The Forgotten Story of How IBM Invented the Automated Fab

In 1970, Costs Harding pictured a completely automated wafer-fabrication line that would certainly create incorporated circuits in much less than someday. Not just was such an objective gutsy 54 years back, it would certainly be strong also in today’s billion-dollar fabs, where the construction time of a sophisticated IC is gauged in weeks, not days. At that time, ICs, such as random-access memory chips, were generally created in a monthlong stop-and-go march via loads of manual labor terminals.

At the time, Harding was the supervisor of IBM’s Production Study team, in.
East Fishkill, N.Y The job he would certainly cause make his vision a fact, just about unidentified today, was called Job SWIFT. To accomplish such a remarkably brief turn-around time called for a degree of automation that can just be completed by a standard change in the style of integrated-circuit manufacturing lines. Harding and his group completed it, accomplishing advancements that would become shown throughout the worldwide semiconductor sector. A number of SWIFT’s groundbreaking technologies are currently prevalent in today’s very automated chip construction plants, yet SWIFT’s exceptionally brief turn-around time has actually never ever been amounted to.

Quick balanced 5 hours to finish each layer of its construction procedure, while the fastest modern-day fabs take 19 hours per handling layer, and the sector standard is 36 hours. Although today’s incorporated circuits are constructed with much more layers, on bigger wafers the dimension of little pizzas, and the handling is a lot more intricate, those variables do not completely shut the space. Harding’s automated production line was actually, really, swift.

A Semiconductor Production Policy

I came across Harding for the very first time in 1962, and wished it would certainly be the last.
IBM was preparing to create its very first entirely solid-state computer system, theSystem/360 It was a rather rough experience. “What the heck good is that?” he shouted at me as I showed exactly how little, unpackaged semiconductor dice can be immediately dealt with wholesale for screening and sorting.

Writer Jesse Aronstein [at far right, in top photo] paused from handling the tools team of Job SWIFT to play French horn one night a week with the Southern Dutchess Pops Band. One more crucial supervisor, Walter J. “Wally” Kleinfelder [bottom left], standing at right, headed the procedure team of Job SWIFT. William E. “Costs” Harding [bottom right], seen right here in 1973, was a brusque WW II fight professional and imaginative trendsetter. He developed and routed IBM’s Job SWIFT, which prospered in producing incorporated circuits in someday. Clockwise from top: IBM/Computer Background Gallery; IBM (2 )

William E. (“Bill”) Harding was an ingenious thinker and innovator. He had actually been establishing semiconductors and their production modern technology at IBM for 3 years when the firm’s brand-new Components Division was created in 1961. Harding ended up being a midlevel supervisor in the brand-new department, in charge of establishing and creating the tools called for to make the System/360’s solid-state gadgets and circuit components.

He was harsh around the sides for an IBM supervisor. Yet probably it was to be anticipated of a person that had actually matured in Brooklyn, N.Y., and was injured 3 times in fight in The second world war while offering as a whole George S. Patton’s Third Military. After the battle, Harding made bachelor’s and master’s levels in maths and physics and ended up being a participant of IEEE.

I signed up with IBM in 1961, originating from rocket-engine advancement at General Electric. Like many designers at the time, I understood absolutely nothing concerning semiconductor production. 5 years prior, I had actually participated in a vacuum-tube electronic devices training course in which the teacher explained the transistor as “a research laboratory interest, which might or might never total up to anything.”.

Job SWIFT inhabited a little area, revealed right here in yellow, in structure 310 at IBM’s vast East Fishkill semiconductor center. IBM

Harding’s harsh and unrefined way appeared each time I went across courses with him. If he ever before mosted likely to IBM.
“charm school” (monitoring training), there was no noticeable proof of it. However, he prospered in his objective. By 1964, solid-state reasoning components for System/360s were streaming from the Parts Department’s brand-new center on a previous ranch in East Fishkill.

In July 1970, I went back to IBM after 3 years of graduate research. I was a first-level supervisor for 4 years before that instructional break, and did not desire one more monitoring task. I desired a totally technological profession, and I signed up with East Fishkill’s Production Study (MR) team intending to obtain one.

Harding and I after that went across courses once again. In mid-August of 1970, he ended up being MR’s leading supervisor. Before that, he invested a year establishing an IBM business technique for the future production and use.
very-large-scale integrated (VLSI) circuits. He was provided command of MR to show the stability of his production principles.

A setting up of MR workers was assembled to introduce the monitoring adjustment. After being presented, Harding explained his sight of future VLSI applications and production. These were his bottom lines:.

