You should do a special video on just keeping the air clean inside a fab. Everything for the building envelope to the HVAC equipment. These places are thousands of times cleaner than any operating room
a video about clean room air would be nice... and perhaps one about the vibration/seismic challenges of wafer production (perhaps there is one for the later, perhaps point to it.)
This is seriously my favorite channel. I know NOTHING about any of this stuff, and after watching your videos I still don't know anything, but it sure is interesting.
You know nothing about the Soviets and Americans bonding in Germany during WW II on the Elbe River? Now, at least you know that your teacher is a communist.
One of the things I like most about the Semiconductor industry is how everything is so directly named. “What is Metal Interlayer bonding?”Well it’s the previously discussed method of bonding but this time…. wait for it ….. it has a metal layer in between the two bits we’re bonding.😂
Yes! I put it down to the structuring of the industry. It's so fragmented and modular. The naming has to be intuitive for the fabless design engineers to understand fab technologies.
Love the video! I can’t speak on semiconductor wafer bonding, but in the MEMS field, wafer to wafer bonding is commonly used to package MEMS components in hermetic environments. The EVG tools that you mention can apply force, temperature, and vacuum, the last one sucks any air out of the MEMS package. This is really important for high Q factor MEMS resonant devices, where any damping, including any air molecules, can severely degrade performance. One subset of thermocompression bonding is eutectic bonding, which is super cool because if you heat up two different element thin films together, the alloy you create will have a lower melting point than each of the individual elements themselves. This can let you bond wafers together with really low temperatures. Indium-tin has a melting point of just over 100C, but most others are higher. Gold-tin is below 300C. When you’re choosing materials for your bonding, it all depends on what you’re designing. In-Sn has a low Young’s modulus, which might be undesirable. But Au-Sn’s eutectic temperature might be too high, and your devices can’t take that much heat. So you need to pick and choose materials to suit your needs
@@jintsuubest9331 It depends on how well the known theory works, but sometimes it is a guessing game with a lot of trial and error, because there are numerous effects and factors that can influence the result, making it quite challenging.
This video gave me new appreciation of our AML wafer bonding system, which has seen little use for some time. It’s not as nice as the EVG bonding machines, but at least it can do in-situ alignment. Thanks
Same - I have a few wafers from various career stops that I should carry with me to the grocery store to keep me from getting upset by the price of beer. How about a necklace of 57mm 4kbit DRAMs....
High quality image sensors for digital cinema cameras have back side connections, rather than interconnects between the pixel sites on the face of the chip. This allows rear side cooling, that helps to increase dynamic range. This way, interconnects have more space, and need not be so thin, and help to lessen signal degradation.
Most advanced sensors are since 10 years 2 chips (sensor and image processor) face to face connected by waferbonding. The sensor chip is thinned down to ~10um and illuminated from the backside. So light does not have to pass through any metal layer. Image sensors were the real driver for direct wafer bonding technologies.
The smoothness and planarity of silicon wafers which are dirt cheap in small sizes, is a good reason to use them in applications not resembling semiconductor manufacture in the least. I've used them as planar mirrors in a UV (IR won't work) spectrometer, and learned pretty quickly that stacking them even in normal pressure and temperature is not a good idea. For a short time you can easily get away with it, but leaving them overnight lets the interstitial air slowly escape, and now you are left with a stack of nearly inseparable quite brittle wafers.
AMD's 3D V-Cache technology was the first product to market using the direct-copper method. It was, apparently, a rather deep collaboration with TSMC that stretched several years. The Zen 2 based chips (Ryzen 3000 series) were equipped with the receiving pads for the V-Cache, but the yields were absolutely atrocious so it was never marketed, but by Zen 3 TSMC and AMD had figured it out. By making the V-Cache on N6 and the chips on N7 they had a match that yielded right, and the rest is history.
@@foobarf8766 The operative word was "direct-copper method" that Jon described in this episode. Yes, MCM/Chiplets have been used for a long time, but that was not what I was referencing. And no, the first wafer-wafer bond consumer product was the HBM on the Radeon Fury cards. But that was solder-bump bonding, not direct-copper.
Bullshit, Sony and Samsung using copper direct bonding for their image sensors much longer. TSMC was for sure not the 1st one. It is crazy how much false claims just focused on CPU are done by wannabe semicon experts.
Thanks for bringing the new technology information coming. Been out of the industry for 11 years now and don't get the trade journals anymore. I kind of miss reading about the cutting edge technologies.
One aspect that the taller silicon packages will have to deal with is heat dissipation. Eventually, thermoelectric effects such as the Seebeck and Peltier effects will have to be engineered in the electrical power supply of these devices in order to carry away the heat from the device onto a surface that can handle the heat.
