Several speakers at this conference will discuss analysis techniques that are based on phenomena that occur when the low energy ions impact on solids. These include the techniques of low energy ion backscattering and secondary ion mass analysis, both of which utilize beams of low energy ions to bombard the surface of interest. In addition to the backscattering of the ion projectile and the secondary emission of positive and negative ions, one also observes the production of intense optical radiation in the low energy particle-solid collision process. In collaboration with Norman Tolk and Douglas Simms, we have been studying the radiation that is produced in collisions of low energy ions and neutral atoms with solids. These studies give fundamental information on the low energy particle-solid collision phenomena and at the same time provides us with the basis for a sensitive technique to identify the constituents of the surface and near surface regions. The radiation that will be discussed is produced as a result of the sputtering process and, therefore, if sputtering is being used for any purpose these optical signals and the optical information can then be used to serve as a monitor for the sputtering process.

Figure 1 schematically illustrates the type of experimental apparatus which has been used in this work. This apparatus produces beams of ions or neutral atoms of well-defined energy in the energy range from about 50 eV to about 8 keV. The neutral beam is used to impact on insulating surfaces to avoid any problems that are associated with charge accumulation on the nonconducting target. In the target chamber, the target to be bombarded is oriented with the surface normal making an angle of from 45 to 60 degrees with respect to the beam direction, and photons that are produced in the collision process pass through a quartz window and are focused by a quartz lens into a small, fast monochrometer. The monochrometer and the photomultiplier are used to record the spectral distribution of radiation that is produced in the collision process and single photon counting techniques are used for sensitive radiation detection.

Table 1 lists at least three possible sources or mechanisms for producing optical radiation in the low-energy collision region.

First of all, there is radiation from excited states of atoms and molecules that have been sputtered off the surface. Basically, interaction of the beam with a solid results in sputtering of atoms and molecules off the surface. A significant portion of the sputtered fragments leave the solid in excited electronic states which then decay and give rise to lines and bands that are characteristic of surface constituents including surface contaminants. Secondly, there is radiation from excited states of backscattered beam particles. This obviously gives information on the quantum mechanical states of the backscattered projectiles. Then, finally, there is radiation which results from the excitation of the electrons in the solid. These, in general, are observed to be very broad continuum of radiation which are produced most efficiently by high velocity projectiles impinging on insulating targets. The radiative continuum that are observed here are localized to the solid and result from the transfer of projectile energy to the electrons in the solid. This gives rise to the creation of electronhole pairs which then produce the radiative continuum upon recombination. The first mechanism provides the basis for a technique to identify surface constituents. This is the technique which we call the SCANIIR technique which stands for Surface Composition by Analysis of Neutral and Ion Impact Radiation. As an analysis technique this method is very sensitive to the first few monolayers of the solid. The spectrum of radiation that one obtains is relatively simple and uncomplicated, but there are effects due to nonradiative de-excitation processes that one has to be aware of.

Figure 2 shows examples of the spectrum of radiation that are produced when argon ions at an energy of 4 keV impinge on surfaces of copper and nickel. The prominent lines that are shown in these two spectra arise from low lying excited states of neutral copper and neutral nickel. Therefore, we are detecting the sputtering of neutral atoms from the surface by detecting the radiation that they emit when they leave the surface. In addition to radiation from copper and nickel, we also see radiation from common surface contaminants such as sodium, from the sodium D lines, as well as organic contaminants from a molecular band that arises from an

excited state of the CH molecule. There is also radiation from hydrogen, which presumably results from either the fragmentation of the organic contaminants or possibly water vapor. The molecular band from the CH molecule is the most intense radiative feature that is seen from any of the common organic contaminants, and it serves as a unique fingerprint to identify specifically the fact that one has organic contamination on a surface. We have used this on numerous occasions to look for such things as photoresist residues on elemental surfaces.

Figure 3 shows results that were obtained using the neutral beam on three insulating targets: sapphire, lithium fluoride, and fused quartz. The optical lines that are shown here arise from low lying excited states of aluminum, lithium and silicon, respectively. The widths of all of these lines, at least the prominent ones, have been measured to be less than or equal to the instrumental resolution (~1 Å) and this reinforces the conclusion that these lines originate from atoms which are radiating after they leave the surface. Experimentally we find that the photon production efficiency is substantially greater for these insulating targets than in the case of metal targets. In the case of metals, a typical prominent line will be produced with an efficiency of

approximately 10-4 to 10-5 photons produced

per incident projectile. In the case of insulators, these same optical lines are produced with an efficiency that is measured to be some two to three orders of magnitude higher.

