Researchers discover fastest light-driven process

Dec 05, 2012

(Phys.org)—A discovery that promises transistors – the fundamental part of all modern electronics – controlled by laser pulses that will be 10,000 faster than today's fastest transistors has been made by a Georgia State University professor and international researchers.

Professor of Physics Mark Stockman worked with Professor Vadym Apalkov of Georgia State and a group led by Ferenc Krausz at the prestigious Max Planck Institute for Quantum Optics and other well-known German institutions.

There are three basic types of solids: metals, semiconductors, used in today's transistors, and insulators – also called dielectrics.

Dielectrics do not conduct electricity and get damaged or break down if too high of fields of energy are applied to them. The scientists discovered that when dielectrics were given very short and intense , they start conducting electricity while remaining undamaged.

The fastest time a dielectric can process signals is on the order of 1 femtosecond – the same time as the light wave oscillates and millions of times faster than the second handle of a watch jumps.

Dielectric devices hold promise to allow for much faster computing than possible today with semiconductors. Such a device can work at 1 petahertz, while the processor of today's computer runs slightly faster than at 3 gigahertz.

"Now we can fundamentally have a device that works 10 thousand times faster than a transistor that can run at 100 gigahertz," Stockman said. "This is a field effect, the same type that controls a transistor. The material becomes conductive as a very high of light is applied to it, but dielectrics are 10,000 times faster than semiconductors."

The results were published online Dec. 5 in Nature. The research institutions include the Max Planck Institute for , the Department of Physics at the Munich Technical University, the Physics Department at Ludwig Maximilian University at Munich and the Fritz Haber Institute at Berlin, Germany.

At one time, scientists thought dielectrics could not be used in signal processing – breaking down when required high electric fields were applied. Instead, Stockman said, it is possible for them to work if such extreme fields are applied at a very short time.

In a second paper also published online Dec. 5 in Nature, Stockman and his fellow researchers experimented with probing optical processes in a dielectric – silica – with very short extreme ultraviolet pulses. They discovered the fastest process that can fundamentally exist in condensed matter physics, unfolding at about at 100 attoseconds – millions of times faster than the blink of an eye.

The scientists were able to show that very short, highly intense light pulses can cause on-off electric currents – necessary in computing to make the 1s and 0s needed in the binary language of computers—in dielectrics, making extremely swift processing possible.

Explore further: Symphony of nanoplasmonic and optical resonators produces laser-like light emission

More information: The first paper, "Optical-field-induced current in dielectrics" is available through dx.doi.org/10.1038/nature11567 . The second, "Controlling dielectrics with the electric field of light," is available through dx.doi.org/10.1038/nature11720 .

Related Stories

Electron ping pong in the nano-world

Apr 25, 2011

(PhysOrg.com) -- An international team of researchers succeeded at the Max Planck Institute of Quantum Optics to control and monitor strongly accelerated electrons from nano-spheres with extremely short and ...

Light oscillations become visible

Aug 28, 2004

The human eye can detect changes in the intensity of light, not however the wavelength because light oscillates too fast (approximately 1000 trillion times per second). An international collaboration led by Ferenc Krausz an ...

New light at the end of the tunnel

Oct 17, 2011

(PhysOrg.com) -- An international team of scientists successfully concentrated the energy of infrared laser pulses using a nano funnel enabling them to generate extreme ultraviolet light pulses, which repeated ...

Restraint improves dielectric performance, lifespan

Oct 25, 2011

Just as a corset improves the appearance of its wearer by keeping everything tightly together, rigidly constraining insulating materials in electrical components can increase their energy density and decrease their rates ...

Flashes of light out of the mirror

Jun 12, 2012

(Phys.org) -- A team of the Laboratory of Attosecond physics at the Max Planck Institute of Quantum Optics developed an alternative way of generating attosecond flashes of light. 

Recommended for you

Awakening the potential of plasma acceleration

21 hours ago

Civil engineering has begun for the new Proton Driven Plasma Wakefield Acceleration Experiment (AWAKE) at CERN. This proof-of-principle experiment will harness the power of wakefields generated by proton ...

Magnetic memories on the right track

21 hours ago

Computer hard drives store data by writing magnetic information onto their surfaces. In the future, magnetic effects may also be used to improve active memory in computers, potentially eliminating the need ...

User comments : 13

Adjust slider to filter visible comments by rank

Display comments: newest first

Eikka
1.4 / 5 (5) Dec 05, 2012
Such a device can work at 1 petahertz, while the processor of today's computer runs slightly faster than at 3 gigahertz.


Wouldn't signal propagation delays become problematic at 1 PHz? At 3 PHz clock rate, it would take approximately 200,000 clock cycles just for the signal to travel 1 cm distance, or the lenght of a typical CPU chip.

