Films and coatings in the nanometre thickness range are increasingly critical in a wide range of industries and applications. Accurate knowledge and control of thickness is a frequent requirement. We report on an inter-laboratory comparison of the accuracy of silicon oxide thin film thickness measurement.

Traceability is conveniently achieved using transfer artefacts. These are available for a limited range of particularly well-characterised materials systems, such as silicon oxide on silicon. Importantly, such artefacts also allow the accuracy of different measurement techniques or instruments to be compared, assessed and monitored.

A set of three silicon wafers with certified oxide thickness values, traceable to NIST, of 47.0nm, 191.2nm and 1055.7nm, have been measured by several laboratories using five different non-destructive measurement techniques:

• X-ray Reflectivity (XRR) providing absolute thickness measurements (traceable to x-ray wavelength) for flat samples with layer-substrate density differences.
• Spectroscopic Ellipsometry (SE) applied across the UV-visible spectral range
• Single-wavelength Ellipsometry (SWE) at 633nm and Optical Reflectometry (OR) instruments designed for rapid multi-point-clean-room process measurements.
• White Light Interferometry (WLI) with the capability for imaging layer and substrate surface roughness, giving local film thickness variation as well as mean thickness.

SWE, OR and WLI techniques require refractive index values to be assumed for the layer and substrate. This is not a source of variance in the present study, because the artefacts are certified with specific fixed SiO₂ and Si refractive indices. However, refractive index uncertainty is a common source of measurement error and uncertainty when optical techniques are applied to less well-characterised materials.

Thickness measurement results obtained illustrated very good agreement with the calibration values for all three thicknesses. Agreement within ±1.2% was found except for one SWE measurement, which deviated by +1.7%. Results from the thinnest layer are shown below:

Figure 1. Measurements of a 47nm SiO₂ layer sample (FTS4-500) made by six laboratories using five different techniques, showing agreement with the calibrated value to within 1%

A secondary observation was a consistent trend for XRR values to be slightly lower than those from the visible wavelength optical techniques. A similar effect observed with ultra-thin (sub-10 nm) oxides on silicon was attributed to the different sensitivities of these two technique families to surface-absorbed moisture and organic species. This underlines the fact that perfect agreement between different techniques is not always to be expected for sound physical reasons relating to the probe-material interaction and hence the need for an understanding of these when making technique comparisons.

For further information on the techniques described here and the wide range of other metrology capabilities and services available through CEMMNT please contact us today

Further Information:

For background information on the NIST-traceable SiO2/Si standards see: Belzer B, Eberhardt K, Chandler-Horowitz D, Ehrstein J R, Durgapal P 1998 "Thin film reference materials development final report for CRADA CN-1364", NIST Gaithersburg, MD 208990001US Department of Commerce; and http://www.vlsistandards.com 

Earlier results from this comparison are reported in: Leach R K and Giusca C "Results from a comparison of optical thin film thickness measurement" NPL report ENG 18 (August 2009)

Results of ultrathin oxide measurements are given in: "Critical Review of the Current Status of Thickness Measurements for Ultra-Thin SiO2 on Si: Part V, Results of a CCQM Pilot Study" by M P Seah, et al., Surface and Interface Analysis, 36 (2004)1269-1303

For information on the WLI thin film thickness measurement techniques and applications see Talysurf CCI Sunstar article

For more information on QinetiQ please see QinetiQ.com