A plausible explanation for the large difference in the excitation efficiency on metal and insulating targets can be found by considering the effects of nonradiative deexcitation. When an atom is in the vicinity of a solid, there are nonradiative electronic processes which result from the interaction of the excited atom with the solid which can very efficiently compete with radiative decay of the excited atomic states. In general, there are two basic types of nonradiative processes, and they are illustrated schematically in Fig. 4. This shows a potential well diagram that is appropriate to a metal with an excited atom located a distance S away from the surface. If the excited atomic state lies above the Fermi level, then the excited atomic electron can tunnel through the potential barrier into one of the unoccupied levels. This is a one electron resonance tunneling process and is known as resonance ionization. However, if the excited atomic state lies below the Fermi level (Case B), then tunneling is prohibited since

the levels are filled. Even in that case the excitation energy can still be transferred nonradiatively to one of the available electrons in the solid and the electron from the solid may or may not appear as a secondary electron; it depends on the excitation energy of the atomic state and the depth in the well from which the upper electron was drawn. This is a two electron nonradiative process and is known as Auger de-excitation.

For both of these processes the nonradiative de-excitation rate is a very strong function of the distance of the excited atom from the surface, and it can be shown that these processes will preferentially de-excite those excited atoms which leave the surface with low velocities. This leaves only the small fraction ejected at higher velocities to efficiently contribute to the radiation. However, the band structure of a typical insulator is such that in many cases these nonradiative processes are not energetically allowed. Under these conditions, then, we expect to see radiation from all excited sputtered atoms even those that leave the surface with low velocities. We believe then that this accounts, at least in part, for the very large difference in the optical excitation that one commonly observes on insulating targets as compared to metal targets.

A good example of the effects of these nonradiative processes is illustrated in Fig. 5 where we have profiled a composite structure consisting of a 1200 angstrom thick film of Si02 on an infinite thickness of silicon. We have done this by measuring the intensity of a prominent silicon optical line as a function of the bombardment time. These measurements have been normalized to unity and then plotted on a logarithmic scale. The location of the Si02 silicon interface region is determined by that time at which the very rapid decrease in the intensity of the silicon optical line occurs. We see a factor of 50 reduction in the intensity of this optical line as we sputter from the Si02 film into the silicon substrate. The explanation for the results of Fig. 5 follows from the consideration of Fig. 6. Figure 6 shows the same type of potential well diagram, this time appropriate to an excited silicon atom in the vicinity of a silicon target and an Si02 target. As illustrated here, only in the case of the silicon substrate are the nonradiative de-excitation processes energetically allowed. Therefore in the silicon case, we expect to see radiation contributed efficiently only by those atoms which are leaving the silicon target with high velocities; only those atoms that lie in the high velocity tail. In the case

of Si02, because of the very large 8 or 9 eV bandgap it is energetically impossible for these nonradiative processes to take place. Under those conditions we expect to see radiation from all excited sputtered atoms, even those that leave the surface with low velocities. And we believe that this accounts for the large difference in the optical intensity from Si02 as compared to silicon.

This model basically says there should be a substantial difference in the velocity distribution of the radiating atoms, and you might expect to see this reflected in the width of the optical emission line due simply to the Doppler effect, i.e., the Doppler broadening of the emission lines. Experimental results are shown in Fig. 7 where we have measured the emission line profiles for the same optical line in the case of the Si02 target and in the case of the silicon target. These measurements were done at higher bombarding energies, a bombarding energy of 80 keV, and they were done in second order to enhance any possible differences in the emission line profiles. In the case of the SiO2 target we measure an emission line that has a full width at half maximum

of about 1 Å, which is the instrumental resolution that we are using for these measurements. Under the exact same experimental conditions from the silicon substrate, we measure a full width at half maximum of about 5 A. Since the instrumental resolution is the same 1 Å then most of the width of the optical line from the silicon target is real and this definitely indicates that there is a substantial difference in the velocity distribution of the radiating atoms. Experimentally, we find that the effects of these nonradiative processes can be minimized by either using oxygen as a bombarding projectile or by purposely backfilling the target chamber with oxygen gas. These, of course, are techniques that historically have been used in secondary ion mass analysis for essentially the same purposes and they drastically reduce the effects of the nonradiative processes in the optical emission.

