DISCUSSION OF THE PAPER

Participant: What was your primary beam current density?

Phillips: The primary ion beam intensity was approximately 250 nanoamps rastered out to cover an 800 micron square.

Cock: For one of your samples you postulated that the sodium motion was enhanced by dislocations. Would you could elaborate on that model.

Phillips: I would say that at the 200°C anneal, you have not removed the channels that are caused by the ion implantation process itself. At the 900°C anneal you have removed some of the channels but you have also mobilized the sodium. Data, which we did not present today, using the picture mode in the Cameca IMS 300 to study the location of the sodium after a 200°C anneal and after a 900°C anneal shows the difference. At 900°C you have mobilized the sodium and you actually have it diffusing back through the channels to the surface and escaping into the furnace atmosphere, making your total sodium content in the entire oxide half as large as it is with the 200°C anneal.

Cock: I see, so this is a dislocation in the oxide structure.

Phillips: Yes. If you implant sodium to a high enough intensity in the oxide structure, somewhere around 1 x 1015 atoms per square centimeter, you can provide a state in the SiO2 which is of such a high energy that the silicon and sodium, - in fact, all ions - coming off the specimen from a section extending from the surface to close to the depth at which the maximum of the implantation peak should occur, are of such a high energy compared to the secondary ion band pass of the instrument that you do not see any secondary ions. So you know that you

are producing an extremely strained state in the Si02 film that does shift even your secondary-ion energy distribution.

Participant: The temperature of 180°C that you quote is, I gather, a bulk sample temperature.

Phillips: That is correct. The microtemperature at the sputtering spot has not been determined. The temperature at which the sputtering actually takes place, or the plasma in which the ions are produced, would be significant for the theory devised by Dr. C. A. Andersen. The bulk temperature has been determined to be 180°C.

Participant: Although the temperature at the sputtering location has not been determined do you have any idea of approximately how large of a difference you might expect? Phillips: Well, in the plasma itself, or at the place at which the sputtering is occurring, the temperature is several thousands of degree Kelvin.

Evans: You are talking of the plasma or electronic temperature, not the thermal temperature. The thermal rise at the surface due to ion bombardment is only of the order of 5 to 50°C above the bulk sample temperature. The plasma "temperatures" are thermal spikes: they last the order of 10-12 seconds; they are not a temperature in the true sense, they are an electronic temperature.

Phillips: The temperature rise is extremely small because we are dealing with a fairly heat-conductive medium, silicon, and a fairly thin film of Si02 with a massive heat sink compared to the size of the sample. So I would say that even going through the Si02 there would be an extremely small temperature rise in the local area surrounding the sputtering; of the order of a few degrees at the most.

INTRODUCTION

The presence of undesirable impurities had detrimental effects on both the electrical and radiation performance of silicon-on-sapphire (SOS) integrated circuits. Conversely, device electrical and radiation characteristics can be improved by minimizing the number and types of impurities in SOS films.

Ion microprobe mass analysis (IMMA) techniques are being utilized for impurity analyses in a program associated with the development of a radiation-hardened CMOS/SOS (complementary MOS on sapphire) technology. From an applications point of view, four goals have been established for this program. The first is to improve lot-to-lot repeatability; the second is to improve wafer-to-wafer repeatability; the third is to improve the electrical stability of the devices (i.e., the bias-temperature stability of threshold voltage); and the fourth is to improve the radiation stability of the devices.

IMPURITY SOURCES

One source of contaminants is the sapphire substrate. Impurities are sometimes added during the growth process to minimize crystal defects such as twinning.

A second source of impurities is the epitaxial film itself. During the epitaxial film growth process, impurities come from silane, hydrogen carrier gas, and dopant gases.

A potential major source of impurities is the processing steps that are used during the wafer fabrication process.

MASS-ANALYSIS SILICON IMPURITY DATA

Figure 1 shows recent data comparing siliconon-sapphire films with an epitaxial silicon layer on bulk silicon. The shaded areas indicate the impurity concentration measured in silicon films on a sapphire substrate grown using the Czochralski process. The

* Partially sponsored by USAF/AFCRL (Contract F19628-75-C-0108) and by Rockwell International Corporation (IR&D).

unshaded lines on the bar graph indicate the impurity concentrations that were measured on epitaxial films grown on bulk silicon. The data are plotted as a function of elemental impurities versus the relative concentration, where each increment on the horizontal axis indicates an order-of-magnitude change in impurity concentration.

These data show that aluminum concentrations in silicon-on-sapphire are orders of magnitude greater than aluminum concentrations in bulk silicon. In the case of sapphire (A1203), there is more aluminum present in the system to act as a potential contaminant, either from top-side diffusion or from backside diffusion out of the sapphire substrate.

Another very important observation from these data is a consideration of sodium content. Sodium and potassium are known sources of electrical instability in MOS transistors. Compared to bulk silicon, typical silicon on sapphire contains much higher sodium concentrations. The data in Figure 1 show very little difference between silicon-on-sapphire epitaxial films and bulk silicon epitaxial films, in terms of potassium impurity concentration.

