[rael-science] Breaking the noise barrier: Enter the phonon computer

 

 

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The Raelian Movement
for those who are not afraid of the future : http://www.rael.org
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Source: New Scientist
http://www.newscientist.com/article/mg20827801.500-breaking-the-noise-barrier-the-age-of-phonon-computing.html

Breaking the noise barrier: Enter the phonon computer

Noise is a chip designer's worst enemy. But handled properly it could become a powerful ally – and usher in the age of phonon computing

In 2001, Pat Gelsinger, then the chief technology officer of Intel, made a striking prediction about the future of microchips. If current design trends continue, he said, microchips will be running at 30 gigahertz by the end of the decade. However, he added, at this speed they will be generating more heat per cubic centimetre than a nuclear reactor.

Sure enough, by 2003, Intel and other chip-makers had found that their plans for faster processors were running into trouble. For a chip to speed up, its transistors need to be shrunk, but smaller transistors must consume less power or they overheat. With chip-makers unable to keep to the reduced heat budget, the race for faster chips hit a wall (see diagram).

At best, today's microprocessors can operate at just 3 GHz or so. To deliver a major performance boost, chip-makers have resorted to putting several processors, or cores, on the same chip. This keeps heat at manageable levels. Just.

Designing transistors that need far less power is, it turns out, no easy task. One of the main reasons is that microchips still require plenty of power to overcome electrical noise, which tends to flip the 1s and 0s in digital data, destroying information. The codes that computers rely on to transmit information have built-in checks to combat this.

The effects of noise become more serious in chips that run at low power since the actual signals being handled by a chip become smaller and can be more easily overwhelmed. The worry is that low-power chips will never be possible because a sea of noise will always drown the information they are trying to process. Noise, it seems, is a fundamental barrier for chip designers.

Yet a growing number of researchers and chip-makers are exploring ways to overcome noise, and the signs are encouraging. New insights into the role of noise are allowing them to build processors that can not only tolerate noise but actually exploit it to perform calculations.

Unexpected boost

This could lead to a new generation of nanoscale devices that can manipulate noise, just as existing devices manipulate electrons. In fact, it may one day be possible to use noise itself to store, carry and process information, opening up a new era of ultra-low-power computing. A chip designer's worst enemy could turn out to be a powerful ally.

Electrical noise - essentially any unwanted signal - can arise from a number of sources. One of the most common is heat, which can increase the random motion of atoms and electrons in the metals or semiconductors of circuit components. This can lead to unwanted fluctuations in current or voltage. If the magnitude of this noise is large enough it can turn an electronic circuit to junk.

Though digital circuits are more resistant to noise than analogue designs, they remain vulnerable. To see why, think of a digital switch such as a transistor. It works by producing either a low or high voltage as an output, the equivalent to a 0 or 1. If the input voltage rises above some threshold, the output jumps between these states. So if the magnitude of the electrical noise exceeds this threshold, it can flip the state of the switch, leading to errors. No wonder, then, that microchips are fitted with heat sinks and fans.

Yet electrical noise isn't always bad news. We have known for decades that the presence of noise can improve the performance of certain switch-like systems. If the noise level is just below the threshold needed to flip the state of the system, even a tiny input voltage is enough to change the system's state. In effect, the noise increases the sensitivity of the switch - a phenomenon called stochastic resonance. When it occurs in neurons, for example, the process can sharpen our senses (New Scientist, 21 June 2008, p 42).

Vital component

This raises the curious prospect that noise could actually be used to improve the performance of electronic circuits. For example, Luca Gammaitoni, a physicist at the University of Perugia in Italy, recently unveiled a type of switch known as a resonant tunnelling diode that benefits from noise in exactly this way. His device can tolerate noise levels that are as large as the input signal itself (Applied Physics Letters, DOI: 10.1063/1.3302457).

This kind of assistance needn't be limited to conventional electronic devices. Raj Mohanty at Boston University and colleagues have harnessed noise using a bar of silicon just 20 micrometres long and 300 nanometres wide. This bar naturally vibrates at 3.145 MHz, but apply an alternating voltage and it vibrates more quickly or slowly.

