required to operate dislocation sources within a grain, at a grain boundary, and at surfaces is found to be larger than the intrinsics tresses observed in polycrystalline films. It is therefore unlikely that a dislocation flow mechanism can relieve stresses in films. Grain boundary sliding and diffusional creep can, however, relieve stresses in films and equations describing the kinetics of stress relaxation are derived. It is suggested that stress relief occurs primarily by a diffusion-creep mechanism. Growth of hillocks during annealing of a film is briefly discussed in terms of the diffusion-creep mechanism.


Journal ArticleDOI
E. Sawatzky1, Eric Kay1
TL;DR: In this article, a two-target sputtering system for stoichiometric, polycrystalline gadolinium iron garnet material is described, which allows an expeditious study of and subsequent compensation for the observedcation deficiency.
Abstract: RF sputtering of stoichiometric, polycrystalline gadolinium iron garnet material resulitns films significantly deficient in iron content. The cation deficiency is shown to be quite sensitiov ep reparatory conditions and reflects itself markedly in the magnetic and structural properties of the resultant films. Temperature dependent sticking probabilities and selective resputterig at the growing film surface are thought to be the most likely causes for the observed deficiencies. A simultaneously operated rf-dc two-target sputtering system is described in some detail, which allows an expeditious study of and subsequent compensation for the observedcation deficiency. Films with essentially the bulk garnet composition and bulk structural and magnetic properties were prepared in this two target system.

Journal ArticleDOI
B. S. Berry1, I. Ames1
TL;DR: In this article, a simple solder immersion test was used to study wetting and dewetting effects on Cr films overlaid with films of Cu and other metals, and it was concluded that the basic cause of solder dewetting is the excessive loss of intermetallic from the underlying Cr film.
Abstract: The thin film metallurgy used for SLT chip terminal contacts has been studied with respect to the soldering procedure used for chip-to-module joining. A simple solder immersion test was used to study wetting and dewetting effects on Cr films overlaid with films of Cu and other metals. It was found that the initial soldering reaction (consisting of the conversion of the Cu film to a Cu-Sn metallic layer) proceeds to completion in only a few seconds. Thereafter, the intermetallic layers tarts to disintegrate and become thinner by a mechanism identified as solution-assisted spalling. Removal of the intermetallic layer by this mechanism is not limited by simple solubility considerations. From metallographic observations and the inability to produce direct wetting of Cr films by solder, it was concluded that the basic cause of solder dewetting is the excessive loss of intermetallic from the underlying Cr film. Dewetting is accelerated if the Cu film is deposited on an oxidized Cr surface. These observations underscore the importance of the manufacturing practice of overlapping the Cr and Cu depositions so as to obtain an adherent and interlocked structure which is resistant to spalling. Other studies have shown that Al films are relatively inert to molten pure Pb-5%, Sn solder, but are susceptible to rapid attack if gold is added to the solder. Appreciabled elay of such attack is afforded by an overlying film of Cr, provided both surface and edge coverage are achieved.

Journal ArticleDOI
W. E. Donath1
TL;DR: A new suboptimal intermediate-speed algorithm which use n2 In n steps is developed for the assignment problem and lower bounds are derived, using this algorithm and other methods, for the average values of three classes of n × n assignment problems.
Abstract: A new suboptimal intermediate-speed algorithm which use n2 In n steps is developed for the assignment problem Upper and lower bounds are derived, using this algorithm and other methods, for the average values of three classes of n × n assignment problems: 1 When the elements of the matrix are random numbers uniformly distributed over the range 0 to 1, the average optimal value is smaller than 237 and larger than 1 for problems with large n Experimentally the value is about 16 2 When the elements of the matrix are random numbers such that the probability of being less than x is xk+1 (k ≠ 0), asymptotic expressions for the upper and lower bounds of the average optimal value are Cknk/(k+1) and Ck[(k+1)/k]nk/(k+1) respectively 3 When each column of the matrix is a random permutation of the integers 1 to n, asymptotic upper and lower bounds are 237n and 154n, respectively Experimentally the value is about 18n

Journal ArticleDOI
T. C. Raymond1
TL;DR: New heuristic algorithm that has been programmed for a digital computer and that obtains optimal or near-optimal solutions to the classical traveling-salesman problem is described.
Abstract: The classical traveling-salesman problem is to determine a tour that will minimize the total distance or cost involved in visiting several cities and returningt o the starting point. This paper describes new heuristic algorithm that has been programmed for a digital computer and that obtains optimal or near-optimal solutions to the problem. The author's general approach was derived from an existing algorithm developed by Karg and Thompson in 1964. Computational results for five multi-city tours are presented and the algorithm is shown to be competitive with other existing heuristic techniques.

