Hacker News new | past | comments | ask | show | jobs | submit login
Memristors built with 2-nanometer-thick parts (arstechnica.com)
81 points by TanakaTarou on Nov 17, 2018 | hide | past | favorite | 20 comments



To address the memristor existence questions: if you define a memristor as a device which "remembers" its resistance via a pinched hysteresis I-V curve, then we have had ReRAM/ memristor devices for some time now. On the other hand, if you define memristor as a device with coupling between charge and flux, then no.

Practically speaking, it's unlikely that memristors will replace transistors because they're about 500X too slow and have about a billion times less endurance. Same issue with memory replacement. They could in theory make excellent flash memory replacement, but currently sneak paths, thermal crosstalk and ion crosstalk remain challenges


Are memristors real? This paper[0] in addition to a handful that have appeared on HN in the last few months asserts that "the 2008 memristor is not the 1971 postulate and neither of them is fundamental" and moreover that "[t]he ideal memristor is an unphysical active device and any physically realizable memristor is a nonlinear composition of resistors with active hysteresis."

I'm confused by something like this being an open question, and even more confused about why this debate isn't lighting up the electrical/computer engineering community as far as I can tell.

[0] https://www.nature.com/articles/s41598-018-29394-7


I asked this question several years ago[0], and had some very interesting responses. The long and short of it was that yes, they are real, and you can make them yourself. The basic concept being that different metal oxides have different resistances; so if you make a "sandwich" like metal_1-oxide-metal_2 you can use current to move the oxygen ions from one metal to the other. The resistance of the sandwich changes as you do so, and stays set in between applications of charge, as well as staying set if the probing current is not too high. I recall making one was a fun afternoon project.

[0]https://electronics.stackexchange.com/questions/76097/what-a...

[1]http://sparkbangbuzz.com/memristor/memristor.htm


If you don't like the PR, just call it ReRAM. Few people actually care about HP's grandiose claims, a lot of people care about the actual performance of the devices.


It’s also worth pointing out that all ideal circuit devices are non physical.

But yes I think the name “memristor” is mostly PR.


> It’s also worth pointing out that all ideal circuit devices are non physical

This is not a useful statement - in this framing, all of circuit theory is unphysical.

Other ideal circuit elements could exist in the sense that their existence would not violate known physical laws. The criticisms of the memristor is that its existence would violate fundamental thermodynamic principles (esp. Landauer's principle).


I mean, an ideal voltage source also cannot exist, and I'm sure their existence would violate several laws of physics. For example what happens when you put two ideal voltage sources in parallel? or when you connect one to a short circuit?

Despite that though we can use a combination of ideal elements to represent real, non ideal, physical devices. So in that sense I disagree with your statement that "in this framing, all of circuit theory is unphysical."


I don't have a clue in general, but I thought hysteresis was memory by definition.


It's 95% marketing and 5% engineering. There's not much to get excited about. (yet?)


Can’t wait to have some of these memristors in my phone, powered by graphene batteries, running on clean electricity from the regional fusion plant.


To be faire, fusion plants lag on because either they are designed by people without enough resources, or by complex red-tape inducing collaboration.

ITER hoped to be done in 10 years starting in 2006, but they could only start in 2010 because every country wanted it to be built somewhere different.

Then, poor cooperation from Domestic Agencies caused frequent changes that were merged without hierarchical validation, causing the need for a full-time team whose only job was to detect inconsistencies post-hoc. Therefore, the planned completion in 2020 was pushed to 2025.

As a result, certain countries like the US made themselves begged to keep their promised investment.

See this analysis: https://www.newyorker.com/magazine/2014/03/03/a-star-in-a-bo...

I don't see how those reasons would apply to memristor, though: development is controlled by a single company, and that company has resources.


And the year after that will be the year of the Linux desktop.


I understand the joke, but I've been running Linux on my desktop for something like 15 years. ;-)


... yes because that will be after the apocalypse and what remained of working pre-war technology were ancient "desktop computers". Big hulking things you couldn't put in your pocket, much less implant and connect to your brain. But they worked, and some source code was available. It was called "Linux".


> regional fusion plant

You mean the sun, right?


People don't realize that solar is not a "fair trade" energy, that celestial body is being ruined to supply our egotistical needs ...


The future sounds beautiful.


And only about 30 years away for as long as I can remember.


So, 11 atoms thick layer of platinum is major cost consideration? No, just bad journalism.


I can just picture William Devane telling everyone to buy Platinum now...<g>




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: