Precise thickness measurements of thin SiO2 mask films
EPFL The Laboratory for Physics of Nanostructures (LPN) belongs to the Institute of Physics (IPYS) at the Ecole Polytechnique Fédérale de Lausanne (EPFL), one of the leading Swiss technical universities. Our main study is photonic nano-devices, based on the integration of position- and energy-controlled semiconductor quantum dots (QDs) with photonic crystal (PhC) optical cavities and waveguides (WGs).
To obtain the required 1 nm accuracy, we now use the 3D optical profilometer S neox in its spectroscopic reflectometry mode as a quick and easy way to measure SiO2 layer thickness.
We study the optical properties of QDs that are embedded in PhC cavities (Figure 1a), leading to enhanced emission (the Purcell effect), or in WGs (Figure 1b), producing photon multiplexing devices. A typical device is fabricated from a multi-layer GaAs/Al0.7Ga0.3As/GaAs epitaxially grown stack, where the top 250 nm thick GaAs layer contains the active part of the device, and the 1 μm Al0.7Ga0.3As sacrificial layer is eventually etched away to produce a floating membrane device.
This membrane contains one or several 20 nm In0.3Ga0.7As QDs in specific positions, and an etched PhC structure (large array of 100 nm holes, with several holes missing in a specific configuration), precisely aligned with the QDs. These devices require accuracies of 1-20 nm in several processing steps, so they all involve high-performance electron-beam lithography. To withstand wet or plasma (ICP) etching of GaAs, we have to use a hard SiO2 mask, to which the lithography pattern is transferred by dry etching (RIE).
The accuracy of this transfer depends on knowing the precise thickness of the mask layer. As the SiO2 layer is 40-80 nm thick, we need 1 nm precision in this measurement. The purpose of this study is to obtain high-precision (1 nm) measurements of thickness of thin (40-80 nm) SiO2 films used as hard masks.
So far, only a stylus profilometer could be used on a daily basis to measure the SiO2 mask thickness, and this was after a global calibration using calibration standards and a Sopra GES 5E spectroscopic ellipsometer in another laboratory. However, profilometry demands the wet etching of a “step” in the SiO2 layer, so it is time consuming and not practical when full wafers are processed. Moreover, typical noise in profilometry is 5 nm RMS, so a lot of averaging is needed to obtain the required 1 nm accuracy (Figure 2a).
We now use the 3D optical profilometer S neox in its spectroscopic reflectometry mode as a quick and easy way to measure SiO2 layer thickness. Using simple GaAs as a reference, we obtain 1 nm accuracy using this technique, which takes seconds to perform.
Using a bare GaAs substrate as reference, we first measured the reflectivity spectrum of SiO2 layers on GaAs using the built-in single-layer model. As seen in Figure 2b, the resulting reflectivity spectrum is very flat, showing no oscillations typical of thicker (d>λ) films. Still, the model fit is very good, giving the film thickness (84 nm) with 1 nm accuracy.
The same type of measurement is shown in Figure 3, where a 38 nm film is measured. In the case of the GaAs membrane structures, we have a 3-layer (SiO2/GaAs/Al0.7Ga0.3As) structure on the GaAs substrate. In such cases, we enter this full structure into the model, still using GaAs as a reference.
As a first test, we measure the reflectivity of the bare semiconductor multilayer structure without SiO2. The resulting spectrum (Figure 4a) shows a good fit, producing the correct (verified by X-ray diffraction) GaAs and Al0.7Ga0.3As layer thicknesses, as well as zero thickness for the top SiO2 layer in the model. This time the reflectivity curve shows the typical oscillations, as usually appears in thicker (d>λ) layers.
Once we verified the bare semiconductor, we measured another sample, which was coated with SiO2 (Figure 4b). The resulting spectrum fit showed not only the semiconductor layers’ thicknesses, but the correct SiO2 layer’s thickness (79 nm) as well.
As a final test, we tried to measure a sample coated with both SiO2 and PMMA layers (Figure 4c). This time, the spectrum showed more complex oscillations, and the model fit was not as good as in the previous cases. Still, the fitted values were correct, showing this method’s power and speed.
We have checked the values obtained with this method using both stylus profilometer and the Sopra GES 5E spectroscopic ellipsometer as controls, as well as with commercial calibration standards (SiO2 layers on Si, from Micromasch), and found that the Sensofar system gives an accuracy of 1 nm, which is what we need.
Although we have shown very good results using multi-layer fitting, whenever possible we try to fit a model where we can fix all known parameters (e.g. semiconductor layer thicknesses) and vary only the top layer’s thickness (usually SiO2).
In order to process complex photonic nanostructure devices, we need high-accuracy (1 nm), fast thickness measurements of thin (usually < 100 nm) SiO2 layers deposited on top of GaAs or multi-layer semiconductors. To this end, reflective spectroscopy provided by the S neox 3D optical profilometer from Sensofar was the perfect tool, providing the high accuracy, high measurement speed, and ease of use that we needed.