418 Chapter 408 True and False 1 Micron (2)

Semiconductor processing technology is essentially a process of continuous exposure and etching on a silicon wafer.

The process of upgrading this process is a process of continuous densification of the film pattern used in exposure.

In the traditional impression of everyone, the densification of the film is the process of improving the accuracy of the film.Is there no other way to increase the density of the film pattern except to improve the accuracy of the lithography machine?

In our daily life, there is an inappropriate example, that is, color printing (or color printing).

With three-color ink, the accuracy of each printing is the same, but when three-color overlap printing, monochrome becomes color!

The color accuracy has risen from 8-bit monochrome to 256-bit!

After 2005, due to the improvement of the process, the minimum resolvable feature size (MRF) has been far smaller than the wavelength of the light source, and it is no longer possible to etch and shape at one time with the DUV lithography machine.

Since it is impossible to etch and shape at one time, etch several times, etch a part each time, and then piece together the final pattern.

From the processing of each part of the graphics, the original processing methods and equipment are used, but it can achieve higher precision chip processing.

It is "Multiple Patterning Technology"!

The "Multiple Pattern Method" separates a figure into two or three parts.Each part is made according to the usual manufacturing method.The entire graphic is finally merged to form the final layer.

————————

According to this theory, the accuracy of graphics can be divided infinitely.

But in fact, this scheme also has its limitations.

The lithography machine has reached its limit because of the wavelength of the light wave.

Pattern segmentation, in the end, will also have this problem.

When the line width of the pattern on the mask is close to the wavelength of the light source, diffraction will be very obvious.

The internal optical path of the lithography machine has a limited ability to capture light. If there is not enough energy to reach the photoresist, the photoresist will not be able to fully react, making its size and thickness unable to meet the requirements.

It cannot play its due role in subsequent development and etching processes, resulting in process failure.

So using this method, stepping to 7nm, you can't do it anymore.Because there is a problem in principle.

After 7nm, EUV (deep purple) lithography machine must be used. That is the reason for the lithography machine embargoed on China.

At this stage (1 micron), it is not yet a problem.The main reason that hinders the progress of wafer process comes from production equipment and process, rather than principle.

————————

There are pros and cons to everything.

The advantages of this technology are very prominent.That is, there is no need to change the existing equipment or make very few changes to meet the requirements of improving the wafer process.

But the disadvantages are also very prominent.

The first drawback is trouble.

The prototype of this technology first appeared in the 130nm stage, and the first complete appearance was in the 30nm stage.

Why did it appear so late?

Each layer must be decomposed, which is very troublesome to think about.This method is absolutely no alternative.

Change to a high-precision lithography machine and its supporting technology, it will be solved all at once!This is also the reason why no one thinks in this direction before 30nm.

Secondly, cost.

The number of processing steps required to process a chip has increased.It used to be a single processing step, but now it takes two or even four times.

This is fatal in the manufacture of commercial chips.

For example, if only one processing is used, the yield rate is 70%.This is an acceptable number.But when one processing is changed to four processing, the yield rate of the whole process will drop to 20%.

The core of the multi-pattern method is to decompose a picture into multiple ones.There will also be the problem of aligning the sub-pictures with each other.Therefore, in the actual production process, the yield rate will be greatly reduced after this process is adopted.

Using the data from the example just now, the yield rate will drop from 70% to less than 10%!

The reason why Intel spent close to 5 years at the 10nm node is due to their low SAQP yield rate.

For a chip factory, the yield rate is their job.

If it is at 14nm, the chip cost is $300.The purpose of upgrading the chip production process is naturally because the higher the process, the smaller the wafer area occupied.After adopting the new technology, the production cost of the chip is naturally reduced.For the same functional chip, its cost should be reduced to $150 in the 10nm era.

However, this process has increased the number of processes and has actually increased the processing cost of the chip.Coupled with the problem of good product rate, the chip produced by the new method may not cost more than 300 US dollars.

In this case, why should mass produce 10nm?

When intel occupies a monopoly position, the performance is even more prominent.This is also the fundamental reason why the speed of the PC's CPU has not improved much for many years.

————————

But this reason is invalid for full color and invalid for photoelectric.

Quan Cai, and even Optoelectronics, is not a chip company. This 10nm process solves the problem of whether there is any problem, the problem of life and death.

This is the same as Huawei.Is Huawei selling chips?

Not!He is selling 5G systems!

With this $300 chip, a system worth tens of thousands of dollars can be sold!No, a dead end!

The process upgrade here does not save the $150 cost.Its value is tens of thousands of dollars!

The same goes for full color.With these dozen dollars of chips, thousands of dollars of monitors can be sold!This chip represents more than a dozen dollars in processing costs.

Under this circumstance, the yield rate, not to mention 30%, is 10%, or even only 1%.

Of course, the yield rate still needs to be improved.

The improvement of the yield rate is a process of patience and time. There is nothing great.

Everyone has experienced the LED era, and LCD is now experiencing.Everyone has confidence.

————————

"There are really good things! Once this project is completed, you will be awesome!"

Seeing the fellow making gestures on the blackboard, both of them understood.

If this project is completed.Its influence is too great.

TFT-LCD was not available in China before. You have made it and filled the domestic gap. Others will praise it.But he definitely didn't realize that the difficulty here is so great.

But is the wafer the same?

How long has China stuck in a micron?

The 1 micron technological breakthrough was proposed by the country during the 531 plan period (1986).It was already in 1997 that the 908 project was put into production.

The two national plans before and after, a full 11 years!Even now, in 1992, it has been more than 6 years.

And more importantly, the limit of this technology is not just to use 2 micron equipment to produce 1 micron chips.In theory, it can even achieve 0.5 microns!

You don’t even need the 909 plan!

Three national plans!

Is it possible to continue to increase, such as 0.25 microns?

Ha ha!

With this technology, China can surpass the world's advanced level in semiconductor technology!

More importantly, is there any development cost for this technology?

All the equipment is existing equipment!

As long as this project is completed, not only Cheng Yongxing, but also Yan Liang and Ye Jing will be famous.There is absolutely no problem with producing three academicians of the Academy of Engineering at once.