New Microchip Breakthrough: Scaling Beyond 1nm

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
Researchers have developed a new kind of  transistor that hardness Quantum effects   and this is very interesting because they  were able to shrink this device down to the   size of a single molecule which is about 1 to 2  nanometers and it appears to be a possible way   to scale beyond 1nm technology in the Physical  Realm not just in the performance metrics. As a   chip designer I'm beyond excited you know because  the future of transistor is our future. Nowadays   microchips are simply at the heart of every single  electronic device you can imagine and each of them   is built of transistors the tiny devices that  are used to build everything from simple logic   gates to CPUs and GPUs for the last 40 years  we've been shrinking the size of transistors   to be able to fit more computing power into  a smaller chip right now we are at about 200   million transistors per square mimer of silicon  and we are at about 200 billion transistors per   GPU at the same time we already have AI chips  built of 4 trillion transistors can you imagine   that that's an impressive number however lately  it was getting even harder and more expensive to   pack more transistors into a silicon die. and  there are several limiting factors here one   of them them is manufacturing process we need to  engineer more advanced lithography tools as well   as improve other manufacturing steps but I'm  certain we can solve it however as transistors   are getting smaller and smaller another huge  problem pops up we bump into the effects of   quantum mechanics and there are actually many  challenges that pop up in this intersection   between classical electronics and quantum physics  in dealing with transistors under 7-5nm in size   Quantum effects start to become apparent which  cause unusual and sometimes unexpected changes   in how electronic devices behave and it's already  an issue for the devices under 10 nanometers but   this problem becomes huge when we approach 2 to  3 nanometers one of the main effect that appears   is quantum tunneling this is a phenomenon when  electrons can pass through barriers which would   usually block their flow in classical physics.  here is what a classical planner transistor   looks like and when we try to scale it down this  means in practice the scaling of the gate length   as the gate length is reduced and the gate oxide  layer becomes thinner electrons can tunnel through   the barrier even when the transistor is switched  off and that's a big problem because practically   it means you can't switch it off completely it  still leaks out you you know it's a so-called   leakage current which basically wastes energy  and degrading the performance and do you know   why this is happening this is happening because of  the Dual nature of electrons because they exhibit   characteristics of both particles and waves at the  same time and for the gate length under 1nm their   wave nature becomes prominent and here we step  into the realm of quantum physics and this part   is super interesting and we will dive deeper into  that later on in the video but this practically   means that we have this leakage current and we are  losing power even when the transistor is switched   off but this story is getting worse and worse as  we continue shrinking transistors because it's an   inevitable effect of miniaturisation and here you  might think well Anastasia what are you talking   about we already have transistors of 3 nm for  example from TSMC and Intel already has a road   map towards 20 Angstrom transistors by the way  an angstrom is 1/10 of a nanometer which means   it's about the size of a single atom you might be  wondering how is it possible considering all the   challenges which we just discussed and you would  be right when it comes to nanometers it's really   not that simple today in the today's video we are  talking about literally shrinking the transistor   so shrinking the classical planner transistor you  know this Quantum-based transistor we discussed it   really has a the channel of the gate length below  one nanometer but in general to be honest this   nanometer concept nanometers haven't meant much  for quite a long time already you know originally   we had this classical planner transistor where  we have a drain a source and the gate and all   arranged in a single two-dimensional plane and  when we apply a certain voltage to the gate the   channel becomes conductive and current flows  through the channel and if we take a planner   3nm transistor its channel length will be around  32 nanometers but as a planner transistor were   scaled down we had many problems with it and this  excessive leakage it just one of them for example   instead of being on and off it was bright and  dim and eventually the solution was to change   the transistor structure completely to move from  the planner transistor structure to a more complex   3-dimensional structure and as a result of this  the first FinFET 22 nanometer transistors were   developed by Intel and they called it FinFET  because the transistors aource drain structure   were stucking out of the wafer like a shark's  fin ever since then the nm term has increasingly   lost its meaning now in a 7nm FinFET transistor  you won't find anything that is of actually 7   nanometers it's actually not 7 nanometers across  that's just a marketing term that indicates that   you would need a planner transistor that small to  achieve the same performance and logic density as   this 7nm FinFET device but now what happens when  we shrink the actual transistor gate to 7nm 5nm   2 or 1 nm, so that a single molecule can fit in  there well so far it didn't work and there were   a lot of researchers working on mitigating this  Quantum effects but just think for a moment what   if instead of fighting it we can use these Quantum  effects to our advantage and this is exactly the   idea behind this new transistor the new transistor  is built from two pieces of graphene connected by   a single molecule and it uses Quantum effects  to switch between two states on and off and the   most interesting part the fact that they utilize  here to switch it on and off is called Quantum   inference before I explain how this new transistor  works and where I see the future of this   technology did you know that every time you go for  a coffee or travel and connect to a public Wi-Fi   network your personal and your banking data are at  risk it's a public access with no security control   so everyone nearby can access it and monitor what  you're doing that's where surfshark VPN comes in   handy it encrypts your personal data and keeps you  safe on public Wi-Fi networks another cool feature   of surfshark VPN is that you can use it to access  content that may be restricted in your location   by swapping the IP address for example when you  travel you can still connect to your home server   and enjoy your streaming collection if you want  the freedom to access content from anywhere in   the world I highly recommend trying surfshark  VPN and for you