Spectroscopic Ellipsometry: The Swiss Army Knife of Optical Metrology

View the inaugural presentation from Covalent Academy and learn about one of the most versatile optical characterization techniques available. Spectroscopic Ellipsometry (“SE” or “spectral ellipsometry”) can be used to investigate a wide range of properties in advanced materials and coatings. Experts combine powerful modeling with raw optical parameters to extract measurements of surface roughness, crystal strain, defect densities, chemical bonding states, and more.

In this webinar event, Covalent speaker Dr. Maxwell Junda will be introducing the fundamentals of ellipsometry, and discussing several case studies relevant to industry. Dr. Junda will guide you through core concepts of ellipsometry analysis and demonstrate just how powerful the technique can be. Make sure you don’t miss this swiss army knife in your metrology toolkit!

This Webinar Will Answer

  1. What is spectral ellipsometry and how does it work?
  2. Why is modeling important for spectral ellipsometry, and what are good strategies for modeling SE data?
  3. How is spectral ellipsometry used to analyze optical properties of materials?
  4. What kinds of problems can you solve with spectral ellipsometry?

Q&A Session


Can ellipsometry be done on coatings on machined metal surfaces? What are the limitations?

In general, yes. In order to measure coating thickness requires semi-transparent and transparent coatings because the measurement beam must be able to transmit through the film, reflect off the metal substrate, and return back through the film to the detector. There aren’t generally any limitations for this specific type of sample beyond the normal limitations for ellipsometry. For thicker films (>0.5 microns) reflectometry is typically a better option.

You start with linearly polarized light but the reflected beam could become circularly polarized, right? Can you characterize nk of birefringent materials?

Yes and yes. To be more specific, linear polarization and circular polarization are actually just two special “end point” cases of polarization states. The more general polarization state is usually somewhere between these and is called elliptical polarization. This is where ellipsometry gets its name. In fact, ellipsometry does not actually require the incident beam to be linearly polarized. It can have any polarization as long as this polarization is known since the measurement is really just detecting the relative change in polarization state from incident to reflected beams. Birefringent materials can be absolutely be characterized, however, the measurement and analysis is often more complex than it is for isotropic materials.

How can we use the ellipsometry to measure the thickness of polymeric film it should be around 7 to 10 nm? We are having problem while building the model and the data we got is eccentric.

It is difficult to address the root of this issue without more information. From your description, it sounds like it can be done, however. Perhaps the films are not homogeneous throughout their thickness? Another approach is to make a couple of samples of different thickness and attempt to create a model that can consistently model them all. Or you could use another technique (i.e. TEM) to verify and fix the thickness in your model so that you can focus the fitting entirely on determining the correct optical properties. If you’d like assistance building a model, feel free to reach out to us directly – we offer SE modeling as a service at Covalent.

What are the units of k?

The extinction coefficient (k) is unitless (so is n). By contrast, describing light absorption in terms of the “absorption coefficient” (usually denoted alpha) is related to k by [alpha] = 4*pi*k/[lambda] where lambda is the wavelength. The absorption coefficient is typically reported in units of inverse cm.

Regarding the instrumentation, could you speak more on what information you extract in the IR range? What changes are made to measure anisotropic samples?

In general, there is no difference in the basic principles and modeling of IR ellipsometry from any other spectral range. There are some practical considerations about sample size, thickness sensitivity, etc. that are different, mostly due to the longer wavelenths used, however. In IR ellipsometry, the models still consist of optical spectra (n,k) and layer thickness. However, the physical phenomena that are dominating the shape of the n&k are often from a different origin than the phenomena that influence the shape in other spectral regions. In the IR, we are usually looking at molecular vibration modes, phonon modes, and free carrier absorption. We will likely discuss this further in an future episode on more advanced ellipsometry topics.

What are the limitations of the multilayer samples in terms of thickness and meterial?

Unfortunately, there isn’t a single simple answer to this because it often depends on the details of the layers and materials. However, ellipsometry excels with samples consisting of layers that are about 500 nm or less. It can handle layers that are thicker, but the accuracy of the optical property results begins to be reduced at larger thicknesses. An additional limitation is that the materials must be either fully transparent or partly transparent in order to have sensitivity to any underlying layers.

How do you consider roughness of thin films in trying to model the delta and psi curves?

