A curated technical Q&A for evaluating TFM reflectance-based thin-film thickness measurement, sample conditions, recipes, fit quality, and integration.
TFM measures reflected light from a film stack and compares the measured reflectance spectrum with a calculated optical model. Thickness is estimated by finding model conditions that reproduce the measured spectral shape.
The reference measurement normalizes the optical system response, while the dark measurement removes detector and background offset. Stable reference and dark conditions help routine measurements remain comparable.
A normal-incidence reflectometer is practical for fast thickness checks, recipe-based operation, and compact equipment integration. However, the result depends on the film stack, optical contrast, known n,k data, backside reflection, and selected fitting wavelength range.
Yes, oxide and nitride films on silicon are representative TFM applications when the film stack and wavelength range provide sufficient spectral contrast. Recipe setup should match the actual layer order, substrate, and expected thickness range.
It can be possible, but transparent substrates may introduce backside reflection. The recipe should consider substrate thickness, backside reflection conditions, and whether the film signal can be separated from the substrate contribution.
The workflow is similar, but the substrate optical response is different. Copper is absorbing and can change spectral contrast compared with silicon or glass, so the recipe and fitting range should be verified with the actual sample structure.
TFM-100 is a compact manual system for spot checks, feasibility tests, and routine sample measurements. TFM-500 adds motorized XY mapping for multi-point measurements on wafers, glass, panels, and larger process samples.
NIR and XT configurations are useful when the target film or transparent substrate is too thick for the standard wavelength range, or when longer-wavelength fringes provide a more practical fitting condition. The usable range must still be checked against the actual sample structure.
TFM-Remote is suitable when reflectance measurement needs to be controlled by external equipment, host software, or an automation sequence. USB or Ethernet-based control can be selected according to the integration environment.
Very thin films may produce only small reflectance changes over the selected wavelength range. Film material, substrate, surface condition, and optical contrast determine whether the thickness can be separated reliably from baseline and model uncertainty.
As transparent films become thicker, interference fringes can become dense in the standard wavelength range. Longer wavelength ranges can spread the fringe pattern and make the fitting condition more practical for µm-range films.
Layer order, material n,k, initial thickness, thickness range, and substrate response all affect fitting stability. A simple and physically reasonable sample-stack setup is often more important than using a wide search range without constraints.
A recipe stores the sample-stack model, material selection, fitting wavelength range, thickness conditions, and related measurement settings. It helps users repeat routine measurements without rebuilding the setup each time.
A sample-stack setup organizes the substrate, film layers, material data, and fitting ranges in the same order as the real sample. This makes recipe creation more accessible for users who do not want to build an optical model from scratch.
If reliable n,k data are not available, start with a practical model such as Cauchy where appropriate, compare fit quality over a reasonable wavelength range, and validate the result using known samples or independent process information.
Repeatability can be affected by reference condition, sample positioning, focus or working distance, surface uniformity, spot size, integration settings, and recipe consistency. Reusing the same recipe helps reduce operator-dependent variation.
RMSE is a residual indicator between the measured reflectance spectrum and the calculated model spectrum. GOF is a helper indicator for judging overall fit quality in a more intuitive way.
Neither value should be used alone as an absolute pass/fail criterion. They should be interpreted together with spectral shape, fringe matching, fitting range, layer model, and n,k settings.
A low RMSE can occur with an incorrect layer model, wrong n,k data, an unsuitable fitting range, or a local minimum. The result should remain physically reasonable and consistent with the known sample structure.
TFM-Remote is intended for host-controlled workflows such as loading a recipe, acquiring reference/dark/measurement spectra, running calculation, and returning thickness or fit-quality results to an external system.
Ethernet or socket interfaces are useful when the measurement system must be controlled from an external PC, equipment controller, or automation sequence over a stable communication link.
A mapping workflow can save coordinates, spectra, recipe information, thickness results, and fit-quality indicators so that each measurement point can be reviewed later.
Send the film stack, substrate, expected thickness range, and measurement purpose for feasibility review.