Those who follow our negative reviews will definitely remember that last month we discussed the glue chip made by Apple - M1 Ultra.

Friends who follow our negative reviews will definitely remember that last month we discussed the glue chip made by Apple - M1 Ultra.

At that time, we said that Apple's glue chip was a compromise, because the larger a chip, the more "scrap materials" was wasted around wafer after cutting.

is like a knife 999's cut cake. If this thing is wasted, it will definitely hurt all over.

At this point, I found that some friends began to discuss in the comment area:

Then why do chips need to be made into square? If it is made into triangles or hexagons, will it not cause waste? ? ?

~, I have to say that everyone’s ideas are very creative, , but in fact, doing so will increase the difficulty of cutting and affect the chip’s yield rate.

Once the cost is calculated, it is better to waste some scraps of materials.

So the idea of ​​"special chip" is actually not very good.

But at this time, some friends may continue to find another way:

chip In order to ensure that it is easy to cut and keep the square shape and cannot change, What if the silicon wafer underneath is directly made into a square shape?

Turn "wafer" into "crystal square", and there is no waste when cutting the chip!

yes. . . This method is OK, and it is OK to say that it is OK, and it is OK to say that it is not OK.

But if you want to explain it clearly, I have to ask everyone this question first:

Have you ever seen square cucumbers?

/ First there is a crystal rod to get the wafer

As we all know, the chip is etched by a wafer, while the wafer is made of high-purity sand. . .

Oh no, it is composed of high purity silicon elements.

performs a series of high-temperature reduction reactions on ordinary quartz sand. After chemical purification reactions, we can obtain the high-purity silicon rod as shown in the figure below.

But this is only the first step. Such a silicon rod is composed of polysilicon and cannot be used to produce wafers at this time.

is as shown in the middle of this picture. Because after various rough chemical reactions before, the internal silicon crystal structure framework is uneven.

single crystal polycrystalline ▼

is full of various asymmetric structures.

. In order to eliminate these asymmetric internal disorders, we also need to process polysilicon in to convert it into single crystal silicon that can be used for chip production with stable structure and good electrical performance.

At this point, it is actually about producing serious wafers.

Currently, the commonly used process in the industry is called Chaiklasky method , and it also has a very vivid and popular name - direct lafa , with a market share of about 95%. Our common logic chips, memory chips, , are basically produced using this method.

We can see a rough look at the specific changes in this picture.

First, the polysilicon just obtained just now is placed in quartz and heated to a molten state, and then a single crystal silicon "seed" is implanted.

This melted silicon solution can start orderly growing at the tail of the silicon seeds when it touches the single crystal silicon seeds.

By controlling the rotation speed and the lifting rate, we can obtain cylindrical single crystal silicon rods of different widths and lengths.

Note! It's a rotation!

is still a friend who has no idea. let’s imagine the marshmallow stall on the roadside . The process of pulling the crystal rod with

is similar to the starting method of turning marshmallows.

That is to say, the current mainstream single crystal silicon production process determines that the silicon rod is likely to be round. Anyway, I have never seen square marshmallows.

As for further back, pinch the crystal rod head and remove the tail, and polish the sides smoothly.

Then slice it bit by bit like cutting sausages, and the wafers are born one by one.

/ Cut it out, but it's not over yet?

In fact, since this step, the question about the wafer shape is not determined.

Because although the straight-pull method is the mainstream single crystal silicon manufacturing solution, in fact, in addition to it, there are also solutions such as zone melting method. Theoretically, square crystal rods are still possible for .

But why do you have to use a circle? In fact, the problem with also involves more subsequent design technology to .

After we polished and polished the sliced ​​silicon wafer, in order to formally lithography, we also need to apply photoresist on it.

Generally speaking, the thickness of the optical adhesive film varies from 0.5 to 1.5 um, and the uniformity must be within plus or minus 0.01 um to . The accuracy of

definitely cannot be solved manually. . .

The commonly used solution in the industry now is "swallowing". That is, add photoresist at the center and then rotate the wafer.

Then we can shake the gum away by continuously controlling the speed, and finally we can get a uniform flat layer of photoresist.

That is to say, if we make the silicon wafer into a square shape and then rotate it, some corner glue may accumulate more, while some corner photoresist will be less.

The uniformity of the corners is suddenly greatly reduced. Even if it becomes a square, the corners may still have to be thrown to the end. . .

And, besides the photoresist problem, there is another even more fatal thing.

That is, under everyone's "conventionality", circular silicon wafers have long been the industry standard . The lithography machine, , automatic production lines, etc. corresponding to

are all designed based on the premise of "wafer".

If anyone thinks of the entire square crystal rod now, not only has to face the problem of "pulling out the square crystal rod", but also has to renovate the entire subsequent production line.

So, can the wafer be made into a square shape? OK, but not worth it.

/ Strictly speaking, this does not waste

Look at this, those "waste" areas on the wafer seem to have to exist.

But we just say, is there a possibility that the concept of "waste" is our preconceived idea. Is it possible that the edge part of the wafer itself should be wasted?

In fact, this is the case. In the cutting, chamfering, grinding and other processes of silicon wafer production, a lot of edge stress will accumulate on the edge of the silicon wafer.

This leads to the relatively fragile structure of the wafer edge.

Even if all the edge areas are used to make chips, the yield rate cannot be guaranteed.

So let's say it. . . Don’t look at the wafers being round, the chips being square.

But in a sense, these two pairs of combinations are still very suitable for .

But of course, even after explaining this, there is indeed some waste in the area inside the edge of the wafer - but the waste area is not as big as we imagined.

If you really want to solve this small part of the waste, in fact, the methods you are using are almost the same now.

That is to align with calculus.

Make the wafer bigger and smaller, and the chip becomes smaller and smaller. This square shape is not enough. . . It looks very round when put together?

As for how to make the wafer large, and how to make the chip small while ensuring performance, this kind of big problem should be left to the engineer to find a way. . .