A plot of the bar's vibrational frequency against the applied voltage reveals a strange effect, however. Raise the voltage and the bar vibrates faster. Now lower the voltage to its previous value and the bar's vibrations slow down, but not to the original frequency. This phenomenon, known as hysteresis, means that for a given voltage there are two corresponding vibrational frequencies for the bar. Which one is actually observed depends on whether the voltage has been increased or lowered to the required value. That makes Mohanty's silicon bar a kind of nanomechanical switch in which the two voltages are inputs representing a 0 and a 1, and the output depends on whether the input is rising or falling.

By simultaneously applying two high-frequency voltages, each of which can be taken to represent either a 0 or a 1, Mohanty devised a way to ensure that the bar's output is 0 unless both inputs are 0, in which case the output is 1. He had built a logic gate - to be precise, a NOR gate.

But there's another important factor in this process: background noise. The exact shape of the hysteresis curve is very sensitive to it. Add slightly more or less noise and the characteristics of the switch change, turning it into a different kind of logic gate. By carefully controlling the noise level, Mohanty can make the bar behave as a NOR gate, an OR gate, an AND gate or a NAND gate - all the logic gates necessary to build a computer (Nano Letters, DOI: 10.1021/nl9034175).

Although nanomechanical logic gates are slow compared with conventional switches, they consume two orders of magnitude less power than their conventional cousins. "There's huge potential for these switches, especially in power-hungry devices," says Mohanty. He has in mind remotely operated sensors and processors that may need to work for years without fresh batteries. Nanomechanical versions of these devices could be ready in five to seven years, he says.

In the meantime, noise-based logic may be possible on a much grander scale. For the last couple of years, Laszlo Kish at Texas A&M University in College Station and colleagues have been working on the theoretical properties of a logic system that uses large-scale random noise signals. Their idea is to represent the 0s and 1s of digital signals not using voltage levels as in conventional computers, but using the presence or absence of noise.

Because noise is random, it's easy to imagine that noise created by different sources is identical. In fact the opposite is true - noise has a pattern that is characteristic of its source. It is this that makes it possible to keep track of different noise signals and compute with them, says Kish. What's more, any background noise will be different from the noise signals we are working with, making it possible to subtract its effect.

In Kish's scheme, representing a single bit of information requires two independent sources of noise, one representing a 0 and the other a 1, while a string of n bits requires 2n sources. That's not a problem, Kish says: transistors can be a good source of noise when operated at low voltage, and we can already fit billions of them on a chip.

Earlier this year Kish and his collaborators published a claim more significant still: that noise signals can be superimposed and sent through a single wire without losing their identity. By operating on a composite signal, or superposition, it becomes possible to carry out two or more calculations simultaneously - much the same trick that quantum computers exploit to speed up calculations. Kish says this kind of logic is especially suited to certain types of calculation, but he and his colleagues are still trying to quantify exactly how this can be realised in practice. So far they have simulated basic circuits for generating composite noise signals, as well as noise-based AND and OR logic gates, among other components. Eventually these could provide fast, low-power processing, they say.

By superimposing noise signals, it becomes possible to carry out two or more calculations simultaneously - much the same trick that quantum computers exploit

Kish's noise-based logic scheme raises another intriguing prospect. Since natural systems have evolved not only to cope with noise but to exploit it, as neurons do with stochastic resonance, could it be that nature has also learned to compute with noise? Kish and his collaborator Sergey Bezrukov of the National Institutes of Health in Bethesda, Maryland, think so. They say their logic scheme could help explain some of the features of neural activity in mammals, such as the delays that seem to occur in certain neural signals. They also suggest a hypothetical scheme by which the brain could efficiently route and encode information using a superposition of noisy neural signals (Physics Letters A, DOI: 10.1016/j.physleta.2009.04.073).

Kish's noise-based logic scheme raises another intriguing prospect: could it be that nature has also learned to compute with noise?

