The material in this article comes from the Internet, if it is inconsistent with the actual situation or there is infringement, please contact to delete.
(This article was first published on the public account "Liu Run", please subscribe to the public account "Liu Run" and follow the connection to me to understand the essence of the business)
Huawei held a press conference today.
Congratulations to Huawei.
This full launch did not mention mobile phones or chips.
However, concerns about the Mate 60 series of mobile phones and their "7nm chips" still exist, which has made many people cry.
On the barrage of the press conference, it flashed for a while: What about mobile phones? What about chips? And what about lithography?
Time goes back 28 days.
A thunderous sound
On August 29, Huawei Mate60 Pro suddenly went on sale without any publicity.
Immediately after, from the major hot search lists to the circle of friends, a word flooded the screen: 7nm chips.
Many of the first to take the phone apart did the same thing: take it apart.
Take the Kirin 9000S chip out of your phone, track the score, test the performance, and see what level it achieves.
The bottom line is: this may indeed be a 7nm chip.
Thunder is loud.
Many people lamented: "The most difficult period has passed, and the ship has passed through ten thousand mountains."
Why do you say that? How hard is it to make a 7nm chip? Is it really amazing? What does it matter to me what I do or don't do? Is it just a question of whether to buy a new phone?
Coincidentally, some time ago, I invited Yu Sheng, the author of "Chip Wars", to my live broadcast room. I took this opportunity to look up some information and ask some friends for advice. More and more, I felt:
Building a 7nm chip does require crossing thousands of mountains and rivers.
It would be great if we could overcome that.
It's amazing, and it's really worth knowing.
So let me help you solve it today.
The information is a bit difficult, so I will try to explain it to you in Mandarin.
Let's start with the "7nm" that makes many people stand up and marvel.
7 nm
The first question: what does 7nm really mean?
Why are you interested in this number? Is this great?
It has to start with you.
When you buy a phone, do you want it to be strong, long-lasting, thin, and sized?
When these three requirements are introduced into the chip world, they become the three "ultimate KPIs":
Power Purchase Agreements.
Performance, performance, area size.
This PPA fell to the chipmaker and became a "small target":
Put more transistors into smaller chips.
The main goal is to give employees more work to do, which can help you complete more and bigger projects, use less electricity and take up less space.
But what if there are too many employees to accommodate?
The solution is outrageous: get employees to lose weight. There is a "groove" in the structure of the transistor, leaving a lot of room for weight reduction.
So note that when we first talk about chips and say "yours is a 28nm chip" or "this is a 14nm chip," 28nm and 14nm do not refer to chip size, transistor size, or the difference between transistors and transistors. The distance between them is the "channel width" in the transistor.
But then when chatting, it rolled. 28nm, 14nm, 7nm...
When it comes to "7nm chips", whether the "channel width" will actually shrink to 7nm is no longer an issue. Everyone has their own opinion, but the bottom line hasn't changed:
Smaller nanoprocesses mean better PPAs that can accommodate more "employees" in smaller "offices."
How much is enough?
Creating a 14-nanometer chip means packing more than 30 million transistors per square millimeter.
Making 7-nanometer chips means that nearly 100 million transistors must be packaged per square millimeter.
Double the capacity. But also redouble your efforts.
And this is just the beginning of "Over the Ten Thousand Mountains".
Because it is not enough to block light, many transistors must be arranged according to a specific circuit diagram to work together.
So the question arises: how do you etch a circuit diagram as fine as the nanometer scale and as complex as a map of Shanghai onto a chip smaller than a fingernail?
photoetching
That's right, it relies on an expensive-sounding method: lithography.
How to engrave with light?
They say it's complicated, but it can also be very complicated. A lithography device with more than 100,000 parts costs hundreds of millions, not including freight. It is more expensive than the Boeing 737, just to make it possible.
But in simple terms, it is also very simple. Have you seen this movie?
