text stringlengths 1 330k |
|---|
Optical lithography, the most important and technologically demanding aspect of chipmaking, is a pillar that won’t be easily toppled. But the technology is at a critical point. The technique, which uses radiation with about half the wavelength of purple light, is fast approaching steep, if not insurmountable, obstacles... |
Nevertheless, one thing is clear. From now on, the relationship between chips and lithography will be two-way. Not only will the fate of chips depend on the future of lithography, but also the reverse will be true. |
Let’s start by defining our terms. Today’s most advanced microprocessors use a 32-nanometer process, and thus are said to be at the 32-nm node. To get a sense of how infinitesimal 32 nm is, consider that to span the width of the lowercase letter l on this page, you would need to bunch together more than 9500 32-nm obje... |
In both logic and memory chips, each of the vast profusion of transistors acts like a switch that allows electrons to flow through the device. A metal-oxide semiconductor field-effect transistor (MOSFET), the kind found on virtually all chips, has three main parts: a source, a drain, and a gate. A voltage applied to th... |
On a chip, that translates to the distance between the parallel metal lines, called interconnects, that carry the electrons through the chip. These interconnects are stacked today in multiple levels, and as many as 10 can populate a chip (a cutting-edge chip could have 10 kilometers of interconnects). The distance betw... |
The half-pitch of the metal lines on the first and densest level is special, because that distance was what once defined not only the half-pitch but also the gate and, consequently, the node. But by 2000, it was a dicey relationship. The half-pitch was becoming bigger than the node. |
So, for example, in 2005 the gate width on an Intel micro-processor was 32 nm. The node was called ”65 nm,” but the half-pitch for the first level of wires was 105 nm. Confused yet? |
The node’s dilemma |
The half-pitch of the first wiring layer is the defining feature for memory chips, while the gate length is the gauge for logic manufacturers. Neither is entirely representative of the node. |
The trouble with the terminology started in the early 1990s, when these gate widths fell down a steep slope [see table, ” Pitch Counts”]. For logic devices, the gate length became the smallest feature, but for memory the half-pitch remained the smallest feature. Those were simpler, happier times. The industry sold micr... |
So it was that, in the second half of the 1990s, the market’s endless appetite for better-performing logic devices drove microprocessors—which used to lag two to three years behind memory in half-pitch—to start closing the gap. Because microprocessor speed was largely determined by the dimensions of the gate, by 2000 t... |
While speed is also a key parameter for memories, there was no similar war between memory manufacturers seeking to drive up the frequency with which transistors execute their instructions, known as clock frequencies. Memory makers focused on reducing the size of each memory cell on their chips so that they could squeez... |
So let’s recap. Today’s cutting-edge 32-nm-node logic chips are actually at a 50- or 56-nm half-pitch . Today’s cutting-edge memory chips, if they were described in logic terms, would be at the 22â¿¿nm node , but they are in fact at about a 34-nm half-pitch, putting them ahead of logic development by a full chip genera... |
The unifying factor is that both memory and logic have always been made with optical lithography. But because memories and logic have different shrink rates, memory makers will have to be the first to make the transition from optical lithography to the lithography technology that will succeed it, called extreme ultravi... |
The death of optical lithography, chipmakers have been known to say, is always seven years away. That joke was new 30 years ago, when chips were being fabricated at the 10-micrometer node. Today’s chips squeeze 4 billion transistors into a space smaller than a postage stamp. The technology that brought about that stunn... |
Lithography is why Moore’s Law endures after 44 years. Nevertheless, for the last 20 years, experts have been uneasy about lithography, which projects the fabulously complex patterns of a modern chip onto a semiconductor wafer using electromagnetic radiation with wavelengths shorter than those of visible light. |
Fundamentally, optical lithography hasn’t changed much in almost 50 years. It has become more sophisticated, but its kinship to old-time film-and-chemistry photography is still discernible. Microchips start out as small blank patches on a silicon wafer about the size of a dinner plate. The virgin wafer is shuttled thro... |
First, the wafer is covered with a thin insulating layer and then with a light-sensitive material called photoresist. Light streams onto the resist through an opaque mask with holes that let light through to form a pattern. This system projects postage-stamp-size patterns onto the wafer until the entire wafer is covere... |
The process has gotten more complicated with each generation of shrinking features. Lithography toolmakers have had to reduce the wavelengths of light they use to project chip patterns through the masks. They’ve also had to find heroic optical tricks to finesse the light into depositing patterns far smaller than the wa... |
