Why are silicon wafers round?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

alimoalem

Diamond Member
Sep 22, 2005
4,025
0
0
just a question cause i'm truly a noob on this subject...can't u have it a circle then cut it into squares? then melt down the excess and make more? or it doesn't work once cut?
 

Wingznut

Elite Member
Dec 28, 1999
16,968
1
0
Originally posted by: alimoalem
just a question cause i'm truly a noob on this subject...can't u have it a circle then cut it into squares? then melt down the excess and make more? or it doesn't work once cut?
No, you definitely cannot melt a semiconductor down. Besides there are many reasons a square piece of silicon would not work in the manufacturing process.

Photolithography is one of them. During this process, photoresist is spun onto the wafer in which the thickness has just a few angstrom tolerance. A pattern is then exposed (and subsequently implanted or etched) into the resist. When the resist thickness changes, it changes the dimensions that the structures end up after exposure. Therefore, it is critically important that the thickness of the resist is very consistant not only across the wafer, but from wafer to wafer. So, anything but a round wafer would just not be very effective.
 

Concillian

Diamond Member
May 26, 2004
3,755
8
81
Originally posted by: alimoalem
just a question cause i'm truly a noob on this subject...can't u have it a circle then cut it into squares? then melt down the excess and make more? or it doesn't work once cut?
You already did the hard part (also known as the expensive pary), which was to create a gigantic single crystal of silicon.

You may as well use as much as you can before having to create another expensive piece of single crystal silicon.

I mean, why cut off and throw away a portion of your (expensive) substrate when it comes "for free"? The better solution is to use what you can out of the substrate by fitting as many chips into the round shape as possible.

It's like if you have a choice of a 12" round pizza or a 9" square pizza for the same price... who's going to choose the 9" square pizza?

Originally posted by: Jeff7181
As seen here. Why round?
Invariably the answer is because that's the cheapest and easiest way to get consistent product. Here is a more extensive link that has significantly more details on the process for those who are interested and have powerpoint (or download a powerpoint viewer):
http://www.usna.edu/EE/ee452/LectureNot...5-Processing_Technology/18_Silicon.ppt
 

route66

Senior member
Sep 8, 2005
295
0
0
So if a glob of silicon is shaped like a Hershey's Kiss, is one part of the glob better then the other? I mean, can it handle higher voltages/frequencies, etc.
 

TuxDave

Lifer
Oct 8, 2002
10,577
3
71
Originally posted by: route66
So if a glob of silicon is shaped like a Hershey's Kiss, is one part of the glob better then the other? I mean, can it handle higher voltages/frequencies, etc.
Yes, there are variations from wafer to wafer, die to die and within the same die.
 

route66

Senior member
Sep 8, 2005
295
0
0
Originally posted by: TuxDave
Originally posted by: route66
So if a glob of silicon is shaped like a Hershey's Kiss, is one part of the glob better then the other? I mean, can it handle higher voltages/frequencies, etc.
Yes, there are variations from wafer to wafer, die to die and within the same die.
Let me clarify -

Is a certain part of the blog always better? For example, is the 'top' always better than the 'bottom' of the goop - or are the flaws evenly distributed? Or do you only use the bottom?
 

Concillian

Diamond Member
May 26, 2004
3,755
8
81
They try to make the "blob" a single crystal with no stacking faults. You will get some errors in the crystal structure, but it will not be predictable where exactly you get those errors.

Making a single crystal that big is not a simple task. Natural single crystals are on the order of fractions of a cubic millimeter, while the single crystals they make for semiconductors are greater than 12" in diameter and several feet long.

My guess is that the variations in processors are probably not going to be primarily due to the base silicon substrate, but on the hundreds of steps of processing that happen beyond the step of making the single crystal silicon. Photolithography defects or distortion, doping issues, unfavorable tolerance stackups, etc... There may be some issues due to the quality of the silicon, but my guess is that on a pie chart of issues, it would be more like go/no go issues than slight variation type of issues. Just a guess though, I work in disk manufacturing, not semiconductor manufacturing.
 

alimoalem

Diamond Member
Sep 22, 2005
4,025
0
0
wow thanks for the answers...so basically this is some hardcore technical stuff with a 2nd grade answer...it would be cool if u could melt it down and reuse the excess u cut off though if it was possible
 

Concillian

Diamond Member
May 26, 2004
3,755
8
81
My guess is that they probably have a recycle loop for their scrap if it's economically feasible. Once they cut the chips out, there's going to be leftover. If it's viable to recycle it (like if they get money back if they do) then it's highly likely that they do.
 

route66

Senior member
Sep 8, 2005
295
0
0
So if you have raw silicon material, how do you create P an N junctions? I thought they were two different kinds of metals.
 

dmens

Golden Member
Mar 18, 2005
1,934
209
106
The silicon is doped to create n-type or p-type regions. Basically, ions are injected into the silicon then annealed to repair the damage.
 

Wingznut

Elite Member
Dec 28, 1999
16,968
1
0
Originally posted by: dmens
The silicon is doped to create n-type or p-type regions. Basically, ions are injected into the silicon then annealed to repair the damage.
Exactly. Picture this...

The wafer is patterned with photoresist (as I mentioned a few posts above), leaving empty spaces. The wafer will be implanted with ions, but the ions won't penetrate through the resist, just where there are spaces. Those spaces are now N wells. The resist is cleaned off and another pattern is put on, this time leaving different spaces clear of resist, and then the wafer is implanted again. This time, those new spaces are now P wells.

This is a pretty good (very short and basic) demonstration of how the chipmaking process works:
Inside the Intel Manufacturing Process: How Chips are Made.
 

ASK THE COMMUNITY