A TECHNIQUE FOR THE MEASUREMENT OF HOT SPOTS AND ISOTHERM PROFILES AT THE SURFACES OF THE ELEMENTS OF HYBRID MICROCIRCUITS

A simple technique for measuring surface temperatures with high sensitivity and spatial resolution, which is particularly useful for examining temperature distributions of microcircuits, is described. The invention exploits the well-defined "Isotropic-point" transitions of nematic liquid crystals from optical birefringence to anisotropy, using a polarising microscope to detect temperatures in microscopic areas. Thus hot spots appear black against a bright background. The temperature sensitivity is better than 0.5C and the spatial resolution better than 5 microns. The technique has distinct advantages over alternative methods of measuring microcircuit temperatures. Examples of applications in support of design evaluation and fault diagnosis of monolithic circuits illustrate the benefits of visual information obtained. The ready application of the technique to the large planar areas of thick film resistors has enabled the location of hot spots confirming their occurrence at the end of laser cuts and the measurement of thermal resistances of resistors either singly or in arrays showing that substantial heat conduction through the substrate can outweigh local non-uniformities in dissipation. Applications extend beyond the supportive role to reliability evaluation illustrated in this paper.


INTRODUCTION
The temperatures of microelectronic components can significantly influence their performance and reliability; but their internal temperatures are rarely uniform, instead being distributed according to the power dissipation and heat conduction pathswhich can follow the microscopic dimensions of some active microcircuits. Therefore techniques have been sought to measure temperatures and thermal profiles on a microscopic scale.
One of the early innovations was the infrared microradiometer, which measured the radiation emitted from microscopic areas. ,2 In principle the technique had a spatial resolution down to 8/am and a temperature sensitivity of 0.5C. In practice the technique rarely performed to its limits and suffered the drawback of needing considerable calibration and computation to translate the measured radiation into temperature. Liquid crystals were examined following reports of the use of cholesterogens to measure surface temperatures, but the technique was found to have a poor spatial resolution (>20/am) and the required non-reflecting coating impaired the temperature profiles. Instead, a property of nematic liquid crystals (nematogens) has been successfully 177 exploited 4 ' into a technique for the measurement of surface temperatures with a spatial resolution of better than 5/am and sensitivity of better than 0.5C. The technique is briefly described in this paper which is devoted more to illustrating its application to active and passive components that could comprise the elements of hybrid microcircuits.

THE TECHNIQUE
In its liquid state, a nematogen is birefringent at temperatures up to a well defined critical temperature (Ti) called its "Isotropic Point," above which it is isotropic. Thus, plane-polarised light is doubly refracted when transmitted through a nematogen below Ti, but is unaltered by a nematogen above Ti. In the application of the technique to microcircuits, a metallurgical microscope is employed in the experimental arrangement shown in Figure 1   Ti appear dark because light is absorbed by the analyser. The boundaries between dark and light regions are then isotherms at Ti, which has been found to be reproducible to within 0.5C, whilst the spatial resolution has been determined to be better than 5/am. 4 The high resolution is simply illustrated in the example shown in Figure 2 of a hot spot of about 5/am diameter, produced by dissipation in a thin-film nichrome resistor.  In order to measure temperatures below Ti, for a particular dissipation, the component ambient temperature (Ta) is increased until light is extinguished in the area of interest. Then T Ta is the temperature rise for that dissipation. The distribution of temperatures between. 7'/and Ta is obtained by raising Ta by small increments to  Figure 9 shows the hotter operation of one of the two elements of the emitter due to persistent unbalanced operation, which was found to contribute to eventual failure by thermal runaway.  Figure 7 shows that one entire segment of the transistor was malfunctioning requiring excessive dissipation in the other three segments to get the fourth up to operating temperatures.