Maybe the processor would have a series of channels with cooling fluid [IE mercury]. Also, diamond is both a good conductor of heat and electric current.
@@ArtificialDjDAGX Diamond has the best heat conductivity of any material, also very good insulator (if pure), see Thermal Conductivity chapter on Wikipedia: en.wikipedia.org/wiki/Material_properties_of_diamond
@@Martinit0 I was literally referring to the wikipedia article for diamond when calling them electrical insulators. You should've made your reply to the OP of this comment chain >.>
@@kj_H65f We were doing three-layer stacks for MEMS accelerometers in the late 80's in Milpitas on equipment we made ourselves. It was gold-gold bonding with heat and pressure. We found out how good the bonds were during dicing. Yeld could be a real problem!
AMD was the first company to introduce a product using wafer bonded HBM- the Radeon 300 series, or "Fiji" back in 2015. The Radeon R9 Nano was perhaps the most popular card in this series. TSMC and Amkor co-developed the silicon interposer technology necessary to make these products a reality. Amkor was the one who provided the silicon interposer in the first production cards. The high mechanical bonding strength makes HBM extremely reliable, far more so than GDDR where the memory is packaged separately from the GPU.
AMD was also the first to market with the direct-copper bonding method: The 5800X3D. It was originally planned for Zen 2, which even had the connecting pads, but the yields were terrible, so it wasn't brought to market.
Really interesting. Thanks. My understanding of what "TH-camrs" are has evolved over the past few years. I appreciate your contribution to my education.
That concluding remark about Moore's Law reminds me of how Ray Kurzweil defined the idea behind it in The Singularity is Near, where the overal idea of increasing computational power was already being looked at on a more holistic level. He talked about the computational capability we could cram into a fixed volume and not on a per die basis. Part of the reason for doing so was so that he could trace Moore's Law backwards in time, to before we used semiconductors, but another reason was that this way you can think beyond the current technological paradigm.
Wringing is not the same as cold welding, it looks similar, but is not permanent. It is used with gauge blocks. They hold very well together, but can be separated without leaving any trace. I suspect that some of the bonding here is not actually cold welding, but wringing, and behaves similar enough.
It's not done in a clean room or in a vacuum, or it would be. These forces are subject to inverse square law, so the slightest film of oil, or speck of dust prevents them from welding.
I am certain wringing is a fluid/capillary action phenomenon. It is advised not to leave two gage blocks wrung together for too long or it will become increasingly difficult to separate them.
I'm sure induction occurs between steel blocks. In a vacuum (like space) there is no air, or air _pressure_ to hold blocks together, but they can get closer by virtue of that fact.
The "bonding" you mention, I also seen a version of this, in HS when making a set of parallel blocks and grinding the faces smooth. If you did it right when you put the blocks face to face they stuck to each other,. You couldn't easily PULL them apart you had to slide them off each other. I think it's a version of the same animal.
@Asianometry I'd join your Patreon if I could, but I'm literally saving every cent so that I'm not living in my car anymore, but I'll join eventually, but pronbably shouldn't be a priority at the moment 😂 I'll make a request for a video anyway. You've covered so much in terms of lithography and a fair bit about packaging etc, but what I'd really like to know is some of the finer details about how the silicon ingots are sliced into wafers, and made almost perfectly flat, then how they cute the wafer into individual dies so cleanly, then the mind boggling bit for me, how they then take that die and attach those stupidly tiny wires so accurately during packaging. If it follows the subject enough, also how they maintain such accuracy across so many lithography steps and somehow maintaining such obscenely tiny features. Thos bits really blow my mind. Also if there's enough development (or available information) about the approaches being taken to develop cooling solutions that will be required in the near to mid future.
I like that i have to watch most parts of the video twice to fully understand them. Your videos are very information dense - as they should be, considering the subject. A nice change of pace from the usual ai slop adhd shortform filler "content" that takes over the algorithm these days
Seriously love your videos. They keep my curiosity very much alive & well. Do you think you would ever consider making a video on Bubbles lol Like all the unique ways bubbles are being utilized in things? Such as ultrasonic cleaning processes and other things. Bubble science actually seems pretty fascinating. Last thing, I saw these robots that mimic roots. They use compressed air, are remote controlled, have a camera at the front end of them etc, etc. It looks like they could be extremely useful and I hope we see a ton of them getting utilized in all sorts of ways, soon.
Great Video!! Thank you! Can you do a deeper dive into the surface roughness of the wafers and how it affects the bonding? What are the current measurement techniques and their limitations and what is the latest ongoing research to improve these techniques?