Journal ArticleDOI
TL;DR: In this paper, the downshifting in the frequency of maximum acoustic intensity (FMI) for high flux domains in piezoelectric semiconductors was studied, and it was shown that the signal domain is narrower than the pump domain because the parametric growth is exponentially dependent on pump strain.
Abstract: Mechanisms for the downshift in the frequency of maximum acoustic intensity fmi for high flux domains in piezoelectric semiconductors are reviewed. For the simple case where an externally introduced acoustic wave (pump) produces a single-frequency domain in photoconducting CdS, clear evidence is givtehna t the downshift in fmi is due to parametric amplification of thermal acoustic noise. For a pump of 990 MHz, after some initial growth (vd = 1.14 v8), the pump is found to be depleted In the pump depletion region, signals in a 200 MHz band about the even subharmonic (445 MHz) are found to grow. At pump strains of about 10-6 the signals propagated at angles to the pump equal to those that give phase matching according to the dispersion of linear theory. For higher pump strains, however, the collinear process is dominant. The signal domain is narrower than the pump domain, as expected, because the parametric growth is exponentially dependent on pump strain. The downshifting of fmi in the region where deviations from linear theory are still small is discussed in terms of a parametric interaction model, with the initial acoustic strain distribution considered as an incoherent pump.

Journal ArticleDOI
A. D. Pearson1
TL;DR: In this article, the role of phase changes in the mechanism of device operation is discussed, and new evidence in support of a filamentary conduction hypothesis is presented, and the laboratory operation of simple diodes and the methods of inducing transitions among the various states are described.
Abstract: Semiconducting glass diodes can exhibit at least three conducting states: a high-resistance, or "off" state; a low-resistance, or "on" state; and a negative resistance state. When appropriately pulsed they can also display a memory function. The laboratory operation of simple diodes and the methods of inducing transitions among the various states are described. In addition, the possible role of phase changes in the mechanism of device operation is discussed, and new evidence in support of a filamentary conduction hypothesis is presented.

Journal ArticleDOI
R. Bray1
TL;DR: In this article, a perspective on the acoustoelectric instabilities is provided in terms of experiments on the III-V semiconductors, selected to limit or control the variables, such as circuit conditions (constant current or constant voltage) which determine the degree of feedback between buildup of flux and gain everywhere in the sample, boundary conditions, including destruction or reflection of flux, inhomogeneities in resistivity of the samples, to which the current oscillations and initial stages of domain formation are particularly sensitive.
Abstract: The acoustoelectric instabilities appear in diverse forms and seem complex because of the large number of factors which influence them. A perspective on the instabilities is provided in terms of experiments on the III-V semiconductors, selected to limit or control the variables. Detailed evidence is presented that the source of the acoustic flux is the thermal equilibrium phonon spectrum. The individual contributions of various factors to the instabilities are identified and discussed; e.g., 1) circuit conditions (constant current or constant voltage) which determine the degree of feedback between buildup of flux and gain everywhere in the sample, 2) boundary conditions, including destruction or reflection of flux, 3) inhomogeneities in resistivity of the samples, to which the current oscillations and initial stages of domain formation are particularly sensitive, and 4) the form of the acoustoelectric gain curve, which determines whether the feedback on the gain is positive or negative when the current is changed by the flux buildup. Also, several aspects of deviation from small signal theory are identified. In strong flux, 1) the peak of the frequency spectrum of the flux is shifted down, 2) the acoustoelectric gain coefficient is changaendd, 3) the carrier distribution function may become hot in the high field of the domain. It is concluded that the strong flux effects modify but do not basically determine the form of the resistance instabilities.

Journal ArticleDOI
TL;DR: An improved "subintervals" technique and a series approximating the Green's function have been combined to yield a single practical computer algorithm to achieve a balance between impedance variations, propagation velocities, and crosstalk coupling coefficients.
Abstract: In order to design an interconnection system for nanosecond-risetime logic circuitry, it is necessary to obtain a balance between impedance variations, propagation velocities, and crosstalk levels so as to achieve the best system speed as well as system speed control. To accomplish this, it is necessary to relate the electrical material properties and physical dimensions of the connections to characteristic impedances, propagation velocities, and crosstalk coupling coefficients. Two practical transmission line configurations: the microstrip line, which is coated for physical protection, and the offset or unsymmetrical triplate line, are being fabricated by mass production techniques. Because of the close control required and the many factors affecting impedance and coupling, these configurations require accurate means for predicting their characteristics. An improved "subintervals" technique and a series approximating the Green's function have been combinedto yield a single practical computer algorithm. Excellent agreement has been obtained in comparing the results of computations with large scale-model transmission line measurements. The method is quite general because dimensions, conductor shapes, and dielectric properties may vary widely.

