Roger • March 12, 2006 11:09 PM

@Anonymous, krunch:
The point being made on the movie physics site was not that the attack wouldn’t work at all, but that it would have taken much longer than depicted. In the movie, the attacker took only 15 minutes to fill a safe with an internal capacity of close to 1,000 litres, using a very narrow bore pipe, and with no air outlet so it had to contend with back pressure. With a pipe of the same diameter but still no back pressure, the movie physics guys got a flow of eight litres per minute, so it would have taken at least 125 minutes, or over 2 hours. While back pressure might have escaped in various ways, it would have increased the time even further.

Something they don’t point out is that the back pressure gets you two ways: if the safe is hermetically sealed, the back pressure will soon build up to the point where it seriously restricts the inflow of water. On the other hand, if the safe is not airtight some of the water will be leaking out while you try to fill it. Most likely you will fall between two stools: the safe will be sufficiently close fitting for some pressure to build up and slow down the inflow without stopping it, but this increased pressure will also cause water to exit through the leaks even faster! In fact, if all the leaks are at the bottom so only liquid can escape, eventually it will reach an equilibrium state where water escapes exactly as fast as it is added.

@Everyone:
the question of whether you could blow open a safe by an internal gas explosion is not hard to answer. Assume you have the inside of the safe completely filled with a mixture of a fuel gas and oxygen. Upon ignition, the temperature will (very briefly) reach an elevated high temperature, and then will cool (largely adiabatically) as it expands. For the sake of argument, let’s assume that peak temperature is 3,000°C (although it could be somewhat higher with some fuel gases and perfect oxygenation). Then from Guy-Lussac’s gas law we can estimate the pressure inside the safe as (3000 + 273)/(20 + 273) atmospheres, or 11.2 atmospheres (164 psi). ATM safes sizes vary, but the closest one I could get a look seemed to have a door size of 460 mm x 850 mm, more or less. Therefore the initial force on the door will be about 443 kN (99,400 lbf), or about 45 tonnes force. It seems quite likely that this is enough force to significantly deform the safe. Note that this force is felt by all internal faces simultaneously, so for example the sides will tend to bulge outward (and thus tend to release the door bolts) as well as the door being pushed outward directly. Note also the ATM safes (at the least the one I just examined) are not actually all that thick walled (appears to be only about 5 mm of steel). By the time all walls have expanded 25 mm away from their initial position (which I wildly guess is enough to release the door and allow it to accelerate freely), the internal volume will have increased from about 0.18 m^3 to 0.21 m^3, so assuming adiabatic expansion and complex combustion products, the pressure will only have dropped by 17%, and the force on the door is still about 38 tonnes.

In short, a simple gas explosion inside an ATM safe is certainly capable of tearing the safe to pieces.

As for counter-measures, many possibilities are obvious, but here is the best one to strike me: fill all unoccupied parts of the interior with blocks of an elastic closed-cell foam. This will occupy volume in the safe and prevent more than a very small amount of fuel gas being admitted. When the ignition occurs, the foam will rapidly collapse with very little resistance, causing the pressure to drop very rapidly, before damage is done to the door. Also, depending on the material used, some of the foam may melt and mark the cash with globs of coloured molten plastic, thus ruining it even if the safe does open.

This has some similarities to the proposed empty chambers, and also other space-filling, very soft materials might be used, but a closed cell foam has several advantages. Inflated plastic bags, thin-walled boxes or other hollow objects could be easily punctured by the attacker through the drill hole, and thus rendered useless. Open cell foams would delay the admission of the fuel gas, but not prevent it. An inelastic foam would prevent one attack, but would then be crushed and allow the attacker to try again; and elastic foam would probably take many attacks to collapse completely. By that time the cash would be ruined, even assuming the police had not yet arrived.

Finally, the elastic foam would be very simple and inexpensive to retrofit to existing ATMs. Once the best shapes have been determined, a kit of suitable shapes can be mass produced, taken to each ATM when it is being refilled, and simply slipped into place. The procedure would be much cheaper and quicker than fitting vent holes or new hollow sections or the like.