To illustrate design of experiments on mixtures at its simplest level, I blew up a plastic film canister Tuesday evening – not just once, but at least a dozen times. It was really nerve-wracking, but I was prepared to have a headache because the explosive power came from Alka Seltzer® — an amalgam of citric acid, sodium bicarbonate (baking soda) and aspirin.
You can see the apparatus for my experiment pictured: launching tube, container with water, the tablets, plastic film canister (Fuji’s works best), a scale and stop-watch. Research via the Internet produced many write-ups on making Alka Seltzer “rockets.” They generally recommend using only a quarter of one tablet and advocate experimentation on the amount of water, starting by filling the canister half way. I quickly discovered that the tablets break apart very easily, so it was most convenient and least variable to simply put in a whole tablet. Unfortunately, my assistant Katie is a troublesome teenager who discovered that the canister would blow up a second time when I went over to record the first shot’s results in Design-Expert. That’s what frazzled my nerves.
However, I must say that Katie was incredibly cool under the pressure of the plop, plop, fizz, fizz. It took a steady hand to quickly snap on the top of the canister, over which I then placed the launching tube and prepared to press my stop watch. After some seconds the explosion occurred – propelling the lid nearly to the roof of our two-story house. See (and hear!) a launch video from Mad Physics Dot Com (“…where knowledge becomes dangerous”). They flipped the film canister over, whereas I left it upright, thus shooting off the lighter lid.
Before designing my experiment, I did some range finding to discover that only 4 cc of water in the 34 cc canister would produce a very satisfactory explosion. However, it would not do to fill the container because the Alka Seltzer effervesced too quickly and prevented placement of the lid. After some further fiddling, I found that a reasonable maximum of water would be 20 cc’s – more than half full. I then set up a user-defined mixture design with Design-Expert that provided the extreme vertices (4 to 20 cc of water), the centroid (12 cc) and axial check blends at 8 and 16 cc’s. I replicated the vertices and centroid to provide measures of pure error for testing lack of fit.