Growing Oxides on Silicon Substrates
by theglassman in Workshop > Science
108 Views, 0 Favorites, 0 Comments
Growing Oxides on Silicon Substrates
.jpg)
This is a guide on how to grow simple oxide layers on a silicon wafer.
Supplies
Silicon Wafer - If you are making transistors, choose a wafer with the correct doping (n/p), if not, then any kind will do (I am using <111> orientation).
Furnace - A tube furnace is probably most appropriate. You will need to reach a minimum of 800-900°C for the oxide layer to grow at a reasonable rate. Here is a link for the furnace I made to grow oxides for semiconductor devices.
https://www.instructables.com/1200C-Tube-Furnace-Using-a-CJD-9000P-PID-Controlle/
Push/Pull Rod and Quartz Boat - You'll need a rod for retrieving your samples from the furnace. I fashioned one with a hook on the end from spare thick copper wire, however I have also used quartz rods which work really well too. The quartz boats come in handy when dealing with lots of fiddly small samples.
Steam Source - I used a conical flask on a hotplate with a glass right angle adapter. This generates a stable source of steam that pipes smoothly into the furnace. There are a million different ways to accomplish something similar to this and all of them will be fine, you will get away with using a kettle if you need to.
Distilled Water - Nice to have, if you can't get it then use branded water marketed as 'pure' or 'mineral-free'.
Other Bits - Wafer tweezers (get a bag of them as they will break/get contaminated/melt), a craft-knife/razor blades and acetone are all nice to have. You will need a timer of some kind, I used a mobile phone.
Background
Thermally growing oxides is common in industry, the process is extremely simple; silicon is exposed to oxygen at a high temperature and the two react to produce silicon dioxide (i.e. glass). This is called dry oxidation growth, and it is characterised simply by the following equation:
Dry:
Si + O₂ → SiO₂
Dry processes are best for growing thin layers of high quality oxides. The main downside of the dry process is the large thermal cost, as the oxides grow extremely slowly and hence the process furnace must stay hot for longer. Dry oxides take a long time to grow since the large O₂ molecules diffuse very slowly through the already existing SiO₂ layer. A wet process resolves this by introducing a steam atmosphere, which can diffuse through the SiO₂ faster (H₂O molecules are significantly smaller than O₂). A wet process is governed by the following equation:
Wet:
Si + 2H₂O → SiO₂ + 2H₂
This is around 10x faster than a dry process. It is important to note that the hydrogen remains within the oxide layer as a contaminant, this has important implications for the chemical structure and electrical properties of the SiO₂.
In the semiconductor sector, these two processes lend themselves to different oxide types. Gate oxides are usually grown in a dry environment (since the dielectric properties of the gate oxide is critical, and they are very thin), while field oxides are usually grown in a wet environment (as they need to be at least 10-100x taller than gate oxides for most transistors).
This guide will focus on wet oxides. If you have a need to specifically characterise a dry process, the same steps will apply except without the steam.
Experimental Setup
.png)
.png)
Setup your furnace with your steam source at one end. You must be able to have access to at least one side of the tube in order to place and retrieve your samples. Heat the furnace to 1000°C and ensure you can see steam flowing properly through the furnace. I would recommend having your thermocouple on the same side of the furnace as the steam input, this will allow the furnace's controller to compensate for the low temperature of steam cooling the furnace tube.
Prepare your wafer by cleaving it into appropriately sizes pieces. For my experiment, I wanted one long strip of around 7cm and 11 smaller pieces of approximately 4x8mm. To cleave a wafer reliably, use sharp metal tweezers (a razor blade or similar will also work well) and press at the edges where you would like the wafer to fracture. Remember that <111> wafers will fracture along a square grid, so only attempt to fracture pieces along a perpendicular to an edge, otherwise the wafer will shatter or split unreliably.
My wafer pieces were to be used in two experiments to characterise the performance of my DIY furnace, these were:
> To verify that the furnace tube was being heated evenly, by baking the long silicon strip for 30mins and checking that the oxide growth on it was uniform.
> To characterise wet field oxide growth in my furnace, by baking the small silicon pieces incrementally at 10min intervals over a 100min period, with an extra data point taken at 120mins.
I assumed a native oxide was of a negligible height. Before putting the wafers in the furnace I rinsed them with distilled water, and arranged the small pieces in a quartz boat. When the furnace was up to temperature, I used my push rod to 'float' the quartz boat into the middle of the furnace, waited 10mins, pulled the boat out, and extracted a sample, and repeated until all the samples were used. The silicon bar was simply baked for 30mins and extracted using the rod. Your experiment will vary depending on your objective.
Results
.jpeg)
When looking at your oxides, try and use a light source with a broad spectral output (e.g. sunlight, candlelight etc). You can refer to the oxide depth chart to determine the thickness of the oxide layer.
From my results you can see that my furnace was of a uniform temperature in the centre, at the edges it was much cooler resulting in a shallower oxide growing (this can be seen by the colour gradient in the thick bar, the blue side is nearer the edge and hence has thinner oxide). The small brown stains visible on the silicon bar are from particulate matter. Dirt will have been removed by the rinsing stages, so I must have reintroduced some contaminants before placing the bar in the furnace (somehow!). I redid this test using a different sample, this time for 2 hours at 1000°C, and the temperature drop-off towards the furnace edges is clear, (evident in the gradient from purple to blue, with the purple end being much thicker).
Wet oxide growth can be characterised using the Deal-Grove model, which is a simple parabolic relationship. This simplifies further to a linear relationship when the pre-existing oxide layer is shallow. The little samples shown were taken at 10min intervals from 0min (left) to 100min (right) at 1000°C, the 120min datapoint is not shown. I have plotted the results, from which we can see that the oxide growth rate in the linear region is ~6nm/min (or ~60nm/10min as can from the samples using the chart). Note the oxide growth tends off when t is large, this is accounted for by the Deal-Grove model (i.e. we can see the x^2 term in the model ceasing to be negligible, hence the 'square-root-esque' behaviour).
Sources
Bit of background on oxide growth < this is part of a very good series of articles about semiconductor process stuff