It turns out that "Lord Rayleigh" refers to two separate men in the world of Physics: 1. the more famous physicist John William Strutt (1842-1919), aka 3rd Baron Rayleigh. 2. his son, physicist Robert Strutt (1875-1947), aka the 4th Baron Rayleigh. Very confusing. I wish that Nature magazine (and the physics world at large) would have referred to Robert Strutt as "Lord Rayleigh 2", or "Lord Rayleigh Junior". (Anything to distinguish them meaningfully.)
Nice, i was searching to learn about wafer bonding since i might need it for my Phd, and this video came at the right time! and as always it's super interesting
His name is Richard Feynman. The surname is pronounced like "Fineman", not "Feignman". Understandable mistake to make if you've only seen it in writing, though. Cheers, and thanks for another interesting video.
Comprehensive quarter-hour dive into wafer bonding. Sounds somehow like there is only one company leading this field… 😉 The only three points of critique I found in this quite relaxed and fun-to-watch quick-dive are: First, the comparison to die2die-bonding and the challenge of efficiency due to the parallel nature of bonding thousands of devices in parallel, which makes the step intriguingly attractive as well as precision alignment required when adjusting the two ‘pizzas’ together. In fact, achieving the nm-scale lateral precision of bonding for 300mm HBM-wafers on a pizza where all pairs of 5cm Peperoni-slices would meet, would allow to expand the pizza to the size of our planet - funnily enough this ratio matches the difference between walking speed and the speed of light… 🍕🌍 The second one is the advantage to avoid risks that every wafer processing step carries along. Wafer bonding drastically reduces the accumulated risk that the 60…100 typical processing steps introduce by joining two or more stacks of pre-testable and therefore verified operational, ultrathin, nm-scale, 3D-structured surfaces in only one process. Thereby eliminating 99% of this risk for the complex stack that otherwise would see thousands of individual steps that each might destroy the delicate sandwich and thus rendering it as a delicious interconnected structure that outperforms the taste of a ‘double-whopper’ 🍔 And third - bonding also allows to stresslessly join heterogenous materials that the common process of epitaxial growing would not permit to do, due to intercristaline tensions when slowly building up new layers with different atomar lattice distances rather than joining the fully grown lattices.. - But hey, they mentioned that there is still more to come that we can look forward to! 😁
I hope we can make Chips as retinas and a complete artificial eye for blind people, and noses for the ones that are born without 1 or lost it in an accident/war; and other parts of the body too, like printing teeth from the person's DNA or something difficult like that. I hope future gens. do better than us, much better
The Rayleigh of the 1936 article can't be THE Rayleigh who died in 1919. It is probably his son who inherited the "Lord" title and was also a physicst (died in 1947).
@@stevengill1736Sure has! Most of it ties back to its in-house development of air pollution monitoring instruments in the 1970s, plus an acquisition of nuclear radiation monitoring instruments. Then on to spectroscopy and many other analytical technologies from the '80s to now.
Fun fact, you don't need the special conditions to experience this, if you stack two wafers they stick together and it's a lot of fun to play with haha
Richard Feynman pronounced his last name fine-man, not fane-man. It looks like it should be the latter, but for whatever reason he used the former. Don't ask the internet "how is feynman pronounced", ask "how did richard feynman pronounce his name".
Some may see it as a gaffe, but I see it as someone who's dived deep into reading and researching, actively seeking knowledge instead of passively receiving it from lectures or videos. Hooray for people who encounter so many words by reading them first they're not aware of pronunciation!
@@michaelturner2806weak argument. It means someone has read a book slo e but never talked to anyone else. This is an indication of shallow knowledge. Also the name is German, and that pronunciation has stuck apparently.
@@lookoutforchris It can often be difficult to find someone to talk to through voice about niche or technical things. First you have to find someone that shares your interest, which can be constrained by geography, class divides, schedules, all manner of different things. I was such a kid, having interest in hard sci-fi and learning all kinds of technical words from reading books in my public library, but having no one to discuss them with except whichever friend I happened to have at the time who was semi-interested as well but similarly had not been educated on pronunciation.
These constant mispronunciations of words worry me. Even more so when they are technical terms relevant to the field than famous physicists. Unfortunately we also know he doesn't read the comments because people keep pointing it out and he keeps making the same "freshman marker" mistakes.
@@michaelturner2806if all your lectures have been passive, you've only had bad professors. And it's not difficult to find people to talk to if you're active in a field.
