Thin Film X-ray Reflectometry

Our goal is to develop Standard Reference Materials (SRMs) and quantitative, reproducible X-ray reflectometry (XRR) data analysis methods to enable accurate measurement of film thickness, roughness, and density in thin, multilayer structures used predominantly in the microelectronics industry. Our approach is to measure robust, temporally stable, uniform thin films on a NIST-constructed X-ray reflectometer that is traceable to the International System of Units (SI), and analyze the measured data with statistical analysis methods capable of quantifying uncertainties in structural information.

State-of-the-art microelectronic devices are patterned from nanometer-scale thickness films. Process development and manufacturing of these devices relies heavily on accurate thickness determination. XRR measurement tools provide the semiconductor industry with internally reproducible film thickness determination; however, different instrumentation and analysis software often produce divergent modeling results. NIST will address this problem by providing the community with SRM thickness standards that can then be used to calibrate XRR laboratory and Fab-line instrumentation.

There are three steps in linking an XRR SRM with SI-traceable quantities. First, industry partners will produce a robust, inert, and homogeneously deposited layer structure as an SRM feedstock. Second, NIST will use our SI-traceable X-ray reflectometer to measure multiple wafer areas for certification of physical properties. Third, NIST will develop metrological, first-principles statistical analysis approaches for estimating structural parameters and uncertainties from NIST XRR measurements.

Impact and Customers:

• XRR SRMs will provide the $43B U.S. semiconductor industry with SI traceability for thin film thickness for ISO compliance and internal fabrication line standardization 

• XRR SRMs will provide absolute thickness calibration for material structural modeling in nanotechnology and related fundamental research fields

• NIST provides technology transfers of metrological XRR characterization approaches to SEMATECH

• NIST has collaborations with key XRR instrument vendors (BEDE, Bruker AXS, Jordan Valley, PANalytical, and TECHNOS)to enhance calibration methods for XRR instruments in the field

• NIST has informal collaborations with semiconductor manufacturers for SRM materials development and acquisition

Several recent accomplishments have enabled us to make preliminary measurements; we plan to circulate pre-standards within the next 18 months.

First, because the coupling of wafer translation and rotation in the XRR system’s original configuration made alignment of the wafer surface to an X-ray line source impractical, we designed and built a novel spherical air-bearing base for the Ceramics Division Parallel Beam Diffractometer (CDPBD), NIST’s SI traceable X-ray reflectometer. By allowing rotation through a spherical axis, the surface of a wafer can be aligned to an X-ray line source while the wafer’s translational position in the beam-path remains unchanged.

Spherical surface of air bearing

Second, we fashioned a novel silicon analyzer crystal with two extended channels, each oriented along a (220) crystal plane, to achieve highly-accurate and SI-traceable determination of the CDPBD’s monochromator wavelength. This crystal has been cut from SRM 640d feedstock, whose lattice spacing is SI-traceable through the Physics Laboratory’s lattice comparator (“Delta-D” ) tool.

Spherical air bearing during assembly

Si analyzer crystal (top-view) showing channel cuts

Third, because the optimization methods typically used in analysis software supplied by instrument vendors cannot provide statistically legitimate estimates of structural parameter uncertainties, we have collaborated with an XRR software firm to develop rigorous statistical approaches using Monte Carlo (MC) techniques. Applying MC techniques reveals that XRR data often results in complex and multi-modalprobability distributions for structural parameters. Though using these statistical methods can be time-consuming, this approach not only provides estimates of parameter uncertainties, but also can allow the validity of structural models to be compared and tested.

Posterior probability of two-layer thickness model.

In 2007, we presented results from our collaboration on XRR data analysis at the Frontiers of Characterization and Metrology for Nanoelectronics, the Spring meeting of the Materials Research Society, and the Denver X-ray Conference.