Traditional motion picture film is projected with a beam of light that passes through a lens like a magnifying glass, and then passes through a layer of film to project the pattern on the film onto the screen.
Lithography technology is similar. It also shoots a beam of light that passes through the lens array and then through the mask, which then projects the circuit diagram etched on the mask onto the chip-fabrication substrate (wafer).
The only difference is that when you show a movie, you use a "magnifying glass" to project a small image into a large image. Lithography uses a "magnifying glass" to project a large image into a small one.
How clever it is to use light projection as leverage. However, at this point I just clearly drew the line and knew where to start next.
But how to get started?
The circuit diagram of a 7nm chip must clearly arrange tens of billions of transistors and other electronic components.
Moreover, from the transistor to the wire connecting the transistor, it is fine to the nanometer level, which is a hundred thousand times thinner than the blade of your kitchen knife. Some insiders once said: This is equivalent to carving the entire Shanghai in an area the size of a fingernail. And you won't miss the room or get lost.
It's crazy, how to carve? How to engrave the grooves of such wiring diagrams "quickly, accurately and stably"? Laser?
It's not that no one tried it in the first place.
However, laser direct writing, nanoimprinting... I tried one method after another. Some are very expensive, some are very slow, and some are easily scrapped and difficult to commercialize. Whoever does this loses money.
Until someone found a very ingenious method:
It took so much to save the land. Use photoresist.
Photoresist
What is a photoresist?
Photoresist is something that is very sensitive to light.
After exposure to a certain wavelength of light, a chemical reaction occurs.
It is originally tough, but becomes timid after exposure and is easily washed off with chemical solvents.
Grasping this, lithography has a completely new way to solve problems:
It does not rely on physics to engrave one at a time, but on chemical etching layer by layer.
Although there are many processes, the idea is roughly similar to "locking the elephant in the refrigerator", with four main steps:
The first step is to apply glue. A layer of photoresist is evenly applied to the raw material of the chip, the wafer.
The second step is lighting. Let a specific beam pass through a mask with a circuit diagram drawn.
When there are lines covering the area, there is no light. Photoresist is temperamental.
If no lines cover the area, the photoresist will have a different temperature when exposed.
The third step is to wash off the glue. Wafers coated with two photoresists are placed into a specific chemical solution for reaction.
After the changed photoresist is dissolved, the circuit diagram is displayed on the photoresist layer.
Step 4: Etching. Place the wafer in the etching solution.
The undissolved area of the photoresist is equivalent to covering a protective film, while the dissolved area of the photoresist will come into direct contact with the corrosive fluid and be etched "quickly, precisely and relentlessly" to match. Wiring diagram. Matching ingress.
Light, mask plates, photoresists, wafers, various chemical solutions, a physical problem that was originally considered difficult to solve suddenly became a simple chemical problem and was solved.
This is currently the main method of lithography:
First, as with film projection, the circuit diagram is projected onto the substrate by a "zoomed-out mirror" projection;
Similar to developing photographs, circuit patterns are etched onto the chip by partial exposure photoresist.
From this point of view, lithography is not very difficult.
It doesn't look like it.
But there is a key problem, which is the wavelength of light.
wavelength
At least the knife in your hand must be thin enough to etch out nanometer-scale wiring diagrams.
How do I get a thinner knife?
When your knife is stainless steel, all you have to do is sharpen the blade. But what do you do when your knife is a beam of light and you can't sharpen anything?
From the source of the material of the knife: the shorter the wavelength of light, the sharper the natural blade of the knife.
Because the shorter the wavelength of light, the smaller the diffuse diffraction angle. That is, the light will walk more obediently in a straight direction, without blurring or running around. It will hit where you are aiming.
It's not easy. Just turn on the spectrum and look for the light with the shortest wavelength. Very simple.
Spectrum (Image source: www.asml.com/en)
Not simple. Because short-wavelength light is not something you want to use.
Do you have the ability to continue issuing steadily and continuously while controlling costs? Will your photoresist react? Are your other processes compatible with it?