There is a bit of stress induced, but the gap between the dielectric and the copper the so caller dish coming from the CMP process is typically only 1-5 nm and therefore the induced stress is very low and overlay accuracies of below 100nm on a 300mm are possible.
That's a choice pronunciation of Feynman. And sure, Google says thoughts how "American" pronunciation of the word is, and he's American, so not a bad guess, but please respect the man and his work enough to know he went by the British pronunciation (i.e. FINEman), and use that.
It would be also great to know what are the benefits of bonding multiple wafers together. The title of the video suggests that this would be answered, but its nonetheless great video
![⭐️( ลอยกระทง 2024 ) Ver. แดนซ์ Remix BY [ ดีเจกิต รีมิกซ์ ]](http://i.ytimg.com/vi/rMWOLhgXUHc/mqdefault.jpg)
"No Mr Die, I expect you to Bond". . .
I am definitely stealing that one!
Haha nice 😂
Brilliant!
😃
😂, got a laugh out of my wife. We were trained as electronics engineer and she understood. Not sure who else will understand. 😅
You win some large part of the internet today.
You should do a special video on just keeping the air clean inside a fab. Everything for the building envelope to the HVAC equipment. These places are thousands of times cleaner than any operating room
a video about clean room air would be nice... and perhaps one about the vibration/seismic challenges of wafer production (perhaps there is one for the later, perhaps point to it.)
That and the water... everything has to be sooo clean!
@@stevengill1736
He already did a video on the super clean water iirc.
@@jintsuubest9331 Yes, the Ultra Pure Water video. He has stated good references for further study as well in that video
As a refrigeration guy I 100% support this
This is seriously my favorite channel. I know NOTHING about any of this stuff, and after watching your videos I still don't know anything, but it sure is interesting.
You know nothing about the Soviets and Americans bonding in Germany during WW II on the Elbe River?
Now, at least you know that your teacher is a communist.
YES!
Why you don't understand it ?
just the basics are too hard ?
@@lucasrem Yes, why are you so condesending about it?
I agree. Since I found this channel, it has quickly moved to the very top of my list of favorite TH-cam channels.
LOL! "I'm sorry Dave, the pod bay door have cold welded together."
true story, though.
One of the things I like most about the Semiconductor industry is how everything is so directly named. “What is Metal Interlayer bonding?”Well it’s the previously discussed method of bonding but this time…. wait for it ….. it has a metal layer in between the two bits we’re bonding.😂
Indeed, CMOS is not a fancy name, its just Complementary Metal Oxide Semiconductor
Well, imagine if more abstract names are being used...it will get really confusing really quick ¯\_(ツ)_/¯
The hardest thing about any industry is naming things, at some point you gotta give up coming with catchy names
Yes!
I put it down to the structuring of the industry. It's so fragmented and modular.
The naming has to be intuitive for the fabless design engineers to understand fab technologies.
Love the video! I can’t speak on semiconductor wafer bonding, but in the MEMS field, wafer to wafer bonding is commonly used to package MEMS components in hermetic environments. The EVG tools that you mention can apply force, temperature, and vacuum, the last one sucks any air out of the MEMS package. This is really important for high Q factor MEMS resonant devices, where any damping, including any air molecules, can severely degrade performance.
One subset of thermocompression bonding is eutectic bonding, which is super cool because if you heat up two different element thin films together, the alloy you create will have a lower melting point than each of the individual elements themselves. This can let you bond wafers together with really low temperatures. Indium-tin has a melting point of just over 100C, but most others are higher. Gold-tin is below 300C. When you’re choosing materials for your bonding, it all depends on what you’re designing. In-Sn has a low Young’s modulus, which might be undesirable. But Au-Sn’s eutectic temperature might be too high, and your devices can’t take that much heat. So you need to pick and choose materials to suit your needs
Just make sure no Helium is around during this MEMS package process.
Do the material science people just sit around and mix different element to see what works what doesn't?
@@jintsuubest9331 It depends on how well the known theory works, but sometimes it is a guessing game with a lot of trial and error, because there are numerous effects and factors that can influence the result, making it quite challenging.
Chips are going from pizza style to lasagne.
This video gave me new appreciation of our AML wafer bonding system, which has seen little use for some time. It’s not as nice as the EVG bonding machines, but at least it can do in-situ alignment. Thanks
9:35 - emotional support wafer had me balling 🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣
Same - I have a few wafers from various career stops that I should carry with me to the grocery store to keep me from getting upset by the price of beer. How about a necklace of 57mm 4kbit DRAMs....
Not sure what E is doing there, but it does support the other wafer gegen motion.
So it doesn’t feel left out
*bawling
Came for the industry info, stayed for the low-key nerd humor.
Keep being awesome!
High quality image sensors for digital cinema cameras have back side connections, rather than interconnects between the pixel sites on the face of the chip. This allows rear side cooling, that helps to increase dynamic range. This way, interconnects have more space, and need not be so thin, and help to lessen signal degradation.
Most advanced sensors are since 10 years 2 chips (sensor and image processor) face to face connected by waferbonding. The sensor chip is thinned down to ~10um and illuminated from the backside. So light does not have to pass through any metal layer.
Image sensors were the real driver for direct wafer bonding technologies.
The smoothness and planarity of silicon wafers which are dirt cheap in small sizes, is a good reason to use them in applications not resembling semiconductor manufacture in the least. I've used them as planar mirrors in a UV (IR won't work) spectrometer, and learned pretty quickly that stacking them even in normal pressure and temperature is not a good idea. For a short time you can easily get away with it, but leaving them overnight lets the interstitial air slowly escape, and now you are left with a stack of nearly inseparable quite brittle wafers.
AMD's 3D V-Cache technology was the first product to market using the direct-copper method. It was, apparently, a rather deep collaboration with TSMC that stretched several years. The Zen 2 based chips (Ryzen 3000 series) were equipped with the receiving pads for the V-Cache, but the yields were absolutely atrocious so it was never marketed, but by Zen 3 TSMC and AMD had figured it out. By making the V-Cache on N6 and the chips on N7 they had a match that yielded right, and the rest is history.
IBM been making the 'MCM' for around for 20 years AMD not the first, but probably the first place most consumers could get a wafer-wafer bond?
@@foobarf8766 The operative word was "direct-copper method" that Jon described in this episode. Yes, MCM/Chiplets have been used for a long time, but that was not what I was referencing. And no, the first wafer-wafer bond consumer product was the HBM on the Radeon Fury cards. But that was solder-bump bonding, not direct-copper.
Bullshit, Sony and Samsung using copper direct bonding for their image sensors much longer. TSMC was for sure not the 1st one. It is crazy how much false claims just focused on CPU are done by wannabe semicon experts.
Thanks for bringing the new technology information coming. Been out of the industry for 11 years now and don't get the trade journals anymore. I kind of miss reading about the cutting edge technologies.
2:01 forbidden oreo
next time use oreo picture as well
I like peanut butter crackers myself 🤣
One aspect that the taller silicon packages will have to deal with is heat dissipation. Eventually, thermoelectric effects such as the Seebeck and Peltier effects will have to be engineered in the electrical power supply of these devices in order to carry away the heat from the device onto a surface that can handle the heat.
Maybe the processor would have a series of channels with cooling fluid [IE mercury]. Also, diamond is both a good conductor of heat and electric current.
@@stevenpace892 diamonds are insulators, though?
@@ArtificialDjDAGX Diamond has the best heat conductivity of any material, also very good insulator (if pure), see Thermal Conductivity chapter on Wikipedia: en.wikipedia.org/wiki/Material_properties_of_diamond
@@Martinit0 I was literally referring to the wikipedia article for diamond when calling them electrical insulators. You should've made your reply to the OP of this comment chain >.>
-Asianometry : wafer bonding is the future
-Me working in an old fab with 20+ years old bonders : wait what
Those are wire bonders, not wafer bonders
@@paulds65 you mean those in the video ? Because he does talk about wafer bonders.
@@paulds65no we've been doing molecular bonding for well over 10 years in my factory. Wafer to wafer bonding is new but chip to wafer is old
@@kj_H65f We were doing three-layer stacks for MEMS accelerometers in the late 80's in Milpitas on equipment we made ourselves. It was gold-gold bonding with heat and pressure. We found out how good the bonds were during dicing. Yeld could be a real problem!
16:18 ah there's nothing like casually violating your employers NDA eh?
AMD was the first company to introduce a product using wafer bonded HBM- the Radeon 300 series, or "Fiji" back in 2015. The Radeon R9 Nano was perhaps the most popular card in this series. TSMC and Amkor co-developed the silicon interposer technology necessary to make these products a reality. Amkor was the one who provided the silicon interposer in the first production cards. The high mechanical bonding strength makes HBM extremely reliable, far more so than GDDR where the memory is packaged separately from the GPU.
AMD was also the first to market with the direct-copper bonding method: The 5800X3D. It was originally planned for Zen 2, which even had the connecting pads, but the yields were terrible, so it wasn't brought to market.
I think you mean first consumer/retail product? IBM call it MCM, been used in Power arch for 20 years but you won't find that in any gaming rig
@@foobarf8766 See my response in the other thread. You're mistaking Multi Chip Modules for wafer bonding.