Journal ArticleDOI
TL;DR: In this article, a stroboscopic strain-birefringent methowde was used to observe the off-axis domains of CdS crystals oriented with the electric drift field parallel to the c axis, and the angle at which the gain is a maximum depends on the angular dependence of the electromechanical coupling coefficient and the component of the electron drift velocity along that angle.
Abstract: In CdS crystals oriented with the electric drift field parallel to the c axis, acoustoelectric domains consist of off-axis shear waves. This is because there is no acoustoelectric gain for shear waves traveling along the hexagonal axis, while the gain may be large in an off-axis direction. The particular angle at which the gain is a maximum depends on the angular dependence of the electromechanical coupling coefficient and the component of the electron drift velocity along that angle. These factors combine to make the angle of maximum gaina function of drift velocity along the c axis. Using a stroboscopic strain-birefringent methowde, observed the off-axis domains directly. The domain tilt angle has been found to depend on drift velocity in roughly the same way as predicted from the small-signal angular dependence theory. Discrepancies may be the result of large-signal effects or of angular dispersion.

Journal ArticleDOI
TL;DR: In this article, it was demonstrated by Kikuchi that two modes of acoustoelectric domain oscillation occuirn InSb in a transverse magnetic field.
Abstract: It was demonstrated by Kikuchi that two modes of acoustoelectric domain oscillation occuirn InSb in a transverse magnetic field. Using lithium niohate transducers on an acoustic amplifier, we have measured linear acoustic gain as a function of electric and magnetic field and frequency. At high magnetic fields (B ≥ 3000 gauss) the results arein good agreement with White's theory. However, at low magnetic fields, the wavelength of the sound waves is lesst han the mean free path of the electrons, and the macroscopic theories break down. We have extended a microscopic theory of magnetacoustic interactions, duet o Spector, to include electron drift. We find excellent agreement between theory and experiment over the whole range of magnetic field. Moreover, the results account very clearly for the two modes of acoustic domain formation.


Journal ArticleDOI
A. M. Barnett1
TL;DR: In this paper, current filaments have been studied in devices exhibiting the current-controlled negative resistance associated with specific cases of two-carrier space-charge-limited current double injection.
Abstract: A current filament is a non-uniform radial distribution of current in the presence of a uniform electric field ina uniform sample. These filaments can have diameters in the 0.005 inch range. The current density at the center of the filament can be several orders of magnitude higher than the background current density in the rest of the sample. Filaments have been studied in devices exhibiting the current-controlled negative resistance associated with specific cases of two-carrier space-charge-limited current double injection. Electrons and holes are injected from opposite contacitns to a semi-insulator in which the lifetime of carriers of one sign is much longer than the lifetime of carriers of the opposite sign. When forward biased this device exhibits a high-voltage, high-impedance (108 ohms) pre-breakdown region and a low-voltage, low-impedance (1 to 103 ohms) post-breakdown region. In the post-breakdown region current increases ata nearly constant voltage followebdy a high-current power-law region. Current filaments have been studied throughout the post-breakdown region by recording the recombination radiation observed through one of the injecting contacts. Experimental and analytical studies of current filaments in silicon at 77°K and GaAs at 300°K are reviewed.


Journal ArticleDOI
D. P. Kennedy1, R. R. O'Brien1
TL;DR: In this paper, a two-dimensional mathematical analysis of the steady-state mechanisms of operation within a planar type junction field-effect transistor (JFET) is presented, showing that the potential distribution within the source-drain channel follows from solutions of Poisson's equation rather than from Laplace's equation.
Abstract: A two-dimensional mathematical analysis is presented of the steady-state mechanisms of operation within a planar type junction field-effect transistor (JFET). This analysis showst hat the potential distribution within the source-drain channel follows from solutions of Poisson's equation rather than from Laplace's equation. In particular, velocity-limited carrier transport produces a region of carrier accumulation in a region of the source-drain channel previously assumetod be depleted of carriers by the gate junction space-charge layers. The results of this two-dimensional mathematical analysis are presented in graphic form.

Journal ArticleDOI
M. Glicksman1
TL;DR: In this article, the experimental observations of microwave emission from InSb and the theories proposed to explain these have been reviewed, and two sources for some of the radiation, the acoustoelectric interaction and a collision-induced plasma instability, appear reasonably well established.
Abstract: This paper reviews the experimental observations of microwave emission from InSb and the theories proposed to explain these. Two sources for some of the radiation, the acoustoelectric interaction and a collision-induced plasma instability, appear reasonably well established. Experiments are proposed to clarify a number of still unanswered questions.