Index of content:
Volume 4, Issue 5, September 1986
4(1986); http://dx.doi.org/10.1116/1.583475View Description Hide Description
Siliconhomoepitaxialfilms have been grown by ion beamsputter deposition using an ultrahigh vacuum (UHV) apparatus with i n s i t u diagnostic equipment. The deposition conditions are characterized and the beginning of single crystal growth occurs at 250 °C. Films of high crystalline and morphological quality are obtained at deposition temperatures above 700 °C, where good doping element transfer efficiency from the target to the film is observed. Room temperature bulk mobility is found for film thicknesses as low as 0.5 μm and deposition temperatures near 700 °C.
4(1986); http://dx.doi.org/10.1116/1.583476View Description Hide Description
This paper reviews some aspects of and recent advances in plasma assisted deposition of thin films for microelectronic applications. The plasma deposition process and some properties of deposited films relative to integrated circuit fabrication will be briefly discussed. Possible trends in future research and development of plasma deposition for microelectronic applications will also be presented.
4(1986); http://dx.doi.org/10.1116/1.583477View Description Hide Description
Schottky barrierdiodes were fabricated by sputter depositing Ti–W on n‐type Si in a dc magnetron sputtering system at voltages ranging from 460 to 780 V dc. The diodecharacteristics were determined by current–voltage (I–V) and capacitance–voltage (C–V) methods, while deep level transient spectroscopy(DLTS) served to characterize the sputter induced defects and investigate their dependence on sputtering voltage. Barrier heights of the devices and the concentrations of the sputter induced defects strongly depend on the sputtering voltage. Devices deposited at 660 V dc had the highest barrier height (0.60 eV) and contained the lowest concentration of defects. Several electron traps were measured in the as‐deposited devices. Their concentrations varied by a factor of up to 20 and depended upon the sputtering voltage. Isochronal annealing for 10 min up to a temperature of 400 °C increased the barrier height of all devices, except those deposited at 780 V dc, to a value approximately equal to that of the 660 V dc deposited devices. The isochronal annealing showed that some defects could be removed and that others were again introduced. When annealed at temperatures above 400 °C, defects appeared to be distributed deeper into the substrates than before annealing.
4(1986); http://dx.doi.org/10.1116/1.583478View Description Hide Description
Thin selective chemical vapor depositedtungstenfilms are evaluated as a barrier metal for reaction between Al–1%Si electrode and Si substrate. Change in contact resistance and junction leakage current after annealing at 500 °C in N2 are measured. It is found that an increase of contact resistance at a very small contact hole of 2.25 μm2 is very slight. On the other hand, contact resistance in Al–1%Si electrode increases to one or two magnitudes with annealing of 210 min. Junction leakage current in a tungstenelectrode is almost the same as that in Al–1%Si electrode. And it is also found that a thin tungsten layer protects the diffusion of aluminum into the Si substrate. From these results, it is concluded that thin selective chemical vapor depositedtungstenfilms have a good barrier effect for the reaction between Al–1%Si electrode and Si substrate.
4(1986); http://dx.doi.org/10.1116/1.583479View Description Hide Description
A completely nonconformal Al step coverage was observed on via structures with almost vertical sidewalls using a nozzle jet expansion deposition technique. For vias of dimensions 0.8 and 1 μ, deposition of Al by this technique resulted in a uniform layer of Al at the bottom of the vias with no sidewall coating. Possible use in several very large scale integrated (VLSI)metallization applications is suggested.
4(1986); http://dx.doi.org/10.1116/1.583480View Description Hide Description
The growth of siliconfilms by low pressurechemical vapor deposition in an unique vertical‐flow reactor and in a conventional tube reactor is studied, with emphasis on the vertical‐flow reactor. For a hydrogen‐carried silane process in the vertical‐flow reactor, the growth rate depends on both the partial pressure and the flow rate of silane. A growth rate expression incorporating both of these parameters is derived which accounts for a nearly linear dependence on either parameter for small values and saturation at large values. No dependence on hydrogen partial pressure is observed. An Arrhenius‐type dependence of growth rate on temperature is observed for both type of reactors, with an activation energy of 1.5 eV. Thickness uniformity of the films deposited in the vertical‐flow reactor is compared to that found in the conventional reactor. For high temperatures (above 650 °C) depletion effects degrade the thickness uniformity of the deposited films. Such effects are more pronounced in the conventional reactor than in the vertical‐flow reactor. These effects can be attributed to the shorter gas flow path in the vertical‐flow reactor. The vertical‐flow reactor can be operated without a temperature gradient along the deposition chamber so that uniform filmproperties can be obtained over the entire load.