These are difficult problems. Everything should be explored.
After research, there are two main "light knives" that people can adopt with stable efficiency and controllable cost:
DUV and EUV.
DUV is the name for one type of light: deep ultraviolet. Wavelengths can be as short as 193 nm.
Many people believe that with this "light knife" lithography equipment, basically only chips with more than 20nm process can be engraved.
EUV is also the name of a type of light: extreme ultraviolet light. As can be seen from the name, this disc is wound more closely, and the wavelength can only be 13.5nm.
Whoever has this knife will be able to go further and cut more advanced chips, such as 7nm, or even 5nm, 3nm, 2nm.
Really good. So isn't the problem of finding short light waves solved?
Chips using EUV for 7nm process are produced.
Technical problem solved. But more problems followed.
Someone got stuck in the throat.
Stuck neck
Currently, only one company worldwide is able to manufacture EUV lithography equipment: Dutch ASML.
In 2018, SMIC spent 120 million euros, equivalent to its full-year profit, to order China's first EUV lithography equipment from ASML.
What's the big deal. ASML is also satisfied, even the export license is ready.
However, the United States spoke up. EUV lithography equipment is said to contain 20 percent U.S. components, and if it wants to be exported, it must be approved by them. They disagree.
Paper ban. What to do? If we can't use EUV that can engrave 7nm chips, can't we make 7nm chips?
Can you try a DUV that can only etch chips above 20nm?
He has hope.
There are two techniques that may offer promise: immersion lithography and multiple exposure.
Immersion lithography
What is immersion lithography?
It's simple. Translated as: soak water cut.
It is known: the shorter the wavelength of your "lightsaber", the better.
DUV light waves are known to be as short as 193 nm. An idea for engraving more advanced chips emerged: Is it possible to shorten the DUV wavelength?
Yes, add water.
A layer of ultrapure water is added between the surface of the plate and the lens, which does not contain any impurities such as minerals, particles, bacteria and microorganisms, and contains only hydrogen ions and hydroxide ions.
Then let the light refract in the water.
The refractive index of 193nm deep ultraviolet light in water is 1.44, and the wavelength can be further shortened to 134nm.
The "blade" became sharper.
Ingenious.
This method allows DUV lithography equipment to go directly from the dry era of "air sculpting" to the immersion era of "water sculpting".
But that's not enough.
By repeating the "blade" like this, you can get nominated in the class and increase the production level from the 28nm process to the 22nm process. However, it is still difficult to be admitted to Tsinghua University and master the 7nm process in one fell swoop.
What to do?
You can also add another method: multiple exposure.
Multiple exposures
What is Multiple Exposure?
It's also simple. Translated: engrave several times.
If I just want to cut a more precise chip, but I don't have a scalpel, only an ax, then ... Can I try chopping more with an axe?
Also.
What is the specific cracking method?
I'll give you an example, cut a cake.
When you go to buy pie cuts, the boss will cut you with an axe instead of a scalpel or kitchen knife.
When the axe falls, it will be cut at least 1 finger width and charged 20 yuan.
You don't accept. You just want to buy a knife for $7, but you don't have a better knife. What should you do?
Move the cutting board and continue slicing.
Cut once, move the board to the side, and press another 1 finger width piece to cut.
With a few more splits and more precise movements, you can split a finger-wide cake into narrower ones, and earn another 7 yuan out of 20 yuan.
Cut once, pan once, cut again until you get a narrower strip.
The same goes for multiple exposures.
The previous set of masks exposed the photoresist at the corresponding location and machined lines as thin as 134 nm on the wafer.
Then fold the base where the wafer is placed once.
Then apply a set of masks, expose again, and you can process 67nm lines.
Repeat a few times and the lines will become thinner and thinner.
This is multiple exposure.
The so-called LELE process, LFLE process and SAPD process are basically multiple exposure and multiple engraving methods.
Then continue to expose, expose a few more times, won't you be able to get a 7nm chip?
