This simulation shows atomic layer deposition (ALD) of aluminum oxide from trimethylaluminum (TMA, Al(CH3)3) and water on a flat, hydroxyl-terminated silicon surface. The view is a cross section of the reactor; each column on the surface is one reactive site.
Press Play to run the sequence TMA pulse → N2 purge → H2O pulse → N2 purge, repeated for the number of cycles you choose. Gas molecules enter from the inlet on the left and leave to the pump on the right. The plot shows the mass gained per surface site (as a quartz crystal microbalance would measure it), with the pulses and purges shaded.
Use the pulse time slider to see self-limiting growth: long pulses saturate every site and extra precursor bounces off; short pulses leave unreacted sites and the film grows less per cycle. Uncheck N2 purge to see what happens when TMA and water meet in the gas phase: they react to form particles that land randomly (CVD-like growth, shown in darker gray). Check pause after each step to stop after every pulse and purge; press Play to continue. When the run finishes, you can increase the number of cycles and press Play to keep growing the same film; press Reset to start over with a clean surface.
ALD grows a film by exposing the surface to two precursors one at a time, separated by inert-gas purges. Each precursor reacts only with the surface groups left by the other, so each half-reaction stops by itself once those groups are used up.
A (TMA pulse):
$$\|\!-\!\mathrm{OH} + \mathrm{Al(CH_3)_3(g)} \rightarrow \|\!-\!\mathrm{O\!-\!Al(CH_3)_2} + \mathrm{CH_4(g)}$$B (H2O pulse), which occurs twice per aluminum atom:
$$\|\!-\!\mathrm{Al\!-\!CH_3} + \mathrm{H_2O(g)} \rightarrow \|\!-\!\mathrm{Al\!-\!OH} + \mathrm{CH_4(g)}$$Overall:
$$2\,\mathrm{Al(CH_3)_3} + 3\,\mathrm{H_2O} \rightarrow \mathrm{Al_2O_3} + 6\,\mathrm{CH_4}$$TMA does not react with a methyl-terminated surface and water does not react with a hydroxyl-terminated surface, which is why growth is self-limiting.
If each molecule that strikes an unreacted site reacts with probability \(s\), the fraction of sites that have reacted, \(\theta\), follows Langmuir-type kinetics:
$$\frac{d\theta}{dt} = \frac{s\,F}{n_0}\,(1-\theta)$$ $$\theta = 1 - \exp\!\left(-\frac{t}{\tau}\right), \qquad \tau = \frac{n_0}{s\,F}$$where \(F\) is the flux of precursor to the surface and \(n_0\) is the number of reactive sites per area. Pulses much longer than \(\tau\) saturate the surface. The simulation uses \(s = 0.8\) for both precursors and treats each molecule's motion as a random walk. During a pulse the flow to the pump is throttled so the precursor has time to reach the surface; during a purge a fast N2 flow carries gas to the pump.
The plot uses the mass of the surface group on each site. Starting from –OH (17 amu), TMA adds –Al(CH3)2 minus H, a gain of 56 amu. Each CH3 → OH exchange in the water pulse adds 2 amu. One complete cycle adds 60 amu per site in this idealized model; the large step comes from TMA and the small step from water, as in QCM measurements.
Each drawn layer is one cycle of growth, about 1.1 Å for Al2O3 at 150–300 °C. In real films, the bulky methyl groups block neighboring sites (steric hindrance), so one cycle deposits roughly a third of a monolayer. The gas-phase particles formed without a purge are drawn as whole layers on one site; real CVD deposits are rough and nonuniform and amorphous unless grown at 800 to 900 °C. The drawing is not to scale: the gas region is compressed and molecules are enlarged.
| \(F\) | flux of precursor molecules to the surface (molecules/(cm2·s)) |
| \(n_0\) | surface density of reactive sites (sites/cm2) |
| \(s\) | reaction probability: fraction of collisions with an unreacted site that react |
| \(t\) | exposure (pulse) time (s) |
| \(\theta\) | fraction of surface sites that have reacted |
| \(\tau\) | characteristic time to saturate the surface (s) |
| \(\|\!-\) | surface-bound species |
This simulation was created in the Department of Chemical and Biological Engineering at University of Colorado Boulder for LearnChemE.com by John L. Falconer using Claude AI. It was prepared with financial support from the National Science Foundation (DUE 2336987 and 2336988). Address any questions or comments to LearnChemE@gmail.com.