The 300mm Silicon Wafer Transition

at the turn of the century the 200 billion dollar semiconductor manufacturing industry across the globe joined hands and underwent a massive transition maybe the last of its kind that transition they made their wafers larger sounds simple right but the 300 millimeter wafer transition started in 1994 took nearly a decade and cost the industry billions of dollars in this video we're going to look at the semiconductor industry's momentous transition from 200 to 300 millimeter wafers but first the asian armature patreon i'll make it quick early access members get to see new videos and selective references for them before their release of the public it's not a lot of money and i appreciate the support thanks and on with the show throughout the second half of the 20th century the industry sought to grow their wafers about 50 percent each decade without compromising on productivity and cost in the early 1960s the industry used wafers with diameters of about half an inch or 13 millimeters in the 1970s the industry transitioned to three and four inch diameter wafers for the 1980s five inch and six inch wafers then in the 1990s the industry went to the metric system with 200 millimeter wafers or about 8 inches i will generally stick to the metric system when the 2000s came around the 50 growth prediction meant going from 200 millimeters to 300 millimeters or 12 inch wafers it had almost been like moore's law for wafers but why do this for many years intel led and paid for the industry's previous transitions up to larger wafers their primary reason for doing so has been cost a bigger waiver makes advanced semiconductor manufacturing more economically viable to allow for a doubling of transistors on an integrated circuit every two years required a rising amount of work and investment increasing the size of the wafer helped dilute that cost by allowing manufacturers to put more dyes on a single wafer at least theoretically moore's law requires the entire industry to increase its productivity by 25 to 30 percent each year it has been calculated that way for size transitions have historically accounted for four percentage points of that 25 to 30 percent intel wanted another wafer transition to help with the yet again worsening cost situation yet by then new concerns had emerged to give people pause about whether to do this next wafer transition the move to 150 millimeters had been relatively smooth but when the industry left to 200 millimeters in the 1990s things got a little out of hand the transition promised an over 20 reduction in cost per die but arguably failed to live up to those expectations first it took a very long time the longest of any prior wafer transition 200 millimeter processing tools first emerged in the late 1980s but it took five years before the industry produced two hundred millimeter wafers in a significant quantity previously it took three and second the cost got so high that it sapped away almost all of the savings the transition cost an estimated five to 10 billion dollars in 1990 across the whole industry or 11 to 22 billion dollars today that includes the cost of new tools factory investments losses and productivity and so on things were only going to get worse early estimates found that would cost the industry about 15 to 50 billion dollars in 2006 dollars to complete the 300 millimeter transition that is about 20 to 70 billion dollars in 2022 far more than any single company can handle even intel so despite the fact that a 200 millimeter wafer can hold 1.78 times as many dies as a 150 millimeter wafer companies didn't get 1.78 times as many dies per single dollar spent kind of sounds like a riddle right when does 1.78 not equal to 1.78 nevertheless a 300 millimeter wafer can hold 2.25 times as many wafers as a 200 millimeter wafer intel and the rest of the industry's leading manufacturers hitachi ibm motorola samsung and so on still felt a 300 millimeter transition was worth it one of the big reasons why the 200 millimeter transition had not delivered on its promises was a lack of consensus believe it or not the industry could not agree on how thick the wafer should be because of that for a period of time wafer manufacturers had to produce wafers of two different thicknesses 725 and 735 micrometers an unnecessary cost this time the industry hashed it out early in early 1994 nearly a decade before the transition scheduled 2001 completion the industry set up two international consortiums i-300i representing the american european and taiwanese device manufacturers and selite for the japanese manufacturers in july 1994 those two met up at the large wafer summit in san francisco there they quickly decided on the wafer size and thickness they then moved on to the other technical challenges which are formidable moving an entire multi-billion dollar industry from 200 millimeters to 300 millimeters means changing and retooling every part of the semiconductor manufacturing process let us start with the wafers figuring out how to grow cut polish and deliver a pure silicon wafer crystal so much larger than the previous generation of crystals turned out to be one of the biggest challenges of the industry transition today we grow semiconductor-grade silicon wafers using something called the chokroski method i spoke more about it in a previous video but how it works is that you dip the seed crystal into a pool of molten silicon called a melt then you carefully pull it out of the melt while being rotated to essentially grow the crystal maybe it's just me but i can't help but think about an ice cream bar dipped into chocolate a 300 millimeter wafer like texas is bigger a 200 millimeter wafer crystal weighs about 90 kilograms a 300 millimeter wafer crystal clocks in at over three times that 300 kilograms or 660 pounds that's heavier than an adult female polar bear and leads to new headaches for wafer manufacturers crystals are grown with thin necks just a few millimeters wide in order to reduce the occurrence of defects but for the new ones when engineers tried to grow these absolute units with prior methods the crystals consistently broke their necks the larger volume also affects the distribution of heat throughout the crystal if one part of the crystal gets too hot while another part stays too cold then that causes thermal stress and possibly cracking so to be safe you can only pull the crystal out of the melt half as fast as you would for a 200 millimeter crystal this lower pulling rate as it is called limits the wafer manufacturer's throughput meaning you can't make as many wafers in the same amount of time fortunately the half as fast pulling rate is not slow enough to cancel out the crystal's 2.25 times size benefits the 300 millimeter wafer transition occurred at the same time as the semiconductor