Characterization of SiO2 Thin Films Using
an SEM and Other Imaging Techniques
by Christopher Bailey
-- Samples were imaged for this project using a wide variety of imaging techniques learned in class lectures as well as the practical setting --
Silicon dioxide is is widely used in the microfabrication domain as an insulating layer. This can be used to shield conductive layers from shorting as well as a protective top-coat for complete IC chips. Because of its insulating properties, silicon dioxide is inherently non-conductive. This presents an obstacle for imaging when using high magnification microscopy. Typical SEM systems use an electron beam's surface scattering properties to image samples with sub-micron (or less) features. With this setup, conductivity of the sample is desirable (almost necessary) for quality images. Nevertheless, the obstacle of non-conductivity is surmountable. Various methods of coating samples with conductive materials exist. Carbon sputtering, gold sputtering and carbon painting are among a few. These can also have negative effects on samples, the magnitude of which scales directly with magnification. Gold coating was used for this imaging project although precautions were taken regarding coating thicknesses, since the objective included high magnification.
My project goals were to use three different types of Silicon Dioxide (SiO2) with the variable being method of deposition and film thickness. The three types of deposition techniques were sputtered, thermally grown, and PECVD (TEOS method) deposited films. It is my intention to image them the best I could given the inherent set-backs, and compare what I could of them. It turned out to be very difficult to image at the molecular level and any other attempts to obtain crystal information were fruitless due to SiO2's lack morphology. Many of the images on display below represent all attempts at showing properties of thin films using micrographs and other imaging technology that my samples allowed.
Above is a depiction (left) and a low mag SEM (In Lens Det.) picture of my sputtered and thermal samples. These were originally made so they could be used in the TEM as well as any other imaging device. The top face is bulk silicon (in blue) and continues down sides in octagonal (due to wafer cleaving) and pyramidal (due to etching) respective patterns. At the center of the pyramidal etch pit, is a smaller square (depicted in green, and at a different angle on the left), is the actual thin film of SiO2. The samples I imaged were of the following thicknesses:
Sputtered -- 20 nm
Thermal -- 540 nm
TEOS sample -- 1 um
The TEOS sample I had left over from a previous experiment and never needed it for the TEM, so I left it as just a piece of silicon with the SiO2 coating.
To show what kind of imaging issues I had, the next two pictures are of charging:
The picture on the left is the best image I could get of just bare TEOS SiO2. We can begin to see small particles (about 1 um), but the image is horrible. On the right, we can see the path the beam took as I zoomed and attempted to focus. This is similar to the raster box of charging, except that it is being dragged along and changes size.
To correct for this, I sputtered 100 nm of gold onto my samples, and found that at the imaging magnification I needed for this project, the only image I would be able to obtain would be of the gold coating itself:
So solve this problem, for the sputtered and thermal samples, I flipped over the film so that the etch pit faced up towards the electron gun. This arranged things so that the film (usually broken from handling), was placed directly onto the carbon tape used to mount the sample. So in theory, the e-beam would completely penetrate the sample and be grounded through to the carbon. This way I wouldn't need to coat the samples, and I could still get some decent images. The theory worked fairly well and below are some of the more interesting of the pictures under significant magnification (description below pictures):
The first of these three is a nice depiction of how thin these oxides are. This is the sputtered sample w/ 20 nm of film. The film itself has folded over onto itself and has stuck to itself due to static friction. As you see it folded over, you can see directly through to the silicon side wall. The second picture shows the ripping of one of the films (during handling) and how it stayed due to static friction during mounting and is now stuck down to the carbon tape. Here, you can see charging occuring slightly on the tips curling upward away from the carbon tape, since there is no grounding. The third of these images shows the stress that can be induced upon drying after etching (also enhancing the rip occurances). Small ripples can be seen throughout the film and finally failure where the stress was presumably too strong. The last two were of the thermal oxide sample.
The carbon tape seemed to give off its own image possibly indicating a non-uniformity in thickness. Below, we see the first three pictures of this occurance while all four display the magnitude of electron-transparency this film exhibits:
Next we look at two samples of TEOS prepared SiO2 samples. These in particular have a peculiar particle abundancy on their surface. Since these were taken at a bit higher mag, and did not have the suspended membrane feature, they were coated lightly with about 20 nm of gold which did not in fact show up in the image this time.
As shown in the picture to the right, we estimate these globules to be between 1 and 1.5 microns. This proves to be a fairly rough surface, since they are very frequent and were viewed every where on the surface, and since their size was on the same order (if not bigger) than the film thickness.
As previously stated, it is expected that the e-beam will penetrate the thin films and move through to the carbon. This was proven in the pictures above. However, I wanted to see if the 1 um thick TEOS sample would produce similar results. Since we could not image it to prove this, I used the Electron Flight Simulation program to predict the paths of the incoming electrons. Below are my results & discussion:
These balls of scatter lines represent the predicted electron paths. The white bar at the top of the charts represents the TEOS layer, and the gray under-portion, represents the bulk silicon upon which it is deposited. The chart on the left is modeled with a 10k accelerating voltage and the right chart is done with a 20k. It is clear that as you increase the accelerating voltage, you can much more easily pass through insulating layers.
This technique lead me to show some other interesting results. Seeing that most of the electrons in the the right chart are passing through to the silicon, as opposed to the left (10k) chart, I wanted to see how this would effect the x-ray spectroscopic data. Below are the two same conditions metered with the actual sample at the two respective accelerating voltages:
As we may be able to (the font is a bit small), the left most spectrum shows us a more normal Si to O ratio as expected to be in SiO2 films. As we increase our accelerating voltages, however, Si becomes the predominant element in our sample, since more electrons are imaging the Si.
So show this more clearly, pie graphs of the quatitative elemental percentages are displayed below:
Summary & Conclusion
Novel steps were taken to overcome charging since typical methods were employed and failed. Thin films have unique properties including stresses and flimsiness shown by folding and stiction. Ripping and tear can occur without much effort and with as little movements as light handling. TEOS oxide is not necessarily very uniform. If insulators are thin enough, electrons can pass directly through it, and an image can be captured with much success and improvement without coating. This can be shown for non-membranes by using an electron path-predicting program such as the one used here. It can be showen through x-ray spectrometry & its complimentary quantitative analysis, that the higher the accelerating voltage, the deeper we image.
7 Techniques Employed:
Coating (showed undesirable results)
TEM (showed no results)
STM (showed no results)
EDAX X-ray Analysis
Electron Flight SImulator
Image Enhancement (each picture was sharpened using Adobe Acrobat filters)
Special Thanks to BRIAN MCINTYRE and CHRISTOPHER STRIEMER for their constant source of faith in my project and seemly endless advice & ideas.
Course OPT 507, U of R
(c) 2004 Christopher Bailey/Brian McIntyre