Slashdot Mirror
Single Molecule Memory
techtrend linked us up to a paper from Mark Reed and James Tour on
single molecule memory which, if it comes about will pretty much make space irrelevant. They say the technology is 3-5 years off.
← Back to Stories (view on slashdot.org)
I'm unsure of the *quality* of manufactured diamonds, but we do have the ability to make 'sheet' diamond atop metal filaments.
It's essentially, under high pressure, using a tungsten heating element run under some methane gasses in a non-combustible environment(nitrogen and stuff), and diamond will start to grow upon the surface.
I may have a few technical details wrong, but I think that's the process.
-AS
-AS
*Pikachu*
The problem is that most of the really exciting research results don't work above liquid nitrogen temperatures, and some don't even work above liquid HELIUM (4K) temperatures!
But I actually saw the quantum dot working, and helped perform some of the analysis of it (on some good old VAX hardware!) I also helped construct a custom I-V trace unit which used a wiggle voltage to produce better curve traces of the results. Some of these novel quantum semiconductor devices (see, for instance, the I-V trace of this one are actually capable of operating in more than just one single state -- the multiple plateaus in the 9 T graph show that this device can operate as a trinary logic device if you know what you're doing. Then again, it requires a 9 Tesla field to bring out these characteristics...
As I've said before on /., we need to solve the temperature and interconnect issues. Interconnect may have a new solution, per that article on molecular computing posted a few days ago here. Our materials science friends, though, need to keep making progress on materials which possess these unique characteristics at room temperature.
"But always she's the spectre of uncertainty I first endured, then faded, then embraced..."
I agree that "speculative science" usually ends up yesterday's "science fiction" and last week's "mad ramblings". On the other hand, no speculation, no progress. Without trying to reach forward, people have a habit of sliding backwards.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Actually, it's "peta" which is then followed by "exa" and then "zetta" and finally "yotta". There is a short breakdown of the meanin of the names at "http://www.ccsf.caltech.edu/~roy /dataquan/ety.html" which tells where the names come from. There is a much larger listing of magnitudes at "http://www.mcs.csuhaywa rd.edu/~malek/Mathlinks/Billion.html".
World Beach List, my latest project.
for anybody.
Can your IM do this?
But nothing in the article says that anything in this particular paper will be implimented any time soon.
this would be cool, except:
;-)
we still dont have a decent way to transfer information through 1 molecule sized pathways...
(keep in mind, they have flipped the gate, and watched with a microscope... they did not do anything useful with it)
the gate/transistor can be that small, but if the path to get there is not, who cares
(electricity can't work well at that size, if the pathways are that small and at all decently near each other you will get massive electron tunneling, where they hop over to the next pathway ) (this is bad
optical pathways ahve not been gotten to work yet AFAIK, and even they would have problems at that level
on a more holistic level, fusion was supposed to be done 20 years ago, those incredibly large harddrives that are the size of my pinky were supposed to be done by now....
this is cool and all, but it is research that will not bear fruit for a LOOONG time
-RS
We are all in the gutter, but some of us are looking at the stars --Oscar Wilde
Grrr. my nick is "Forward the Light Brigade"...
Electron spins would be a decent place to store bits...
The ultimate reeally should be not storing bits. Seriously, whatever happened to fuzzy logic. I know that all conventional logic still holds true under the fuzzy rule sets so all software could be emulated.
For example::
-Analog music sounds better than sampled music
-People don't really use 1 bit logic in everyday life.
-fuzzy machines perform much better in real world tasks than traditional logic
Where are the attempts at hardware fuzzy logic? I know all of the obsticles in voltage regulation are staggering but.... you would think that there would be more research. The main reason we use the binary number system is to emulate a switch. I would like to see how fast a variable switch processor would be.
-Pos
The truth is more important than the facts.
The truth is more important than the facts.
-Frank Lloyd Wright
1) How fast can a computation happen (in a physical system) in theory?
A: The answer depends on the physical system. There is no hard theoretical limit on computation speed (only unreliable estimates based on current and developing technologies).