Really interesting. Thanks. My understanding of what "TH-camrs" are has evolved over the past few years. I appreciate your contribution to my education.
I love the way you pronounced Diderik. From now on it shall be pronounced this way. Greetings from the Netherlands.
😂 I actually had 2 300mm wafers stick together under vacuum and it never occurred to me that this kind of bond was actually at play
Have to say I am sad that you did not cover more of the UHV or SAB part.
PS. I am a bonding process engineer 🤓😂
That concluding remark about Moore's Law reminds me of how Ray Kurzweil defined the idea behind it in The Singularity is Near, where the overal idea of increasing computational power was already being looked at on a more holistic level. He talked about the computational capability we could cram into a fixed volume and not on a per die basis. Part of the reason for doing so was so that he could trace Moore's Law backwards in time, to before we used semiconductors, but another reason was that this way you can think beyond the current technological paradigm.
Very cool! I've used plasma bonding for microfluidics when sticking PMMA to slides, works ridiculously well.
Wringing is not the same as cold welding, it looks similar, but is not permanent. It is used with gauge blocks. They hold very well together, but can be separated without leaving any trace.
I suspect that some of the bonding here is not actually cold welding, but wringing, and behaves similar enough.
Engineers talk of "wringing together" gauge blocks, the highly machines surfaces "stick" together, but not as strongly as cold-welding!
It feels like a form of magnetism...
It's not done in a clean room or in a vacuum, or it would be.
These forces are subject to inverse square law, so the slightest film of oil, or speck of dust prevents them from welding.
I am certain wringing is a fluid/capillary action phenomenon. It is advised not to leave two gage blocks wrung together for too long or it will become increasingly difficult to separate them.
I'm sure induction occurs between steel blocks.
In a vacuum (like space) there is no air, or air _pressure_ to hold blocks together, but they can get closer by virtue of that fact.
@@stevengill1736 it is! all chemical bonds are electrostatic effects.
As a semi engineer this channel seriously has helped me in my career by broadening my scope of knowledge on past and future tech.
Hopefully with enough videos you’ll become a full engineer
@@Gpig20it's means he's in university
@@Gpig20 one day :']
@@NadeemAhmed-nv2br Finally free of duochorustown
@@Gpig20semi is short for semiconductor.
The "bonding" you mention, I also seen a version of this, in HS when making a set of parallel blocks and grinding the faces smooth. If you did it right when you put the blocks face to face they stuck to each other,. You couldn't easily PULL them apart you had to slide them off each other. I think it's a version of the same animal.
😋 Pizza, cheeseburger, Cuban sandwich: very tasty episode, Jon! 😎✌️
Speaking about wafer bonding one should not forget how much EV Group contributed to this field since decades.
you are really sparking my internet in the semiconductor industry. i would love learning this stuff through
@Asianometry
I'd join your Patreon if I could, but I'm literally saving every cent so that I'm not living in my car anymore, but I'll join eventually, but pronbably shouldn't be a priority at the moment 😂 I'll make a request for a video anyway. You've covered so much in terms of lithography and a fair bit about packaging etc, but what I'd really like to know is some of the finer details about how the silicon ingots are sliced into wafers, and made almost perfectly flat, then how they cute the wafer into individual dies so cleanly, then the mind boggling bit for me, how they then take that die and attach those stupidly tiny wires so accurately during packaging. If it follows the subject enough, also how they maintain such accuracy across so many lithography steps and somehow maintaining such obscenely tiny features.
Thos bits really blow my mind. Also if there's enough development (or available information) about the approaches being taken to develop cooling solutions that will be required in the near to mid future.
I like that i have to watch most parts of the video twice to fully understand them. Your videos are very information dense - as they should be, considering the subject. A nice change of pace from the usual ai slop adhd shortform filler "content" that takes over the algorithm these days
Mmmmmmm wafers..... 🤤
Loacker, the hot firm for the future! 😂
I hear Homer Simpsons voice in that comment.
And the channel image shows a deer.
I like chips too much to care about wafers
Cold bonded to perfection
Seriously love your videos. They keep my curiosity very much alive & well. Do you think you would ever consider making a video on Bubbles lol Like all the unique ways bubbles are being utilized in things? Such as ultrasonic cleaning processes and other things. Bubble science actually seems pretty fascinating. Last thing, I saw these robots that mimic roots. They use compressed air, are remote controlled, have a camera at the front end of them etc, etc. It looks like they could be extremely useful and I hope we see a ton of them getting utilized in all sorts of ways, soon.
