HP makes memory from a once-theoretical circuit
It's the tale of the lost circuit.
Thirty-seven years ago, Leon Chua, a professor at the University of California at Berkeley, mathematically theorized that scientific symmetry demands that there should be a fourth fundamental circuit element. Engineers were already familiar with resistors (which resist the flow of electricity), capacitors (which store electricity), and inductors (which resist changes to the flow of electrical current), which can be combined to build more complex devices. The fourth circuit, which Chua called a "memristor" for memory resistor, would register how much current had passed.
"He looked at fundamental circuit equations and noticed there was a hole," said Stan Williams, who heads up the Information and Quantum Systems lab at HP Labs, "There should be a device that remembers how much current flowed through a device."
An atomic force microscope image of a circuit with 17 memristors in a row. The memristor consists of two titanium dioxide layers connected to wires. When a current is applied to one, the resistance of the other changes. That change can be registered as data.
(Credit: J.J. Yang, HP Labs)Williams and other scientists at Hewlett-Packard are publishing a paper in Nature on Wednesday demonstrating that that these things actually exist. HP has a few discrete memristors as well as a silicon chip embedded with memristors. It's a first, according to HP.
If memristors can be commercialized, it could lead to very dense, energy-efficient memory chips. Scientists have made devices that function like memristors, but it took a good number of transistors and several capacitors, Williams said. Memristor chips would function like flash memory and retain data even after a computer is turned off, but require less silicon, consume less energy, and require fewer transistors.
A memristor effectively stores information because the level of its electrical resistance changes when current is applied. A typical resistor provides a stable level of resistance. By contrast, a memristor can have a high level of resistance, which can be interpreted as a computer as a "1" in data terms, and a low level can be interpreted as a "0." Thus, data can be recorded and rewritten by controlling current. In a sense, a memristor is a variable resistor that, through its resistance, reflects its own history, Williams said.
Varying resistance is the same principle at work with phase change memory. The difference in phase change memory, which will come to market later this year, is that changes in resistance are accomplished through a substantial amount of heating. A bit on a CD-like substrate is heated rapidly a few hundred degrees and then cooled. Depending on how rapidly the bit cools, the material becomes crystalline or amorphous. The different states--crystalline and amorphous--exhibit different states of resistance.
"We can get it (resistance changes) with less energy," Williams said. "It is a large amount of resistance change with a small amount of memory."
The secret sauce in HP's memristors is two layers of titanium oxide, a crystalline material consisting of one titanium atom and two oxygens, sandwiched between two metal wires. The bottom layer consists of standard, consistent titanium dioxide. The upper layer is missing a few oxygens--less than 1 percent--which creates voids. When a current is applied (via the wire) to the upper layer, the vacancies are pushed into the lower level of titanium dioxide. That changes the resistance of the lower level. Subsequent bursts of current can then reverse it.
"All we have to do is push around a very small number of vacancies in a crystalline material," Williams said. "We can switch it very fast, faster than we can measure."
Pushing the voids into the consistent layer of titanium dioxide does not change its characteristics otherwise. He likens it to bubbles in beer. "You can have bubbles in it, but it's still beer," he said.
Memristors in green. The wires in this image are 50 nanometers wide, which comes to about 150 atoms.
(Credit: J.J. Yang, HP Labs)Memory and storage are the new frontier for chip designers. The explosion of data will require new ways to retrieve and store it. Cloud computing? It's a big hard drive, if you think about it. Numonyx, the Intel and STMicroelectronics joint venture, is leading the effort to commercialize phase change memory. IBM is working on ways to store data through magnetic charges on a wire. Seagate Technology, Hitachi, Zettacore, Grandis, and others are working on different memory and storage concepts.
HP has largely exited the chip business, but it has increased efforts to license the intellectual property inside its labs. The company, for instance, will likely try to commercialize the crossbar latch technology, which allows molecular grids to perform calculations. (Williams also works on that.)
While memristors can be made on silicon chips, memristor devices will require engineers to learn a new circuit design discipline.
"The technology is in good shape. The big barrier is not whether you can make it," Williams said. "It is the effort to design new circuits."
I don't see any logical grouping that includes resistors, capacitors, inductors, and memristors, but not diodes. From what it sounds like, the memristor acts almost like a diode, only the holes can migrate substantially based on current flow, thus creating the hysteresis. If anything, it's a more complicated device than a diode.