The shorter the wavelength, the finer the resolution of the features you can print on the chips and the more transistors you can squeeze onto the chip. The history of semiconductor lithography is essentially the history of the search for stronger and shorter-wavelength sources of light. The first commercial lithography... |
When semiconductor lithography began in the 1960s, the feature sizes of transistors were much larger than the wavelength of the exposure light. To print his original transistors, Gordon Moore actually cut patterns into Rubylith and projected them onto chrome-covered glass plates, or masks, using 16-mm movie-camera lens... |
Lithography had to keep up with feature sizes as they shrank to the size of the wavelengths of light itself—in the hundreds of nanometers—and then, more recently, vanished into mere fractions of the exposure wavelength. |
As each image is exposed onto the semiconductor wafer, the lenses reduce the images 75 percent. Such sophisticated systems can expose more than 140 wafers per hour. Advances of this sort have improved resolution by a factor of about a thousand from the days of Moore and Noyce and the movie-camera lenses. |
Impressive though it is to be printing today’s 38-nm features with 193-nm light, what the industry would like more than anything else is to get back to printing features that aren’t any smaller than the wavelength being used to print them. Here’s why. Light diffracts when it shines through the mask, spreading out and b... |
Because of tricks like water-immersion lithography—which increases resolution by replacing the standard air gap between the lens and the wafer surface with a liquid—chipmakers have been able to print sharp images within these parameters. But that ability has come at a great cost. ASML, Nikon, and Canon have all pushed ... |
Most industry experts agree that features of complementary metal-oxide-semiconductor (CMOS) silicon transistors will continue scaling to below 20 nm. But that won’t happen without extreme ultraviolet lithography. |
Extreme ultraviolet lithography uses a wavelength of 13.5 nm, right near the point where the deep ultraviolet becomes X-rays. If we could harness light at that wavelength, we could continue shrinking features without many of the resolution enhancement tricks we have developed to push 193-nm lithography to the limit. AS... |
ASML expects to ship its first commercial EUV production lithography systems next year. We have already installed two EUV development tools—one at the Albany NanoTech Complex, in New York state, and the other at IMEC, in Leuven, Belgium. These machines produce patterns with a 28-nm half-pitch, better than the so-called... |
These EUV developments are welcome, but they shouldn’t be interpreted as proof that EUV has arrived. The systems are experimental, capable of turning out chips at a rate of a few wafers per hour, much slower than would be needed for a commercial system. EUV still faces significant technical challenges. Consider the con... |
First we boil the tin, and then we drip the liquid tin in carefully timed droplets, synchronized to the firing of the CO2 laser so that it hits each tin drop as it falls. When the laser hits the droplet, the tin is vaporized, and 13.5-nm photons are released. A spherical reflector mirror takes this radiation and channe... |
And that’s just the first challenge. At 13.5 nm, your optics can’t be made of glass, because glass—and air, and just about everything else—absorbs 13.5-nm radiation. You need a good vacuum to prevent EUV light from being absorbed by stray molecules. You need to use mirrors rather than lenses, and that brings up the nex... |
The mirrors in our EUV lithographic system are based on Bragg reflection, a concept used for optical fibers and other waveguides. To create a strong Bragg reflector, you start with a rigid substrate, then coat it with several dozen alternating layers of molybdenum and silicon. The layers must be spaced with a uniform t... |
The EUV development tools use a different light source. They still use the tin droplets, but instead of a CO2 laser, they vaporize the tin drops with an electrical discharge—a small lightning bolt, basically. These have proved rather inefficient. |
One of the many technical challenges of making EUV work is figuring out how good the vacuum has to be to prevent EUV light from being absorbed by stray molecules. |
Who are the potential customers for EUV? Anyone who wants to stay on the road map implied by Moore’s Law: big memory makers that have to shrink relentlessly, the largest microprocessor companies, and chip foundries. |
But the majority of semiconductor companies don’t and won’t need to be at the cutting edge. Consider the chip-foundry business, which accounts for a large and growing share of the overall chip business worldwide. |
Taiwan is home to several chip foundries, the two biggest being Taiwan Semiconductor Manufacturing Co. and United Microelectronics Corp., both headquartered in Hsinchu. A chip foundry fabricates anyone’s chips, on a contract basis. Unlike integrated device makers like Intel, Samsung, and Toshiba, which design and build... |