Great Video!! Thank you! Can you do a deeper dive into the surface roughness of the wafers and how it affects the bonding? What are the current measurement techniques and their limitations and what is the latest ongoing research to improve these techniques?
This some transcending stuff. I searched for lithography; the literal act of hewing signs into rock, and found this. I feel like I'm in a time machine
It's a shame you missed one really important electronics cold bonding use - wire wrapping which was a thing before PCBs became common.
It turns out that "Lord Rayleigh" refers to two separate men in the world of Physics:
1. the more famous physicist John William Strutt (1842-1919), aka 3rd Baron Rayleigh.
2. his son, physicist Robert Strutt (1875-1947), aka the 4th Baron Rayleigh.
Very confusing.
I wish that Nature magazine (and the physics world at large) would have referred to Robert Strutt as "Lord Rayleigh 2", or "Lord Rayleigh Junior". (Anything to distinguish them meaningfully.)
Under rated task - correct electrical bonding. Hi resistance bonds faults even worse with AC signal. Nice video on the topic.
Nice, i was searching to learn about wafer bonding since i might need it for my Phd, and this video came at the right time! and as always it's super interesting
His name is Richard Feynman. The surname is pronounced like "Fineman", not "Feignman".
Understandable mistake to make if you've only seen it in writing, though. Cheers, and thanks for another interesting video.
Great content, I hope you continue to do such topics and present them in depth as you do.
Excellent introduction
Did anyone else glance at the title and think it said "water boarding"
Making the video before lunch, spawned the visual food analogies ? I approve !
Getting two gauge blocks stuck together is so annoying but it does show the principle very clearly.
Perfect material for during my shave
First thing that popped into my head was gauge blocks.
2:19 Feynman, rhymes with "lineman".
Comprehensive quarter-hour dive into wafer bonding. Sounds somehow like there is only one company leading this field… 😉
The only three points of critique I found in this quite relaxed and fun-to-watch quick-dive are:
First, the comparison to die2die-bonding and the challenge of efficiency due to the parallel nature of bonding thousands of devices in parallel, which makes the step intriguingly attractive as well as precision alignment required when adjusting the two ‘pizzas’ together. In fact, achieving the nm-scale lateral precision of bonding for 300mm HBM-wafers on a pizza where all pairs of 5cm Peperoni-slices would meet, would allow to expand the pizza to the size of our planet - funnily enough this ratio matches the difference between walking speed and the speed of light… 🍕🌍
The second one is the advantage to avoid risks that every wafer processing step carries along. Wafer bonding drastically reduces the accumulated risk that the 60…100 typical processing steps introduce by joining two or more stacks of pre-testable and therefore verified operational, ultrathin, nm-scale, 3D-structured surfaces in only one process. Thereby eliminating 99% of this risk for the complex stack that otherwise would see thousands of individual steps that each might destroy the delicate sandwich and thus rendering it as a delicious interconnected structure that outperforms the taste of a ‘double-whopper’ 🍔
And third - bonding also allows to stresslessly join heterogenous materials that the common process of epitaxial growing would not permit to do, due to intercristaline tensions when slowly building up new layers with different atomar lattice distances rather than joining the fully grown lattices..
- But hey, they mentioned that there is still more to come that we can look forward to! 😁
I hope we can make Chips as retinas and a complete artificial eye for blind people, and noses for the ones that are born without 1 or lost it in an accident/war; and other parts of the body too, like printing teeth from the person's DNA or something difficult like that. I hope future gens. do better than us, much better
12:24 "Hughes Aircraft". The "G" is silent. More like "huse".
Loved the food analogies, but I had to pause for a lunch.
Feynman is pronounced "Fineman" FYI
Makes me wonder how they get wafers in and out of a vacuum without "losing" too much vacuum.
Pre-pumping and large sliding gate valves. Very common in electron microscopy.
you finish the bond before unloading
Room temperature wafer bonding is needed for bonding a silicon wafer to a compound semiconductor wafer.
Sorry, couldn't hear you over the images of burgers
Just a dip in HF, without mentioning how nightmarish that stuff is...
On the other hand, backside power delivery sounds like fun.
The Rayleigh of the 1936 article can't be THE Rayleigh who died in 1919. It is probably his son who inherited the "Lord" title and was also a physicst (died in 1947).
3:30 The cited company should be "Hilger." Eventually was acquired by a subsidiary of what's now Thermo Fisher.
Thermo Fischer seems to be into EVERYTHING.....
@@stevengill1736Sure has! Most of it ties back to its in-house development of air pollution monitoring instruments in the 1970s, plus an acquisition of nuclear radiation monitoring instruments. Then on to spectroscopy and many other analytical technologies from the '80s to now.