Topographical limitations to the metallization of very large scale integrated structures by bias sputtering: Experiments and computer simulations4(1986); http://dx.doi.org/10.1116/1.583482View Description Hide Description
Aluminum deposition by radio‐frequency bias sputtering on grooves and holes with aspect ratios (depth to width) in the range of 0.5–1.0 was studied experimentally and by computer simulation. The structures with an aspect ratio of up to 0.5 were planarized. Aspect ratios larger than about 0.6 however led to the buildup of cavities within the grooves and holes. This limitation for narrow structures with vertical sidewalls was observed experimentally and was in excellent agreement with the results of the profile simulations. The critical aspect ratio may be increased by using a higher bias potential and slope angles of less than 90°.
4(1986); http://dx.doi.org/10.1116/1.583483View Description Hide Description
We have investigated a method for passivation of InSb by vacuum deposition of SiO x on native oxide layers grown by wet anodization. We show that this multilayerdielectric approach results in improved passivationproperties. Results of high resolution Auger spectroscopy reveal important information on the layer structure and composition of this passivationfilm. Specifically, we report the experimental observation of SiO2 formation at the SiO x anodic oxide interface. The interfacial reaction is limited to a thin layer, about 10 nm thick. The SiO x oxidation proceeds by reduction of the native oxide and formation of elemental In and Sb. The electrical features observed in the C–V curves (such as flat‐band voltage, hysteresis, low‐frequency‐like response in the inversion region, and other deviations from the ideal curves) are explained in view of the oxidation states of In and Sb, observed at the oxide layers and at their interfaces. These correlations were used for characterization of the desired interlayer parameters.
4(1986); http://dx.doi.org/10.1116/1.583484View Description Hide Description
This paper presents a study of the ion‐enhanced etching of InP by Cl2 using a modulated ion beam technique, in which pulses of rare gas ions impinge on an InP surface with coincident Cl2 and the resultant products are detected with a mass spectrometer. The stoichiometry of these products and the dependence of the ion enhanced etching on InP temperature and chlorine pressure are compared to the thermodynamically predicted chemical etching of InP by Cl2. The dependence of the ion‐enhanced etching on ion current, ion energy, and ion mass are compared to the dependence on these parameters observed in the physical sputtering of InP. The experimental results are discussed in relation to a model in which the incident ion creates a transient thermal pulse (∼100 ° K) which causes the desorption of the relatively volatile indium chlorides from the surface. Of particular technological importance is the fact that large chemical enhancements over physical sputtering (>100/1) are possible with such a mechanism.
Chemical etching of GaAs and InP by chlorine: The thermodynamically predicted dependence on Cl2 pressure and temperature4(1986); http://dx.doi.org/10.1116/1.583485View Description Hide Description
This paper reports a detailed thermodynamicanalysis of the GaAs/Cl and InP/Cl chemical systems, which are of interest in the technological processing of these III–V materials. The thermodynamically predicted dependence of the steady state chemical etching on both Cl2 pressure and temperature is derived assuming Langmuir free evaporation from the surface. The chemical potential data base used in this thermodynamicanalysis has been checked for accuracy against all available vapor pressure measurements in the literature. The thermodynamically predicted chemical etching is compared to the etching observed in a Cl2 plasma. This approach shows promise in semiquantitative modeling of the dependence of these reactions on both the temperature and Cl2 pressure. In addition, changes in the surface morphologies resulting from plasma etching appear to be correlated with the thermodynamically predicted transitions of various compounds on the surface.