industry was starting to move on to the 180 and 130 nanometer process node as always this means new design rules and defect requirements for this particular node generation the first micrometer of silicon on the wafer surface had to be completely free of all wafer defects so now the wafer industry found itself with the challenge of making crystals that were simultaneously bigger and purer and a bigger crystal is inherently susceptible to having more defects there are generally two types of silicon defects the first is a vacancy defect this is where you have an empty hole called a void or a vacancy in the silicon crystal structure the second is when extra silicon atoms jam together mess up their crystal structure and create defects these dislocations as they are called tend to be larger which is more of a pain to deal with they might even spread across the crystals with which is a failure mode these defects happen because of the local conditions at the point where the silicon melt solidifies onto the crystal keeping control of those conditions oxygen most importantly of all was extremely challenging and led to the industry considering what are called epitaxial wafers epitaxy is a fancy word that describes a fancy process here you use chemical vapor deposition or physical vapor deposition to produce a layer of flawlessly pure crystal silicon onto the wafer so you don't have to worry so much about wafer growth purities or the sawing possibly causing dislocation defects but the downside is that it is even slower and costs a whole lot too all of these are technically difficult challenges and the manufacturers have to figure them out while also achieving scale and equal or better cost than the status quo semiconductor equipment suppliers faced several expectations when it came to this transition first and foremost they had to handle and process more wafers without proportionally increasing the tool's size to do this many of them started redesigning their products to put non-critical items underneath the clean room floor tool makers also had to align their tools and practices with new industry standards and larger government regulations and the lithography makers on top of all that had to do this while producing new more sophisticated systems for the 180 and 130 nanometers process nodes after the industry standards were set in 1995 the tool vendor started development in 1996 with an eye to rollout in 1998 at the start the industry's big tool vendors estimated it would add about 40 to 70 percent more cost than with a 200 millimeter system this turned out to be an underestimate a 2006 survey found that the equipment industry spent 11.6 billion dollars on the transition nine times more than the last transition this is a lot even for some of the richest companies out there a 300 millimeter fab is just bigger architects need to design for higher floor and ceiling loads utility piping have to be rerouted to give workers more access to items tools have to move some of their guts under the floors but i think the biggest change was how they automated wafer handling wafers are never carried around raw they are stored in things called a front opening unified pod or foop which preserves them in a pristine environment the industry standard 300 millimeter food carries 25 wafers fully loaded it weighs about 9 kilograms companies had already started to notice a rise in worker compensation claims and worker discomfort complaints with 200 millimeter wafer hoops so the far heavier 300 millimeter version necessitated the automating of basically all wafer transport inside the fab in their 200 millimeter fabs intel moved their hoops around using rails that vehicle could either be autonomously guided or manually pushed to the right equipment bay the 300 millimeter transition brought forth a new technology the overhead hoist transport system or oht moving these to the ceiling allowed intel to squeeze together the equipment base from 9.5 feet to 6 feet freeing up 3.5 percent more space inside the clean room the downside was that these overhead hoists required new software and routing algorithms specifically for the clean rooms later case studies found that manufacturers struggled with the big software requirements involved causing some aberrant behavior by setting industry standards on everything from fab architecture to logistics to software integration the industry was able to eliminate duplicative work the japanese managed to cut 11 different food variants to just one and accelerate wafer fab startup times 200 millimeter fabs at the time used to take 20 months to first hit wafer production after the wafer transition that number was cut down to 18 months a substantial savings the planning and coordination paid off first started in 1994 the industry largely completed its transition by 2001. it has been 20 years since and the industry is still using 300 millimeter wafers talks to move to 450 millimeters the supposed next generation haven't panned out as i talked about in the first wafer video the industry doesn't seem to have the energy to push towards it equipment development costs alone are estimated to be 25 billion dollars and the economic benefits are still murky furthermore it's a way different industry now at the start of the transition the major industry players had been in the united states europe japan korea and taiwan since then china singapore thailand and malaysia had entered the game intel lost his position as the driver of the entire industry there are many other companies now with significant say and their own incentives coordinating between all of them gets harder i don't know how popular this video is going to be but i am fascinated by the entire notion of this decade-long international effort between billion dollar companies who are usually competitors recently there has been a shortage in 200 millimeter wafers fabricated knowledge picked a beautiful quote from a recent earnings call by some co-management one of the big wafer manufacturers there is no way to increase 200 millimeter volume the equipment is no longer available and older facilities that have been idled are obsolete so increasing volume is not possible at the time i had thought this a little weird the market is willing to pay for this why dismiss it so completely but after doing this video i feel more sympathetic to the wafer maker's plight the 300 millimeter wafer transition was a monumental effort these companies moved heaven and earth to make it happen asking them to roll back the clock 20 years is impossible alright everyone that's it for tonight thanks for watching subscribe to the channel sign up for the newsletter and i'll see you guys next time

Channel: Asianometry

Views: 243,521

Keywords:

Id: 0s5TO9h6fco

Length: 14min 59sec (899 seconds)

Published: Thu Aug 04 2022