Semiconductor battery will make electrical vehicles more viable.
17 minutes of sweet, sweet heaven.
Thank you this is interesting I have no idea how I could apply this at the moment.
I learned long ago everything however seemingly random is valuable.
Fun fact, you don't need the special conditions to experience this, if you stack two wafers they stick together and it's a lot of fun to play with haha
The food analogues are weird but they sure do work. Very visual about what happens.
Wafer bonding is truly the backbone of future semiconductor innovations-fascinating overview!
Richard Feynman pronounced his last name fine-man, not fane-man. It looks like it should be the latter, but for whatever reason he used the former. Don't ask the internet "how is feynman pronounced", ask "how did richard feynman pronounce his name".
Some may see it as a gaffe, but I see it as someone who's dived deep into reading and researching, actively seeking knowledge instead of passively receiving it from lectures or videos. Hooray for people who encounter so many words by reading them first they're not aware of pronunciation!
@@michaelturner2806weak argument. It means someone has read a book slo e but never talked to anyone else. This is an indication of shallow knowledge. Also the name is German, and that pronunciation has stuck apparently.
@@lookoutforchris It can often be difficult to find someone to talk to through voice about niche or technical things. First you have to find someone that shares your interest, which can be constrained by geography, class divides, schedules, all manner of different things.
I was such a kid, having interest in hard sci-fi and learning all kinds of technical words from reading books in my public library, but having no one to discuss them with except whichever friend I happened to have at the time who was semi-interested as well but similarly had not been educated on pronunciation.
These constant mispronunciations of words worry me. Even more so when they are technical terms relevant to the field than famous physicists. Unfortunately we also know he doesn't read the comments because people keep pointing it out and he keeps making the same "freshman marker" mistakes.
@@michaelturner2806if all your lectures have been passive, you've only had bad professors. And it's not difficult to find people to talk to if you're active in a field.
So, in hybrid bonding, does the copper of the interconnects become stressed when the temperature comes back down?
There is a bit of stress induced, but the gap between the dielectric and the copper the so caller dish coming from the CMP process is typically only 1-5 nm and therefore the induced stress is very low and overlay accuracies of below 100nm on a 300mm are possible.
chosen pictures are phenomenon
9:40 - that's gold Jerry, gold!
we're rich
Thanks for my latest tech update!
backsidepower delivery
all that bonding has to come with benefits at some point...
Would love to see a piece on ATE companies like Teradyne and Advantest. I can work in this industry and would love your take on it
Wafer bonding? *_OREO_* _Has entered the chat_
The skyscrapers of the semiconductor industry - up up and away to a better day - lets go for it and thanks for sharing.
So you were in the Dutch park with tiny buildings (Madurodam)?
I wasn't ready for assault of food imagery
Correctly pronouncing “Desaguliers” wins the internet
Hello from France, "good not great" - A for effort tho :)
Great video as usual
Wait till I explain this at the water cooler/BBQ/footie game night/card game
That's a choice pronunciation of Feynman. And sure, Google says thoughts how "American" pronunciation of the word is, and he's American, so not a bad guess, but please respect the man and his work enough to know he went by the British pronunciation (i.e. FINEman), and use that.
Speaking of bonding and TSMC… does this have something to do with the latest NVDA Blackwell delivery delay?
Any insight?
Nice work again Jon!
😃
TCB: taking care of business
For a minute I read, "wafer bombing...!"
I didn't know glass cold welded, wow - I've seen polished aluminum do that - it's amazing....
Are you gonna need to cut channels so you can pump cooling fluid between the wafers?
bonder man here. we do not use glue or solder to bond W2W or D2W. I can’t disclose what we actually do, but that part is not correct.
It would be also great to know what are the benefits of bonding multiple wafers together. The title of the video suggests that this would be answered, but its nonetheless great video
fwiw, Feynman is pronounced "fine-man"
15:15 I love the historical analogies lol
I guess somebody was hungry when they were editing this video. 😂
Is this how they bond aluminum and stainless steel for Alclad cookwear?
When does the Emotional support wafer shirt go on sale?
Have you done Johansson blocks yet?
Please Mr Asianometry, I beg of you! Talk about micro channels !
I know some folks who worked for many years on those "emotional support layers"...
So,
Pizza silicon is bad
Panini silicon is good.
"We introduce FUCKING PLASMA to the bonding process"
Science is so metal man..
We’re waterboardng silicon wafers now?
Always have done. The biggest part of an in-production wafer's life is spend in cleaning stations.