4(1986); http://dx.doi.org/10.1116/1.583486View Description Hide Description
An etch process for SiO2 in a C6F1 4+N2plasma at 1 Torr pressure and with a power of 10 W/cm3 was examined for its etch rate dependence on gas composition, pressure, and electrode separation. Rates larger than 1 μm/min were obtained. The selectivity with respect to silicon could be made infinitely high by choosing various gas mixtures. An axial magnetic field increased the etch rate by 20%. The process was highly anisotropic. Ion and neutral mass spectroscopy was applied to illuminate the differences between this process and the conventional cathode coupled process (RIE) at pressures of 0.05 Torr.
4(1986); http://dx.doi.org/10.1116/1.583487View Description Hide Description
The reactive ion etching (RIE) of GaAs/Al x Ga1−x As heterolayers has been examined as a function of alloy composition (x=0.00–0.31) and the etching parameters of pressure, rf applied power, gas flow rate, and time using CCl2F2 as the etching gas. The information derived from this study was used to determine the GaAs to Al x Ga1−x As etch selectivities. The use of CCl2F2 promotes the formation of nonvolatile reaction products on the surface of Al x Ga1−x As, so that even for small amounts of aluminum the etch rate of Al x Ga1−x As is significantly reduced compared to GaAs. This compositional factor favors the highly selective etching of GaAs over Al x Ga1−x As. High gas pressure and flow rate and a low rf applied power also contribute to high selectivities in this system. A selectivity (GaAs to Al x Ga1−x As etch ratio) of 300 was attained for a pressure of 20 mTorr and an rf power of 500 W.
4(1986); http://dx.doi.org/10.1116/1.583488View Description Hide Description
The following electrostatic e‐beam deflectors are simulated numerically and compared: (1) Box deflector; (2) modified box deflector; (3) dokekapole deflector; (4) simplified octupole deflector; (5) 20‐pole deflector; (6) conventional octupole deflector. Two types of aberrations arise in the deflection of an electron‐image of negligible size: raster distortion and spot blurring. For each deflector, these aberrations are computed and plotted as functions of deflection amplitude. The present computations make no use of perturbational aberrationtheory but rely on the method of ray tracing in 3D fields. The full effect of fringe fields due to finite gap widths and finite lengths of electrodes is accounted for. The surface‐charge density method has been utilized for the accurate numerical calculation of the fields.
The Cramer–Rao accuracy bound for optimum processing of the edge registration mark signal in electron beam lithography4(1986); http://dx.doi.org/10.1116/1.583489View Description Hide Description
Experimental data from the IBM direct‐write electron beam lithography machine EL‐3 are used to evaluate a lower bound on the estimation error resulting from the application of optimal signal processing in determining the mark edge position for registration. The e‐beam is swept using a 20 kV accelerating voltage in a raster fashion across a bare rectangular groove of dimensions 5 μm wide×100 μm long×0.5 μm deep. Waveforms for backscattered electrons detected by diodes are processed statistically. The mathematical and numerical techniques are developed to handle a large but sparse noise correlation matrix. The bound obtained decreases exponentially with the sampling rate (at least throughout the range of measurement explored), and it represents the ultimate accuracy achievable with any signal processing technique. The implications of having multiple records of a waveform are included. Comparisons with other approaches are discussed.
4(1986); http://dx.doi.org/10.1116/1.583490View Description Hide Description
A design method for a low aberration matrix lens is developed. The matrix lens consists of an electrostatic lens array and beam limiting aperture array located outside of the lens array. It is shown that off‐axial aberrations, which are dominant in the outer lenses, can be reduced by shifting the aperture from the optical axis of each lens to its optimum position. Using this aperture shift effect, a low aberration matrix lens consisting of einzel lenses is designed for submicron electron beam exposure systems. An aberration of 0.1 μm is predicted for a 40 mm2 matrix lens with a beam half‐angle of 5 mrad.
4(1986); http://dx.doi.org/10.1116/1.583491View Description Hide Description
A novel scheme is introduced which allows spatially resolved x‐ray photoelectron spectroscopy (XPS) studies to be performed in device applications by taking advantage of the ability to bias the various components of the device. The application of a bias voltage produces a shift in the Fermi level of the biased component and there is a corresponding shift in the binding energy of electrons emitted from this component. In this manner it is possible to discriminate between the XPS signals that are due to a given element but which originate from different regions of the sample. By analogy with the corresponding biasing scheme employed in SEM studies, this technique has been called voltage contrast XPS.