To use this optical technique to identify the constitutents in the near-surface region of unknown solids, the procedure that we use is to simply impact the surface with the projectile beam, record the spectral distribution of the radiation, and identify the prominent lines and bands. A typical result is shown in Fig. 8. This shows the spectrum of radiation that was produced when nitrogen molecules impinged on two silica targets containing various oxide impurities. These two spectra were produced using an equivalent

neutral current of about 1x10-6 amps/cm2 and it took about 20 minutes to accumulate this data. We are using a scanning monochromator and therefore the full wavelength range must be scanned. By identifying the prominent lines that are shown in Fig. 8, one can show that there are common oxide impurities such as sodium, aluminum, and calcium in both of these. However, in one sample we have substantial optical lines from elements such as magnesium and iron, and those are not seen with any intensity in the other sample.

An indication of the sensitivity of the optical technique was obtained by using Si02 samples containing oxide impurities which were distributed homogeneously at known concentrations in the Si02 matrix. These samples were impacted with a beam of neutral atoms and the signal to noise ratio of the most prominent optical line from each impurity was measured. The detection limit for each impurity was estimated from the known impurity concentration and the measured signal to noise ratio. Table 2 summarizes results for some 10 or 12 oxide impurities in Si02. The left hand column of Table 2 lists the oxide impurities which were present in several different Si02 samples, and the right hand column gives the detection limit for the corresponding impurity. The numbers in the right hand column of Table 2 refer to the weight fraction of the oxide impurity which is necessary to be seen with a signal/noise ratio of one to one. An integration time of 10 secs was used and the particle flux was approximately 1x10-6 amps/cm2. These results were obtained using a monochromator to isolate the prominent optical line characteristic of each impurity. However, if one is willing to use a narrow band interference filter to isolate the prominent optical lines, then 2 orders of magnitude improvement can be expected simply because this increases the photon collection efficiency by 2 orders of magnitude.

One other way that we have used to estimate detection limits is to use substrates which contain impurities distributed on the surface at known concentrations. Figure 9 shows results that were obtained using silicon samples which had chromium purposely deposited on the surface. For these measurements chromium was deposited on the surface of three silicon wafers at different concentrations and each of these samples was measured by 2 MeV Rutherford backscattering to give the chromium concentration. These range from approximately 2x1016 to 2x1014 chromium atoms per square centimeter. The samples were then put into the apparatus and the intensity of the prominent chromium

optical line was measured as a function of bombardment time. The sample with 2x1016 chromium atoms/cm2 was almost infinitely thick during the time scale of these measurements. For the sample that had 2x1014 atoms/cm2, the chromium optical line was observed with a peak signal to noise ratio of approximately 200 to 1. The noise level was approximately 20 counts per second. Since the chromium coverage is about one-tenth of a monolayer on the silicon surface then one might expect to see 5 parts in 104 of a monolayer of chromium on silicon, and that could be improved also by using a narrow band interference filter.

One can also use this technique to provide depth profile information. Figure 10 shows the profile of a composite structure consisting of a layer of Si02, on Al203, on Si02, on an infinite thickness of silicon. These results were obtained by measuring the intensity of a silicon and an aluminum optical line as a function of bombardment time. In Fig. 10 you can easily distinguish the three different interface regions, but this is not a particularly good example because we obviously have problems related to the cratering phenomena. Presumably that is the reason that the aluminum optical intensity does not decrease to the noise level. Nevertheless, I think this example illustrates the potential for the optical technique and it certainly shows, for example, that if ion etching is being used for any purpose and you are interested in milling away a given layer to an interface region, then detection of the optical radiation produced in the collision process provides an in situ method to know when to turn off the beam.