DIFFUSION PROFILES IN SILICON ON SAPPHIRE

=

IMMA analysis techniques are being used in the analysis of intentional dopants. Figure 2 shows phosphorous concentration as a function of diffusion drive-in time. The drivein time at 1,000 degrees centigrate was varied from 0 to 15, 30, 60 and 80 minutes. Initially (at T 0), phosphorous was concentrated close to the surface. Following high-temperature processing, the profiles changed as shown in Figure 2. We typically use silicon-on-sapphire material with a silicon epitaxial thickness in the range from 0.8 to 1.0 micron. For typical wafer-processing thermal cycles, phosphorous diffuses completely through the silicon-on-sapphire film. The diffusion coefficient in sapphire is very low and phosphorous does not diffuse through the silicon-sapphire interface. The sapphire interface acts as a diffusion stop causing phosphorous to remain in the silicon film.

MASS ANALYSIS DATA ACQUISITION

Figure 3 illustrates the instrument used to obtain impurity data presented in Figures 1 and 2. The ion microprobe mass analyzer generates an ionized beam that is swept electrostatically both in the Y direction and the X direction in much the same manner that a TV raster is generated. The ion source is allowed to impinge on the sample of interest. Beam diameters can be adjusted from 2 microns to 500 microns. The rate at which the surface is eroded depends on several experimental factors, i.e., the ion source, the intensity of the ion beam, the energy of the ion beam and the nature of the sample that is being analyzed. Erosion rates vary over a range from less than one angstrom per second up to 1000 angstroms per second. The secondary ions are analyzed electrostatically (momentum) and magnetically (mass-to-charge ratio). The ions impinge on a target where an electron beam is produced. The electron beam generates light in the scintillator, and this light is coupled to a photomultiplier tube. The signals from the photomultiplier tube can be recorded and displayed in several ways. One way is to record them on

a CRT which utilizes the signal from the photomultiplier tube for Z axis modulation. CRT Y and X signals are derived from the waveforms used to deflect the primary ion beam. Data can also be recorded on a stripchart recorder by using the signal from the photomultiplier tube to modulate the Y axis of the recorder. Data can be recorded by using a scaler to count pulses from the photomultiplier tube. Data can be recorded in the MHz range without significant data loss due to scaler dead time.

SPECTROCHEMICAL-ANALYSIS SAPPHIRE IMPURITY

DATA

Table I lists data showing impurity levels in the sapphire substrate, rather than the silicon film. These data were obtained using spectrochemical analysis techniques rather than the ion microprobe mass analysis technique just described. These data were obtained by using Crystal Systems sapphire which was grown using the gradient furnace technique. The gradient furnace technique is one of three techniques that is used commercially to grow sapphire crystals that can be used to fabricate silicon-on-sapphire integrated circuits. Table I compares pre1971 sapphire impurity data with data for the improved process that is being used at this time by Crystal Systems in the growth of gradient furnace sapphire. The number of impurities present to a significant extent are greater for the improved process, but

the impurity concentrations, in general, are much less than for the sapphire-growth process used by Crystal Systems prior to 1971.

MATERIAL COMPARISON AND DATA CORRELATION

Considering the different analysis techniques that have been used to analyze both sapphire and silicon films, data correlation is needed. Figure 4 indicates some of the work that is being done to correlate the data that have been accumulated. Data are being obtained for three types of sapphire substratesCzochralski-grown sapphire substrates fabricated by Union Carbide, ribbon sapphire substrates fabricated by Tyco, and gradientfurnace sapphire substrates fabricated by Crystal Systems. These sapphire substrate growth techniques are quite different, and preliminary indications are that the impurity concentrations are quite different for these

three techniques.

Figure 4 describes a correlation study that is being performed as a cooperative effort with NBS and Rockwell participating in the data acquisition activities. This study includes the analysis of silicon-on-sapphire wafers from three sources. Each wafer will be divided into halves. One-half of the sapphire wafer will be analyzed using the flame emission spectrometry techniques in use at the National Bureau of Standards. The other half of the wafer will be analyzed by Rockwell using the ARL-IMMA analysis technique. An attempt will be made to correlate the results of these two measurement methods with the objective of learning more about the sapphire substrates and the silicon epitaxial films, and using that information in our attempts to improve the material characteristics and the device electrical and radiation characteristics. Wafers from the same lot will be used in integrated circuit fabrication. These integrated circuits will be characterized to determine the electrical characteristics and the radiation performance of SOS integrated-circuit devices. The data from the integrated circuit fabrication and test activities will also be correlated with the analysis of the SOS starting material using the flame emission spectrometry and IMMA analysis techniques. At the same time, we have other silicon-onsapphire integrated-circuit fabrication and characterization programs that are continually providing additional data that can be used for correlation.

SUMMARY AND CONCLUSIONS

Unintentional impurities in silicon-on-sapphire films and substrates are relatively

high, based on mass analysis data. Concentrations of the unintentional impurities are equal to, or greater than, the dopant concentration of the most lightly doped region of the semiconductor.

There is a significant difference between the impurity concentrations in typical SOS devices or SOS wafers as compared to bulk silicon wafers.

Preliminary data indicate that there is a good deal of impurity-concentration variability depending upon the sapphire source and the sapphire growth technique. In addition, considerable lot-to-lot and wafer-towafer variability has been observed in silicon films and has been attributed to growth conditions and device fabrication methods, as well as impurity concentration variations.