In situ x-ray reflectivity for thin-film deposition monitoring and control
05/01/1999
Grazing-incidence x-ray reflectivity shows potential for
in situ deposition and etch process monitoring and control,
as well as operator visualization of the film growth process.
X-ray reflectivity is based on interference of x-rays reflected from two interfaces (i.e., vacuum-film or film-substrate). For angles above the critical one, the expression for the x-ray reflection coefficient from a film-substrate system is described by Fresnel theory. In semiconductor manufacturing, existing x-ray instruments for thin films are designed for ex situ measurements.
The most common implementation of x-ray reflectivity rotates the sample being studied by some angle in the range from zero to several degrees while the detector is rotated at double angular velocity; this is "Q-2Q" scanning.
 |
The equation describing the variation of the reflection coefficient shows that interference can occur with changing angle of incidence (AOI), wavelength, or film thickness. In principle, variation of the wavelength is not different from variation of the AOI. However, variation of film thickness (e.g., as occurs during thin-film deposition) with constant AOI and wavelength provides a fundamental new approach to the application of x-ray reflectivity. By recording the time dependence of the reflection coefficient, one can find the same set of parameters as that determined from angular dependence. Practically applied, this can be used as in situ measurement providing real-time information on film thickness, density, and roughness during a deposition process.
 |
In situ x-ray reflectivity (Fig. 1) is similar to that of in situ laser interferometry, except that the wavelength is <1/4000 that used in laser interferometry, providing some essential differences:
- The imaginary part of the dielectric constant at this wavelength is small, and the method can be applied to metals, semiconductors, or dielectrics (Fig. 2).
- Since the wavelength is comparable to the interatomic distance, one can measure film thicknesses down to 10?.
- The reflection coefficient (R) essentially depends on the root mean square roughnesses of the film surface and interface, the roughness amplitude being comparable to the wavelength of the incident radiation. This permits calculation of roughness magnitude.
- For the wavelengths considered, the refraction coefficient or, more accurately, refraction decrement (df) of the material is directly related to its density (rf).
In situ thin-film control
In situ control during thin-film deposition is very attractive because deposition technology is at a point where further improvements require stricter control of both the gaseous atmosphere and layer parameters during deposition, particularly when considering the ever-widening variety of films being used in wafer fabrication. Because film thickness is a main parameter that determines the function of a film, and considering that film thicknesses in active devices are decreasing, an in situ monitoring method that detects physical properties (especially roughness) and film operation parameters (i.e., electrical, optical and magnetic) of the film is most important.
We believe that measurement of thin-film parameters during deposition, and even during plasma etching, will become one of the most important factors for improving the quality of these processes. The need is for universal monitoring systems capable of monitoring any material on any substrates. Perhaps ideally suited, in situ x-ray reflectivity has potential for simultaneous determination of thickness, surface, and interface roughness, and density as a measure of film quality and structure.
In situ x-ray reflectivity is very well suited for monitoring and controlling wafer processes. In a practical application, in situ x-ray reflectivity can provide feedback to the system it is monitoring, even enabling termination of a given process run when failures are detected. It provides:
- direct monitoring of the parameters of the growing layers,
- simultaneous measurement of the most important film parameters (thickness, growth rate, density, and surface roughness), and
- direct measurement of layer thickness without performing preliminary calibration for each new material.
Using in situ x-ray reflectivity, it is possible to change the organization of today's processes. Generally, development of a multilayer growth process (e.g., structures consisting of six or seven metal layers) takes a lot of time and requires performing and repeating many experiments. In situ x-ray reflectivity can dramatically shorten this development period.
The capabilities
Thickness, density and roughness determined from in situ x-ray measurements are not the key parameters; rather, magnetic, optical or electrical film properties are more important. This seems true only at first glance. Thus, the operation of a spin valve sensor (e.g., Ta/NiFe/Co/Cu/Co/NiFe/FeMn/Ta) is based on quantum size effect. For this effect, thickness of each layer and interface roughness play a decisive role. Indeed, film density characterizes its porosity, roughness its electron scattering. Generally, it is impossible to obtain a thin-film structure with good parameters if it has a rough surface or its layer density is substantially different from the bulk value.