Congratulations! This is great and it is getting back to the HP research of past.
But seriously, fundamental circuits do not require memory for change, because when voltage is applied, they perform their function until the voltage is turned off. If we wanted another function, we would build another circuit. (three position switch)
Having a component that would remember the current to enter a state of on and off... well... it might be useful to indicate a button has been pressed, but there is a time constant that must be taken into account. How useful is it to know what the previous state was without an oscillating clock to count the ticks that have transpired since the last state change?
Now we are in the realm of ICs, sampling a data bus with every oscillation of the circuit, and the memristor is no longer needed.
I guess the challenge is to throw out the IC and arrange the memristors to duplicat the logic. But for what gain?
Since it uses less power, it could also drive the amount of time a laptop has on a battery.
And let's not forget particular testing industries. This could make some testing equipment cheaper and smaller, and even some diagnostic equipment in everything from autos to medical devices. I'm particularly looking at a better controller for pacemakers.
And HOPEFULLY that means both lower prices for customers and higher profits for technology companies, not just the latter.
It means nothing until you use it to build things like diodes, bjts, mosfets, etc.
Now, most resistors, capacitors, and inductors in the microelectronics industry are made of semi-conductors. . . so what is your point?
It is a fundamental circuit element if it cannot be modeled using any of the existing three.
How would this new memristor, compare to that, in the overall scheme of things. Not in function, but rather in the dual, inverse, opposite, mirror, etc. sense of fundumental circuits.
I think it is amazing, as a metrologist (one who specializes in measurement), that they can actually change the state of it faster than they can measure that change, truly innovative!
Never underestimate the brain power behind that company started in the garage!
1. Non-linear component with negative differential resistance region on its VI curve (such as tunnel diode); or
2. A circuit with positive feedback, whose effective output resistance is negative. Such circuit is unstable and will oscillate, unless loaded appropriately.
If you read "memristor" description more closely, you'll see that it doesn't have anything common with "flux plus charge" phrase describing the hypothetical fourth element.
http://www.hp.com/hpinfo/abouthp/histnfacts/
1939, really ...
Have a nice day.
- by neurdon July 8, 2009 8:49 AM PDT
- First the facts: SyNAPSE is a project supported by the Defense Advanced Research Projects Agency (DARPA). DARPA has awarded funds to three prime contractors: HP, HRL and IBM. The Department of Cognitive and Neural Systems at Boston University, from which the Neurdons hail, is a subcontractor to both HP and HRL. The project launched in early 2009 and will wrap up in 2016 or when the prime contractors stop making significant progress, whichever comes first. ?SyNAPSE? is a backronym and stands for Systems of Neuromorphic Adaptive Plastic Scalable Electronics. The stated purpose is to ?investigate innovative approaches that enable revolutionary advances in neuromorphic electronic devices that are scalable to biological levels.?
- Like this Reply to this comment
-
(27 Comments)SyNAPSE is a complex, multi-faceted project, but traces its roots to two fundamental problems. First, traditional algorithms perform poorly in the complex, real-world environments that biological agents thrive. Biological computation, in contrast, is highly distributed and deeply data-intensive. Second, traditional microprocessors are extremely inefficient at executing highly distributed, data-intensive algorithms. SyNAPSE seeks both to advance the state-of-the-art in biological algorithms and to develop a new generation of nanotechnology necessary for the efficient implementation of those algorithms.
The vision for the anticipated DARPA SyNAPSE program is the enabling of electronic neuromorphic machine technology that is scalable to biological levels. Programmable machines are limited not only by their computational capacity, but also by an architecture requiring (human-derived) algorithms to both describe and process information from their environment. In contrast, biological neural systems (e.g., brains) autonomously process information in complex environments by automatically learning relevant and probabilistically stable features and associations. Since real world systems are always many body problems with infinite combinatorial complexity, neuromorphic electronic machines would be preferable in a host of applications?but useful and practical implementations do not yet exist.
SyNAPSE seeks not just to build brain-like chips, but to define a fundamentally distinct form of computational device. These new devices will excel at the kinds of distributed, data-intensive algorithms that complex, real-world environment require. Precisely the kinds of algorithms that suffer immensely at the hands of the Von Neumann bottleneck.