News & Analysis
Comment
resistion
Inspection-free doesn't count as defect-free!
Hephaestus
Technical solutions exist now for defect-free masks. The litho industry is just ...
EUV litho era begins, but is it ready?
Mark LaPedus
2/1/2011 2:34 PM EST
SAN JOSE, Calif.—ASML Holding NV has recently shipped the world's first ''pre-production'' extreme ultraviolet (EUV) lithography tool to a customer, reportedly Samsung Electronics Co. Ltd., sources said.
It's unclear if EUV is ready for prime time, as the power source and other technologies remain behind schedule. The throughput remains a problem, as the tool is running only 10 to 12 wafers per hour—far below what is needed in high-volume production fabs, sources said.
In its recent results, ASML said it shipped the first of its second-generation EUV systems, the NXE:3100, to an undisclosed customer manufacturing site. The NXE:3100 will offer a resolution of 27-nm with a numerical aperture (NA) of 0.25, overlay of less than 4.5-nm and a throughput of 60 wafers per hour.
A ''pre-production'' EUV scanner from ASML runs about 60 million euros, or $86.9 million, per unit. Some say that price tag could hit $125 million when ASML ships a production-worthy tool.
Five additional NXE:3100 systems are in various stages of buildup in ASML's cleanroom in Veldhoven, the Netherlands. All of these machines will have been shipped to customers by mid-2011.
Samsung reportedly obtained the first tool. Future customers include Intel, Toshiba, TSMC, Hynix, and IMEC. ASML declined to comment on the customer and power source issues.
''We did ship the first NXE:3100 system at the beginning of Q410. It was installed at the customer's R&D site and did expose wafers before the end of the year. We haven't disclosed the customer,'' according to a spokesman for ASML. ''Exactly where customers choose to insert EUV is of course their decision and based on a number of factors.''
''Updates on source power will be given by all parties (ASML, Cymer, Ushio/Xtreme, Gigaphoton) at SPIE,'' the spokesman said.
For its part, Samsung hopes to produce DRAMs at the sub-20-nm node with EUV. Hynix and Micron are also looking at EUV for similar reasons. On the logic front, IBM Corp.'s ''fab club'' is looking to insert EUV at the 14-nm node.
If EUV is not ready, IBM also has the option to move to 193-nm, double patterning, plus various computational lithography techniques, said Gary Patton, vice president of IBM's Semiconductor Research and Development Center, in a recent interview.
Patton said there are still concerns about the power source, resists and defect-free mask technology for EUV.
ASML's rival, Nikon Corp., believes EUV will not be ready for the 22-nm node. Nikon has devised two EUV alpha tools. One is installed at the company's headquarters, while one is running in Selete, a Japanese R&D organization.
Nikon believes the world will use 193-nm immersion and double-patterning at 22-nm. It is shipping the NSR-S620D for 32-nm double patterning, with extendibility to 22-nm applications. It competes with ASML's 193-nm immersion tool, dubbed the TWINSCAN NXT:1950i.
It's unclear if EUV is ready for prime time, as the power source and other technologies remain behind schedule. The throughput remains a problem, as the tool is running only 10 to 12 wafers per hour—far below what is needed in high-volume production fabs, sources said.
In its recent results, ASML said it shipped the first of its second-generation EUV systems, the NXE:3100, to an undisclosed customer manufacturing site. The NXE:3100 will offer a resolution of 27-nm with a numerical aperture (NA) of 0.25, overlay of less than 4.5-nm and a throughput of 60 wafers per hour.
A ''pre-production'' EUV scanner from ASML runs about 60 million euros, or $86.9 million, per unit. Some say that price tag could hit $125 million when ASML ships a production-worthy tool.
Five additional NXE:3100 systems are in various stages of buildup in ASML's cleanroom in Veldhoven, the Netherlands. All of these machines will have been shipped to customers by mid-2011.
Samsung reportedly obtained the first tool. Future customers include Intel, Toshiba, TSMC, Hynix, and IMEC. ASML declined to comment on the customer and power source issues.
''We did ship the first NXE:3100 system at the beginning of Q410. It was installed at the customer's R&D site and did expose wafers before the end of the year. We haven't disclosed the customer,'' according to a spokesman for ASML. ''Exactly where customers choose to insert EUV is of course their decision and based on a number of factors.''
''Updates on source power will be given by all parties (ASML, Cymer, Ushio/Xtreme, Gigaphoton) at SPIE,'' the spokesman said.
For its part, Samsung hopes to produce DRAMs at the sub-20-nm node with EUV. Hynix and Micron are also looking at EUV for similar reasons. On the logic front, IBM Corp.'s ''fab club'' is looking to insert EUV at the 14-nm node.
If EUV is not ready, IBM also has the option to move to 193-nm, double patterning, plus various computational lithography techniques, said Gary Patton, vice president of IBM's Semiconductor Research and Development Center, in a recent interview.
Patton said there are still concerns about the power source, resists and defect-free mask technology for EUV.