One other possible application for this optical technique is to measure the range of implanted ions in silicon dioxide. Figure 11 shows some very preliminary results for an aluminum implant into silicon dioxide. The implantation dose was 1015 aluminum ions per square centimeter at an energy of 25 keV. This sample was then profiled by measuring the intensity of the aluminum optical line and the silicon optical line as a function of bombardment time. The time at which the rapid decrease in the silicon optical line occurs is a measure of the Si02/silicon interface region. If the initial film thickness is known, then the film thickness and time can be used to measure the sputtering rate for the conditions under which the experiment was performed. Knowing the sputtering rate, you can then determine the depth of the peak in the distribution of the implanted aluminum, which in this case is about 350 Å. From

the full width at half maximum you can estimate the projected standard deviation which in this case is about 160 A. The gaussian profile that one would expect is not indicated in Fig. 11. It has been fit and there is very little deviation until you get into the wings of the distribution and then it does start to deviate somewhat from the gaussian profile presumably due to effects associated with recoil implantation.

I would like to conclude by indicating the results of an experiment that was done to determine how impurity profiles of a mobile species such as sodium, which we have already heard about this morning, can be modified as a result of several different types of heavy particle bombardment. These experiments were done in collaboration with Richard Kushner and D. V. McCaughan. Previous work by these two had shown that if one took SiO2 films on silicon, purposely coated on the outside surface with sodium, and bombarded the contaminated film with positive ions, then a large fraction of the sodium which was initially on the surface of the Si02 film was transferred as a result of the bombardment to the Si02-silicon interface region. They observed that up to 10 to 20% of all the sodium could be driven through the oxide film to the interface, simply as a result of the low energy ion bombardment. They attributed this transport to effects associated with the neutralization of the incoming ion at the Si02 surface. This, of course, gives rise to a positive charge on the outside surface which then provides a source of driving electric field to move the sodium to the interface. The experiment we did was to compare sodium transport in S102 films subjected to ion and neutral particle bombardment. The results are shown in Fig. 12. The films that were used in this work were 5,000 angstrom S102 films on silicon, purposely contaminated on the outside surface with approximately 1014 sodium atoms/ cm2; the sodium contained 22Na as a radio tracer. The slices were then bombarded with nitrogen ions at 2 keV, nitrogen neutrals at 2 keV, and a third case was to use nitrogen ions at 2 keV but to flood the surface of the Si02 film with thermal electrons from a heated filament during bombardment. This was done in an effort to keep the surface neutral during bombardment. Each film was impacted with a particle dose of ~1015 heavy particles/cm2. Following bombardment, the sodium profile in the film was determined using planar etch and radio tracer counting techniques, i.e., counting the etching solutions for the 22Na removed in a given etching step. The profile results are given in Fig. 12. In the case of ion

bombardment we observed that sodium is left distributed throughout the oxide and is piled up very near to the interface region, certainly within the last 500 A of the interface region. In the ion bombardment case, approximately 10% of all the sodium that was initially on the surface was moved through the oxide film to the interface. Under the conditions of neutral particle bombardment, we cannot detect any sodium in the Si02-silicon interface region. After removing a few hundred angstroms of the S102 film, there is very little or no sodium observed above the detection limit for the radio tracer technique which is approximately 7x1014 Na atoms/cm3. Therefore in neutral bombardment there is no detectable sodium at the interface, and this is substantially less, by at least 3 orders of magnitude, than in the case of the ion bombardment. The case of using nitrogen ions with the surface flooded with thermal electrons is an intermediate case. The sodium at the interface is down by almost two order of magnitude compared to the case of positive ion bombardment, but this is substantially above that which was observed in the case of the neutral particle bombardment.

This work was done using 2 keV ions or neutral molecules of nitrogen and the targets were at room temperature. Other work, using argon as a bombarding projectile, gave essentially the same results. This work therefore suggests that using neutral particles for bombardment may substantially reduce problems associated with ion induced mobile impurity transport in Si02 films. However, a word of caution is in order. In our work a particle dose of ~1015/cm2 was used. However, much greater doses, 1017-1018/cm2, are required to profile a film of a few thousand angstroms, and it is dangerous to extrapolate our results to that dose range.

In conclusion, I want to emphasize that the study of optical radiation produced in particle-solid collisions provides fundamental information on particle-solid interactions and serves as the basis for a technique to identify constituents and contaminants on surfaces. This radiation is produced in the sputtering process and the optical information is available for a variety of diagnostic purposes in any type of sputtering experiment.