of an oxide-silicon (or other) interface, the electron beam is slowly swept in a line scan across the portion of the crater edge which exposes the interface. The Auger spectrum of an element is recorded. In this manner elemental profile shapes essentially identical to those of conventional depth profiling are obtained. The abscissa now becomes the position of the electron beam on the crater edge, which, from a knowledge of the ion beam current density and the associated sputter rates, can be transformed to a distance normal to the specimen surface. Although a number of elements can be recorded simultaneously by using a multiplexer, repetitive scans can be made across the interface, so there is no absolute need to multiplex.

This procedure was tested by carrying out experiments on specimens with 50- to 150-nm thick oxides thermally grown on phosphorusdoped silicon [61]. The experimental geometry is shown in figure 21a; the rastered 1kev ion beam (current density≈ 40 μA/cm2) formed an approximately elliptical crater as shown in figure 21b. The ion beam current density as measured along the x-direction through the focal point of the analyzer by using a small Faraday cup is shown in figure 21c.

In an experiment to compare conventional depth profiling with crater edge profiling

the specimen was depth profiled to the interface as determined by monitoring oxygen, and then for an additional time equal to 20 percent of the time that it took to reach the interface. The additional sputtering time was chosen such that the position of the interface was along a relatively linear part of the ion current density distribution of figure 21c. The width of the crater at the interface along the x-direction should then be given by the width of the current density distribution at five-sixths of maximum. This was checked and the widths were found to be identical within the experimental error of 0.1 mm, giving confidence that the slope of the crater wall may also be obtained from the slope of the current density distribution provided that the sputtering rate is known and that it remains unchanged across the interface.

Line scans for oxygen, phosphorus, and elemental silicon were made along a 530-μm length of the crater edge which included the oxide-silicon interface (C to c1 in fig. 21b). The results are shown in figure 22. From the slopes of the oxygen curve and the crater wall (fig. 21c) the interface width was found to be 5.0 ± 0.2 nm as compared with the value of 4.9 ± 0.2 nm obtained from the initial depth profile. Of chief significance, however, is the relative ease with which the measurement could be repeated.

a.

b.

c.

The axis, Y1,

X-SCAN (mm)

0.5

igure 22. Line scans for oxygen, phosphorus, nd elemental silicon across the interface beween the points C and C1 on the x-axis.

he crater edge interface analysis technique an be used as an adjunct to the conventional echnique but in many cases it can be used

quite independently since the positioning of the electron beam during sputtering is much less critical when the monitoring of Auger peaks is mainly to determine the time to an interface. In many cases the sputtering may be rapid and fears of bulldozing too rapidly through the interface are no longer important, since the sputtering and analysis are now separated. When there is difficulty in measuring details of the ion beam shape, the slope at an interface can be calibrated by making a good conventional depth profile from which the interface width can be measured.

Obviously not all interface analysis lends itself to this technique since in a macroscopic sense the films must be uniform, parallel to the surface over a distance of a few millimeters. However, films of this degree of uniformity are not uncommon. (N. J. Taylor, J. S. Johannessen #, and W. E. Spicer

5. MATERIALS WAFER

5.1. Sodium Contamination in Oxidation Furnaces

*

Additional experiments verified the previous conclusion (NBS Spec. Publ. 400-25, pp. 29– 30) that despite its great sensitivity for free sodium, the resonance fluorescence method [62] is not adequate to detect sodium contamination in furnace tube atmospheres at 1000°C which produce oxide films with sodium contamination of the order of 1011 cm-2.

To investigate the relationship between free sodium density and the density of sodium compounds, thermodynamic calculations were made to determine the equilibrium partial pressures of various species at 1300 K in oxygen flowing at 4 cm3/s. It was assumed that the source of sodium was the contamination in the fused silica tube wall which generally has an average sodium content of about 10 ppm by weight [63]. Because of errors and approximations in thermodynamic data involved in these equilibrium calculations an absolute error of ±50 percent could be expected, although the relative error of these results is believed to be much smaller.

The results of these calculations [64] are summarized in table 8 which lists both partial pressures, P, of appropriate compounds and density of sodium, N, for several levels of water content in the oxidation atmosphere. For dry oxygen, free sodium is the predominant species but as water vapor is added the total sodium density increases significantly because of the formation of sodium hydroxide. The effect of the added water vapor is illustrated in figure 23 which also includes the results of calculations made for 1200 and 1400 K.

Calculations were also made to establish the effect of cleaning the furnace tube walls with hydrogen chloride or chlorine. The results of these calculations for the case of a 10-percent mixture of hydrogen chloride or chlorine in an inert carrier gas at atmospheric pressure are summarized in table 9. The calculations indicate a pressure of NaCl higher than the vapor pressure for NaCl (liq), but under the actual conditions, equilibrium is probably not reached. Micro-droplets of liquid sodium chloride, if formed on the tube wall hot zone, would be removed by the gas stream and transported to a cooler tube zone where sodium chloride precipitates on the wall (sodium chloride melts at 801°C). This

At a typical flow rate of 4 cm3/s (0.5 scf/h), this quantity of chlorine or hydrogen chloride is transported through the oxidation tube in about 1/2 s. Thus the reaction rate is regulated by diffusion of sodium in the fused silica. When sodium in the wall bulk diffuses to the wall surface, the reaction with chlorine effectively takes place immediately. In practice, a cleaning period of about 20 h is generally used to deplete the tube wall. The longer the cleaning period, the longer the oxidation tube can later be operated under acceptable oxide growing conditions. (S. Mayo and W. H. Evans)

5.2. X-Ray Dose in Electron-Beam Evaporators" Additional calculations were made to estimate the x-ray dose received by the oxide films in MOS structures with aluminum or chromium gate electrodes deposited by electron-beam evaporation [40]. Results of a preliminary experiment (NBS Spec. Publ. 400-17, pp. 20-21) suggested that an absorbed dose of the order of 1 kGy (0.1 Mrads (SiO2)) is received during the evaporation of a 1-um thick aluminum film over silicon dioxide with 10-keV electrons.