In

The planar silicon technology for semiconductor device manufacturing was first described in 1960. Since then, planar silicon technology has become the principal method for fabricating semiconductor devices and integrated circuits (ICs). The technique has now been developed to the extent that 3,000 to 10,000 MOS components can be manufactured on chip areas which only fifteen years before could hold no more than a dozen bipolar components. To meet today's component density requirements, on chips barely millimeters on a side, design rules have been met that call for micrometer linewidths and gate oxides less than 0.1 um thick. addition, most of the device physics is found to occur within nanometers from the silicon surface. The continuing trend toward larger scales of integration and microminiaturization has consequently increased the need for quantitative measurements in extremely shallow multilayer device structures and accounts for the growing interest in surface analysis for silicon devices. Over the years, manufacturing techniques for control of the IC process have not substantially changed. Yet, it has been estimated that only three percent of the silicon entering the manufacturing process now ends in acceptable devices. New techniques are clearly called for to improve production yield and assure the quality of completed devices. Modern surface analysis can provide the IC manufacturer with the potential to identify, count, and locate, with monolayer precision, trace impurity atoms. the device scientist, these analytical methods also allow determinations of chemical bonding, carrier trapping levels, and densities of states information. Several of the surface analysis techniques are already in successful use for IC failure analysis.

For

From among the myriad of surface spectroscopies which have been spawned over the past decade, a half dozen of the most promising techniques for the analysis of silicon device structures and associated materials were selected for workshop discussion. The determination of impurity profiles, surface contamination, and interface characteristics for IC process control were held foremost in mind. The instrumentation had also to be

commercially available or capable of readily being constructed.

Referring to figure 1 we see that each spectroscopy employs a beam of primary particles to probe and interact with the impurity atoms on the device structure. A flux of secondary particles is detected which may, or may not, be the same as the primary particles. The difficulty in detecting neutrals has been reason to exclude this species as a viable secondary particle spectroscopy. The ion and neutral excitation spectroscopies on the right half of the diagram form a group distinct from the electron detection spectroscopies on the left. Of course, there are other spectroscopies, but they do not appear to meet the objectives of the workshop.

The ion and neutral excitation spectroscopies may be ranked in energy. At the low end, between 0.5 and 2 keV, is Ion Scattering Spectroscopy (ISS), while at the high end, from 0.5 to 3 MeV is Rutherford Backscattering Spectroscopy (RBS). Both spectroscopies measure the energy loss of the primary ion after it experiences an elastic binary collision with an impurity atom. The energy loss serves to identify the impurity atom through its mass. Because of its low energy, ISS is extremely surface sensitive. A most important aspect of RBS is that it is the only quantitative and essentially nondestructive spectroscopy for depth profiling available today.

RBS faces problems detecting light elements in heavier substrates, but here the complementary accelerator technique known as Nuclear Resonance Profiling (NRP) is of aid. The chief disadvantage of all these accelerator techniques is the cost and size of the equipment. However the increasing usage of ion accelerators for impurity implantation in device structures may eventually offset this.

At intermediate beam energies, from 1 to 20 keV, the impact of inert or reactive ions is effective in removing and ionizing surface material in a controllable manner. The method of Secondary Ion Mass Spectroscopy (SIMS) consists of determining the mass to

charge ratio of these sputtered ions. SIMS is the most sensitive of all the depth profiling methods, but it suffers from orders of magnitude variation in sensitivity from one element to the next.

SCANIIR, an acronym for Surface Composition by Analysis of Neutral and Ion Impact Radiation, is the optical analogue of SIMS. SCANIIR utilizes the characteristic radiative relaxations of sputtered metastable ions to identify and quantify the amount of an element present on an investigated surface. The major advantage of this technique is its ability to use neutral beams to analyze insulating films under field-free conditions.

Auger Electron Spectroscopy, AES, is the most popular of the surface analysis methods. In conjunction with simultaneous sputteretching it can be used to determine compositional depth profiles. Impurity elements are identified by the characteristic energies of their detected Auger electrons. Sensitivities vary by less than a factor of ten for the various elements, except for hydrogen and helium which cannot be detected. In addition to having superb depth resolution, AES is also capable of excellent lateral resolution in the plane of the specimen, e.g., when the focussed beam of a scanning electron microscope is used as the primary excitation.

The UV and X-ray Photoelectron Spectroscopies (UPS and XPS*) are becoming increasingly important for electronic materials analysis. UPS has been applied to study the effects of contaminants on surface potentials, and to study the trapping levels in the energy bandgap of insulators. Since UV photoelectrons originate from valence band states, or thereabouts, they are not element specific. On the other hand, the core level binding energies measured by XPS not only uniquely identify the parent atoms, but also contain information concerning its chemical environment. The chemical information contained in the XPS spectra is what makes this technique so useful.

experimentalist. There is also a certain transmutability of the instrument parameters themselves, e.g., the ability to trade sensitivity for spectral resolution. It may be well to consider the definitions of sensitivity and detection limit used in evaluating a spectroscopy. These definitions are illustrated in figure 2 for a typical impurity calibration curve.

By

The sensitivity of an analytical technique refers to the slope of a calibration curve, i.e., the change in the output signal to the increment in material concentration. It is, so to speak, the gain of the instrument. the output signal is meant the amplitude of the spectral peak less the background level. The background consists of all unwanted information which is correlated with the desired signal, and it is different from noise which is not correlated and, in principle, can be reduced to any desired level by averaging the signal over an ensemble of identical specimens or over a sufficently long period of time. The data acquisition time is, of course, limited in practice by the rate at which the surface deteriorates under the action of the probe or by contamination. Even under ultrahigh vacuum conditions, at pressures of 10-8 pascals (~10-10 Torr), a surface can become completely covered by deposited contamination in less than two hours.

The detection limit of a surface analytical technique for a specified element is usually stated as being a fraction of a monolayer. A surface measurement should ideally sample only the very outermost layer of surface atoms, a depth of less than one nanometer. Silicon has a bulk atomic density of

5 × 1022 cm 3 and therefore a monolayer of coverage would contain about 1015 atoms per square centimeter of surface. Most surface measurements today sample depths of several nanometers with the result that impurity atoms residing below the surface, but within the sampling volume, are interpreted as belonging to the surface. Consequently, the detection limit for a surface analysis may appear to be either better or worse than it

sheet to realize how complex the choice of a technique or an instrument can be.

In summary, the evolution of high component density, planar silicon technology has forced us to look for new techniques to assist or replace the traditional process control methods. Microprobes of extreme sensitivity are being called for which can identify, quantify and locate impurity atoms at surfaces, interfaces and within thin films. Only a few commercially available surface analysis techniques have demonstrated their usefulness for studying silicon device structures and related material properties. These techniques will be examined at this workshop in light of their performance for analyzing silicon, silicon dioxide, and related materials.