TimeDomain CVD, Inc. 

In the previous section we exhibited a couple of important solutions to the diffusion equation. In both solutions, the distance x was divided by ("scaled by") a particular combination of the other parameters in the problem: the time t and the diffusivity D. This quantity has the dimensions of length (as it must) and is known as the mass diffusion length.


The diffusion length is the characteristic length scale for diffusion problems. It increases as the square root of the time. For semiinfinite regions, the solution looks the same for all times if the length is scaled by the appropriate diffusion constant for that time. 

However, for real finite systems (like chemical vapor deposition reactors) the behavior of solutions changes qualitatively depending on how large the system is in comparison to the diffusion length.


If the diffusion length is much longer than the system size, the profiles of concentration must be essentially linear, and are independent of time (if diffusion is the only thing happening). A change in one part of the reactor is reflected throughout the reactor in a diffusion time (L^2/4D). Note that the relevant time used to find the diffusion length will often be the residence time of gases in the reactor. As we'll discuss below, in this case fluid velocity is generally not important: diffusion dominates transport. 

If the diffusion length is much shorter than the system size, concentrations can differ drastically from one part of the reactor to another. Gradients can be very large, and concentrations are highly timedependent. Large changes in concentration in one region in the reactor will have no effect on other regions if the time involved is short enough that the diffusion length is small. 
The diffusion length is our first introduction
into a characteristic scale for a problem. Chemical engineers like to
divide a characteristic length (like the reactor size, for example) by
such a characteristic scale for a physical process, to obtain a dimensionless
number. Instead of asking whether the length scale is large or small
relative to the system, they ask the entirely equivalent question, is
the relevant dimensionless number large or small compared to 1?
This approach is mathematically fine, but I personally find it very hard to remember! It is much easier to keep things straight when a physical quantity (like a length or time) is calculated and compared to the relevant parameter for your system. Therefore in this tutorial we'll usually use length and time scales, but we'll try to remember to translate them into Chemical Engineerese to help the reader access the technical literature. 
An important example is the ratio between the system length and the diffusion length, typically calculated with the time set to the residence time (the average time gases are in the reactor). In order to avoid having square roots in the calculation, it is traditional to calculate the SQUARE of this ratio (with a constant factor): the Peclet number Pe, which works out to be the product of system length and fluid velocity, divided by diffusivity. If Pe >>1, large gradients can exist. If Pe << 1, diffusion dominates transport and things are linear. 

So what do you think? Lengths or bare numbers? 

Let's consider the case where the Peclet number is much less than 1  that is, the diffusion length is much longer than the system size. In this case, all transport is dominated by diffusion: convection is insignificant within the chamber. If deposition (or any other reaction) occurs at the surface, there is a loss of species to the wall. Thus we must consider simultaneous transport by diffusion and loss by deposition, as shown below for the case of a long, thin region where concentration is approximatey constant in the ydirection and we consider only transport in the xdirection. 
[This relatively simple situation is actually of practical relevance, for example in a singlewafer reactor at low pressures (and low molar flows  more later) or transport in between wafers in a tube furnace. ] 
If we balance the moles entering and leaving a control volume dx wide we find: 

From which we find the transport equation: 

Solutions in one dimension are exponentials 

with x scaled by the Thiele length:
(naturally, we could also define the dimensionless Thiele modulus, the ratio of the system size to the Thiele length) 
We can see how transport scales: larger
separation between the surfaces encourages transport; at lower pressures D
increases while the reaction rate stays the same, so the characteristic
length increases: LPCVD.
This problem was originally studied in connection with catalysis in porous media, in which the consumption of species occurs not through deposition but through conversion into products. 