they're currently offering an   exclusive deal go to surfshark.deals/anastasi to  get three free extra month to the subscription   using the code anastasi check out the link in  the description below and thank you Surfshark for   sponsoring this video you know as we discussed in  a very small scale electrons exhibit both particle   and wave properties and when an electron travels  through a very small 10 nanometers channel its   wave nature dominates and when it acts as a wave  there is always inference as for what inference is   you know waves can interact and also overlap if I  try to simplify it to the most basic level imagine   you have two sinusoids and you add them together  when you add them and both are in phase the   result is twice the amplitude this is so-called  constructive inference for example imagine you   have a friend who is very motivated and brilliant  and he or she is an expert in a bit different   domain than you are but together as a team your  strength are multiplied and that's beautiful on   the other hand if you add two sinusoids that are  180° phase shifted they will cancel each other   out and this is so-called destructive inference  what's interesting you don't often hear about   inference when we talk about Quantum Computing  we mostly talk about entanglement tunneling and   superposition but inference is not there we don't  talk about it but it's quite easy to comprehend   because we face it in our everyday lives you  know when someone is using noise cancelling   uh headphones that's exactly how it works because  it's one wave cancelling another you know in the   headset there is an integrated mic that listen to  the noise record it invert it and place it back   to your ear so they built a transistor based on  these two states and the states are controlled by   the voltage applied to the gate and this applied  voltage changes the phase of the electrons that   behave as waves through the channels how it  actually works when the transistor is switched   on electrons interfere constructively so they  can flow from this source to the gate through   this molecule when it's off electrons in the  channel interfere with each other destructively   meaning that they're canceling each other and  the transistor is off in this case completely   off and there is no linkage current and actually  this is one of the main claims of this work that   they were able to develop a device that has almost  no leakage current and that's a big deal and you   know usually destructive inference is considered  to be a bad guy but here they managed to turn it   into something really useful so to me this work  is super interesting and if you know about my love   for transistors and quantum physics no surprise  right but the surprising fact here is that that   they've managed to build a really good device so  what do we call a good transistor you see this   curve showing the current versus applied voltage  is very steep which is not usual for small process   nodes so technically from this perspective it's a  very good transistor but it comes with the strings   attached first of all at the moment they have  no clue how to connect these transistors and of   course a single transistor is of no use because  you can't do much with it you need to connect   many of them into logic gates like AND, XOR, Etc  and they are figuring this out right now however   there is another huge elephant in the room and at  first I didn't quite get it I was like wait why do   we need to cool it you know intuitively I thought  maybe because Quantum effects behave better and   also better absorbed at cool temperature that's  why they have to cool it down to 30 Kelvin which   is what minus 240° C or about 400 something  Fahrenheit depending on where are you in the   world however from the paper it seems like this  limitation comes from the material itself in the   case of graphene at higher temperatures there is  noise fluctuation and charge trapping happening   at the ages of graphene and deteriorating the  performance of the trans resist that's why all   the measurements are in the range from 30 to 100  Kevin another thing that I was very concerned   about is switching frequency because it's quite  important you know and in this works I mentioned   the switching frequency of 7kHz and that's of  course not useful modern transistors work at   tens of gigahertz range but as I understood in  this case this is limited just by the current   setup and these transistors can work up to  terahertz range which is reasonable number   but in practice this THz switching frequency can  never be achieved because of the interconnect   parasitics this is the main limitation I talked  about this in the previous video that's why we   see even in the smaller process nodes like in 7,  5nm the operating frequency doesn't increase and I   imagine only computing with light can get around  this all in all it's a very interesting attempt   and as transistors will be getting smaller and  smaller these Quantum effects will dominate and   this new technology seems like a possible way to  scale Beyond 1 nm in a physical real dimensions   you know and this is really exciting the future  of transistors is our future because these   transistors are the most fundamental building  blocks of the modern electronics and our success   in its continuous Improvement is fundamental for  the advancement of technology and for the future   of our civilisation let me know what you think  in the comments and I will leave some links in   the description below so check it out I'm pretty  confident that we will see more and more research   that tries to take advantage of this Quantum  effects and using inference in this case is not   a new approach they have been similar works in the  past this one appears to me like the first real   transistor with decent characteristics those there  is obviously a long long way to go until they can   enter commercial production another thing I'm not  entirely sure about the materials they used here   to be specific about this single molecule that  they trapped between the electrodes honestly I   don't know much about this because I don't have  a degree in chemistry maybe you guys can help me   to determine if this material is something viable  or not so if you're an expert please let me know   in the comments so this was quite a geeky video  right but quantum physics is a fascinating field   that has many implications and you know this idea  that quantum entanglement can be responsible for   the interconnectedness of all the things including  the connection between our thoughts and the world   around us and I would love to understand it  way better than I do now let me know what you   think and if you want let's connect on LinkedIn  all the links are in the description below ciao
Info
Channel: Anastasi In Tech
Views: 235,681
Rating: undefined out of 5
Keywords: new transistor, transistor breakthrough, new quantum, quantum transistors, beyond 1nm
Id: Gzkb3Zc8pGE
Channel Id: undefined
Length: 16min 10sec (970 seconds)
Published: Wed Apr 10 2024
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.