2. Thick@)'lm Resistors
In support of current reliability studies, recent attention has been devoted mainly to thick-film resistors whose large planar areas are ideally suited to examination by the technique. The examples presented here are for resistors in arrays in various dual-in-line packages, which were used as test vehicles in the studies. In order to relate thermal ageing during stress tests to the temperatures generated within the resistors, it was necessary to refer to a common parameter such as thermal resistance (Ro) which is defined as the temperature rise per unit power dissipated. Because the measurements showed that the temperature distributions were distinctly non-uniform, and ageing is fastest in the hottest parts, it was decided that the FIGURE 10 Hot spots on thick-film resistors. thermal resistances should be calculated for the hottest spots on the resistors. Single resistors gave the most straightforward results because the hot spots occurred at the known sites of maximum dissipation. Illustrated in Figure 10 is a typical observation, that hot spots always originate at the ends of laser "plunge" cuts-the hotter spots occurring at the longer cuts. The values of hot spot thermal resistance for resistors of similar geometry were found to be confined to a small range regardless of their electrical resistivity. Typical thermal resistances were 100C]W to 125 C/W for 2 mm x 2.2 mm resistors and 55C/W to 65C/W for 2 mm x 4.5 mm resistors in the range 50 ohm to 50 kohm. Generally, resistors located in the middle of substrates had the lower thermal resistances because of the substantial heat sinks surrounding them; but significant variations of Ro also arose because of differences in the extents of laser cuts in. nominally similar resistors.
When power was dissipated in arrays of resistors, the thermal resistances of individual resistors ceased to be meaningful because of the mutual contribution to the generation of heat in the substrate. Ro then depended on the total power dissipated in the substrate. Hot spots still occurred at the ends of laser cuts in individual resistors (Figure 11), but their location depended on the thickness of the substrate and the configuration of the array. For example, the resistors located in the middle of the substrates were now surrounded by heat sources. Thus, the resistors in which most power was dissipated were not necessarily the hottest. To be consistent with the earlier choice, the thermal resistances were referred once again to the highest temperatures, wherever _ 7 7 t substrate thickness 2.6 mm resistor size 2 mm x 4.5 mm ambient temperature 21 C dissipation/resistor 215 mW 75 74 7 resistor number FIGURE 12 Temperature distribution in an array of identical resistors. the conduction of heat between resistors and the finite boundaries for heat loss from the package. 6 Even where power was dissipated non-uniformly in an array likely in applications requiring various resistances in the array-the temperatures were relatively uniform about the centre of the array (Figure 14), endorsing the argument presented earlier of the mutual influence of the dissipations.
The technique has also been used to aid the diagnosis of faults in resistors. Low value resistors, which increased after exposure to voltage surges, were found to have developed hot spots at locations remote from constrictions due to trimming ( Figure   15), implying that new constrictions had developed

DISCUSSION AND CONCLUSIONS
The foregoing illustrations show that the liquid crystal technique is simple yet provides high spatial resolution, giving it distinct advantages over even the more advanced of the alternative methods of temperature measurement of microcircuits. In principle, the method may be applied to any planar horizontal surface and has a ceiling of about 300 C. The ready application of the technique to both active and passive components that may be used in hybrid microcircuits, and the obvious benefits of simultaneous visual examination of entire components have been illustrated by the examples presented.
Thick-film resistors have been shown to be eminently suitable components for examination by the technique, which was used in support of reliability evaluation. The observations that were made are consistent with heat conduction between the resistors and substrates. This was particularly notable when significantly non-uniform dissipation within a resistor array still produced a fairly uniform temperature distribution.
It was only in single resistors that local dissipation played a significant part the variation in thermal resistance due to different extents of trimming giving warning that families of resistors could age at different rates because of trimming cut variations within the design rules of some manufacturers.
The applications of the technique extend beyond the supportive role to reliability evaluation. For instance, it is clear that measurements of thermal resistance and temperature distributions can be used to estimate resistance changes associated with the temperature coefficient of resistance, and in calculating any derating adjustments that are rlecessary. As the technique becomes more widely used, still more applications may be discovered.