Soon only a few foundries, many of them in the Far East, will remain to run all these companies’ production lines. Others include Chartered Semiconductor in Singapore and Semiconductor Manufacturing International Corp. in Shanghai, as well as IBM Microelectronics in East Fishkill, N.Y., which makes processors for its o... |
About half the foundry customers need chips that can be fabricated with the previous generation of process technology—currently 65 nm. Another 40 percent of foundry customers use even larger, earlier nodes. At most, only 10 percent have orders that require the most up-to-date chipmaking technology, to keep up with the ... |
The standard sedative to EUV anxiety is the assurance that the gap will be bridged by double-patterning lithography, a technique of last resort that improves the resolution possible with 193-nm light by making two or more exposures, slightly shifted with respect to each other, and with two or more different masks [see ... |
But double patterning is more cumbersome than many people realize. The trouble comes down to design restrictions, cost, and yield. To understand why, let’s go back to the foundry customer. With double patterning, these companies face a horde of new design restrictions. For example, with double patterning at 193-nm wave... |
So, first, companies painstakingly rework their designs to meet these onerous restrictions (no elbows! no zigs! no zags!). Then they’ve got to split the design into two or more parts. There is just no easy way to do that. Different electronic design automation companies are working on that problem, but none of them has... |
Double patterning also demands more process steps, which again adds to the cost by cutting effective throughput, increasing fab cycle time, and adding additional defects. If your original process could yield 100 wafers an hour, using double patterning will leave you with much less. If you’d been turning a profit at 100... |
Then there’s depreciation. Most fab owners keep their tools running 24 hours a day because they depreciate at the staggering rate of several thousand dollars per hour. That’s one of the major problems with double patterning. Cutting your throughput means you’re losing thousands of dollars an hour. Pretty soon you’re ta... |
Memory-chip manufacturing doesn’t require as many masks as logic does. Where memories have only a few wiring layers, state-of-the-art logic has a complicated stack of as many as 10 metal layers. Consequently, a set of masks needed to make a logic chip is going to cost more than a set for a memory chip. Both processes r... |
Memory makers don’t care as much about the cost of the masks because they don’t use as many. For them, a single mask set will make up to 100 million chips. A $1 million mask set, even doubled, works out to about a penny per chip. But for logic chipmakers, the economics are completely different, because the vast majorit... |
If we meet our targets for EUV throughput, EUV is preferable to double patterning because it lets layers be exposed with single masks. Double patterning begins to fail at about 20 nm. Lens designers at Carl Zeiss (ASML’s supplier) believe we can build EUV optics capable of reaching at least an 11-nm half-pitch. |
The principal logic makers are spending hundreds of millions of dollars on R&D every year to keep scaling their transistors, in speed as well as in physical size. Memory chipmakers, on the other hand, are now in glorious pursuit of the ”Grand Unified Memory”—one technology that will do it all, easily taking over for th... |
Is it worth it? That’s the billion-dollar question. Scaling won’t continue forever, if only because we will eventually be down in the atomic realm. |
Right now we’re at 34-nm features. Let’s assume no features can be smaller than the spacing between one atom and the next, which in silicon is 0.546 nm. Within a few years, progress won’t depend so much on making transistor parts smaller. Instead, it will increasingly depend on new transistor designs and on materials t... |
The enormous capital investments required by optical lithography and silicon manufacturing mean that no new technologies will easily displace these workhorses. To take over, any contender technology must build on, and incorporate, the incumbents. |
About the Author |
BILL ARNOLD has been chief scientist at ASML since 1998, which means he’s helping to shape the next generation of lithographic processes. His article, ”Shrinking Possibilities” [p. 26], is about the future of chipmaking, the inevitability of extreme ultraviolet lithography, and how the continued realization of Moore’s ... |
Location: Tigray province, parts of Wollo and Bergemir provinces, northern Ethiopia |
Population: about 5 million |
% of population: about 12% |
Religion: Christian, Muslim |
Language: Tigrinya, minority languages |
The Tigrayans are the chief inhabitants of Tigray province in Northern Ethiopia and in some adjoining areas in Wollo and Bergemir provinces. Seventy per cent of its estimated population of five million are Christians, members of the Ethiopian Orthodox Church, while one-and-a-half million people are Muslims. Eighty per ... |
Background and history |
