Measuring Transparent Layers

After several days of work extracting and demetallizing the bubble memory wafer described in the section on Chaos (page 7) I thought that it would be interesting to determine the thickness of the YIG layer deposited on the GGG substrate. Unlike the opaque layer measurement challenge described on the previous page, now I wanted to determine the thickness of a transparent layer deposited on a transparent substrate.

Measuring transparent thin films

There are a number of ways to measure uniform, transparent thin-film coatings. Assuming access to a scanning electron microscope, it is possible to prepare a section of the wafer and examine it under magnification. Fortunately, there are nondestructive approaches.

One nondestructive technique is performed with a sophisticated optical device called an ellipsometer. Unlike the magneto-optic Kerr effect (MOKE), no magnetic field is needed since coated surfaces can produce elliptically polarized reflections under incident linear polarized light under certain conditions. The technique involves very sophisticated software that creates mathematical models of the thin films and substrates to compare to the test sample. Ellipsometry is very accurate, enormously powerful (e.g., its ability to handle complex situations involving multilayer films, varying refractive indices, extinction coefficients, etc.) and very expensive, of course.

A less expensive and somewhat less accurate way to measure uniform, transparent thin films is to use a device called an optical reflectometer. It really involves producing a channeled (also channelled) spectrum and fringes of equal chromatic order (FECO). If you have access to a prism or diffraction-grating spectrometer, you can see a channeled spectrum using methods described in various optical lab experiments manuals. The thicker the film, the more absorbing bands visible across the observed spectrum. It's straightforward and interesting, but you wouldn't want to measure thin films with an optical spectrometer more than once.

A homemade optical reflectometer

The easiest way to perform thin film measurements is with a PC-based spectrometer. I happen to have an Ocean Optics PC2000 plug-in card spectrometer intended for use in the visual spectrum. It came with a single fiber-optic cable and standard transmission/absorbance software. Usually, to convert the instrument into a reflectometer would entail adding an expensive reflection probe cable, a light source, and specialized software. The reflection probe is comprised of a bifurcated fiber optic assembly that joins one light-carrying cable hooked up to a tungsten halogen lamp to another cable used to transfer the optical signal to the PC2000. The two cables become one and irradiate a sample sitting on a stage. The light is reflected from the sample and transferred to the computer-resident spectrometer. In principle, the instrument can measure film thickness ranging from 50 angstroms to 250 microns. The specific range will depend on the particular configuration of the instrument.

To bypass having to purchase a reflection probe, I cut a short length of aluminum rod 1/2" in diameter. One end was cut at a 45° angle and that end was ground and highly polished to act as a mirror. A tiny hole was drilled along the axis to carry the reflected signal to the computer. At the other end, I attached an appropriate coupler for the fiber optic cable.

A bright light was aimed at the mirror at a 90° angle to the rod axis as shown in the photo at right of the homemade reflectometer probe. The light reflects down to illuminate the sample, in this case the Intel 7110A-1 YIG/GGG wafer sitting on a black paper-covered stage. You can see a hint of the fiber optic cable connection at the top of the aluminum rod.

The picture to the right of the reflectometer probe is a photo of the computer screen showing the waveform actually obtained from an extracted wafer. The waveform is shown as a transmission spectrum using Ocean Optics' standard OOIBASE32 software. Mathematically, the software handles reflection in the same way as it handles transmission. Plugging the appropriate waveform parameters into an equation, along with an assumed index of refraction in the visible spectrum of 2.2 for the YIG layer and 1.965 for the GGG substrate, yielded a value of 3.055 microns for the YIG layer's thickness. There are a number of other assumptions used in the measurement, including the assumption that I was dealing with a single layer rather than with multiple layers. Looking at the waveform, one can detect an underlying low-frequency waveform on which the higher frequency wave is riding. This might indicate a second, very thin layer in contact with the YIG. Fourier transforms could help here.

I could not find layer thickness specifications for the Intel 7110A-1 chip, but I did find specs on a different chip made by Fujitsu. The Fujitsu chip listed a YIG thickness value of 3.1 microns. If the chips are really similar, this would place my measurement within 1.5% of the actual thickness. Promising, but without independent verification the 3.055 micron measurement falls within the category of a definite maybe!

In addition to film thickness, reflectometry can also be used to measure refractive indices and extinction coefficients. It can even work with two or three layers of thin films. At some point, however, as samples get complex, commercial instruments with their powerful software are going to prove irresistible.

Reflectometry measurements are very fast and, if your requirements like mine are modest, the workaround described could prove quite useful. For example, one can set up a system to monitor transparent or partially transparent layers being deposited in a vacuum system or in a sol-gel. Finally, I should mention that reflectometry finds application in microscopy, and it can be employed in a transmission mode as well.

Next up, a look at the possibility that diatoms may have peculiar optical properties.