A challenge facing designers of future computer chips
November 8, 2012
The total conductance per unit area is similar for both tungsten (W) and gold (Au). However, by joining the two highly conducting metals, one finds a conductance density that is about 4 times lower of either material individually. (Credit: David J. Olivera et al./PNAS)
To build the computer chips of the future, designers will need to understand how an electrical charge behaves when it is confined to metal wires only a few atom-widths in diameter.
Researchers at at McGill University General Motors R&D, have shown that electrical current could be drastically reduced when wires from two dissimilar metals meet. The surprisingly sharp reduction in current reveals a significant challenge that could shape material choices and device design in the emerging field of nanoelectronics.
As feature sizes in future chips shrink to the level of atoms, the resistance to current no longer increases at a consistent rate as devices shrink; instead the resistance “jumps around,” displaying the counterintuitive effects of quantum mechanics, says McGill Physics professor Peter Grütter.
“You could use the analogy of a water hose,” Grütter explains. “If you keep the water pressure constant, less water comes out as you reduce the diameter of the hose. But if you were to shrink the hose to the size of a straw just two or three atoms in diameter, the outflow would no longer decline at a rate proportional to the hose cross-sectional area; it would vary in a quantized (‘jumpy’) way.”
This “quantum weirdness” is exactly what the McGill and General Motors researchers observed. The researchers investigated an ultra-small contact between gold and tungsten, two metals currently used in combination in computer chips to connect different functional components of a device.
On the experimental side of the research, Prof. Grütter’s lab used advanced microscopy techniques to image a tungsten probe and gold surface with atomic precision, and to bring them together mechanically in a precisely-controlled manner. The electrical current through the resulting contact was much lower than expected. Mechanical modeling of the atomic structure of this contact was done in collaboration with Yue Qi, a research scientist with the General Motors R&D Center in Warren, MI.
State-of-the-art electrical modeling by Jesse Maassen in professor Hong Guo’s McGill Physics research group confirmed this result, showing that dissimilarities in electronic structure between the two metals leads to a fourfold decrease in current flow, even for a perfect interface. The researchers additionally found that crystal defects — displacements of the normally perfect arrangement of atoms — generated by bringing the two materials into mechanical contact was a further reason for the observed reduction of the current.
“The size of that drop is far greater than most experts would expect — on the order of 10 times greater,” notes Prof. Grütter.
The results point to a need for future research into ways to surmount this challenge, possibly through choice of materials or other processing techniques. “The first step toward finding a solution is being aware of the problem,” Grütter notes. “This is the first time that it has been demonstrated that this is a major problem” for nanoelectronic systems.”
Funding for this research was provided by the Natural Sciences and Engineering Research Council of Canada, le Fonds Québécois de la Recherche sur la Nature et les Technologies, and the Canadian Institute for Advanced Research.
Comments (22)
by Tal
Either a different crystal lattice or a new alloy, a new item on the agenda of nano labs.
very interesting, thanks!
by marty weiss
If mass bends gravity, the weak force, and electromagnetism, the strong force, is reduced to quantum linearity by physical constraints, try employing inharmonic modulation to wield and ameliorate the weak force in computational function. I think gravity is the next frontier. Defining it seems to be the threshold of true comprehension and dominion in physical position. All the other means of positional adjustment are skating on thin ice. Gravitic computation would employ an universal element. No more than the Pope’s boxes fit Galileo’s universe do the old conventions of our time fit an expanding sphere of action.
by Ralph Dratman
I am wondering if this would also affect SiO2-to-metal contacts.
by Marcos Marin
Yes. The exact same problem a solar cell have when transferring current to your appliances. =) Although SiO2 solar cells may not be in use so much anymore…
by GAUSS
This is all assuming we’ll be using electricity as our medium for transfer of information.
by Chaz
I see nano gold. ;P
by mary
Have’nt read it yet. Want to train my mind but can’t make an account!
by Marcos Marin
simply dope the connectors! ugh!
by manicmoose
Doping generally refers to the process of implanting impurities into semiconductors in order to introduce excess charge carriers. In this case, the materials they are using are metals. Metals have excess free charge carriers, hence they are conductors and doping isn’t applicable. Perhaps there is something along those lines that could be done, but it isn’t clear what it would be.
by Marcos Marin
For lack of a better term for metals.. not that “doping” seems very formal anyway:-)
Yes, “something along those lines” (much easier to simply call it doping :-P), I was thinking of a degradé of donors and acceptors for your charge carriers. It WOULD need doubling the wire(diode effect), BUT more than pay for it by going over the aforementioned “limit”.
