Ellipsometer or reflectometer: choosing a film thickness tool

Every deposition, oxidation and etch step needs a thickness measurement, and in most research cleanrooms that measurement comes from an optical tool. The two common choices are a spectroscopic reflectometer and a spectroscopic ellipsometer.

Both are widely available used, from tabletop reflectometers to research-grade ellipsometers such as the J.A. Woollam M-2000. They overlap for many routine films, but they answer different questions, and buying the wrong one leaves either your thin films unmeasured or your budget spent on capability nobody uses. This article explains how each works, where each fits, and what to check on a used metrology tool.

How reflectometry works

A reflectometer shines broadband light onto the sample, usually at or near normal incidence, and measures the reflected intensity across a range of wavelengths. Light reflected from the top and bottom of a transparent film interferes, producing a pattern of peaks and valleys in the spectrum. Software fits a model to that spectrum to find the thickness.

For this fit to work, the tool needs to know the optical constants of the film and substrate. For common materials, such as thermal oxide, LPCVD nitride, polysilicon and standard photoresists, these are in the library. The measurement takes a second or two, and many systems, including the Nanometrics NanoSpec 6100, can map a wafer with a motorised stage.

Reflectometers are at their best with transparent films from tens of nanometres up to several micrometres, depending on the wavelength range. They are simple to operate, so students and operators can measure their own wafers with little training.

How ellipsometry works

An ellipsometer measures how the polarisation of light changes on reflection, usually at an oblique angle near the Brewster angle of the substrate. The result is two parameters per wavelength, Psi and Delta, rather than a single intensity. Delta, the phase difference between two polarisation components, is very sensitive to thin layers, down to sub-nanometre changes.

That extra information lets an ellipsometer:

• Measure very thin films, such as native oxides, ALD layers and self-assembled monolayers
• Determine optical constants (n and k) of unknown or new materials
• Resolve multilayer stacks, interface layers and surface roughness
• Characterise absorbing films, such as thin metals and semiconductors, where reflectometry struggles
• Measure at several angles to separate correlated parameters

Spectroscopic ellipsometers with a rotating compensator and array detector, like the Woollam M-2000, acquire a full spectrum in seconds. Wavelength range, focusing optics, mapping stages and in-situ mounting depend on configuration.

The trade-off is modelling. Ellipsometry rarely gives a number directly. The user builds an optical model of the stack and fits it to the data. For known films, saved recipes make this routine. For new materials, it takes skill and time.

Comparing the two for common lab tasks

The thresholds are indicative. Exact limits depend on the instrument, wavelength range and material.

Matching the tool to your lab

A teaching or general-purpose cleanroom benefits most from a reflectometer next to the furnaces and coaters. It is quick, robust and easy to learn, so it gets used. Most oxide, nitride and resist checks in coursework fall comfortably within its range.

A research lab working on thin films, especially ALD, 2D materials, optical coatings or new dielectrics, needs an ellipsometer. The ability to extract optical constants and measure single-nanometre layers is the whole point of the work. Our guide to thermal and plasma ALD explains why ALD users rely on ellipsometry for process checks.

Pilot lines often run both: an automated reflectometer for routine thickness checks on every lot, and an ellipsometer in the process engineering lab for development and troubleshooting.

If your budget allows only one tool and your users are mixed, an ellipsometer can do almost everything a reflectometer does, but slower and with more training. Make sure someone on staff is willing to own the models and recipes, or the tool will sit idle.

Running costs, training and placement

Day-to-day costs are modest for both tools, but they differ in where the effort goes.

A reflectometer's main consumable is its lamp, and its main overhead is keeping the material library honest. If a user deposits a film whose optical constants differ from the stored values, such as a silicon-rich nitride measured with a stoichiometric nitride file, the tool still reports a number, just the wrong one. A short list of approved recipes, each tied to a process, avoids most of these errors. Some reflectometers also cannot measure films on transparent substrates such as glass, because the back surface adds its own reflection; check this if your users work on glass or quartz.

An ellipsometer's lamp is a consumable too, but the larger cost is staff time. Somebody has to build and maintain models, check new materials and help users interpret fits. Budget training for at least one owner of the tool, and keep reference samples nearby.

Placement matters for both. A reflectometer earns its keep next to the furnaces, PECVD and coater, where users check every wafer. An ellipsometer can sit in a quieter corner of the cleanroom or a characterisation lab, provided samples reach it without contamination.

What to check on a used metrology tool

Optical metrology tools have few mechanical wear parts, but their accuracy depends on calibration and on parts that age quietly.

For both types:

• Light source. Lamps have a finite life. Check the hours and whether intensity is stable across the spectrum, especially at the ends of the range.
• Detector and spectrometer. Look for dead pixels, noise or drift. A measurement on a reference wafer reveals problems quickly.
• Optics. Check fibres, lenses and windows for damage, haze or contamination.
• Stage. Motion, vacuum hold-down and mapping accuracy on motorised systems.
• Software and licences. Confirm that the analysis software runs, that the licence transfers and that the material library is intact. Older software may need a specific operating system.

For ellipsometers, add:

• Polariser, compensator and analyser. These must be aligned and calibrated together. Ask for a recent calibration result.
• Angle of incidence. On variable-angle systems, check that the goniometer moves cleanly and reports the angle accurately.

A short acceptance test on reference wafers of known thickness, such as oxide thickness standards, shows whether the tool reads within its expected accuracy.

What we can do

We inspect and refurbish ellipsometers and reflectometers, replace lamps and optical components, and check results against reference samples. Our calibration service covers periodic verification so results stay consistent between users and over time. Tell us what films you measure, and we will suggest which tool fits.