June 01, 2011 - by Simone Ulmer

Research without supercomputers is unimaginable nowadays. However, they increasingly represent an energy problem. Every single computer operation, especially deleting data, converts electrical energy into heat. For this reason, the latest research results by a team of physicists from Switzerland, England and Singapore deserve careful attention: under certain conditions, cold is generated instead of heat when deleting data. The only condition is that the content of the memory must be known “more than completely” during the deletion process. This is possible provided the so-called quantum-mechanical entanglement is included, since such entanglement carries more information than a classical copy of the data. This is confirmed by a study led by ETH Zurich Professor Renato Renner together with Vlatko Vedral of the National University of Singapore and published today in “Nature”.

Landauer's Principle not valid in all cases

The fact that computers produce heat when they process data is a logistical challenge for computer manufacturers and supercomputer operators. In addition, this heat production also imposes a fundamental limit on their maximum possible performance. According to the so-called Landauer Principle formulated by the physicist Rolf Landauer in 1961, energy is always released as heat when data is deleted. Renner says, “According to Landauer’s Principle, if a certain number of computing operations per second is exceeded, the heat generated can no longer be dissipated.” Assuming that supercomputers develop at the same rate as in the past, this critical limit will probably be reached in the next 10 to 20 years. The physicist stresses that, in principle, the heat output when deleting a ten-terabyte hard disk is less than a microjoule. Nevertheless, if such a deletion process is repeated many times per second, the heat accumulates accordingly.

 

However, the study now shows that the Landauer Principle holds true only if the value of the bits to be deleted is unknown. Erasing a memory is normally an irreversible process, but if the memory content is known, it is possible to delete it in such a way that, in theory, it could be restored. As a result the deletion operation becomes a reversible process to which Landauer’s Principle no longer applies.

Two identical formulae, but understood differently

The scientists proved this mathematically by combining the entropy concept from information theory with that from thermodynamics. Entropy appears differently in these two disciplines, which are, to a large extent, independent of each other. In information theory, entropy is a measurement of the information density. It describes, for instance, how much memory capacity a given set of data would take up when compressed optimally. In thermodynamics, on the other hand, entropy relates to the disorder in systems, for example to the arrangement of molecules in a gas.

The concept of entropy was introduced into science twice, independently of each other. Renner says, “We have now shown that the notion of entropy actually describes the same thing in both cases.” Since the formulae look the same, a deeper connection between the two had already been suspected. “Our study shows than in both cases the entropy can be regarded as a measure of ignorance.” He says that the implication for physics is that an object does not simply have a certain entropy, but that this entropy always depends on the observer: if two persons erase data stored in a memory and one of the two knows more about the memory content, he requires less energy to erase the memory. Consequently, the thermodynamic entropy depends on the observer.

Negative entropy is decisive

According to the scientists, the results lead to the conclusion that, in principle no energy is needed to delete data stored in a classical computer. In the case of a quantum computer, in which the user could know the memory content “more than completely” due to quantum entanglements, the entropy would even be negative, i.e. heat would be withdrawn from the environment – and it would cool down. The process would convert the heat into usable energy. However, Renner stresses that, “This does not mean we have discovered a perpetual motion machine.” He says that, because deleting data is a once-off process, it cannot be used to generate energy continuously.

In practice, that would ideally mean that water supplied to supercomputer processors for cooling would, after the deletion process, return slightly colder – but only if the memory content was known exactly. For this purpose, however, the computer’s processors would need to be designed so as to enable quantum effects to be exploited. This means that each data bit is represented by only a single atom instead of hundreds of atoms.

Renner and his team expect that the new knowledge about the concept of entropy in physics and information theory will lead to the discovery of yet more connections between the two disciplines. With the aid of information theory and its ability to handle partial knowledge, it might be possible to close gaps in our current understanding of thermodynamics and statistical mechanics.

Further reading

Del Rio L, Aberg J, Renner R, Dahlsten O & Vedral V: The thermodynamic meaning of negative entropy, Nature (22011) DOI: 10.1038/nature10123