02.12.2014 Views

Immersion lithographic performance and S610C ... - Nikon Precision

Immersion lithographic performance and S610C ... - Nikon Precision

Immersion lithographic performance and S610C ... - Nikon Precision

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Immersion</strong> <strong>lithographic</strong> <strong>performance</strong> <strong>and</strong><br />

<strong>S610C</strong> production integration at AMD<br />

Rolf Seltmann, Marc Staples, Holger Bald,<br />

Udo Nothelfer, Harry Levinson


Outline<br />

• <strong>Immersion</strong> challenges <strong>and</strong> prior experience at AMD<br />

• Overlay<br />

• Defectivity<br />

• <strong>Nikon</strong> <strong>S610C</strong> in AMD‘s 45 nm manufacturing<br />

• Overlay <strong>performance</strong><br />

• Focus <strong>performance</strong><br />

• Defectivity<br />

• The pattern orientation issue<br />

• Matching <strong>performance</strong><br />

• Product <strong>performance</strong><br />

• Summary/ conclusion<br />

2 | <strong>Nikon</strong> LithoVision | 22 January 2009


Introduction<br />

• The logic 45 nm technology has pitches between 130 nm <strong>and</strong><br />

160 nm.<br />

• For dry tools, this corresponds to a k 1 between 0.31 <strong>and</strong><br />

0.38.<br />

• With immersion we could delay the use of double<br />

patterning <strong>and</strong> a lot of design rule restrictions.<br />

• <strong>Immersion</strong> opens the door to k 1 in the range of 0.40<br />

<strong>and</strong> above!<br />

• But what about defectivity, overlay?<br />

• Some people questioned whether immersion was mature<br />

enough to put into production in 2007!<br />

3 | <strong>Nikon</strong> LithoVision | 22 January 2009


<strong>Immersion</strong> challenges: overlay<br />

Indeed, early immersion tools had noticeable non-linear<br />

signatures:<br />

Advanced dry tool<br />

Early <strong>Immersion</strong><br />

State of the art<br />

immersion<br />

4.0 nm, 3σ 9.0 nm, 3σ 4.5 nm, 3σ<br />

4 | <strong>Nikon</strong> LithoVision | 22 January 2009


But: even with non-linear chuck-signatures we<br />

can achieve great overlay results<br />

• Measure signature difference between scanner for layer 1 <strong>and</strong><br />

scanner for layer 2.<br />

• Signatures are time dependent: S(t 1 ), S(t 2 ), ...S(t n ).<br />

• S = set of overlay error vectors.<br />

• Apply a signature S(t i ) depending on the history of layer 1/ layer.2<br />

exposure.<br />

• This scenario can be extended to match to tools with 2 chucks.<br />

Mixed tool, uncorrected<br />

Mixed tool, corrected<br />

Dedicated tool/chuck<br />

S(ti)<br />

Single digit<br />

numbers in<br />

both cases<br />

5 | <strong>Nikon</strong> LithoVision | 22 January 2009


<strong>Immersion</strong> challenges: defectivity<br />

Is immersion by definition worse than dry?<br />

Not necessarily!<br />

If one has good monitoring in place, <strong>and</strong> if appropriate changes made, even<br />

lower defect counts (at smaller defect sizes) are possible !<br />

The pareto changes but the overall number is even lower!<br />

6 | <strong>Nikon</strong> LithoVision | 22 January 2009


<strong>S610C</strong> <strong>performance</strong><br />

7 | <strong>Nikon</strong> LithoVision | 22 January 2009


Overlay Performance of the <strong>S610C</strong><br />

Local fill technology<br />

promises excellent overlay<br />

However, < 5 nm overlay numbers required<br />

more:<br />

• Interferometer improvements (filter<br />

algorithm).<br />

• Improved airflow (more homogeneous).<br />

• Reduced environmental effects (chamber<br />

ceiling, relocated temperature sensor, ...).<br />

Overlay-residuals, contact to gate<br />

Interferometer, airflow,<br />

lower environmental<br />

effects<br />

7.5nm<br />

5nm<br />

8 | <strong>Nikon</strong> LithoVision | 22 January 2009


Focus <strong>performance</strong>: early S610 data<br />

Edge die leveling is always a challenge!<br />

<strong>Nikon</strong> did a lot of testing <strong>and</strong> optimization:<br />