2) How fast could molecular gates and molecular bits effect a computation?
A: The answer depends on the way these gates work. Who knows? Maybe they will use quantum tunneling to have a gate delay shorter than the time it would take light to pass through the space the gate occupies.
3) How do you estimate these numbers would translate into Teraflops?
A: Teraflops are meaningless out of context, and they are highly dependent on design. There is no limit to the teraflop rating of something made with current technology if you allow SMP.
Sorry, but the questions are just not answerable.
If you had a super-computer the size of a mitochondria, where would you keep it?
"Nobody move, I dropped my Cray!"
Even if memory could be made on the molecular level, and processors could flip electrons instead of bits, do you all think we could afford the Scanning-Tunneling microscopes we'd need for I/O? I mean, hell! I like having a big monitor. There's no way to plug it into a sugar-cube computer..
Every time you sneeze, you'd have to get new hardware.
-- What you do today will cost you a day of your life.
You can see what the Tour research group is up to at Rice by going to his homepage at http://www.jmtour.com/. There is information about this project at http://www.jmtour.com/info.htm. Scroll down the page a bit.
Finally, don't forget that you can see more about the Rice nanotechnology program at The Rice Quantum Institute and The Center for Nanoscale Science and Technology. Don't forget that Rice is where the Buckyball craze started, with Smalley and Curl winning the Nobel for the discovery of its shape.
I have no numbers, but I think there's a section in Applied Cryptography that discusses some theoretical maximums of computing power. My copy's at home - anyone have theirs handy?
Tom Swiss | the infamous tms | my blog
You cannot wash away blood with blood
Accually it takes approximatly 2 petrabytes from what I've read to store all sensery inputs and all other memory (the things that we thought but didn't nessesarly have inputted) for a human for an average lifespan. 67
assuming 1 bit per molecule 6.02e23/2^50/8=
about 67 million
so you could store 67 million lifespans. APPROXIMATLY
in a mole...... shrugs.. thats not that much.. really windows 2050 will take atleast that. (even then.. imagine the load time.. though I guess you would buy memory preloaded with windows just never let it loose its power... assuming of course it works like that..)
I saw many Science articles on molecular memory and computers in 1993. Seems whenever grant money for this thing dries up, people start calling portals and TV stations saying they're on the verge of a magor breakthrough. Why are they only on the verge of major breakthroughs when their grants run out?
This made it sufficiently into public consciousness that Michael Crichton's book, Congo, had this as the "plot point" behind the search for "uniquely pure" diamonds.
19 years have now passed since the book was written; computers are not yet based on diamonds.
Thinking back only to 1998, IBM announced that PPC chips that used a copper-based production technique would provide massive performance increases; it is not clear that this is yet being deployed in present PPC systems.
I suspect that "single molecule" memory elements are more than 3-5 years away.
If you're not part of the solution, you're part of the precipitate.
My roommate had this mentioned in his electrical computing class at Rice today.
n o.html
The prof printed out copies for everyone.
You can find YOUR copy at:
http://www.nyt imes.com/library/tech/99/11/biztech/articles/01na
Usual free registration/login for nytimes.
Mark Covington
It is not just a single bit that can be in a superposition of states, but all the bits of the computation. (A superposition of states can be described as probability distribution over all the possible states the system could be in). Thus, the limit on the number of parallel computations in a binary quantum computer is 2^N, where N is the number of quantum bits (qubits) used in the computation.
The degree of parallelism this implies is staggering. Some problems that are believed to require exponential time to solve on a classical computer could be solved in polynomial time on a quantum computer. This includes factoring (think RSA encryption).
Some people overgeneralize and think that a quantum computer could solve NP-complete problems in polynomial time. Unfortunately, that's not the case (or at least, hasn't been proven). To get an answer out of a quantum computer, you need to be able to get all of the exponentially many wrong solutions to somehow "cancel out," leaving the correct solution. Doing that in the general case is non-trivial and probably impossible. Quantum compilers are a long way off.
