CHAPTER: SPUTTERING RATES
RELATIVE SPUTTERING RATES AND ION YIELDS OF SEMICONDUCTORS, METALS, AND INSULATORS UNDER OXYGEN AND CESIUM ION BOMBARDMENT
We measured the sputtering rates and ion yields of a number of semiconductors, metals, and insulators under identical oxygen or cesium ion bombardment conditions. The materials include Si, Ge, GaAs, GaP, InP, InSb, HgCdTe, CdTe, and diamond, some metals from Be to Au, and some insulators, including SiO2, Si3N4, Al2O3, and LiNbO3. The semiconductor materials and Al2O3 and LiNbO3 are single crystals; the other materials are polycrystalline or amorphous. An example of an application is the ability to determine and plot mixed material depth scales and impurity densities, such as SIMOX SOI (Si/SiO2/Si), Si/Ge layered structures, GaAs on Si, etc. These results are also useful for setting up appropriate SIMS conditions when a desired profiling rate is specified.
We measured the crater depths created by sputtering polished surfaces of a variety of materials under identical conditions using both O and Cs primary ion beams, each group of samples internally calibrated via a Si sample, and divided these depths by the beam current and the sputtering time, and normalized them to the Si results, to give relative sputtering rates compared with Si as a standard. We also calculated the absolute sputtering rates for a specific set of conditions for Si, namely for a 500 mm sq. raster. The results are 0.445 and 1.13 nm/s/mA for a 500 mm sq. raster for O and Cs, respectively, from which all other absolute sputtering rates can be determined via comparison with the Si standard, using the nuclear density of each material. Data are presented in Table 1 for 18 materials that include 9 semiconductors, 5 metals, and 4 insulators. We estimate that the error in these sputtering rate data is about ~10% for the semiconductors and metals, and about ~25% for the insulators. The estimated error in the absolute sputtering yield S is larger because of a lack of certainty in the measurement of the magnitude of the primary beam current and the possibility that secondary charged particle suppression by the detector systems is not complete nor uniform among instruments. This error is probably about 40% for the data measured with O bombardment, and could be a factor of 1.5 to 2.5 for the data measured with Cs bombardment.
It surprised us that the sputtering rates for Au under oxygen bombardment were less than for most semiconductors. Therefore, we repeated the measurements for Au for each bombarding ion at three different times, and obtained consistent results. This result is assumed to be caused by the large mass mismatch between O and Au. The sputtering yield for Au under Cs bombardment is relatively larger; the mass match is better between Cs and Au. The data for GaAs were measured three times, and most other matrices were measured twice. Si was measured seven times, being the standard.
The sputtering rates measured under these conditions are larger for semiconductors than for either metals or insulators (those studied). The incident energy may be more readily absorbed in the covalent bonded metals and less transferred to the surface to result in atom ejection (sputtering). The strong bonds in these insulators may cause lower sputtering rates. For semiconductors, the sputtering rate is dependent not only on the mass of the atom, but also on the "crystal strength" or bond energy in the crystals, which decreases as Si > Ge > GaAs > InP, so the increase in sputtering rate for the semiconductors is probably the result of a combination of crystal strength and mass (atomic number).
Observations or trends that we see in these data are:
1) Cs sputtering rates are approximately 2.5 times those for O.
2) Sputtering rates are larger for semiconductors than for insulators or metals.
3) Sputtering rates increase with atomic number of the matrix more rapidly for Cs bombardment than for O bombardment.
4) Relative sputtering rates (rule of thumb) for Si, GaAs, InP, and HgCdTe vary approximately as 1, 2, 3, and 5.5 for 4-keV O bombardment.
5) Sputtering rates for Si, a-Si, SiO2, and Si3N4 are approximately the same.
6) Using the 0.445 and 1.13 nm/s/mA for a 500 mm sq. raster for O and Cs, respectively, we can calculate the approximate maximum sputtering rates for a CAMECA SIMS instrument for depth profiling. Assuming a 10 mA maximum current for both O and Cs instruments, into a raster size of 100 mm, the results are about 160 nm/s for O and 560 nm/s for Cs on Si, 450 nm/s for O and 1100 nm/s for Cs on GaAs, etc. for the other materials.
TABLE 1. Relative sputtering rates (SR) for 4-keV O and 14.5-keV Cs ion bombardment (relative to Si). (Multiply by 0.445 and 1.13 nm/s/mA for a 500 mm raster for O and Cs, respectively, to obtain absolute rates.) Approximate absolute sputtering yields (S) in atoms/ion.
______
Matrix O Cs Matrix O Cs
SR S SR S SR S SR S__
Si 1.00 0.9 1.00 2.3 Be 0.3 0.7 0.95 5.3
a-Si 0.94 0.8 0.92 2.3 Al 0.51 0.6 1.07 2.9
Ge 1.71 1.4 2.3 4.6 Ti 0.54 0.6 1.03 2.6
GaP 1.74 1.5 3.2 7.2 Cr 0.69 1.0 1.26 4.7
GaAs 1.99 1.6 3.2 6.4 Au 1.38 1.5 4.42 12
GaSb 2.80 1.8 2.9 4.7 Al2O3 0.34 0.4 - -
InP 3.04 2.3 6.6 13 Si3N4 0.82 1.5 0.98 4.6
InSb 3.36 1.8 6.5 8.7 SiO2 0.95 1.2 0.94 3.0
HgCdTe 5.6 3.5 16 25 LiNbO3 0.36 0.6 - -
Diamond 0.42 0.28/0.98___
Results for additional measurements of SR (relative to Si) made for oxygen bombardment are
Be 0.32
Al 0.42
Ni 0.76
W 0.29
Au 1.49
Sputtering rate is determined by:
* mass, energy, and angle of incidence of the bombarding ions
* mass (masses for compound targets) and (weakly) crystal condition, including amorphous, polycrystalline, or single crystal, crystal orientation, and temperature
* current density of the sputtering ion beam, which is the beam current that impinges the crater area divided by the rastered area
The first two of these are generally contained in the sputtering yield S (sputtered atoms per incident ion). Curves of S versus ion Z, target Z, incident ion energy, and angle of incidence are known for many conditions. We are primarily concerned here with the third bullet; the primary current and rastered area are major variables for the SIMS process, and are those that allow the sputtering rate to be varied for depth profiling and impurity analysis.
Sputtering rate in depth profiling then varies as the primary ion current and inversely as the square of the raster dimension. For example, if one should desire to increase the sputtering rate by four, the primary ion current could be increased by four (if possible). the raster dimension could be decreased by a factor of two (from 500 to 250 mm, e.g.), or by a combination of both to cause the same increase of four.
Considerations:
* Decreasing the raster size too much can decrease the profile quality as the result of crater side-wall effects (cause a tail on the deep side of the profile).
* Increasing the primary beam current could cause charging difficulties for a dielectric matrix
* Increasing the sputtering rate may decrease the depth resolution because of greater spacing of the data points (in depth).
* Increasing the sputtering rate nearly always improves the background count rate and the detection sensitivity.
* Increasing the sputtering rate decreases the profiling time.
A balance must be reached among profiling time. detection sensitivity (background), depth resolution, and profile shape quality.