• VLSI circuits would certainly be based upon field-effect transistor modern technology (at the time, bipolar-junction transistors were leading);
• Defect-free high returns would certainly be vital;
• Production would certainly be completely automated;
• Finest outcomes would certainly build up from refining one wafer at once;
• Brief turn-around times would certainly provide crucial advantages;
• Quantity would certainly scale up by reproducing effective assembly line.

After the instructional lecture, Harding altered from teacher to leader, General Patton– design. MR’s single objective was to show Harding’s concepts, and continuous tasks not lined up with that said objective would certainly be moved in other places within IBM or deserted. MR would certainly show that an automatic system can be created to refine concerning 100 wafers a day, individually, with high return and a one-day turn-around time.

What? Did I listen to that right? One-day turn-around from bare wafer to complete circuits was what we would certainly currently call a moon shot. Keep in mind, at the time, it generally took greater than a month. Did he actually indicate it?

Harding understood that it was in theory feasible, and he was established to accomplish it. He stated that IBM would certainly have a significant affordable benefit if model speculative IC layouts can be created in a day, as opposed to months. He desired the circuit developer to have testable circuits the day after sending the electronic summary to the assembly line.

One-day turn-around from bare wafer to complete circuits was what we would certainly currently call a moon shot.

Harding promptly arranged a tools team and a procedure team within MR, calling me to take care of the tools team. I did not intend to be a supervisor once again. Currently, unwillingly, I was a second-level supervisor, in charge of establishing all the handling and wafer-handling tools for a yet-to-be-defined production line that I had actually hardly begun to imagine. My desire research study task had actually lasted bit greater than a month.

Walter J. (“Wally”) Kleinfelder moved right into MR to take care of the procedure team. They would certainly choose the item to make and specify the procedure whereby it would certainly be made– the in-depth series of chemical, thermal, and lithographic actions called for to take an empty silicon wafer and develop incorporated circuits on its surface area at high return.

Kleinfelder picked a random-access memory chip, the IBM RAM II, for our presentation. This item was being created on-site at East Fishkill, so we would certainly have every little thing we required to develop it and assess our outcomes about those of the existing nonautomated production line.

IBM’s SWIFT Pilot Wafer Fab Had a Monorail “Taxi”

Integrated-circuit production includes very first developing the transistors and various other parts in their appropriate put on the silicon wafer surface area, and afterwards electrical wiring them with each other by including a slim movie of light weight aluminum uniquely engraved to develop the called for electrical wiring pattern. That slim movie of conductor is called the electrical wiring, or metallization, layer.

IC production makes use of.
photolithography to develop the lots of layers, each with an unique pattern, required to produce an IC. These consist of the steel electrical wiring layers, of which there can be greater than a lots for a sophisticated chip today. For these actions, the steel layer on the wafer is covered with a light-sensitive photoresist product, after which a photo of the pattern is revealed on it. The locations where conductors will certainly be created are obstructed from the light. When the picture is established, the withstand is eliminated from the pattern locations that were revealed, allowing these locations to be engraved by an acid. The remainder of the surface area continues to be shielded by the acid-proof withstand. After engraving is finished, the staying safety withstand is eliminated, leaving simply the electrical wiring layer in the called for pattern.

Job SWIFT at a Look

The Job SWIFT construction line was based upon 5 “fields.” Each was a confined automobile-size maker that did every one of the procedure actions in between lithographic mask-pattern direct exposures. Besides fields one and 5, wafers got in a market with pattern-exposed photoresist all set for advancement and left the market with fresh photoresist all set for the following pattern direct exposure. The taxi communicated private wafers from a market to the lithographic-pattern reveal terminal, and ultimately to the following market in the series.

aspect_ratio

The SWIFT construction line was regulated by a three-level power structure of control systems. At the highest degree, an IBM 1800 computer system took care of the line generally. It interacted with controllers for each and every of the 5 handling fields, within which were handling- and wafer-handling subsectors, some with their very own controllers.

The IC procedure likewise makes use of lithography to develop transistors and various other parts on the silicon wafer. Right here, openings are engraved in shielding layers whereby little quantities of particular pollutants can be instilled right into the revealed areas of pure silicon to alter the electric residential or commercial properties. Making the RAM-II ICs called for 4 different lithographic procedures making use of 4 various patterns: 3 for developing the transistors and various other parts, and one to develop the steel electrical wiring layer. The 4 patterns needed to be precisely lined up with each other to efficiently develop the chips.