Well, soon I’ll have a STM microscope(or something of the sort) and will be able to redirect my complaints about the lack of vision of others to myself =)
by Marcos Marin
Also, how much “metalness” is left after you do “something along those lines” =) Eventually might look very much like a semiconductor, warranting the term perfectly, no?
by manicmoose
I personally (clearly) don’t know enough about this problem to know anything for sure. My guess though is that more free charges wouldn’t be enough to solve the problems they are facing. I don’t believe that a lack of electrons is the problem. Perhaps it has something to do with the lattice structures of different metals being dissimilar. When electrons traveling along the metal meet the barrier, they may be sorta ‘colliding” and being forced around a new path since the atomic structure is changing. This would expend some electric potential and too the observer look like a high resistance.
In large wires, it is not an issue since there is just so much space, but when the wires shrink to this size, maybe there just isn’t as much room for new paths to be found. Who knows, perhaps the electron velocity is being saturated in a similar way that happens to short channel MOSFETS, capping the current at a lower than anticipated value.
by Editor
According to the PNAS paper, the conduction of electrons across the intermetallic interface is drastically reduced because of the fundamental mismatch between electron waves in gold vs. tungsten and the presence of defects generated by plastic deformation in the gold. No solution was proposed in the paper.
by manicmoose
Awesome. Well, not awesome, but thanks! haha =)
by Editor
Take that, Moore! :)
by Mr.X
I read Mr.Moore doesn’t believe anymore in his law.
by Dagwood
Not a doping problem, just geometry. They need to try tilting the lattice of the tungsten wire a few degrees (estimate 4 to 15 for starting points). If they can’t get any improvement from that then I would try an intermediate metal like Iridium.
by Editor
According to the PNAS paper, the lattices are already matched. Adding an intermediate metal would significantly decrease the conductivity due to the additional junction and path resistance, especially if using iridium, with a conductivity of 1.891E+07 Siemens/m vs. 4.257E+07 for Au. Doping would of course hugely decrease the conductivity.
As I noted below, the conduction of electrons across the intermetallic interface is drastically reduced because of the fundamental mismatch between electron waves in gold vs. tungsten and the presence of defects generated by plastic deformation in the gold, according the PNAS paper.
The paper further explains that the reduction in conductance through the W–Au junction is affected by the intermetallic interface, defects in the Au, and the orbital mismatch between W and Au, which increases the probability of backscattering at the interface. A larger distance between Au atoms reduces electronic hopping and leads to additional resistance and defect configurations across a set of indents, with the resistance varying according to the electron conduction pathway.
by Dagwood
The article has a very simplified picture of the lattice interface so it is difficult to picture their “optimal” connection. How do you match the lattices when they are so different? Experimenting with the lattice orientation should be easy to do and either the measurements show improvement or they do not. Prove me wrong with experimental results.
I would only use a barrier metal as last resort due to added manufacturing complexity. That said, you compared Iridium conductance with Gold when you should have compared it to Tungsten. Their article also refers to the large orbital mismatch between Au and W, this should be less with Iridium, causing less backscattering.
by Editor
This is beyond the scope of KurzweilAI posts (and my limited knowledge). Best to discuss offline within the context of the PNAS paper. Sounds like you are more knowledgeable on the subject and I welcome your thoughts.
by Editor
Should compare to Tungsten: right; actually, it should be compared to both. Iridium conductivity is 1.821E+07, about the same as Tungsten, so the conductivity loss would be essentially 2x for the added segment, plus the additional loss from the added junction, so not an optimal solution :) — unless I am misunderstanding what “barrier metal” means.
by manicmoose
Just for clarity, you can’t dope metal with the same types of materials that you dope semiconductors with and expect similar effects – maybe not any materials (especially since metals already contain an abundance of charge carriers) The reason it works in silicon is because the silicon naturally forms 4 covalent bonds with surrounding silicon atoms. When those atoms are replaced with a group 3 or 5 element, either a hole (positive carrier) is left from a massing covalent bond or an extra electron (negative carrier) is present. So you can’t convert a metal into a semiconductor by doping, as ‘semiconductor’ is an intrinsic property of certain elements, and metallic bonds work differently. Also, even very highly doped silicon is still vastly mostly silicon. The dopants are only a tiny percentage of the mass.
Cheers!