Surface roughness is almost always described as a separate layer which is an effective medium approximation that consists of a combination of material identical to the underlying film and void.

Are Psi and Delta mirror images of each other?

No. Often they can look to be related because an absorption feature in the material will affect both of them at the same wavelengths, but they are completely different parameters. Delta is related to the phase change, and psi is related to relative amplitudes.

Do you calculate the mathematical area or absolute area between the model and the experimental data to guide the next step of iteration?

Calling it “area” is actually a bit of an oversimplification I used to help make the general concept more accessible. The error functions root mean squares of the difference between model and measurement at each spectral point.

How do we go from k to alpha (or viceversa)?

Alpha = 4*pi*k/lambda (lambda is wavelength)

Does the thickness of the material/sample matters?

In a general sense, no. The modeling can handle a wide range of layer thickness. However, there are some limitations. These have been discussed in other answers, but generally, the accuracy of results begins to substantially degrade for films over ~500nm and the films must be transparent in at least part of the measured spectral range to get thickness results.

Can ellipsometry be used to interrogate surfaces of prisms, or other various optic shapes?

If I understand your question correctly, i.e. can ellipsometry measure planar surfaces of prisms, then the answer is yes.

Sometimes you will get the goodness of fit value, Chi square very low, but the extracted values are unrealistic, like negative loss. So, is it true that we cannot simply rely on Chi square alone find a good fit to experiment data?

Yes. As in most modeling, you need to consider the physical limitations of the model. Sometimes this type of thing can result from having an anisotropic material being modeled by an isotropic model. Or, in other cases, the model is not a good representation of the actual sample characteristics such as if the model uses a single homogeneous layer to model an inhomogeneous layer. Any insufficiencies in the model configuration will then manifest themselves as artifacts in the optical properties that are compensating for the insufficiency.

What maximum thickness you can work with? Does it matter if the material is transparent/opaque in visible(400nm-700nm)?

Maximum thickness is material dependent and there is no universal maximum thickness. The accuracy of the results and ease of modeling decrease with increasing film thicknesses after about 500nm (that’s a rough benchmark). As long as the material is transparent in some portion of the measured spectrum, we can gain sensitivity to the thickness.

What is the extra information from measurements at multiple angles compared to a single angle?

For isotropic samples, all information about the sample is contained in a measurement at a single angle. However, fitting a model to multiple angles ensures that the model is reproducing the measured data at multiple angles which can reduce correlation between fit parameters and ensure that the model result is consistent. With a single angle, it is possible to “coincidentally” fit the measured data well with a model that doesn’t accurately describe the true sample properties. With multiple angles, the likelihood of this case is drastically reduced. For anisotropic samples, multiple angles of incidence can result in the measurement beam taking different paths through the sample which can help provide sensitivity to the anisotropy. Usually, however, a combination of a few strategic measurements are best for gaining sensitivity to anisotropic properties in a sample.

What is the limitaion of the choice of wavelength based on the surface roughness?

The shorter the wavelength, the more roughness adversely affects the signal intensity and results. Both the Woollam NIR-UV and IR instruments that Covalent uses, however, collect the entire spectrum for each measurement automatically. If part of the measurement is deemed unreliable due to poor intensity, it can be removed from analysis following the completion of the data acquisition.

What is the minimum surface roughness that ellipsometry can detect?

Same as the minimum film thickness: sub angstrom but depends on the sample.

What is the max thickness ellipsometry (up to 1000nm wavelength) can measure?

Other answers address this, but in general, it depends on the material and desired result accuracy. The accuracy decreases with increasing thickness. Typically, 500nm or less is a good benchmark to be within the range where ellipsometry is best suited, although it can handle films that are multiple microns thick depending on the application and desired results.

How do you define the optical properties of surface roughness? Some effective refractive index models?

Correct. We use an effective medium approximation that mixes the optical properties of the underlying material with void. It depends on the material, morphology, roughness thickness, and other factors, but sometimes this is fixed at a 50%/50% mixture, other times the relative proportion of material and void can be a fit parameter.

How do you get the composition from the index of refraction?

Composition changes change the optical index and the relationship between the index of refraction and composition can be “calibrated” by complimentary characterization that is sensitive to the composition. Once known, this relationship can be used to directly determine composition in ensuing measurements.