ASML's rival, Nikon Corp., believes EUV will not be ready for the 22-nm node. Nikon has devised two EUV alpha tools. One is installed at the company's headquarters, while one is running in Selete, a Japanese R&D organization.
Nikon believes the world will use 193-nm immersion and double-patterning at 22-nm. It is shipping the NSR-S620D for 32-nm double patterning, with extendibility to 22-nm applications. It competes with ASML's 193-nm immersion tool, dubbed the TWINSCAN NXT:1950i.
|
|
|
|
|
Navigate to related information
resistion
2/2/2011 1:50 AM EST
Can't claim EUV litho era has started, when no infrastructure exists. 2015 doesn't look feasible anymore. Device damage also hasn't been covered yet.
Sign in to Reply
double-o-nothing
2/2/2011 2:19 AM EST
So if I understand it right, the tpt of the preproduction 3100 is 10-12 wph instead of 60, well short of spec.
Sign in to Reply
resistion
2/2/2011 4:17 AM EST
I knew something else read funny. Didn't TSMC get first NXE:3100 before Samsung?
Sign in to Reply
EUV OR BUST
2/2/2011 4:44 AM EST
@t resistion.
Samsung is the first customer of the NXE:3100
Sign in to Reply
resistion
2/2/2011 8:42 AM EST
Ok, then this would be an important data point. Checking the old news, TSMC declared their interest in early 2010 but did not plan to get it in 2010.
Sign in to Reply
EUV OR BUST
2/2/2011 9:14 AM EST
Almost all customers shared interest in euv in 2010:
Delivery of the NXE:3100 is another story and only some people at asml knows when delivery of the next system will start.
All customers want their tools as early as possible
Sign in to Reply
mark.lapedus
2/2/2011 12:01 PM EST
Is industry in trouble, given EUV is late to the party?
Sign in to Reply
double-o-nothing
2/3/2011 9:32 PM EST
Since memory structures are approaching 4F^2 shape factor, this can be accommodated by crossed line patterning with spacers. For logic, the use of incrementally increasing exposures is already the default plan. So the industry will go on as usual. Introducing EUV at 22 nm is already much harder than the original plan to introduce at 100 nm. But it's already too late for 22, and even 15 nm.
Sign in to Reply
SiFarmer (Ret.)
2/2/2011 1:19 PM EST
A bit more details please. What's the NXE:3100's laser source and lens material?
Sign in to Reply
SiFarmer (Ret.)
2/2/2011 4:09 PM EST
'scuse the dumb questions. Answers: plasma source: drops of molten tin+CO2 laser (Wikipedia); no lenses -- mirrors+vaccuum.
Sign in to Reply
EUV OR BUST
2/2/2011 2:49 PM EST
Industry is not in trouble, given the fact EUV is a little bit later than planed.
I think economic situation is a bigger problem for the future, when it's down there are only a few company's that wil invest in R&D.
Sign in to Reply
EUV OR BUST
2/2/2011 2:55 PM EST
http://www.asml.com/doclib/press/asml_20101006_SEMICON_West_2010_ASML_EUVL_final_handout.pdf
Sign in to Reply
SiFarmer (Ret.)
2/2/2011 4:09 PM EST
Thanks.
Sign in to Reply
resistion
2/2/2011 7:20 PM EST
The existing litho technology can be iterated indefinitely as needed. So the motivation for alternatives is only cost, which is tricky.
Sign in to Reply
Vladislav_L
2/4/2011 7:13 PM EST
First. I see, that the future of EUV is adequate beyond 10-14nm resolution. Indeed, if we use short, one-pulse laser many-beams lithography tool plus nonlinear sofisticated resists as a part on wafer image formation process we will have this.
Second. We will use non-traditional logic before charge transmission, Imean tha the wawe function will going between some logic elements based on QD. The nuts of this I see that we can use one step lithography on layer of metallization. Otherwords, we can make fully flat logic. The idea this logic based on researrchers form France, but realisation, as I suppose, fully new.
Sign in to Reply
unknown multiplier
2/4/2011 10:04 PM EST
As mentioned earlier, EUV was originally intended for 100 nm and now can only be targeted for 10 nm or less. This is an order of magnitude or more tighter requirements. But the tools did not improve an order of magnitude. At the same time, the 193 nm wavelength was able to cover this entire range. So the relevance of wavelength selection is disappearing quickly.
Sign in to Reply
resistion
2/4/2011 10:50 PM EST
Indeed, the unpolarized 13.5 nm wavelength is not sufficient for 10 nm half-pitch.
Sign in to Reply
yalanand
2/5/2011 10:21 PM EST
@Mark,
When you say "these machines will have been shipped to customers by mid-2011" you mean production worth tool or tool will be sent to R&D site ?
If its R&D then I think there is still long way to go before EUV era begins.
Sign in to Reply
Hephaestus
2/8/2011 11:34 AM EST
Technical solutions exist now for defect-free masks. The litho industry is just slow (as usual) in recognizing this fact.
Sign in to Reply
resistion
2/8/2011 12:16 PM EST
Inspection-free doesn't count as defect-free!
Sign in to Reply