Specific Process Knowledge/Thin film deposition/Temescal/Ion source in E-beam evaporator (Temescal)
Feedback to this page: source in E-beam evaporator (Temescal) click here
Except where otherwise stated, this page is written by Rebecca Ettlinger, DTU Nanolab staff
Ion source in E-Beam Evaporator (Temescal)
The E-beam Evaporator (Temescal) has an Ar ion source for in-situ substrate cleaning and for ion beam assisted deposition (IBAD).
We don't have much experience with tweaking parameters for this source and have not yet used it for IBAD (2025).
Note: Due to the many different metals in rotation on the Temescal e-beam evaporator and inevitable resputtering of material from the chamber walls when using the ion source, it is possible to end up with unintended cross-contamination of the samples you wish to clean. We have had users come back to us saying they found better results from using an HF dip to remove native oxide on Si wafers than using the ion source.
Etch rate tests
During the 2018 Acceptance Test we tested the cleaning and found an etch rate of around 1 nm/ min using a standard recipe (info from the Acceptance Test here).
In 2025 we tested the ion source again for SiO2 using the standard recipe called "ion_etch_Stephen". We found good agreement with our previous results using 150 mm SSP Si wafers with SiO2 layers in the range of 30-300 nm that had been made as dummies during dry and wet oxidation in the E1 furnace. We measured the SiO2 with the Ellipsometer VASE before and after etching for 4 minutes. Two separate tests (batches) were made. Two wafers were measured in 5 points in batch 1 and 4 wafers measured in 9 points for batch 2.
Using a similar ellipsometry model to that used in 2018 (the Sellmeier model was used for SiO2, with a JAW interlayer and JAW model for the Si substrate, and the model fit the refractive index of SiO2), we found an etch rate of 1.1 +/- 0.1 nm/min for the second batch, with the uncertainty here given as twice the standard deviation for the 9 measurement points on each wafer. Wafer-to-wafer variation was around 2 % within this batch. The etch rate was ~10-15 % higher in the center than at the edge of the wafer.
Etch rate at edge: ~1 nm/min
Etch rate at center: ~1.1 nm/min
In the first batch only 5 points were fit per wafer and the fitted value of n varied >1 % from pre to post measurement, which were made on different days. The resulting etch rate was 0.5 +/- 0.1 nm/min for one wafer and 0.9 +/- 0.1 nm/min for the other, but these results may be less trustworthy than those listed above for batch 2, where the fitted value of n varied <0.4 % between pre and post measurement and measurements and etch were all done on the same day.
A subsequent test a few months later yielded an etch rate of 0.8 nm/min on a 4" wafer. The same model was used in the ellipsometry fits and the fitted values of n for the pre- and post measurements were identical within 0.1 %. Here the etch rate was marginally stronger (4 %) at the wafer edges than the center.
Conclusion: While the rate does vary somewhat from run to run, around or just under 1 nm/min is a good rule of thumb for SiO2 etching with the "ion_etch_Steven" recipe.
Full results are found here:
File:ion etch pre minus post.xlsx
File:Temescal ion etch.pptx
Native oxide removal on Si
To remove the native oxide on Si samples we recommend etching with the standard recipe "ion_etch_Stephen". Exactly how long the etching should be done is not totally clear. Our testing of the recipe was with thermal oxide, which probably has a different composition to native oxide (there is probably a transition region from Si to SiO2). And what is the thickness of native oxide? It's not so easy to measure as ellipsometry measurements are not so accurate in this region exactly because the material properties used for modeling are not well determined.
In this paper the native oxide on a Si wafer is measured by TEM and is found to be 2.2 nm thick (see Figure 6 a).
Other papers mention similar thicknesses and also talk about the sublayer composition, e.g., this 1991 article.
Based on an expected thickness of 2-2.5 nm of the native oxide and expecting it to be etched in a similar manner to thermal SiO2 we recommend 2-3 minutes of etching to remove all native oxide. Be aware that this treatment may result in roughness and heating (not tested).
Ion etching or IBAD in combination with III-V materials
We are a little wary of using the ion source with III-V materials in the Temescal as some (toxic) III-V material or reactants hereof will be removed and eventually pumped away or incorporated in the shields. Additionally we need to limit cross-contamination. For this reason please contact staff if you wish to use the ion source with III-V materials. You may be asked to make a risk assessment.
Some general thoughts:
- We are more worried about GaAs-based materials than InP-based materials
- For GaAs it may be ok to perform a short mild etch to remove native oxide on a few small chips, but we would need to consider carefully if it can be allowed on a routine basis, for longer times or with larger substrates.
- To avoid contamination of the metal sources used by all users we need to think about which source is exposed while the ion beam is on. It may be better to expose, e.g., Ge (commonly used for III-V applications) than for instance Au or Cr. Please talk to staff about how to ensure this.