Next-generation FTIR tools offer enhanced process control and improved yields of compound semiconductor devices such as PIN diodes, avalanche photodetectors, infrared lasers, blue emitters, and LEDs.

Fourier-transform infrared (FTIR) reflectance-spectroscopy methods have recently migrated from the silicon industry to become metrology tools for characterizing compound semiconductors, including indium phosphide (InP), gallium arsenide (GaAs), indium gallium arsenide (InGaAs), alloys, silicon carbide (SiC), and gallium nitride (GaN). The technique allows nondestructive measurement of parameters such as multiple-film thickness, free-carriers concentration, index of refraction, dopant concentration in dielectrics, and alloy composition (see "How FTIR tools work," p. 18).

The use of nondestructive FTIR measurements is a significant advance over standard destructive thickness and free-carriers concentration measurements obtained with chemical etching and electrochemical current-voltage techniques. Fourier-transform IR spectroscopy can measure parameters that other optical tools cannot, because free carriers only absorb in the IR, molecular-bond absorption is only present in the IR, and micron-thick layers are typically opaque iat visible or x-ray wavelengths.

Some recently introduced FTIR tools eliminate limitations of previous FTIR systems. These tools can be used in conjunction with other techniques such as photoluminescence, ellipsometry, and x-ray and standard nondestructive techniques. The latest generation of FTIR metrology can enhance process control and improve yields of compound semiconductor devices such as PIN diodes, avalanche photodetectors (APDs), infrared lasers, blue emitters, and light-emitting diodes (LEDs).

Applications of FTIR
Fourier-transform IR spectroscopy is sensitive to the presence of thin films on a substrate, with a spectral interference pattern that depends on the thickness and refractive index of each film and substrate (see Fig. 1). The reflection mode is favored versus the transmission mode because while semiconductors are typically transparent in the IR, heavily doped wafers are opaque to IR light and cannot be measured in transmission.

Strong free-carrier absorption in the IR is the main mechanism that allows FTIR tools to quantify free-carriers concentration (see Fig. 2). Fourier-transform IR spectroscopy is also sensitive to chemical bonds and has been used for the analysis of polymers, photoresists, dielectrics, and so on for decades. Until now, FTIR was not extensively used with compound semiconductors mostly because of the need for special test wafers. Special test wafers have been required because most semiconductor materials are partially transparent to the IR light and the ghost reflections from the wafer backside result in uncontrolled measurement artifacts that vary from wafer to wafer. To avoid the back-side reflection problem with conventional FTIR, test wafers that are completely IR-opaque (typically very low resistivity) usually must be used.

Another limitation of FTIR tools has been the inability to extract useful parameters on wafers with a complex film stack because model-based spectral analysis had not been implemented in FTIR reflectometry tools. While R&D infrared ellipsometry systems have been using model-based analysis, the long measurement time (minutes per point) in these systems limits their use for production applications. Thus, reflectometry-based systems, with measurement times of seconds per point, are favored.

Advanced FTIR capabilities
At MKS, we have developed an advanced FTIR reflectometry metrology tool (called FilmExpert) that allows accurate, quantitative infrared analysis of multiple optical thin-film properties on product wafers. The unique combination of ghost-image suppression optics and model-based analysis overcomes the two problems associated with FTIR thin-film analysis described previously.

Specialized optics allow measurement of front surface reflections while removing the reflections from the backside. This is important for wafer backside surfaces that are not well controlled, such as those with a rough backside or with glue residues, since artifacts that would otherwise be added to the measured spectrum are eliminated. In addition, ghost-suppression optics provide a significantly reduced spot size (200 x 600 μm compared to 5 x 5 mm in standard FTIRs), which allows for increased measurement capability at the wafer edge.

The integration of model-based algorithms is the other major improvement of advanced FTIR reflectometry tools. While different models exist, these algorithms typically use a multilayer, multiparameter analysis routine to extract film parameters. The first step in the MKS system is a model-based fitting routine that fits the reflectance spectrum with parameters for film thickness and dielectric function to extract film thickness, spectral dielectric function (otherwise known as refractive index n and extinction coefficient k), and graded transition-profile thickness. The second step involves extracting parameters from the dielectric function to obtain free-carriers concentration and alloy concentration, as well as dielectric dopant concentrations from the vibration absorptions.