• Wafer edge autofocus algorithm improvement.<br />

• Stage control mode, data filtering.<br />

• Autofocus mapping route optimization<br />

• + some other small incremental steps<br />

...with big success!<br />

9 | <strong>Nikon</strong> LithoVision | 22 January 2009


Focus <strong>performance</strong> after improvements<br />

before<br />

after<br />

Please note: different<br />

scale on the two maps<br />

3σ = 40 nm 3σ = 24 nm<br />

10 | <strong>Nikon</strong> LithoVision | 22 January 2009


Focus <strong>performance</strong> after improvements<br />

• A 3σ value of 24 nm is an excellent achievement.<br />

• Good mean focus stability (


Defectivity: a basic question<br />

Local Fill Technology shows<br />

excellent overlay, ....<br />

but what about water loss?<br />

?<br />

An unguarded water film is quite challenging <strong>and</strong> requires<br />

a very robust nozzle design.<br />

12 | <strong>Nikon</strong> LithoVision | 22 January 2009


The joint way to low level defectivity<br />

Essential steps toward good defectivity:<br />

• Use material with a high dynamic receding contact angle<br />

• Wafer edge / bevel optimization<br />

• Nozzle optimization is the key!<br />

• Smooth nozzle surface; prevents sticking of particles<br />

• Proper cleaning / nozzle flushing execution<br />

• Strong collaboration between <strong>Nikon</strong> <strong>and</strong> the customer!<br />

13 | <strong>Nikon</strong> LithoVision | 22 January 2009


The joint way to low level defectivity<br />

as it was today target<br />

• At the beginning<br />

of the joint<br />

project,<br />

immersion<br />

defects were<br />

dominating.<br />

• Today, they are<br />

equal or lower<br />

than other<br />

(mostly track<br />

related) defects<br />

14 | <strong>Nikon</strong> LithoVision | 22 January 2009


Pattern orientation <strong>and</strong> its impact<br />

193 nm dry<br />

<strong>Nikon</strong><br />

<strong>Immersion</strong><br />

F<br />

It‘s a one-time effort to get the logistics<br />

done. Not nice but not really a big deal!<br />

<strong>Immersion</strong><br />

Company A • If you also have scanners from Company A<br />

you can‘t share reticles anymore.<br />

F<br />

• Layer dedication requires more complex<br />

factory management systems.<br />

• In a transition phase, there are extra costs for masks.<br />

• Technology start-up, new fabs.<br />

• In high volume manufacturing, dedication by tool type is more<br />

manageable.<br />

15 | <strong>Nikon</strong> LithoVision | 22 January 2009


Can we use tools from different vendors for the<br />

same product? The overlay question<br />

• For manufacturing flexibility it is desirable to run even critical layers<br />

in a mixed tool scenario.<br />

• How much overlay <strong>performance</strong> do we give up if we run in mixed<br />

mode, ultimately mixed between different vendors?<br />

Higher order matching (both field <strong>and</strong> grid) is the key!<br />

• Grids were matched to fab<br />

reference better than 7/6 nm.<br />

• No clear systematics anymore!<br />

Single shot correction as<br />

an additional knob not used<br />

6.8/5.6nm (3σ)<br />

16 | <strong>Nikon</strong> LithoVision | 22 January 2009


Production data in the matched scenario<br />

Field residuals<br />

Wafer overlay<br />

Single machine<br />

S610 – baseline,<br />

initial matching<br />

S610-baseline second<br />

matching iteration<br />

Wafer overlay map uses different scale than field residual maps<br />

17 | <strong>Nikon</strong> LithoVision | 22 January 2009


Matched machine overlay statistics<br />

• RSS addition of the<br />

individual terms<br />

• Field residuals include<br />

the mask contribution<br />

• Other contains wafer to<br />

wafer + correctables<br />

• If matching is done accurately, the <strong>performance</strong> ∆ is just 20% (same<br />

vendor) <strong>and</strong> 30% (mixed vendor) compared to dedicated machines.<br />