Or not. This approach has limitations.
Raising the axe multiple times will cost more effort. The more you cut and re-cut the cake, the more likely it is that the cake will be cut crookedly.
The same goes for multiple exposures. Each exposure will increase a lot of time, consumables and other costs, but also increase the possibility of scrapping, reduce the yield of the chip. Higher cost and lower efficiency are the price to pay for using DUV lithography devices to produce 7nm chips from multiple exposures.
The production of chips is not only a technical problem, but also an economic one. In addition to "maybe", we also have to consider "is it worth it".
Therefore, multiple sources believe that after comprehensive consideration, even with the use of immersion lithography and multiple exposures, the production of 7nm chips is almost the ceiling of DUV lithography equipment.
If you want to produce 7nm chips, or even more advanced 5nm chips, or 3nm chips, EUV lithography equipment must be more reliable.
It's so hard.
EUV lithography equipment cannot produce 7nm chips, but DUV lithography equipment has a cost and ceiling.
What about the future? What does this have to do with me?
Future
Of course, the relationship with you is not just about buying a mobile phone.
There are three laws in the tech industry.
Moore's Law, Anti-Moore's Law, Andy Beale's Law.
Moore's Law, you probably already know.
The law is Intel founder Dr. Gordon Moore:
Every 18 months, the performance of computers and other IT products doubles.
Then there's the anti-Moore law.
Eric Schmidt, former CEO of Google, points out that if we look at Moore's Law the other way around, if an IT company sells the same number of products today as it did 18 months ago, its revenue will be cut in half.
Andy Beale's Law is more of a prophetic blessing.
The original words of this law are "Andy gives, Bill takes." (Andy gives, Bill asks.) ）”
Andy was referring to former Intel CEO Andy Grove.
Bill was referring to Bill Gates, the former CEO of Microsoft.
This law states that the improved performance of hardware is quickly consumed by software.
For example, when the Intel processor in your computer is listed as i5, the Windows system installed in it will also be updated immediately and say goodbye to XP.
Every time you pair software and hardware, it's a leap in computing power.
What happens when you combine these three laws?
First, the industry will repeat itself.
Everyone is shouting 7nm chips, and soon they will shout 5nm, 3nm, 2nm, 1nm...
In addition, the company will grow.
You'll see more hot searches, talking about phone sales, competing phone sales and high-end market share...
You may or may not switch to a new, faster phone later.
However, as hardware options improve, software options also have the opportunity to move upstream.
The leap in computing power is on again:
A clothing e-commerce company in Hangzhou will one day find that once the factory releases a new product in the supply market, downstream customers can choose online and no longer need to bear the pressure of inventory. A customer service in Shijiazhuang discovered one day that he could serve the whole city alone.
An oil and gas exploration engineer in the South China Sea will one day discover that he can calculate 100 terabytes of data at a time and perform CT scans of the entire Earth's surface.
You will find that your world has entered an era in which computing power is once again improved and efficiency is restored.
The story of the 7nm chip is not just about chips, but also about computing power, technological development and competitive games.
This is a fundamental change that has not taken place in a century.
In this changing situation, some people shouted that the ship had crossed the Ten Thousand Heavy Mountains.
From not having a 7nm chip to having a 7nm chip. From DUV to EUV. From new phones to new computing power.
Everything is difficult, everything is possible.
But there are mountains beyond the mountains.
In addition to 7nm, there are 5nm, 3nm, and even 2nm, 1nm...
What to do?
Qingzhou did not answer. It just keeps swimming forward.
Go on, go on.
Blessing.
Disclaimer: The above content and materials are derived from the Internet, and the author of this article does not intend to target or allude to any real country, political system, organization, race, or individual. The above does not mean that the author of this article endorses the laws, rules, opinions, behaviors in the article and is responsible for the authenticity of the relevant information. The author of this article is not responsible for any problems arising above or related to the above and does not assume any direct or indirect legal liability.