We're still a long way from understanding how the brain processes information, says Gammaitoni, who suggests that some radical rethinking may be necessary before we can untangle the details. What is clear is that nature has evolved to cope with noise far more efficiently than conventional electronics can. Whatever new picture of biological computation emerges, it looks as if noise will play a central role.

Circuits run on heat

The latest transistors measure just 22 nanometres across, but chip-makers plan to shrink them further still. That means heading into unfamiliar territory, for the properties of materials change when devices are just nanometres across.

At this scale, noise and heat are essentially indistinguishable, manifesting themselves as vibrations in the lattice of atoms forming the material from which the chip is carved. These vibrations can be thought of as quantum objects called phonons, and the study of noise boils down to understanding their physics.

That's hugely challenging, says Luca Gammaitoni of the University of Perugia in Italy. Phonons interact with atoms, electrons and electric fields, or even with each other, in a complicated and non-linear fashion, so a small change in the lattice can lead to a huge change in phonon behaviour.

Nonetheless, the emerging discipline of phonon engineering could lead to devices in which phonons themselves are the information carriers. Instead of hindering the flow of information, noise could actually carry it.

Physicists have already shown that the phonons in a row of ions can behave rather like a computer. Information can be "written" to one ion by zapping it with a laser to change its state. This also modifies the electric forces between the ions, changing the way the row vibrates. Phonons, representing the collective motion of the row, in effect share the data between the ions, allowing them to "process" it. When processing is complete, the ions can be made to release the "answer" as a photon.

Of course there are important differences between the phonons in a string of ions and those in an ordinary material. The former system is a laboratory creation in which the ions are trapped by an electric field, then cooled and kept isolated from the environment to preserve the quantum information. Phonon behaviour in a solid under everyday conditions is much more complex.

That's not to say that phonon information processing is a lost cause. Far from it. Devices are being built that can manipulate phonons, allowing them to flow in one direction only, for example - the thermal equivalent of a diode.

In 2006, a team led by Alex Zettl at the University of California, Berkeley, showed how this could be done using carbon nanotubes. Zettl embedded the tubes in an asymmetric pile of heavy molecules. The effect was to make each tube behave as if it was more dense at one end than the other. He then measured the way that heat passed along the tubes and found that the flow was greatest towards the "low density" ends. In other words, the phonons moved more easily in one direction than the other (Science, DOI: 10.1126/science.1132898).

These thermal diodes could have a number of applications, such as improving the efficiency of thermoelectric converters. This may make it possible to harvest enough heat energy from the environment to power future generations of low-power microchips, doing away with batteries entirely.

The work also indicates that thermal equivalents of electrical circuits are possible. "But we need to prove we can do logic with these devices," says Clivia Sotomayor Torres, an expert on phonons at the Catalan Institute of Nanotechnology in Barcelona, Spain.

Justin Mullins is a consultant at New Scientist



__._,_.___
Recent Activity:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Ethics"  is simply a last-gasp attempt by deist conservatives and
orthodox dogmatics to keep humanity in ignorance and obscurantism,
through the well tried fermentation of fear, the fear of science and
new technologies.

There is nothing glorious about what our ancestors call history, 
it is simply a succession of mistakes, intolerances and violations.

On the contrary, let us embrace Science and the new technologies
unfettered, for it is these which will liberate mankind from the
myth of god, and free us from our age old fears, from disease,
death and the sweat of labour.

                                    Rael
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-----------------------------------------------------------------

Tell your friends who love scientific news that they can
subscribe to this list !!

They can do it by sending a blank email to:
                  rael-science-select-subscribe@yahoogroups.com

It's free !
-----------------------------------------------------------------
To unsubscribe, send an email to:
                 rael-science-select-unsubscribe@yahoogroups.com
-----------------------------------------------------------------
To change your e-mail address, unsubscribe from the old address and subscribe from the new address (see above).
-----------------------------------------------------------------

.

__,_._,___

0 comments:

Post a Comment