Cheap Metal-Insulator-Metal (MiM) Diode Created

An anonymous reader writes "Progress on metal-insulator-metal diode manufacturing was just reported online in the professional journal Advanced Materials (abstract). For the first time a high-performance 'metal-insulator-metal' diode was created with cheap materials. This is a fundamental discovery. It could change the way manufacturers produce electronic products at high speed, on a huge scale, and at a very low cost, even less than with conventional methods."

Comments

Ground breaking

[Anonymous Coward]

A complete gamechanger, just like memristors!

Re:Ground breaking

[TopSpin]

Whatever happened to memristors?

HP has partnered with Hynix to develop the manufacturing process and commercialize memristor products. Memristors used for storage will eventually appear as ReRAM (resistive RAM.) Meanwhile, other companies are working on memristor designs based on material other than TiO2 as is used by HP.

Would someone with a good grounding in semiconductors please elaborate on why MIM diodes are significant? I have a good handle on basic electronics but not enough experience to deduce how MIM diodes would improve circuit design.

Re:Ground breaking

[kurokame]

Mainly, most immediately, it gives you an additional way to make a diode or diode-based structure when you're designing your fabrication sequence. Fabrication on the foundry / mass-production level occurs through processes which give you pretty much a set sequence of layers (deposited materials, treatments, patterning, etching, etc.). You can make anything you can design within that process...and most anything else usually stays in a research lab.

The extraordinarily common CMOS process involves numerous metal layers "high" above the wafer (numerous layers intervene). These are separated by insulators. Normally, you make diodes at the wafer layer where you're doing your doping.

MiM means you can put diodes in regions of your chip where they couldn't practically be fabricated before without a lot of time doing a one-off chip in a lab. With "a lot" often being several months to a year, assuming everything turns out perfectly, assuming your lab even HAS all the necessary equipment, and assuming you don't have something better to do - which is rare if you're not still a grad student.

Re:Ground breaking

[kurokame]

Silicon is not something we're going to run out of in the foreseeable future. If we do, it would probably be right after we ran out of nitrogen.

Re:Ground breaking

[kurokame]

The insulator is generally treated silicon, e.g. silicon nitride.

Also, metals are something you find pockets of in the Earth's crust. The majority ended up in the core by virtue of its greater density. Silicon, on the other hand, is a key ingredient in the crust itself, and tends to be present in the minerals which you would have to find, extract, and process to get the metals involved in circuit-on-silicon fabrication.

Also, the amount of material in the silicon wafer itself is far, far more than the entirety of all surface features comprising the integrated circuit.

If anything, you would want to be comparing the relative scarcity or value of the metals involved versus the dopants involved, the relative ease of fabrication, and the particulars of what you can fabricate like minimum feature size, chip area per circuit element, and compatibility with other things you want to do on your wafer.

Re:Ground breaking

[John Hasler]

No. It's that MiMs are fast. The best junction diodes run out of steam at a few THz while MiMs work up into optical frequencies and so can be used to rectify sunlight. MiMs have been made before and are used in some exotic lab equipment but those point-contact devices are hard to make and touchy. These guys claim to have produced MiMs using more or less standard planar processes.

Here's a paper that explains MiM theory, though it isn't about this development.

Re:Ugh... yet another paywall stopping innovation

[WarJolt]

Everyone wants information to be free... Until they come up with an idea of their own and publish it.

Re:Dooooood !!

[$RANDOMLUSER]

What about Smoke Emitting Diodes? (or Light Emitting Resistors?)

How does this work?

[Required Snark]

I went and asked Mr. Google how this worked, and I couldn't find any answers. The best clue I got was that these devices use quantum tunneling, but this still does not explain how they exhibit diode behavior. Even the font of all online knowledge, Wikipedia, doesn't seem to know. Someone please post about this.

One thing I did see is that this kind of diode can operate at 100's of THz frequencies, and that this enables nantennas. http://en.wikipedia.org/wiki/Nantenna If these kind of MIM diodes can be made cheaply then a new cost effective class of solar power device may become feasible. So it could be a really big deal.