Lithography is only component of the IC production procedure, nonetheless. In the existing assembly line, it took lots of weeks to refine a RAM-II wafer. Yet the raw procedure time– the moment a wafer invested really being worked with at different thermal, lithographic, chemical, and deposition terminals– was much less than 2 days. The majority of a wafer’s time was invested waiting to undertake the following procedure action. And some actions, chemical cleansing specifically, can be gotten rid of if wafers proceeded swiftly from one action to the following.

It was the obligation of Kleinfelder’s team to identify which actions can be gotten rid of and which can be increased. The resulting raw procedure time was much less than 15 hours. It after that was up to.
Maung Htoo, my supervisor of chemical-equipment advancement, to check the recommended procedure. His individuals rushed 1.25-inch-diameter wafers via a “pots and frying pans” laboratory arrangement to assess and fine-tune it. The shortened treatment efficiently created functioning circuits in around 15 hours, as expected.

The style of an automatic system appeared. It was at first pictured as a collection of connected makers, each carrying out one action of the procedure, like an auto production line. Yet tools downtime for preventative repair and maintenance of failures needed to be fit. This was accomplished by the insertion of temporary storage space “barriers” that would momentarily keep wafers at picked factors while doing so chain when required.

This procedure chain idea was additional interfered with by factors to consider associated withlithographic-pattern imaging Direct exposure of the photoresist on wafers was frequently completed at the time by a procedure comparable to photo get in touch with printing. The lithographic mask, whereby light shone when revealing the photoresist, was the matching of a photo adverse. Any kind of problem or bit on the mask would certainly cause a matching problem on a chip, at the exact same place, wafer after wafer.

The East Fishkill lithography team had actually established a noncontact 10:1 decrease.
step-and-repeat image projector Consider it as a type of photo slide projector that created a contracted picture having the pattern for a solitary layer on a chip. It after that “tipped” throughout the wafer, revealing one chip place at once. About speak to masking, the stepper assured reduced level of sensitivity to particle contamination, since the dimension of the darkness of any type of roaming bit would certainly be lowered by 10:1. Various other benefits consisted of greater optical resolution and longer mask life.

– YouTube

www.youtube.com

Due to the fact that it was sluggish, however, numerous steppers would certainly be required to satisfy the throughput target. Attaining the most effective pattern positioning on each wafer for numerous pattern direct exposures called for that a wafer be transmitted back to the exact same stepper for direct exposure of each layer while doing so chain. That would certainly terminate the result of picture distortions presented by mild variants from one maker to one more. Constructing the RAM-II circuits after that called for that a wafer make 4 different journeys to its designated stepper. That split the direct series right into 5 fields. A monorail “taxi” would certainly take a wafer from one handling market to its designated stepper, and return later on to take it to its following market.

Each of the 5 fields was pictured to be an unit having every one of the automated wafer-processing and managing tools called for to achieve that section of the procedure chain. The market units and the taxi would certainly be made to offer a clean-room-quality neighborhood atmosphere for the wafers. Within a market unit, generally, a wafer would certainly pass straight from a wet-chemistry component to small heating systems to a photoresist application component, and, lastly, to the taxi pick-up port. Inside the wet-chemistry component, as an example, the wafer would certainly undertake cleansing, advancement of the photoresist and its elimination, and etching, to name a few treatments.

Control of the whole line was to be completed at 3 degrees. Total production-line monitoring, recordkeeping, taxi logistics, and procedure surveillance would certainly be dealt with by a main computer-based system. Committed controllers, one for each and every market, would certainly take care of wafer logistics within the market and feed wafer website traffic and handling information to the main system. The private handling and wafer-handling components inside each market unit would certainly have their very own specific controls, as required, for independent arrangement and upkeep.

Lastly set up, our automated presentation line for the RAM-II chips would certainly contain 5 fields, a taxi, and a lithographic-pattern imaging facility, all handled by computer system. 6 months after Harding took command, MR began to make and develop the real system.

The Bold Center Supervisor Found Motivation in Literary Works

Harding made regular journeys to.
IBM’s headquarters, in Armonk, N.Y., to report development, demand sources, rebut obstacles, and encourage the leading brass that the cash being invested was a great financial investment in the future. It was a challenging objective. His prolonged regular team conferences commonly showed the stress he was under. He talked in detail on points he understood we understood, informed allegorical tales, and rotated examples.

