usually shorter [3]. Temperatures as high as 400 °C have been measured in this study at the break-off tab (fig. 2). In the center of the glow area itself, the visible color may vary from a dull red to a bright white, indicating maximum temperatures in excess of 1200-1300° C [1].

The origin of the junction resistance leading to resistive heating in aluminum wire/steel connections is not well understood. Previous work on copper/copper junctions by Suzuki et al. [4] indicated that the formation of filamentous Cu2O between the loose copper conductors was responsible for high junction resistance. This observation suggests the possibility that the formation of aluminum oxide and/or iron oxide might be an important mechanism for the development of high resistance junctions in aluminum wire/iron screw connections. In the present study, the microstructures of components from laboratory simulations of aluminum-iron glowing connections as well as actual wall receptacles tested at the onset and after the full development of glow formation were examined by scanning electron microscopy and x-ray microanalysis to determine the underlying mechanism of high resistance junction formation.

To simulate a glow failure in a controlled location on an aluminum wire/iron screw complex a simple apparatus was constructed (fig. 1). In this device, which was essentially a micromanipulator, an aluminum wire was pressed against the head of a brass-coated steel 8-32 screw taken from a commercial receptacle. The gap between the wire and screw was adjustable. Because of the curved geometry of screw and wire (longitudinal axis perpendicular), contact area was minimal.

Voltage was provided by a variable ac source, and a resistor of about 1 in series with the wire-screw connection was used to control the current level to prevent gross instabilities.

In operation, voltage was first removed from the fixture and the thumbscrew turned until the test wire and steel screw just made contact as indicated by an ohmmeter. Next, voltage was applied and current increased to 15 A. If a

FIGURE 1. Apparatus for controlled wire/screw arc experiments. glow did not appear within 2 or 3 min, the thumbscrew was backed off very slightly. Arcing (defined as a low voltage, high current sustained discharge) would then be observed visually and a sustained glow could frequently be obtained. In order to avoid the effects of residual contaminants on the test wire and screw, both were cleaned with trichloroethylene and acetone and rinsed with distilled water and methanol.

The specimens produced with this simple apparatus were primarily used for examination of the initial phases of the glow phenomenon, since the location of the electrical contact could be carefully controlled.

2.2 Residential Connections

The second set of experiments examined the situation which resembled the conditions in house wiring. Here the goal was to induce a glow under the head of a screw in an actual residential receptacle. To this end 10 duplex receptacles were mounted on a rack and wired in series with #12 gage aluminum wire. All receptacles in the experiment were of a common commercial variety using steel screws with a very thin brass or zinc plating. Figure? shows a typical device removed from the rack for observation. The connections were tightened to a relatively low torque of about 0.4 nt-m (4 in-lb), in order to try to induce glow in a few current cycles. A current of 40 A a was periodically passed through all connections. A complete cycle took 20 min with equal "on" and "off" times. High currents were employed to accelerate the development glow failures in order to obtain suitable samples in 1

Fe screw

Al wire

1cm

GURE 2. Duplex junction box showing wire and screw components and location of tab for temperature measurements.

asonable time. Previous work indicated that glow failures ould be obtained over a wide range of current, 0.8-50 A !].

Some of the connections tested developed a glow failure hich could be sustained for considerable time. Sample A❞ was one such example (see table 1) which developed a low after only 8 h of cycling. This specimen was left in the rcuit and continued to glow, but was finally removed after more hours had elapsed. The total number of "on"-"off" cles at that time was 36. In other tests, e.g., specimens C nd D in table 1, the test was interrupted at earlier stages the development of the glow phenomenon in an attempt O trace the sequence of the events. Temperatures were easured during the test with a thermocouple connected to electronic digital thermometer at the position shown in gure 2.

Samples were prepared for scanning electron microscopy nd x-ray microanalysis. In the case of the simple ire/screw contact experiments, the region of the contact uld be examined directly in the scanning electron

microscope without prior preparation. For the household. fixtures, metallographic sections were prepared through the wire/screw/brass plate junction. Because of the three dimensional nature of the junction, it was not always possible to simultaneously intercept the region of glow interaction at the wire/screw interface and the wire/brass plate interface. An optical microscope incorporated in the scanning electron microscope allowed direct comparison between the optical and electron metallography of the sample.

