Path: physics insights > misc >

## Rust and the Oxygen Content of Air

Rust is iron oxide.

When iron which is exposed to air rusts, where does the oxygen come from?  The air, of course; air is about 20% oxygen and that 20% is consumed in forming rust -- or so we are told, and so we generally assume.  But can we test this, using a few articles from the kitchen?  Yes.

 Figure 1: The setup
Never having done this before, we weren't sure what to expect, so we tried a couple different things.  I knew from experience that in dry air, iron mostly just sits there.  However, if it's wet, it tends to rust, sometimes quite rapidly.  In particular, when ordinary steel wool pads are used to scrub pots, they tend to turn into a disgusting mass of dirty brown icky stuff in no more than a day or two.  So we decided to use wet steel wool for the experiment.

For a small supply of air, we used a couple of wine bottles.  We placed a wet pad, and a little bit of water, in the first bottle.  Since none of us here are too clear on exactly how the rusting process works, we also decided to try completely immersing a pad in water.  So we filled the second bottle half full of water, and dropped a steel wool pad into the water.

To measure how much gas was actually consumed, we used balloons.  We squeezed as much as the air as we could from two balloons, and put one over the mouth of each bottle (figure 1).

 Figure 2: 24 hours
The following evening, we noticed that the balloon on bottle one (the "just wet" bottle) seemed to be partly sucked in.  It looked a little twisted at the base; I wiggled it a little and untwisted it a bit, and much to our surprise there was a sort of schloonk sound, and we rather abruptly arrived at figure 2.

Over the next couple of days the inverted balloon continued to inflate (or unflate?).  On the sixth day, nothing further seemed to be happening, so I decided it was time to measure the volume of the balloon.  I filled it with water by pouring it in the top (in the open neck of the inverted balloon), and poured it out and measured it.  It was about 4.5 ounces.  The bottle, when completely filled, held about 25.7 ounces, so the rusting of the pad seemed to have consumed about 17.5% of the air in the bottle.  Apparently, nearly all the oxygen in the bottle had been used up.

Somewhat to my surprise, over the next few days the balloon again seemed to be getting larger.  Perhaps turning the bottle upside down to get the water out to measure the balloon had gotten things going again; whatever the cause, something was happening.  By the end of two weeks, though, things again seemed to have plateaued and at that point I decided to make one more measurement, and then take the balloon off and measure the volume of the bottle (which I hadn't done to start with).  (See figure 3.) Final measurement, after filling the balloon and emptying it out, was 6 ounces.  With a nominal bottle capacity of 25.7 ounces, that means 23.3% of the air was consumed.  Of course, that's more than the amount of oxygen which is in the air!  We'll have a little more to say about that later on, when we discuss sources of error.

 Figure 3: 2 weeks

### What About the Second Bottle?

So what about the bottle half-filled with water?  There's only about half the air in it, compared with the "empty" bottle, and hence only about half the oxygen.  But none the less, in the empty bottle, apparently all the oxygen was used; if anything of the sort had happened in the bottle with water in it, its balloon should also have been sucked into the bottle.  Why didn't that happen?

24 hours after we started the experiment the difference was obvious, and our first thought was that there might be a leak in the second balloon.  Since apparently nothing had happened with the second bottle, there was no concern about disturbing it by breaking the seal, so I took the balloon off and partly inflated it.  No apparent holes.  On general principles, I replaced it with a new (tested) balloon anyway.  But this made no difference; over the succeeding days that bottle apparently did nothing.

But something was happening in the bottle -- the pad was rusting.  It was getting oxygen from somewhere -- but where?  My next guess was that the rust in that bottle was fueled entirely by oxygen which was dissolved in the water.  But there's only a limited amount of oxygen in solution in tap water; what would happen when that was used up?  My guess would have been that oxygen from the air in the bottle would dissolve in the water, and we'd get exactly the same effect we saw in bottle #1 -- but that didn't happen.  Second guess:  If no additional oxygen dissolved in the water, the rusting process must stop when the oxygen originally in the water was used up.  Did that happen?

Indeed, the rust did seem to stop after a point in the second bottle.  Peering into the bottle, it appeared that the surface of the pad was rusted but the parts lying deeper within were still clean-looking.   Without opening the bottle, I photographed the pad through the glass (figures 4, 5).  The glass is fairly dark green, so the result is a bit ambiguous looking (even after adjusting the color balance) but none the less it makes it seems plausible that the (now de-oxygenated) water may actually be preserving the pad by keeping it from the oxygen in the air.

[As it happens, in the days since doing the original analysis, bottle #2 has apparently started sucking in its balloon.  It's not clear why that should be happening now, more than 2 weeks after the experiment started...]

 Figure 4: Figure 5:

### Sources of Error

Let's revisit the numbers we obtained above.

With a nominal volume of 25.7 ounces and a balloon volume of 6 ounces, it appears that the rusting pad in bottle #1 consumed 23.3% of the air.  But that's not quite right.  The pad itself took up some space, and there was a little water in the bottom of bottle #1 as well.  I measured those items after removing them from the bottle, by filling a measuring cup with 6 ounces of water and dropping them in.  The water level rose to just about 7 ounces.  So, there was actually an ounce of space taken up by other stuff, and the volume of air in the bottle was just 24.7 ounces.  With that value, the actual portion of the air consumed was 6/24.7, or about 24.3% of the volume.

From Wikipedia, dry air is about 20.95% oxygen.  Wet air, which we would expect to be dealing with here, actually has a marginally lower oxygen content.  So, where did that extra 3.35% come from?
• Measurements were made with a plain 8 ounce measuring cup, with markings every 2 ounces.   Never mind the accuracy of the markings; the big problem is the difficulty in making a precise reading when the level falls between two marks.  We could easily be off by a quarter of an ounce.  Since the volume of water the balloon contained was just six ounces, this represents an error of +/- 4%.  What's more, the volume of the bottle was measured by filling it with water and then emptying it into the same measuring cup, several times, and counting up the total; that could have introduced an additional error in the volume measurements.  To get a more precise result, we would have been much better off using a graduated cylinder, but we haven't got one.
• When water was added to the balloon, it stretched.  As we all know from filling water balloons, water is heavy, and balloons sag under the weight.  These are fairly strong balloons we're using but the balloon could still have stretched enough to make a significant difference.
This would have increased the apparent volume of gas consumed, but I don't have a reasonable guess as to how much it would be likely to affect the result.  If we wanted to improve on our result we'd need to do some experiments to try to get an estimate of "balloon stretch" in this situation.
• Some water was spilled when emptying the balloon out again.  It wasn't much, but any loss here would tend to make the apparent volume smaller.
There may be other error sources as well, but the ones we've just described are more than large enough to account for a difference of a few percent between what we measured and the "expected" value.

Page created on 1/20/2008