• Slightly higher number for different vendor is due to different, noncorrectable<br />

grid <strong>and</strong> lens fingerprint.<br />

• Based on great <strong>performance</strong>, we can run all critical layer<br />

combinations in mixed mode (except one super-critical layer<br />

combination).<br />

18 | <strong>Nikon</strong> LithoVision | 22 January 2009


OPC Matching of <strong>S610C</strong> to existing<br />

45 nm process<br />

• Even if different reticles have to be used for resource <strong>and</strong> simplicity<br />

reasons, it is desirable to use the same OPC-model for different<br />

vendors tooling.<br />

• Matching method.<br />

<br />

<br />

<br />

<br />

<br />

<br />

Expose wafer with same nominal illumination settings on the<br />

reference tool <strong>and</strong> the S610; measure 1D <strong>and</strong> 2D patterns.<br />

Simulate the response CD. Take different pupil shapes into<br />

account.<br />

Calibrate the model using the measurements.<br />

Determine a set of optimized illumination settings for the <strong>S610C</strong>.<br />

Simulate effects of small illumination changes <strong>and</strong> find optimized<br />

illumination settings.<br />

Verify optimized settings.<br />

19 | <strong>Nikon</strong> LithoVision | 22 January 2009


OPC matching result example for 45 nm<br />

metal application<br />

Initial Space through pitch deviation of<br />

S610 from reference tool<br />

S610 Space through pitch deviation from<br />

reference tool after matching<br />

2<br />

6.00<br />

0<br />

4.00<br />

Deviation / nm<br />

-2<br />

-4<br />

-6<br />

Deviation / nm<br />

2.00<br />

0.00<br />

-2.00<br />

-8<br />

-4.00<br />

-10<br />

0 100 200 300 400 500 600 700 800<br />

-6.00<br />

0 200 400 600 800<br />

Pitch /nm<br />

Pitch / nm<br />

• Changes up to 20% on individual knobs necessary<br />

• Final result is noise dominated<br />

20 | <strong>Nikon</strong> LithoVision | 22 January 2009


Electrical <strong>performance</strong> of lots processed via <strong>S610C</strong><br />

• We successfully run 45 nm lots on the <strong>S610C</strong>.<br />

• Most critical layers: contact <strong>and</strong> gate (impact on yield, speed).<br />

• We achieved good yield <strong>and</strong> great speed.<br />

leakage<br />

S610<br />

speed<br />

• Other layers with hard requirements on overlay <strong>and</strong> CD control are<br />

running on the tool as I speak.<br />

21 | <strong>Nikon</strong> LithoVision | 22 January 2009


Conclusion /summary<br />

• The <strong>S610C</strong> was introduced into AMD‘s well-established 45 nm technology.<br />

• In close cooperation between NIKON CORPORATION <strong>and</strong> AMD, all<br />

<strong>lithographic</strong> metrics were improved significantly within the project.<br />

• We achieved world class focus <strong>and</strong> overlay data.<br />

• Defectivity, which required process optimization, was improved to a<br />

level that justifies the use of the tool in high end production.<br />

• Cooperation continues to get world class <strong>performance</strong> as well.<br />

• We could prove very good matching <strong>performance</strong> to our FAB-baseline,<br />

both with respect to overlay <strong>and</strong> OPC.<br />

• However, issues related to the pattern orientation must be addressed.<br />

• We have demonstrated great product <strong>performance</strong> (yield, speed).<br />

22 | <strong>Nikon</strong> LithoVision | 22 January 2009

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!