X-ray microanalysis was carried out with an energy dispersive x-ray spectrometer (EDS) attached to the scanning electron microscope (SEM). The elements of interest in the analyses, aluminum, iron, copper, and zinc, have characteristic x-ray energies which are sufficiently separated in energy so that no peak overlap occurs. For the flat metallographic sections, quantitative x-ray microanalysis was possible. The NBS theoretical matrix correction procedure for x-ray microanalysis, FRAME C, was used to reduce the measured x-ray spectral intensities to quantitative compositional values [5]. Pure elements were employed as standards, and the beam energy was 20 keV for all analyses. Under such analytical conditions with well resolved peaks, previous experience in the analysis of known samples indicates that FRAME C analyses lie within 5 percent relative of the correct value in 90 percent of the cases tested when major elements (concentrations greater than 10 wt%) are being considered.

Quantitative x-ray microanalysis of the irregular surfaces observed on the aluminum wire and iron screw couples from the loose-contact simulation was performed with the NBS theoretical matrix correction procedure for rough surfaces, FRAME P [6,7]. This is a recently developed procedure for the analysis of samples which deviate from the ideal sample, e.g., flat, polished, and set at a specific angle to the x-ray detector. The typical error limits for FRAME P analysis have not yet been determined, but preliminary testing suggests that an error range of ±20 percent relative should be obtained in most cases.

3. Experimental Results

3.1 Wire/Screw Loose Contact Simulation of Glow Failures

d-visible glow not detected; the test was interrupted to study the cipient failure.

A cycle consisted of 10 min "on" followed by 10 min "off."
Temperature measured at tab on fixture.

3.1.1 Free surface

The regions on the surfaces of the aluminum wire and iron screw associated with a glow failure were examined by scanning electron microscopy. Examples of the structures observed on the wire and screw are shown in figures 3(a) (wire surface) and 4(a) (screw surface). Craters are observed on both surfaces. These crater surfaces have a

smooth appearance which appears to be a result of melting and solidification. X-ray microanalyses of the areas indicated in figures 3(a) and 4(a) are listed in table 2. The analyses reveal that both aluminum and iron are present in significant amounts within both crater regions. Since the surrounding matrix material is either nearly pure aluminum or nearly pure iron, the intimate mixing of these elements in the crater regions indicates that significant material transport occurred between the wire and the screw during the glow phenomenon.

Close examination of the craters revealed a number of interesting features. Ball-like structures, marked in figures 3(b) and 4(b), found in both craters contain mostly iron,

FIGURE 4(b). Ball-like structures observed in region of figure 4a) (SEM image).

with only 2-4 percent aluminum [table 2, Al(2.7 Fe(4,5,8,9)]. Elongated structures, indicated in figure 4(c are also composed mainly of iron. The spherical an elongated structures appear to be the result of melting an rapid freezing of the metal. The presence of molten ir during the glow phenomena indicates that temperatures excess of 1500 °C are generated in the region of the crater

The imaging process in the scanning electron microscop is sensitive to the presence of insulating layers through th phenomenon of electrical charging. The presence of thi layers of aluminum oxide or iron oxide on the surface the craters would be revealed in the scanning electr micrographs by bright regions characteristic of su

charging. These areas are occasionally observed within the craters, e.g., as indicated by circles in figures 3(a) and 4(c). The regions of charging in the image represent only a small portion of the crater area, and thus the dominant character of the crater is conductive.

3.1.2 Interior

The sub-surface structure of a crater formed during the glow process on the aluminum wire was examined in a metallographic cross section (fig. 5). The electron image of this structure reveals a new phase at the surface of the wire below the crater, which is found to contain principally aluminum and iron with a small amount of copper (table 2).

3.2 Branch Circuit Receptacle Glow Failures

3.2.1 Long term sustained glow failures

Sample A (table 1) underwent a sustained glow, 4 h in duration reaching a measured temperature of 350 °C. The metallographic cross section (unetched, figs. 6(a) and 6(b)) showed evidence of extensive reaction at the aluminum wire/iron screw interface and at the aluminum wire/brass plate interface.

(a) Aluminum wire/iron screw interface. A backscattered electron image (which is sensitive to compositional differences) of the aluminum wire/iron screw interface,

(fig. 7) reveals a complex reaction zone which consists of two distinct layers located between the aluminum and iron. Quantitative analyses (table 3A) performed at the locations indicated in figure 7 reveal that the thin outer layer (near the aluminum wire) consists of 62 wt% Al-38 wt% Fe0.5 wt% Cu while the thick inner layer (near the iron screw) consists of 57 wt% Al-43 wt% Fe-0.1 wt% Cu. Zinc (from the brass electroplated coating) was not detected.