This two-step analysis is done automatically through a fitting routine that varies the model parameters to obtain a match between measured and theoretical reflectance spectra. When combined with the ghost-suppression optics, the result is the ability to measure thin films in product wafers with more accuracy and precision.

Compound semiconductor production
Several applications of advanced FTIR measurement tools have been used successfully in compound-semiconductor production environments to increase product yields and enhance process control. A single tool can typically be used in a variety of applications to provide a cost-effective metrology solution. For most applications, the main interest is to check that the parameters are within specifications and to measure nonuniformity across the wafer.

Epi SiC and other single layer on substrate. For some compound semiconductor structures, such as epi SiC and calibration structures, the final structure consists of only one or two layers (see Fig. 3, top). In this case, the layer thickness, index of refraction, and free-carriers concentration of both the layer and the substrate can be extracted and represent important parameters as final specification criteria. Until now, only destructive testing was available. Advanced FTIR systems can test all wafers to ensure quality and catch problems at an early stage.

PINs, APDs, and multilayer calibration structures. Advanced FTIR systems can measure the thickness and free-carriers concentration of multiple layers in film stacks for PINs and APDs (see Fig. 3, bottom). The fast and nondestructive measurement of PIN or APD product wafers ensures that every wafer is within specification. Calibration wafers with only a few layers can also be measured faster by FTIR than by destructive techniques, providing quicker turnaround time.

Lasers, high-electron-mobility transistors, and heterojunction bipolar transistors. These types of structures can be quite complex, with up to 15 layers in the stacks. While not all layer thicknesses and free-carriers concentrations can be extracted due to parameter intercorrelation, useful information can be extracted, including free-carriers concentration of contact layers, depth of a particular layer (to determine, for example, subsequent etching-process parameters), thickness of some layers, total thickness and free-carriers concentration, and free-carriers concentration of critical layers in HBT structures. The parameters that can be extracted depend on the film-stack structure.

Faster nondestructive measurements
While FTIR metrology has been widely used in the silicon industry, its application to compound semiconductors has been limited. Drawbacks such as requiring test wafers and the inability to extract parameters from complex film stacks have been overcome by advanced FTIR systems designed with the specific needs of compound semiconductors in mind. Specialized optics and advanced algorithms have thrust FTIR technology into the realm of viable, production-worthy metrology for compound semiconductors. Fourier-transform IR spectroscopy tool users can now enjoy the benefits of cost-effective, flexible metrology for the measurements of free-carriers concentration, layer thickness, alloy composition for multilayer stacks, and molecular-bond absorption in dielectric layers. The fast, nondestructive measurements allow for fast turnaround time and help identify process faults and drifts.

SYLVIE BOSCH-CHARPENAY is applications group manager and PETER ROSENTHAL is general manager at MKS On-Line Products, 87 Church St., East Hartford, CT 06108. They can be reached at sylvie_bosch-charpenay@mksinst.com and peter_rosenthal@mksinst.com.

How FTIR tools work
The heart of a Fourier-transform infrared (FTIR) system is a scanning Michelson interferometer, which allows the simultaneous measurement of multiple wavelengths as opposed to slower dispersive infrared spectrophotometers that measure one wavelength at a time. Using a beamsplitter, a beam of light is separated into two beams—one of fixed path length and the other of variable path length. The two beams are then recombined using the same splitter, creating interference between the beams following their optical path difference.

The direct output of the spectrometer is an interferogram, which is a plot of light intensity versus the mirror position (the moving mirror position is related to the optical path difference). All wavelength information is encoded in the time-space interferogram. To obtain a frequency-space spectrum, a Fourier transform of the interferogram is performed. The obtained spectrum is the raw detector response versus wavenumber (inverse of wavelength). Absolute spectrum measurements are obtained by referencing the sample spectrum to a known sample spectrum.

Measuring a sample spectrum typically takes a fraction of a second to a few seconds. The spectrum is then analyzed by various methods, the most advanced involving model-based analysis describing the full interaction of the light with the material. The modeling, which is usually done automatically after having set a "recipe," returns parameters such as (in the case of thin films) thickness, composition, and free-carriers concentration, which can be mapped across the wafer and used for process control.