Tigray is made up of a central highland plateau bordered on the east and west by lowland plains. The highland region has the highest population density in the country owing to its favourable climate, although the western plains have a more fertile soil. The eastern region is the site of the Danakil depression, one of t... |
Tigray has been under the control of various dynasties since the founding of the first Axumite empire in the first century AD. An Amhara emperor led a successful campaign against the Italian invasion of 1896; however the cost of the campaign was high and the country was left in a poor economic state which deteriorated ... |
War with the Dergue |
The Tigrayan National Organization (TNO) was formed in the early 1970s with the aim of improving literacy and promoting political debate, and it also played a part in bringing about the overthrow of the Emperor in 1974. With the ascendance of the new military regime, the Dergue, the TNO opposed military rule, especiall... |
In 1975 the Tigray People’s Liberation Front (TPLF) was formed. Its objective was self-determination for the Tigrayan people. Its ideology had much in common with that of the Eritrean People’s Liberation Front (EPLF); however whereas the EPLF is fighting an anti-colonial battle with Eritrean independence as it goal, th... |
The Ethiopian government launched a series of military campaigns in Tigray. The sixth such campaign, waged in 1980-81, was aimed at disrupting agricultural production and the economy of the densely populated central region. Government troops were heavily armed and were successful in their aim; however the major damage ... |
The TPLF has gained the active support and participation of the majority of the Tigrayan population. Land reforms were implemented and a campaign to promote women’s interests was mounted. The Relief Society of Tigray (REST) has undertaken education, health, agricultural, craft and resettlement programmes and by 1983 wa... |
Drought and famine |
The drought from 1983 in northern Ethiopia was equal to that of 1970-73 in which an estimated 200,000 people died in Tigray and northern Wollo province. According to REST officials, in early 1983 two million people were living in drought-affected areas under the TPLF’s control, of which at least 1.2 million were in urg... |
From early 1988 fighting intensified, leading to large scale casualties. By the end of May the TPLF had gained control of most of Tigray including the historic centre of Axum and the area surrounding the regional capital of Makelle. Government forces counter-attacked, beginning a ruthless aerial bombardment of the main... |
There was now intense pressure on the Addis Ababa regime, by the USSR among others, to make peace with the Tigrayans and Eritreans. There was internal disaffection also which resulted in an attempted coup by military officers in May. On June 5 the government announced that it would be prepared to enter into uncondition... |
(See also Eritreans; Falashas of Ethiopia; Oromo of Ethiopia) |
News: JJT - Jerry needs to commit to Garrett or fire him |
Discussion in 'News Zone' started by Risen Star, Jan 11, 2013. |
1. Risen Star |
Risen Star Likes Collector Zone Supporter |
25,120 Messages |
13,038 Likes Received |
IRVING, Texas -- The sun should not set again with Jason Garrett as coach if Jerry Jones isn't committed to him beyond next season, regardless of whether the Dallas Cowboys make the playoffs. |
Either ride with Garrett or get rid of him. Anything else is a colossal waste of time. |
Hey, the Jacksonville Jaguars fired coach Mike Mularkey on Thursday after one season, so it's not like there's some statute of limitations on firing coaches once the season ends. |
Jerry made it clear he's the person calling all the shots with his soliloquy about the factors that led to defensive coordinator Rob Ryan's dismissal while taping his weekly TV show. |
Monte Kiffin is considered the leading candidate to replace Ryan. Kiffin spent the past three seasons at Southern Cal coaching with his son, Lane, the head coach. Kiffin's claim to fame is creating the Tampa 2 defensive scheme that involves playing the safeties deep and limiting big plays, forcing an opponent to dr... |
The Cowboys' defensive assistants already have been told they will have to interview with the new defensive coordinator. If he chooses to keep them, fine. If not, they'll be unemployed. There's also a possibility Jerry will yank the play-calling duties away from Garrett and hire someone else to call plays. Norv Tur... |
Now, we're talking about the potential addition of two coaches who have considerable more loyalty to the owner than the head coach. |
How dumb is that? |
Disaster lurks in that scenario. The only question is how long it takes before the back-stabbing and infighting begin, especially on a staff led by a dude who's not confrontational. |
2. Aven8 |
Aven8 Well-Known Member |
4,882 Messages |
1,413 Likes Received |
And JJT finally gets it! |
Yes we should have made a play for Payton, and now we are making a bigger mistake not going after Gruden. |
Oh well. At least Jerry isn't boring. I'll give him that. |
3. DallasDW00ds0n |
DallasDW00ds0n Well-Known Member |
3,361 Messages |
356 Likes Received |
You have to kind of agree. But not letting Garrett work these decisions, if thats what is truly happening, what does that say about Garrett-Jones relationship? |
4. Iago33 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.