一、 Diffusion

1. Diffusion doping

The characteristic of semiconductor materials that can be doped with n-type or p-type conductive impurities to adjust resistance values without affecting their mechanical and physical properties is the basis for further creating p-n junctions, diodes, transistors, and even the vast world of integrated circuits (ICs). And diffusion is an important initial process for achieving conductive impurity doping.

As is well known, diffusion is the transport phenomenon of nature; Mass transfer, heat transfer, and momentum transfer (i.e. frictional drag) are all three known phenomena in reality. The diffusion of this impurity belongs to one type of mass transfer, and the effect can only be significant in high temperature environments above 850oC.

Due to the diffusion phenomenon, the impurity concentration C (concentration; the number of conductive impurities or carriers per unit volume) follows the diffusion equation as follows:

This is a parabolic partial differential equation that is related to both diffusion time t and diffusion depth x. In other words, at a certain diffusion moment (t fixed), the impurity concentration will decrease from the surface position with the highest concentration towards the depth direction, forming a concentration curve that varies with depth x; On the other hand, this concentration curve changes its style as the diffusion time increases, moving towards a flat and consistent diffusion concentration distribution at infinite time!

Since it is a diffusion differential equation, applying different boundary conditions will result in different concentration distribution profiles. Constant surface concentration and constant surface doping are two commonly discussed diffusion boundary conditions with analytical exact solutions:

2. Pre position

The analytical solution for the first fixed concentration boundary condition is the so-called complementary error function, and its corresponding diffusion step is called "pre diffusion", which is the diffusion process we generally understand; When the high-temperature furnace tube rises to the working temperature, the wafer to be diffused is pushed into the furnace, and then the diffusion source (p-type diffusion source is usually a solid boron nitride chip in the shape of a wafer, n-type is the heating vapor of liquid POCl3) is released for diffusion. The characteristic of its concentration profile is that impurities are concentrated on the surface, with the highest surface concentration and rapidly decreasing with depth, or in other words, the surface concentration gradient value is extremely high.

3. Post drive in

The second boundary condition with a fixed doping amount is a concentration analytical solution with a Gaussian distribution. The corresponding diffusion processing program is called "post drive in", which is a general high-temperature annealing program; Basically, only the driving working temperature of the furnace tube is maintained, while the diffusion source is no longer released. Or ask: Where does the starting boundary condition for a fixed doping amount come from? The answer is the result of the "pre diffusion" process; The impurity concentration produced by the previous "pre diffusion" is concentrated on the surface, which can approximate the boundary condition of a certain amount of doping!

As for why diffusion should be divided into these two types of steps, of course, it is not for the sake of mathematical analysis, but to meet the needs of resistance value adjustment. The original impurity implantation dose of "pre diffusion" quickly reached saturation, and even if the "pre diffusion" time was extended, the impurity implantation dose could not be significantly increased. In other words, the electrical resistivity characteristics quickly stabilized; However, the "backward drive" reduces the surface concentration and gradient (due to the diffusion of impurities from the surface to the depth), while creating another "forward diffusion" to increase the opportunity for impurity implantation dose. Therefore, through repeated "forward diffusion" and "backward drive", the electrical resistivity characteristics can be adjusted, and the effective cross-sectional area of impurity resistance can be changed. Therefore, according to the well-known resistance formula; Among them, the diffusion program of the required conductive area can be designed based on the length of the resistance.

4. Other key points of diffusion are briefly described as follows:

(1) The diffusion process has the advantages of batch production and low cost, but there is lateral diffusion error at the edge of the diffusion area, which limits its application in sub micron processes.

(2) The resistance measurement after diffusion is usually carried out using the four point probe method, as shown in Figure 2-5. At present, there are various commercial machines available for purchase on the market. (3) The pattern and masking required for diffusion are usually filled with an oxide layer to resist high temperature environments. A micrometer thick oxide layer is sufficient for the general diffusion process.

2、 Oxidation

For silicon semiconductors, as long as oxygen or water vapor is introduced into the furnace tube above or equal to 1050 ℃, as shown in Figure 2-3, the surface of the silicon crystal can naturally be oxidized, and a so-called dry/gate oxide or wet/field oxide layer can be grown, which is used as electrical insulation or process mask for electronic components. Oxidation is the cleanest and simplest type of semiconductor process; This is also one of the advantages that silicon crystal materials can achieve (other semiconductors, such as gallium arsenide GaAs, cannot grow insulation layers using this method because gallium arsenide has dissociated and released arsenic at around 550 ℃!) The silicon oxide layer can withstand the subsequent process environment of 850 ℃ to 1050 ℃ because the oxide layer grows at higher temperatures mentioned above; However, for every 1 micrometer thick oxide layer grown, the surface of the silicon crystal also consumes 0.44 micrometers of thickness.

Here are some key points of the oxidation process:

(1) The growth rate of the oxide layer is not always a constant trend, and the repeatability of process time and growth thickness is a more important consideration.

(2) The later longer oxide layer will penetrate the previous longer oxide layer and accumulate on it; In other words, the oxygen or water vapor required for oxidation must also penetrate the previously grown oxide layer into the siliceous layer. Therefore, the thicker the oxide layer to grow, the greater the obstacles encountered. Generally speaking, it is rare to grow an oxide layer with a thickness of more than 2 micrometers.

(3) The dry oxygen layer is mainly used to make carrier channels for metal oxide semiconductor (MOS) transistors; The wet oxygen layer is used for other less stringent electrical barriers or process masking. The thickness of the former is much smaller than that of the latter, and 1000-1500 angstroms is already sufficient.

(4) For wafers with different crystal plane orientations, the oxidation rate varies: typically, under the same growth temperature, conditions, and time, {111} thickness ≥ {110} thickness>{100} thickness.

(5) Silicon crystals with good conductivity have a faster oxidation rate.

(6) Moderate addition of hydrogen chloride (HCl) results in better texture of the oxide layer; But due to its tendency to corrode pipelines, it has gradually been used less.

(7) The measurement of oxide layer thickness can be divided into two categories: destructive and non-destructive. The former involves removing the exposed oxide layer by soaking in buffered hydrofluoric acid (BOE, a corrosive agent composed of HF and NH4F in a 1:6 ratio) under the definition of light resistance, exposing the surface of the silicon crystal that is not wet, then removing the light resistance, and using a surface profiler or alpha step to obtain the height difference between the presence and absence of the oxide layer, i.e. its thickness.

(8) Non destructive thickness measurement methods are most commonly and accurately used, such as ellipsometers or nanospecs. The former can simultaneously output the refractive index (used to evaluate the quality of thin films), initial thickness b, and step thickness a (total thickness t=ma+b). The actual thickness (which needs to be determined as an integer value of m) still needs to be interpreted in conjunction with process experience. The latter also requires prior knowledge of the refractive index to infer the thickness value.

(9) Different thicknesses of oxide layers will exhibit different colors, and have the characteristic of being cycled once with a thickness of around 2000 angstroms. Experienced individuals can also determine the approximate thickness of the oxide layer based solely on color. However, if the thickness exceeds 1.5 microns or more, the color of the oxide layer gradually becomes less noticeable.