But so far, quantum computers have proven difficult to build. The problem is getting a useful computation without a "collapse" of the wave function. (A collapse is when the system rolls the dice or whatever it does, and settles on a single state to be in. Oops, there goes your parallelism!) The biggest quantum computer I've heard about has 2 qubits. An impressive achievement, but not quite ready to port Linux to.
It just won't be cool anymore by then.
Interesting idea that you can control a single molecule. But can it be done fast? is it expensive? If not it won't be any good for hardrives or internal memory. The article doesn't go into detail on these issues.
Will it beat holographic storage and other interesting techniques?
Interesting times are ahead of us. I could have said that anytime during the last 100 years or so and I hope I can keep saying it for quite some time.
Jilles
Now, for other purposes than individual users, this has interesting implications in terms of computational capabilities, especially if this memory can become quite fast.
--Carl
...if these guys do what they say they will.
You've assumed that molecules are the lowest level at which we can compute;
We can go smaller, into atoms and energy and spin states of electrons in a shell, for example, both of which are different things entirely. So we haven't quite hit the limits of information and computing yet.
So lets say we use a stable lithium ion as a storage 'bit' where we can flip the electron's spin to indicate 1 or 0. We ignore the two inner s orbital electrons and concentrate on the single valence electron. You'd prolly flip it with a single photon of light. How? Beats me. Anyway, you can actually calculate the energy of the photon required to do so, and the time it takes to flip as well( ~instantaneous?) and that is some sort of limit, but there are still levels beyond that with which we could play information games, I'm sure.
You could go into multi-bit storage by including energy level as well as spin; bump up an electron 1, 2, 3 or 4 levels, and flip it's spin in either direction and we get a 3 bit storage out of a single atom. If you play with two electrons in such a system, you could conceivably get a 5 bit system, or something like that. With a complex enough atom, you could prolly get 6 or 7 bits of data off a single atom!
-AS
-AS
*Pikachu*
I may be mistaken, but I thought the iBook used a Cu process G3, and the G4 was also a copper process?
Both are already in the market, with more on the way with future G3 PowerBooks, and quite possibly even SOI and Cu based G3s and G4s.
Computers are not yet based on diamonds because they don't provide any performance improvement over silicon, over the past 19 years. They are definitely part of the research on optical computing, but silicon, in theory, still has another 8 to 10 years of life still at which point an alternative technology may take over. Like optical.
-AS
-AS
*Pikachu*
I am actually somewhat an expert in this subject as I have been doing research on molecular electronics for about two years ago. I have been to most of the conferences so far on the subject. (All this reseach is being funded by DARPA) Anyway, the significance of this research is that it involves passing electrical current through molecules, not just a two-state system created by structural conformations. Although I do not have the specific details on this recent experiment, I know that the past work of Reed has involved synthesizing a molecular structure and then testing its electrical properties (I-V, C, etc.) by using a scanning tunneling microscope (STM) tip to apply varying voltages across the molecule and then measuring the results.
My guess is that they have fabricated a molecule with a high capacitance that can store charge in a similar fashion as conventional DRAM. However, such a molecule cannot be used in a memory array until a switch (molecular-sized transistor) can be fabricated. That should come soon. However, the above is only speculation on my part.
If you want some good introductory information on molecular electronics in general including both memory, switches, and higher level logic architecures (AND, OR, XOR, etc.) in molecules, download this paper:
"Architecture s for Molecular Electronic Computers" by James Ellenbogen and J. Christopher Love
The research in that paper was performed at the MITRE Corporation, which is also in the process of developing molecular electronic architectures. I contributed a large portion of the computational data to the above paper.
As someone has already pointed out, they're still looking for a molecule that will act as a switch, never mind implenting it in any practical way.
Even if we end up with individual memory cells 1 molecule big, we still need to design circuits on about the same scale, not much point having some ultra dense memory array when you don't have an efficient way of connecting to it...
I'm guessing it'll ve using some really low voltages, so shielding out interferance would be tricky.
Kill'em! Kill'em all!
