Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/not-enough-euv-for-intel-7nm.13780/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Not enough EUV for Intel 7nm?

Even if Intel can get EUV tools, TSMC and Samsung implemented EUV in steps. For Intel to think they can go from 0 EUV to full EUV in one node and yield properly is questionable. I certainly hope they can but it is questionable.
 
Even if Intel can get EUV tools, TSMC and Samsung implemented EUV in steps. For Intel to think they can go from 0 EUV to full EUV in one node and yield properly is questionable. I certainly hope they can but it is questionable.
Was gradual increase in EUV layers done because multiple EUVs present a technological challenge or because of the scarcity of the machines?
 
Was gradual increase in EUV layers done because multiple EUVs present a technological challenge or because of the scarcity of the machines?

I would say both. Throughput was also a problem. Those machines do not work right out of the box. A lot of effort was required to integrate EUV into the manufacturing process. There was a lot of yield learning and some secret sauce resulted. EUV is yet another way foundries will differentiate.
 
People are retiring from Intel. We'll see if something happens or not.
 
ASML update from SPIE. More SPIE coverage to come:

ASML shipped 85 systems in 2020. While the NXE:3600 was first announced last year, 2021 presentations will describe further improvements in availability (95%), productivity (2500 wpd) and pellicle transmission (90%), as well as matching as ASML gets ready to ship the system in 3Q21. The modular source vessel also continues to contribute to improvement in availability, while increases in source power continue to improve productivity. Not to mention, consumers can now buy things partially made with EUV from the likes of Samsung and Apple!

  1. Metrology: Akin to big data, it’s about making sense of more
ASML will present a 40x reduction in measurement time using e-beam massive metrology, enhancing customer’s ability to identify the defect-based process window (stochastic process window) when an EUV tool is qualified for a new process before HVM, enabling its continued manufacturability.

Additionally, ASML will discuss ongoing improvements in its alignment sensor technology to deliver further process robustness, as well as improved cross-platform matching. This enables a more complete and accurate picture of EPE to improve manufacturability at the most advanced nodes.

  1. DUV: The overlooked wunderkind
ASML continues to make investments in its DUV systems to further improve productivity and matching. The new NXT:2050 now delivers 295 wph and the NXT:1470 delivers 300 wph. The matched overlay between an NXT:2050 and NXE:3400 is now at 1.2 nm – nearly as good as each individual system.

  1. High-NA: Moves from PowerPoint to hardware
ASML continues to make steady progress in realizing High-NA EUV lithography, presenting new pictures of various modules that are being manufactured and tested.
 
Last edited:
What about pellicles? Are they still being used?
There were several papers on the development of EUV pellicles in their own section at the SPIE virtual conference. Most focused on the use of carbon nanotubes because of their high transmission (98%) and thermal conductivity. ASML has its own pellicle with 90% transmission, it is planning to continue improving with new materials.
 
Never could've imagined it going this way given such effort to move away from multiple patterning. Do you think single patterning still has a chance if people will back off on geometry, or ~500W+ EUV sources will become available?

So at the conference one thing highlighted for high NA was the opportunity to reduce dose, because of better edge contrast, but actually, more importantly, the previous illumination limitation would not be there (unless they scaled proportionally to the same extreme k1). The dose reduction may not be enough to rely on because of stochastic issues. Of course, at the same time, power is also being driven toward 500W.

But single patterning by itself is not a guarantee of better wafer throughput. The clearest example of this is electron-beam (direct write) lithography. It has much higher resolution, is single patterning, and no mask complexity, but all these benefits could not justify the much lower throughput.
 
Back
Top