Re:Cheaper than silicon?

[Tailhook]

the most abundant element in the universe?

The most abundant element in the universe is Hydrogen. Silicon, while plentiful in raw form, must be purified, crystallized, doped, etc. for use in microelectronics. This is an expensive, energy intensive process with less than perfect yield. Copper and aluminum are vastly easier to deal with.

Re:Ugh... yet another paywall stopping innovation

[martin-boundary]

What are you smoking? Sounds yummy.

Just about every scientist who's employed in a university wants to give away their published articles for free to anyone with even a tiny interest. The only ones who like paywalls are publishers.

Re:Ugh... yet another paywall stopping innovation

[mdmkolbe]

As a published academic myself, I concur. I don't get a dime from my published articles so paywalls don't help me. I benefit from people hearing about, reading, being influenced by and eventually citing my work because those things lead to higher academic ratings which lead to better positions, grants, etc.

Horrible horrible public science

[Bender_]

I am a semiconductor scientist, but I completely fail to understand what this news is about. The article does nowhere mention the materials used, the device behavior, the application, the purpose or anything else.
A MIM device as is, is a capacitor. And that is exactly what the picture is showing. When this type of capacitor is scaled to the nanometer regime it starts to get leaky due to quantum mechanical tunneling through the dielectric. The abstract mentions 'controlled quantum mechanical tunneling'... Aha, this could be what it is about. But as long as metal electrodes are involved this will only create a nonlinear resistor. Still no idea what the exact purpose is.

Are nanoscale MIM capacitors new? No, not at all. Right now you have billions of them doing their job in your computers main memory. Depending on the vintage of your computer, these capacitors employ nanolaminates of ZrO2 and Al2O3 at a total thickness of 5 to 10 nanometers. Quantum electrical tunneling is of high relevance in these devices, since it leads to loss of stored information. So, is cheap new? A quick calculation suggests that the manufacturing cost of a single MIM device in a DRAM is approximately 10^(-10) US$.

Re:Horrible horrible public science

[GrepNut]

I agree that the article could be a lot more informative. However, one can actually figure out quite a lot. For instance, since they call it a diode, it means the leaky capacitor you mention is asymmetric, and leaks more easily in one direction than the other. This is born out by the picture, which seems to show two layers of insulator, of different thickness, between the metal layers. And once you have electrical asymmetry, you can start building all sorts of interesting logic. But I agree they do not explain why tunneling through two layers A and B in the order "A then B" should be different from "B then A". Is it because of the direction of applied voltage, or are the insulator materials asymmetric at the molecular level, or is there some deep fact about quantum tunneling that makes it work? I would certainly have liked to see that covered.

Second, in terms of benefits, it seems this device would be far easier, simpler and cheaper to fabricate than a normal PN diode. In a PN diode, you essentially need to arrange a metal wire, a small P-doped semiconductor region, a small N-doped semiconductor region and another metal wire. In addition, the two semiconductor regions need to be insulated from their surroundings somehow. This all requires pretty careful alignment. It looks like the MiM diodes would be self-aligning, in the sense that you could just create a pattern of vertical metal wires on one layer, then overlay the two layers of insulator, followed by a layer of horizontal metal wires. The diodes would form at the points where the wires cross, without any precise alignment being needed. And the fact that the middle layers are insulators would mean no further insulation was necessary. One could probably fabricate giant sheets of these things very cheaply.

Finally, the fact that they can use these for rectifying infrared radiation implies they can operate many orders of magnitude faster than normal CMOS diodes.

Googling a little also hints that MiMs are better at extracting the full energy from incident photons in photovoltaic applications, which could be a useful side benefit, allowing one to efficiently convert optical or infrared radiation into DC current with a single type of device. But I'm just guessing, an expert would need to confirm that.

In summary, I agree this article could be a lot better. However, I have seen a lot worse, and it does seem to be alerting us to something which could turn out to be important.