Because x-ray reflectivity is a direct method of measuring film thickness and growth rate, it captures direct information on deposition kinetics and thickness. Its wavelength is comparable to the interatomic distance, so film thickness can be measured to a few angstroms. Moreover, the method provides a simple way to measure roughness and density of ultrathin layers.
Compared to other methods, an x-ray system is not sensitive to vacuum parameters, such as deposition method, pressure, working gas composition, plasma parameters, etc., nor to the type of deposited material. Overall, this method of metrology has no effect on the manufacturing process, since the x-ray tube and radiation detector are placed outside the working chamber (Fig. 3).
For different applications, several modifications of in situ x-ray monitoring systems are possible. In the simplest version, a weakly divergent x-ray beam falls on the substrate. The detector situated at the same angle as the x-ray tube records the incident radiation. Variation of the film thickness during growth produces oscillations of the reflection coefficient. It is important that the period of oscillations determined by the Bragg condition weakly depends on the type of material, since the refraction coefficient is close to unity, especially for light substances with densities below 2.5g/cm3. Analyzing the resulting dependence, we can calculate thickness, density, and roughness variation. In contrast to the method of small-angle x-ray reflectivity, layer parameters are determined not from the angular dependence of the reflection coefficient, but from the dependence of the reflection coefficient on deposition time (or on film thickness).
 |
The advantage of in situ x-ray reflectivity is most clearly seen in monitoring multilayer nonperiodic structures (Fig. 4). Angular dependence of the reflection coefficient obtained from ex situ measurements on multilayer structures is extremely complicated, because of superposition of reflections from different layers, and generally cannot be interpreted. For in situ measurements, the interference patterns from different layers appear sequentially. Therefore, at each step we calculate parameters of only one growing layer.
For the data in Fig. 4, Ta2O5 was deposited by magnetron sputtering from a tantalum target in an Ar-O2 mixture. The first diamond-like carbon (DLC) layer was deposited by rf-PECVD in an Ar-C6H12 mixture, P = 0.22W/cm2. The Ta layer was deposited by magnetron sputtering from a tantalum target in Ar. The second DLC layer was deposited by rf-PECVD in an Ar-C6H12 mixture, P = 0.14W/cm2.
Accuracy of thickness, density, and roughness measurements is important. The accuracy of thickness measurements depends on the error in the angle of beam incidence, on the beam divergence, and on the layer growth rate. If the growth rate is not greater than 2-3?/min, then the error in the oscillation period is about a fraction of an angstrom. Taking into account the statistical character of the measurements, we can reduce the absolute error in r and s to ?0.05 g/cm3 and ?0.5?.
Calibration of in situ x-ray reflectivity before measurements is simple in comparison with XRF. When calculating parameters, we used only the absolute values of the intensity of the reflected radiation; there is no need to know the number of quanta emitted by the x-ray tube itself.
 |
Apart from the possibility of determining layer parameters and growth kinetics, the system makes it possible to visualize the growth process. An operator can observe the output harmonic signal (a sinusoid) on a monitor (Fig. 5). By its shape, he or she can estimate kinetics and stability of the deposition process without making any calculations. The closer the shape is to a sine curve, the higher the stability of the deposition process. It is our opinion that it is much easier to relate the sinusoid to the growth process than data obtained by ellipsometry and quartz microweighing.
Conclusion
The method of in situ x-ray reflectivity could be a strong alternative to ellipsometry and picosecond ultrasonic laser sonar (PULSE). It is a nondestructive control method sensitive to sub-? variations of thickness and surface roughness and is not limited by material type or plasma luminescence during a deposition or etch process. Unlike PULSE, in situ x-ray reflectivity is more convenient for film thicknesses <100?. In addition, in situ x-ray reflectivity allows definition of film thickness without prior knowledge of film density, the latter being calculated simultaneously with thickness. This results in greater measurement precision.
Via direct measurement, in situ x-ray reflectivity provides simultaneous determination of thickness, growth rate, roughness, and density, making it suitable for the next generation of deposition and etch systems.