1. Thermal oxidation
In thermal oxidation, silicon wafers are oxidized at about 1000 °C in an oxidation furnace. This furnace essentially consists of a quartz tube in which the wafers are placed on a carrier (boat) made of quartz glass, several independently controllable heating coils, and various gas supply lines. Quartz glass has a very high melting point (well above 1500 °C) and is therefore very well suited for high-temperature processes. To avoid wafer warping or cracking, the quartz tube is heated in very small steps (max. 10 °C per minute). Thanks to the separate heating coils, the temperature along the entire tube – over a length of about 1 m – can be controlled precisely to within ± 0.5 °C.
The oxygen then flows as a gas over the wafers and reacts at the surface to form silicon dioxide. A glass-like layer with an amorphous structure is formed. Depending on the process gas, different oxidations take place; a thermal oxidation naturally has to occur on a silicon surface. Thermal oxidation is divided into dry and wet oxidation, the latter of which can in turn be divided into wet oxidation and H2-O2 combustion.
Dry oxidation
The oxidation process takes place under a pure oxygen atmosphere. Silicon reacts with oxygen to form silicon dioxide:
This process usually takes place at 1000–1200 °C. To produce very stable and thin oxides, the oxidation is carried out at about 800 °C.
Characteristics of dry oxidation
- slow oxide growth
- high density
- high breakdown voltage (for electrically highly stressed oxides, e.g. gate oxide)
Wet oxidation
In wet oxidation, the oxygen is passed through a bubbler vessel containing water (about 95 °C), so that in addition to oxygen, water is also present in the quartz tube in the form of steam. This results in the following reaction equation:
This process takes place at 900–1000 °C. It exhibits fast oxide growth at low temperatures and is used, among other things, for producing masking layers and field oxides. The quality of the resulting layer is lower than that of dry oxidation.
Comparison of growth rate for dry and wet oxidation
| Temperature | Dry oxidation | Wet oxidation |
| 900 °C | 19 nm/h | 100 nm/h |
| 1000 °C | 50 nm/h | 400 nm/h |
| 1100 °C | 120 nm/h | 630 nm/h |
H2-O2 combustion
In H2-O2 combustion, high-purity hydrogen is used in addition to high-purity oxygen. The two gases are fed into the quartz tube separately and burned at the inlet opening. To avoid an explosive (Knallgas) reaction with the highly flammable hydrogen, the temperature must be above 500 °C; the gases then react in a silent combustion. This process allows the production of fast-growing oxide layers with only minor contamination. This makes it possible to produce both thick oxides and thin layers at comparatively low temperatures (900 °C). The low temperature also allows for the thermal treatment of wafers that have already been doped (see doping by diffusion).
In all thermal oxidation processes, oxide growth is higher on (111)-oriented substrates than on (100)-oriented substrates (see the single crystal). In addition, a very high concentration of dopants in the substrate significantly increases the growth rate.
Course of the oxidation process
At the beginning, the oxygen reacts at the wafer surface to form silicon dioxide. An oxide layer is now present on the substrate, through which subsequent oxygen atoms must first diffuse in order to react with the silicon. The growth rate depends on the reaction time between silicon and oxide only at the beginning; beyond a certain thickness, the oxidation rate is determined by how quickly the oxygen diffuses through the already grown silicon dioxide. As the oxide thickness increases, growth therefore slows down. Mathematically, this is described by a linear-parabolic model according to Deal and Grove: initially, thickness grows roughly proportional to time, but later only proportional to the square root of time. Thick oxides are therefore disproportionately time-consuming to produce. Since the resulting layer is amorphous, not all bonds of the silicon atoms are intact; there are partly dangling bonds (free electrons and holes) at the Si-SiO2 interface. This results in an overall slightly positive charge at this interface. Since this charge can adversely affect devices, efforts are made to keep it as low as possible. This can be achieved, for example, by using a higher oxidation temperature, or by using wet oxidation, which likewise causes only a very low charge.
Volume increase
In thermal oxidation, silicon is consumed through the reaction with oxygen to form silicon dioxide. The ratio of the grown oxide layer to the consumed silicon is 2.27; that is, the oxide grows into the substrate by 45 % of the oxide thickness.
Segregation
Dopants present in the substrate can be incorporated either into the silicon crystal or into the oxide, depending on which material dissolves the dopant more readily. This so-called segregation coefficient k is calculated as follows:
$$k=\frac{Solubility\ of\ the\ dopant\ in\ Si}{Solubility\ of\ the\ dopant\ in\ SiO_2}$$If k is greater than 1, the dopants are incorporated at the surface of the substrate; if k is less than 1, the dopants accumulate in the oxide.
Where thermal oxidation stands today
The classic role of thermal oxide, forming the gate dielectric, has been discontinued. At layer thicknesses of just a few atomic layers, so much current tunnels through the oxide that it is no longer suitable as an insulator. Its place has been taken by hafnium dioxide, which is deposited via atomic layer deposition and, for the same effect, may be thicker. Even there, however, it is not possible to do entirely without thermal oxide: between the silicon and the hafnium dioxide lies an intermediate layer of silicon dioxide only a few atomic layers thick, since a low-defect interface to the silicon cannot be produced any other way.
What has remained for thermal oxidation are the tasks for which its particular characteristic – growing directly out of the substrate itself – is especially useful: the pad oxide beneath nitride masks, sacrificial oxides to protect the surface from implantation damage, the lining of freshly etched trenches in trench isolation, and the oxides used in power electronics, where thickness and breakdown strength continue to matter.
The equipment technology has also changed. The furnace described above processes an entire batch over several hours. For thin layers, this is too imprecise and thermally too demanding; there, a single wafer is heated to temperature within seconds using lamps and cooled down again just as quickly. This keeps the amount of heat introduced so low that existing doping profiles barely shift at all.