At the time, I did not recognize that he was utilizing his team conferences to create and fine-tune concepts for the discussions at Armonk. He was noting our responses and changing his discussion concepts as necessary. His discussions to the leading brass worked. Throughout of the job, covering concerning 3 years, MR had all the financing and assistance it required to create, style, develop, and run the whole system.

At one team conference, Harding reviewed out loud Heywood Broun’s narrative “.
The 51st Dragon,” to stress the power of a name or motto to encourage individuals to accomplish the difficult. His factor, obviously, was that we required a truly reputation for the job. “SWIFT” was ultimately selected. Harding constantly firmly insisted that it was not a phrase, yet nonetheless individuals figured it was shorthand for “Semiconductor Wafer Integrated Manufacturing Facility Modern Technology.”.

SWIFT’s exceptionally brief turn-around time has actually never ever been amounted to.

SWIFT’s handling and wafer-handling tools was customized made completely within IBM’s Parts Department. The key style purposes were to refine wafers immediately, regularly, and consistently and maintain them tidy and intact. Wafer-handling experiments figured out the cleanest and gentlest methods. Taking care of tools was made to sustain the wafer instead of grasp it. An unique wafer trainer that utilized a circulation of air over the wafer to raise it, without physical get in touch with, was efficiently included for a few of the wafer-transport actions.

There was one exemption to the “tidy and mild” style of SWIFT’s managing device. Monitoring at the Parts Department’s Burlington, Vt., website forced Harding to utilize “air-track” wafer-transport tools that they had actually established. This tools utilized air movement to raise and relocate wafers, just like a puck in a video game of air hockey. Harding required Burlington’s ongoing assistance, so he announced that some air-track tools be utilized in SWIFT. And it was, despite the fact that wafer-contamination and integrity inquiries were unsolved.

One more top-down mandate discusses why SWIFT wound up with 2 various kinds of market control systems– the reverse of great style for maintainability. A customized controller had actually been made, and 5 systems were being constructed (one for each and every market), when HQ called for that we include the freshly introduced.
IBM System/7, which had actually been established especially for factory-equipment and process-control applications. Nevertheless, if IBM itself really did not utilize the computer system in its very own sophisticated assembly line, possible consumers would certainly ask yourself “why not?” Yet if SWIFT utilized a System/7 and the job confirmed to be effective, it would certainly aid market System/7s. Therefore for the 5 fields, SWIFT wound up with 4 customized controllers and one System/7. Both kinds functioned well.

Tools integrity was SWIFT’s Achilles’ heel. To aid accomplish high integrity and convenience of upkeep, specific devices and controls were standard for usage throughout the system, and they were selected for integrity and simpleness instead of uniqueness or beauty. As an example, an individual observing the system in procedure would certainly observe that lots of movements were completed in distinct smooth actions instead of a solitary traverse. Underlying that peculiarity was the considerable use the basic, durable, and trustworthy.
Geneva drive, initially established centuries ago for clocks, today adjusted for direct and rotating movements that needed to be smooth and exactly secured at the end factors. Each conveniently regulated turn of the Geneva drive’s input shaft made one action. Lengthy traverses called for numerous turns of the shaft, causing the odd-looking movements.

Inside a market’s encased chamber, a wafer underwent a collection of completely automated handling actions. 2 of the very early idea illustrations are stood for right here. The wafers entered into the top chamber with a pattern revealed onto the withstand and went through a collection of handling actions that consisted of advancement, solidifying, etching, and others, as suggested.

One more simplification entailed rotating the wafers to centrifugally spread out fluid photoresist that was gone down onto the facility of the wafer. In existing lines, “incorrect spin rate” was regularly pointed out as the root cause of resist-related wafer-processing declines. Rotate rate was gotten rid of as a variable by driving SWIFT’s rewriters with concurrent air conditioning electric motors secured to 3,600 rpm by their 60-hertz air conditioning source of power, equally as phonograph turntables are driven. No rate controllers would certainly be called for. The wanted photoresist movie density would certainly be accomplished by changing the staying variables– temperature level, thickness, and/or rotate time. Ultimately, system integrity was enhanced by the removal of 4 different rate controllers.

As SWIFT proceeded from blue-sky idea to real equipment execution, Harding modified MR’s company and acquired the teamwork of support system. He made sure that his individuals had the sources to do the task and can concentrate on the job. I involved appreciate his business abilities and his capacity to select and hire excellent ability from within the firm.