If your memory is on the other side of the chip from your ALU, that's a bit of a problem. At 3 GHz your wavelenght is about 10 cm in vacuum, and roughly 5cm in silicon, so all parts see the same signal at roughly the same time.
IronhorseA
2.6 / 5 (5) Dec 05, 2012
Wouldn't signal propagation delays become problematic at 1 PHz? At 3 PHz clock rate, it would take approximately 200,000 clock cycles just for the signal to travel 1 cm distance, or the lenght of a typical CPU chip.

If your memory is on the other side of the chip from your ALU, that's a bit of a problem. At 3 GHz your wavelenght is about 10 cm in vacuum, and roughly 5cm in silicon, so all parts see the same signal at roughly the same time.


Keeping important function units next to each other would be important, but as with modern CPU's, cache memory would help alleviate cross chip delays.
Eikka
2 / 5 (6) Dec 05, 2012
cache memory would help alleviate cross chip delays.


That'd blow up the circuit size badly, though.
Lurker2358
1.7 / 5 (6) Dec 05, 2012
That'd blow up the circuit size badly, though.


Once our Flatlander scientists learn how to build in 3-d that won't be a big problem.
Telekinetic
2.6 / 5 (5) Dec 05, 2012
That'd blow up the circuit size badly, though.


Once our Flatlander scientists learn how to build in 3-d that won't be a big problem.

I thought it was decided that we really do live in Flatland.
DavidW
1.8 / 5 (5) Dec 05, 2012
Gotta love Dr. Quantum
mrlewish
4 / 5 (1) Dec 06, 2012
They've been talking about 3D chips for more then a decade... besides minor things in which it is feasible there is the problem of heat having nowhere to go in transistors buried under a 3D matrix of other transistors. You are back to the same problem. It's not the chip speed speed which is the problem, it's keeping the heat generation down.
Eikka
5 / 5 (2) Dec 09, 2012
You are back to the same problem. It's not the chip speed speed which is the problem, it's keeping the heat generation down.


Well, it is somewhat related, because increasing the voltage at which you drive the transistor makes the transitions sharper and it takes less time for the chains of logic gates to settle down into the correct answer, which means you can run the chip faster - and vice versa.

The energy loss in a CMOS switch is proportional to the square of voltage, and only directly proportional to the switching rate, which means that if you have faster transistors with naturally faster transitions, you can afford to run the chip much faster on the same power budget.
Tachyon8491
2.3 / 5 (3) Dec 10, 2012
The 3-D challenge reminds me of the "hairy smoking golfball" syndrome that was already prophecised in the sixties - the smaller and more compact, the "hairier" with interconnects, and hotter. Dielectric processing does appear to hold some promise though, perhaps coupled with photonics.
El_Nose
1 / 5 (1) Dec 12, 2012
Earlier a post was voted down -- but he was correct-- he just didn't state his position with eloquence.

At a certain point it will not matter how fast the processor is because the signal it generates will be seemingly idle.

Let me explain the issue of signal propagation.

If we assume that all electrical signals on a mobo are traveling at c ... which we know from bus speeds that they are a lot slower than this ... but lets say internal to the CPU everything is buzzing around at c, with new signals equal to clock speed.

at c
299792458000 mm per second (millimeters)
and a clock speed of 3 GHz
( 1 / 3000000000 )
a signal can travel = c * 1 / clock speed
99.9 mm

So todays computers a signal travels basically 99.9 mm in one clock cycle.

now look at 1 femtosecond
( 1 / 1000000000000000 )
the answer becomes
0.0002997 mm
or 299 nm (nano meters )

please check my math -- i have been known to make mistakes

But processor designer will have very real constraints at that speed.
johanfprins
1 / 5 (3) Dec 12, 2012
But processor designer will have very real constraints at that speed.


I do not know enough, but please speculate on what can be achieved by zero-resistance faster than light teleportation of charge-carriers through the gate region. Would the availability of such a substrate change the ball-game; and if so, by how much?
Pununquintium
1 / 5 (2) Dec 14, 2012
But processor designer will have very real constraints at that speed.


I do not know enough, but please speculate on what can be achieved by zero-resistance faster than light teleportation of charge-carriers through the gate region. Would the availability of such a substrate change the ball-game; and if so, by how much?


To my limited understanding of electronics, one just need to consider signal wavelengths for EMI or mechanical purposes other than those, it is better to have a solid physical device holding my transistors than an empty void to pulse them ultraviolets.
johanfprins
1 / 5 (3) Dec 14, 2012
To my limited understanding of electronics, one just need to consider signal wavelengths for EMI or mechanical purposes other than those, it is better to have a solid physical device holding my transistors than an empty void to pulse them ultraviolets.


I am a bit stupid and could not follow the intent of this remark. I have asked about a solid-state device.