A memory system comprising uncased integrated circuits, evaporated wires, and magnetic, thin-film storage elements has been fabricated and tested. The 1536-bit memory system is contained on one side of a 3-by 4 1/2-inch glass substrate. All system wiring on the substrate is formed by evaporating aluminum through masks. All system circuits are contained on 182 silicon chips, each 0.050-inch square, and the chips are ultrasonically bonded directly to the evaporated aluminum. Data are presented on the operation of the selection system, of the recirculation loops, and of a portion of the memory system. Operation of the complete memory system was precluded by oscillations which developed in the system as recirculation loops were added. All subsections of the memory system function properly except for the recirculatinn loops. The origin of the oscillations and the techniques for eliminating them are discussed. The data obtained in fabricating and testing partially populated systems indicate that the fabrication techniques are practical.

Techniques have recently been developed for using uncased integrated circuits in electronic systems. The use of uncased integrated circuits in computers will lead to the development of more reliable, more economical, and physically smaller computers than can be fabricated from discrete components or packaged integrated circuits. Before these advantages can be realized, practical techniques must be developed in the following areas:1. Memory design2. Logic design3. Chip testing4. Chip bonding5. Chip passivation

THE DESIGN of an evaporated memory system presently under development and the fabrication of selected parts of the system will be described in this paper. A completely evaporated system is not possible now since suitable evaporated active elements are not available. but the availability of suitable elements is anticipated Chip transistors and diodes and electroplated magnetic films are used in the memory; all other parts memory are evaporated.

Planar, thin film magnetic memory arrays of 4×4 and 64×24 elements have been fabricated from evaporated matrix wiring and insulating layers. Four metallic and three insulating layers were evaporated sequentially in a vacuum chamber without breaking the vacuum. The magnetic element is an electroplated ternary alloy of nickel-iron-phosphorus 0.13 mm (0.005-in.) in diameter and a few hundred Å thick. The film has an anisotropy field of about 1 Oe and switches in 10 nsec with an applied word drive current of 35 mA. Sense and cancellation lines and dummy bit lines are used to cancel interwire coupling. Disturb tests were performed with up to 104 disturb pulses being applied along the easy axis. The outputs from films with Hc/Hk≈2 were decreased by less than 10% when disturbed in this manner. An evaporated memory array made with highly inverted films may be usable as a high-speed, low-current memory storage device.

Calculations and measurements have been made of the shift of the ferromagnetic resonance frequency in the presence of excited magnons. The calculations were carried out by retaining terms in the Hamiltonian up to fourth order in the spin-wave variables and by treating these terms in a random phase approximation. The frequency shift depends linearly on the number of excited $k=0$ and $k\ensuremath{\ne}0$ magnons, with a $k$-dependent proportionality factor. Measurements of this shift therefore can be used to determine the magnon distribution at resonance, and such measurements have been used to study the effect of surface preparation of the sample on the magnon distribution. The results show considerable agreement with the surface pit model; namely, the linewidth and the number of $k\ensuremath{\ne}0$ magnons increase with increased surface roughness, the surface-induced scattering from the $k=0$ mode occurs to a localized group of degenerate $k\ensuremath{\ne}0$ spin waves, and the effective wavelength of the spin wave to which scattering occurs increases with increased surface roughness. The data are not consistent with the theory that there is a rapid equilibration of the degenerate spin waves. The frequency shift was also observed at power levels above the critical value for the onset of nonlinearities. The results cannot be explained simply by invoking the instability of ${\ensuremath{\theta}}_{k}=0$ spin waves. Possible alternate explanations are offered.

The presence of excited magnons results in a shift of the ferromagnetic resonance frequency. Calculations of this frequency shift have been made by retaining terms in the Hamiltonian up to fourth order in the spin wave variables. The shift can be expressed in terms of the number of excited k=0 magnons and k≠0 magnons. Determination of this shift can then be used to study the magnon distribution during resonance. This frequency shift has been measured in four single-crystal yttrium iron garnet spheres at 9 Gc. The shifts are observed to be directly proportional to the rf power absorbed by the sample, positive when the dc magnetic field is in [100] direction, and negative when the dc magnetic field is in the [111] direction. These results require a distribution of k≠0 magnons dominated by magnons with 48°<Θk<60°. If it is assumed that the k≠0 magnons are degenerate with the k=0 magnons, the total number of k≠0 magnons is approximately equal to the number of k=0 magnons, and magnons in the region centered about k=1−2×105 cm−1 dominate. This result is in agreement with a prediction of the surface pit scattering model; namely, the relaxation of the k=0 magnons results in the production of k≠0 magnons with wavelength approximately equal to the pit size. The results are in disagreement with any model requiring a uniform distribution of degenerate modes.

A polished, ellipsoidal, single crystal, yttrium iron garnet sample with rotation axis along a [100] direction and ΔH = 1.0 oe was prepared. X-band measurements of the microwave resonant frequency were made as a function of rf power. The resonant frequency was found to shift at the rate of 41 Mc/w. The shift of the resonant frequency was related to the number of k = 0 magnons (n0) and the number of k ≠ 0 magnons (nk)by means of spin-wave theory. From this relation and the experimental data the value of nk/n0 was determined to be 2.7.

It has been observed that the magnetic anisotropy direction of certain thin Permalloy films prepared by the thermal decomposition of nickel and iron carbonyls can be rotated through 90 deg at room temperature by fields less than 10 oe. The experimental conditions under which these films were prepared suggest that this phenomenon is due to the presence of carbon. The relaxation time for the rotation of the anisotropy axis was calculated from the time dependence of the induction and found to be 10 msec. A simple model has been developed which relates the anisotropy rotation to specific domain wall configurations and to the establishment of a “directional order” among interstitial carbon atoms.

The remagnetization time τ has been measured as a function of external field H for a series of manganese ferrites, Mn1+xFe2−xO4. The physical properties of these compounds which enter into the domain wall model of Menyuk and Goodenough and the rotational model of Gyorgy have also been measured. It is found that α, the loss constant, is critical to the correlation of data with theory, and that neither theory can be unequivocally proved until more insight into the remagnetization process loss mechanism is acquired.