Harding developed a team to create SWIFT’s master control system, which kept an eye on the development on every wafer as it relocated via the fields. This Implementation Control System (ECS) was based upon an.
IBM 1800 Each wafer had an identification number and was tracked at every action via the line. The ECS saved and kept an eye on each wafer’s handling criteria, identifying and responding swiftly to out-of-spec circumstances. Its strike cards and tape cartridges appear enchanting by today’s procedure, yet it was a significant development in manufacturing control and surveillance for a wafer line.

He likewise moved a whole instrumentation division, handled by Sam Campbell, from IBM Endicott to East Fishkill. Campbell’s division ultimately established innovative techniques for real-time, in-situ procedure control for SWIFT.

A Brief Life yet a Long-lasting Heritage in Semiconductor Production

Mockups of heating systems and chemical cpus were constructed and checked.
Robert J. Straub‘s division in East Fishkill’s Production Design team made and constructed the fields and the handling tools components within them. Harding generated Bevan P.F. Wu to take care of the installment, debugging, and procedure of the line. As tools and centers integrated in SWIFT’s devoted 4,000-square-foot area, Rolf H. Brunner, that had actually handled a great section of the market makes in addition to advancement of the vacuum cleaner metal-deposition tools, took obligation under Wu for tools installment, start-up, and debugging.

Just one procedure in the whole procedure was not completely automated. Placement of the wafer for revealing the pattern on the photoresist still relied on a trained driver. In its last type, SWIFT had both a 10:1 optical stepper and likewise a 1:1 contact-mask maker, yet as it occurred, a lot of the chips created were with the 1:1 maker, since the throughput was greater by doing this.

By the end of 1973, IBM HQ was currently encouraged that complete automation of wafer handling can do well. A lot to make sure that this objective was embraced as a main purpose for a brand-new wafer-processing line to create the circuits for IBM’s next-generation computer system, the “FS” (.
Future System). The recommended brand-new line was referred to as “FMS” (Future Production System), and SWIFT was relabelled “FMS Expediency Line.”.

Bevan Wu efficiently took care of the line’s conclusion, trial run, workers training, and improvements of tools, procedure, and treatments. He brought the line to the factor of being certified to create circuits for IBM items. The system made 5 continuous-operation runs in between mid-1974 and very early 1975. In between runs, his team examined outcomes and applied renovations. The lengthiest constant run extended 12 days. Wafer throughput balanced 58 wafers each day, 83 percent of its made optimum. Ordinary turn-around time from bare-wafer input to testable-circuits outcome had to do with 20 hours. The raw procedure time was 14 hours. The return inevitably amounted to the most effective ever before accomplished by East Fishkill’s standard RAM-II assembly line.

A total amount of 135 service technicians, designers, and supervisors from IBM areas around the world were educated on the procedure of the system. They created 600 product-quality wafers with 17,000 RAM-II FET memory chips.

Yet like his WWII leader, General Patton, Harding was bypassed to lead “the huge program”– in Harding’s situation, the production of the brand-new FMS automated line. Leaving the monitoring profession ladder behind, he was advertised to IBM Other, the highest possible nonmanagement degree in the firm.

The FMS Expediency Line, initially SWIFT, made its last constant run in very early 1975. It had actually completed its purposes. Its individuals were currently required to aid develop the FMS line to create FS computer systems. Yet later on in 1975, the FS job was terminated, and FMS ended up being unneeded. A section of the tools predestined for FMS ended up being Eastern Fishkill’s.
QTAT (Quick Turn Around Time) line, a groundbreaking IBM masterpiece that is much better loved than its odd precursor, Job SWIFT.

Although SWIFT’s life was brief, and it was never ever in the spotlight, its lots of technologies are plainly noticeable in today’s semiconductor fabs. Like SWIFT, these fabs are very automated and computer system regulated; have a main transportation system and “Bernoulli” trainers, which manipulate the circulation of air to raise wafers without making physical get in touch with; use withstand promptly after oxide or steel movie development; usage steppers for lithographic pattern direct exposure; and use real-time procedure control. Every one of these were groundbreaking functions of Job SWIFT half a century back.

The experience of functioning under Harding on SWIFT for 3 years was, for me, transformative. What had actually begun with nervousness finished with affection. I have actually involved think about Costs Harding a real wizard, in his very own method. Stimulated on and sustained by his distinct monitoring design, a little team of devoted individuals accomplished much more than any individual at first pictured. Greater than also we ourselves believed feasible.

We consider the very first up-and-comers in a market as the “papas” of the modern-day personification of their developments. Edison, Bell, Ford, and the Wright bros, are frequently mentioned in this manner. Because feeling, William E. Harding is plainly the dad of the modern-day, automated, billion-dollar fab.