Fundamentals of Terahertz Devices and Applications. Группа авторов
Frequencies
In order to achieve the depth accuracy and smoothness of the features on silicon, Silicon‐on‐insulator (SOI) wafers are commonly employed, since thin layers of SiO2 act as etch stops of the silicon. For example, [26] presented an iris of the leaky‐wave antenna at 1.9 THz fabricated on a 2.5 μm silicon membrane. Compared with a thermal oxide membrane, the crystalline silicon provides more robustness to compressive stress. In that case, a SOI wafer with a 2 μm device layer was used because, together with the metallization, the membrane would reach the desired 2.5 μm. And a handling wafer of 400 μm defined the waveguide on the back of the membrane.
The step by step fabrication of the slot iris and waveguide process is shown in Figure 2.23. Thermally grown SiO2 is used as a hard mask in the front and back of the SOI wafer. Since the small size of the double slot iris, an i‐line Canon stepper is used to illuminate the SiO2. And for the rest of the waveguide and alignment features, regular UV photolithography can be employed. The pre‐etching of the features, i.e. the slot iris and the waveguide, on the front and back of the SiO2 layer are performed using photoresist and an inductively coupled plasma (ICP) reactive ion etcher. Then, the silicon on the front backside is etched using a PlasmaTherm DRIE system. While the selectivity of the silicon to photoresist is around 70 : 1, the selectivity of the silicon to SiO2 can be optimized to reach 130 : 1, which allows to etch high depth features with high control. The SiO2 was removed using a hydrogen fluoride HF solution.
Figure 2.23 (a) Sketch of the membrane fabrication process that contains the iris and waveguide of the leaky‐wave feed. (b) SEM of the iris developed at 1.9 THz in [26] using the explained process.
Source: Alonso‐delPino et al. [26]; IEEE.
At the end of the process, the overall wafer was sputtered with gold, used due to its high conductivity, immunity to oxidation, and ease of deposition. The overall results of this process can be observed in the scanning electron microscope (SEM) image in Figure 2.23b, showing a very clean and well‐defined pair of double slots.
The rest of the wafers that define the lens are processed similarly as the procedure described but, because the radiation is going through the wafers, it is necessary the use of high resistivity silicon wafers (the resistivity around 10 kΩ cm) to avoid the introduction of absorption losses.
2.5.1.1 Fabrication of Silicon Lenses Using DRIE
Silicon shallow lens arrays have been fabricated either by laser micro‐machining process or using photolithographic processes based on DRIE. Laser micro‐machining allows the fabrication of 3‐D geometries with accuracy as presented in [33], however, it is a linear process where the cost depends on the laser time, which might not be the most cost‐efficient method for large lens arrays. A novel DRIE silicon process presented in [47] allows the fabrication of arrays of lenses on a single wafer and in parallel. This section will provide an overview of the process and a fast method to estimate the overall fabrication accuracy without needing to test the lens antenna.
Figure 2.24 (a) Sketch of the fabrication process of the shallow silicon lens. (b) Photograph of shallow lens antennas at 1.9 THz of diameter 2.6 and 6.3 mm presented in [26].
Source: Alonso‐delPino et al. [26]; IEEE.
The process to fabricate the silicon shallow lens consists of four steps, illustrated in Figure 2.24a. The first step consists of the patterning of the photoresist on a high resistivity wafer with the desired lens aperture diameter. The thickness and aperture diameter of the photoresist applied onto the silicon wafer defines the thickness and curvature of the lens surface. Multiple photoresist coatings can be applied to achieve the desired thickness. Next, the photoresist reflows by applying heat, i.e. around 110 °C on a hot plate, to the wafer. The surface tension applied to the photoresist by its coating above the glass transition temperature, makes the surface reflow into a spherical shape. Last, the photoresist shape is transferred into the silicon wafer using a DRIE process. The photoresist and silicon etching selectivity, controlled by the CF4 and O2 gas ratio and the DC and RF power applied in the process, defines the curvature of the overall lens. The examples shown below a selectivity of 1 : 1.3 was used to achieve a total height of 475 μm for a 360 μm of photoresist. The surface roughness achieved with the process in the order of hundreds of nanometers.
The last step of the process consists of applying an antireflective coating to the lens which is essential to reduce the high reflection losses that occur by using a dielectric with high permittivity. A coating with the polymer Parylene is usually employed as matching layer at submillimeter‐wave frequencies.It has an index of refraction around 1.64, which is not the ideal for silicon, but it is close enough to considerably reduce the reflections. The coating conformal is deposited using vapor deposition which allows high control of the thickness and uniformity.
2.5.1.2 Surface Accuracy
The fabricated surface of the lens can be analyzed independently of the rest of the antenna by the accurate characterization of the actual fabricated surface. Surface profilometers are used to map the 3D profile of the lens surface with high resolution. The basic parameters of this surface in terms of radius R, diameter of the aperture D, height H are computed with an optimization procedure using, for example, the optimization toolbox in Matlab and a basic cost function to minimize. The cost function compares the fabricated surface and an ideal spherical surface until it finds the best fitting sphere. By evaluating the error between both surfaces we can have a sense of how spherical the fabricated lens is, and identify flaws that can improve the fabrication process. For example, Figure 2.25a shows the profile of a fabricated shallow lens of D = 2.6 mm measured with a profilometer and its error surface obtained when compared with a perfect sphere. In this case, since the edges of the lens did not provide a good fit with the sphere, an improvement of the performance could be achieved by adjusting the illumination to minimize the lens area illuminated. This adjustment can be achieved by decreasing the lens thickness W, which will decrease the illumination of the lens edges, i.e. increase the edge field taper.
Figure 2.25 (a) Surface measured of the fabricated lens of D = 2.6 mm. The error surface defined as the difference between the measured surface and a perfect sphere is shown underneath. (b) Computed radiation pattern of the measured lens surface and the measured radiation pattern of the whole lens antenna at 1.9 THz from [26].
One can have an estimation of the effects of the aberrations in the radiation pattern by translating the error surface into a phase error surface. We can assume that the field distribution