The longer you use a computer, the hotter it will get. That seems like just an everyday fact of life, but it might actually be its own law of physics. And, like all phyiscal laws, quantum mechanics apparently violates it.
The original idea was first proposed by physicist Rolf Landauer, who argued that it is a basic principle that, in order to erase one bit of information, you must increase the entropy of the environment by the same or greater amount. Since the simplest way to increase entropy is for heat energy to build up and dissipate, this means erasing a bit of information raises the temperature by one "bit" of heat. That idea has been around for about 50 years, but there's still sharp disagreement among physicists as to whether Landauer actually described a full-on law of physics.
Working from that principle, physicist Vlatko Vedral and his team have figured out how it would interact with quantum mechanics. As you might expect, the results are counterintuitive and deeply strange. Writing about his work in Scientific American, Vedral explains:
Our new paper argues that in quantum physics, you can, in fact, erase information and cool the environment at the same time. For many physicists, this is tantamount to saying that perpetual motion is possible!
Thankfully, like all other apparent violations of basic physical laws, this bit of quantum weirdness isn't actually opening up the doors to perpetual motion. Vedral goes on:
This, luckily for the second law (though not for would-be inventors of perpetual motion machines), is not the case. Landauer's insight is still fine, and erasing information adds entropy to the environment. What saves the second law is that, in quantum physics, entropy can actually be negative. Adding negative entropy is the same as taking entropy away. The key phenomenon behind it is the spookiest of all quantum phenomena, entanglement.
To understand the connection between entanglement and negative entropy we have to go back to Schrödinger's view of entanglement. When two systems are entangled, we have complete information about their joint state, but have no information about their individual states. If we are erasing the state, as a whole we need not generate entropy (since the state has zero entropy), but if we erase subsystems individually, then each will contribute to entropy generation. The difference between the global and local erasing is negative entropy. To rephrase, if we have to erase some information, it helps to know whether this information arises from the entanglement with another system. Then, by invoking the other system in the erasure, we can actually erase and the environment can lose entropy.
For more, check out Vedral's entire article over at Scientific American.