What is needed is an infrastructure that goes on top of this .... how do you put the molecules into the states you want, how do you sense their state, how do you get that information to the outside world.
And once you get it out what do you do with it? put it into a molecular computer? .... then probably you're using the same technologies you fabbed the ram with .... put it into a 'traditional' silicon computer? then you need a whole other bunch of technologies/infrastructure that allows these molecular structures to be fabbed alongside silicon ones ....
In other words there's a lot of work to be done!
It'll happen one day .... probably not next week, or next year
Oh yeah and noise/cosmic rays/quantum effects etc etc you have to be able to handle all those other things that can cause these molecules to change state when you don't want them to - with ram you can do heavy ECC and scrubbing to get reliability .... for random logic in, for example, a CPU it's a whole different, and much harder, problem
I read the article and came up with more questions than answers... How does it work? What are the 'off' and 'on' states? How do you read/write it? How fast can you cycle it?
I followed the link from the article to 'Mark A. Reed', one of the scientists mentioned. A quote from his personal site (deep breath): "My areas of research are quantum electron device physics; tunneling and transport phenomena in semiconductor heterojunction and nanostructured systems; reduced dimensionality effects in nanostructures; resonant tunneling transistors, circuits, and novel heterojunction devices; investigations into the physics and technology of quantum-confined electronic devices; investigation of resonant tunneling physics in a variety of heterojunction systems and materials, including 0D quantum dots and resonant tunneling transistors; and molecular electronics, nanotechnology."
...wheeze...
OK, now I know as much as I did before, and am buzz-worded to death besides. So I drilled deeper into the site and found some pictures of his current work that do give some clues. Most interesting is the illustration titled "Molecules in nanopores."
And, of course, there is his List of Publications which I probably wouldn't understand anyway. Even if they were online... Perhaps someone more competent can read these, and peruse the 'Break Junction Lab' description for us.
My take at this point is: the guy probably knows what he is talking about, but I still don't have enough information to determine if the end result would work well enough to actually be useful in '3 to 5 years'. The thing is, there are plenty of technologies that work. But only a few of them have survived the true test of fitness in the marketplace.
Jack
- -
Are you an SF Fan? Are you a Tru-Fan?
Is that a SingMolec Module in your pocket containing all Human Knowledge or are you happy to see me?
It will certainly need more than one molecule to store one bit since the flipping the bit due to radiation (cosmic or otherwise) would probably be pretty easy with single molecule memory. It's a real problem with memory today. So, they would probably need triple redundancy - or three bits.
Then there's the problem of wiring up this memory, addressing it, and there's no discussion performance versus current memories.
Also, any speculation of having working implementations in the next few years needs to do a reality check versus the real time it takes to investigate new technology.
I know it may not happen for a long time, but imagine this:
You walk into your local PC parts store.
"I want 96 petabytes of memory, please."
"That's it? That will be $6.28."
It seems that as we start thinking of molecules as bits, we're getting down near the theoretical limits of information and computing. So that brings me to a few questions that I hope some knowledgable person (someone with a background in physics and computation) might be able to answer:
1) How fast can a computation happen (in a physical system) in thoeory.
2) How fast could molecular gates and molecular bits effect a computation?
3) How do you estimate these numbers would translate into Teraflops?
Quite simply, you use the spin of a single electron to determine whether you have a 0 ('down' spin) or 1 ('up' spin) for that given atom. These are read with lasers, and I believe this can be done rather quickly.
If this is the same thing, then the theory has existed for maybe three years, but they seem to have found a practical application for it. Before that, all they could do was use some sort of awkward prototype filled with lens for interferometry.
If this is indeed the same thing, it also leads to a spiffy thing: fuzzy logic. Since quantum mechanics is essentially a matter of statistics, it means an electron may be in a probabilistic state between 0 and 1. For instance, it could be:
How this can lead to more efficient calculations, I have no clue. Still, it's cool to think of a single bit as "maybe 0 but most probably 1".
Again, not sure if this is the same technology. It may just not be; but regardless, the idea remains a really cool one.
"Knowledge = Power = Energy = Mass"