REGISTER | LOGIN
Breaking News
News & Analysis

One-Nanometer Transistor Demonstrated

DoE funds carbon nanotube gate
10/13/2016 10:28 AM EDT
11 comments
More Related Links
View Comments: Newest First | Oldest First | Threaded View
Page 1 / 2   >   >>
R_Colin_Johnson
User Rank
Author
Re: I read the paper
R_Colin_Johnson   10/15/2016 2:25:40 PM
NO RATINGS
Thanks for the added info. Yes, their one-nanometer transistor is certainly merely a proof of concept. (They almost sounded like they just wanted to prove it can be done--and get the credit--rather than committ to the needed development work.)

traneus
User Rank
Author
I read the paper
traneus   10/15/2016 12:26:12 PM
NO RATINGS
I subscribe to Science, and that issue came a few days ago, so I read the paper. In their conclusions, the authors agree that their device is far from practical. They first laid down carbon nanotubes in random locations, then photolithographed the source and drain contacts to match where one nanotube happened to be. They then deposited the gate dielectric film, and dropped a flake of MoS2 on top.

The spacing between source and drain contacts is 500 nM, from measuring Fig.2C in the paper. 500 nM is huge compared to the 1 nM diameter of the nanotube gate, so most of the MoS2 channel is not covered by the nanotube gate, and has to conduct in the absence of the nanotube. Thus, this transistor is a depletion-mode device, as reported in Fig.3A: It conducts heavily at zero bias on the nanotube, and cuts off at negative bias (the device is N-channel).

As a depletion-mode device, this transistor is of little use for logic circuits, as it would need the same level shifters that vacuum-tube logic required.

AKH0
User Rank
Author
Re: Reliability in academic research
AKH0   10/15/2016 12:18:53 PM
NO RATINGS
You do not need a large sample size to draw meaningful conclusions. You need a minimum integrity though but more importantly a sense of purpose. The problem we are facing - and I am not talking necessarily about this paper - is that many publications do not advance either technology or science. They do not teach something that can be either used or improves our understanding of the physics of the structure/material being studied. The problem is not that such publications are useless; they are dangerous. They create a generation of "scientists" that do not know why they are working on the subject but worse they fool the researchers and students in lower tier schools as well as funding agencies.

AKH0
User Rank
Author
Re: "gate insulator with low dielectric constant"?
AKH0   10/15/2016 5:32:54 AM
NO RATINGS
Yes, I think everyone here knows it's a high-k. The problem is that once you scale the oxide thickness to get higher drive current, the fringing region will diminish; Leff becomes closer to 1nm and the claimed good electrostatic control vanishes.

kevinhahn
User Rank
Rookie
"gate insulator with low dielectric constant"?
kevinhahn   10/15/2016 5:14:39 AM
NO RATINGS
Actually ZrO2 is a type of high-k dielectric (k=25) and the paper also mentions "...thus demonstrating the need for TOX scaling and high-k(dielectric constant) 2D dielectrics to further enhance the device performance."

"lower dielectric constant" material here is MoS2(~4 vs. bulk Si 11.7). Its "heavier carrier effective mass, larger band gap, and lower in-plane dielectric constant yield lower direct source-to-drain tunneling currents" than Si-based transistors.

 

resistion
User Rank
Author
Reliability in academic research
resistion   10/14/2016 9:48:44 PM
NO RATINGS
Even if the results in this article were not questionable, there is still a serious problem of how to carry out academic research that can be of practical reliable significance for industry today. Sample sizes really need to exceed millions to be free from drawing conclusions on hand-picked best cases to be published. This hardly seems affordable to most groups. Industrial collaboration has too many binding conditions of confidentiality.

Big data availability and internet information availability are also playing a big part. Do we need schools for research anymore?

GroovyGeek
User Rank
Author
Re: Lg_eff?
GroovyGeek   10/14/2016 7:57:20 PM
The problem being solved is one of academics publishing papers to get funding.  I am surprised that this was accepted in Science, it reflects very badly on the rigor of the review process.  Which does not surprise me too much, after seeing what kind of crap makes it through the review process of Nature, for example

http://www.nature.com/nnano/journal/v8/n10/full/nnano.2013.191.html

It is not that this paper had shady data or questionable interpretation of existing data.  It was full of outright fabrications that were immediately evident.  The paper is still archived at various places online.  Just look at the alleged SEMs of 2nm particles, that is all you need to see before all sorts of alarm bells go off.  How this got past three reviewers is a mystery.

I am not implying that there is shady business going on in the Javey paper.  It is a tour de force in the utilization of cheap graduate student labor to obtain a press-worthy but not very technically meaningful result.

AKH0
User Rank
Author
Re: Lg_eff?
AKH0   10/14/2016 7:08:51 PM
NO RATINGS
Yes, it is an underlapped device and they rely on fringing cap to turn on/off the underlap region (Fig 3 D and F). For this to work, you need the gate oxide to be relatively thick so that the fringe region effectively extend over a larger Leff (a few nanometer). That means the gate voltage needs to be high to get any useful current (as showing in their experimental Id-Vg). In other words, even if one solves all integration challenges and place the device at a meaningful gate pitch, reduce parastic resistance, etc etc, it cannot benefit from a scaled gate oxide to support meaningful operating voltage. So what is the problem being solved here?

GroovyGeek
User Rank
Author
Lg_eff?
GroovyGeek   10/14/2016 12:49:52 AM
NO RATINGS
Based on the published TEM I struggle to see how this device can have an electrical gate length of 1nm. Have to read the actual paper, but I am guessing that the 1nm gate length claim is based on physical dimensions. It is quite possible that the speed of light barrier has not been broken :=)

Yasuda
User Rank
Rookie
really?
Yasuda   10/13/2016 10:20:45 PM
NO RATINGS
Based on the TEM image, the CNT gate is far from the MoS2 channel, how can this device work so well? the subthrehold swing is even 65mV/dec (from actual paper). Is the measured data really from this device?

Again from the TEM image, CNT seems in amorphous possibly because of its intermix with ZrO2. 

Page 1 / 2   >   >>
Most Recent Comments
Tim R Johnson
 
ewertz
 
antedeluvian
 
ewertz
 
perl_geek
 
R_Colin_Johnson
 
perl_geek
 
R_Colin_Johnson
 
dt_hayden
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed