Lithography cost-of-ownership considerations

David K. Lam, Multibeam Corporation

June 27, 2011 -- The semiconductor industry is facing major cost challenges in patterning advanced ICs in high volumes. A cost-effective solution remains elusive. In this article, we address the important issue of lithography cost-of-ownership.

We start with a cost analysis by Burn Lin of TSMC, published in 2009, followed by updates from the last two years. In his study, Lin projected the cost of lithography at the 22nm node, comparing optical, extreme ultraviolet (EUV) and multiple electron beams (MEB) [1].

The table summarizes the major cost components to arrive at projected costs per layer, all normalized to single patterning (SP) with 193nm ArF immersion (193i) system.

As shown in Table 1, at a throughput of 100 wafers/hour (wph) and excluding costs of EUV masks, infrastructure and power usage, etc., EUV was projected to be comparable to MEB in cost goals. 

Cost-per-wafer comparison for three lithography technologies at advanced nodes [1].
DP=double-patterning; WPH=wafers/hour.
All costs normalized to SP exposure cost/layer Immersion SP Immersion DP EUV goal EUV possibility MEB goal MEB possibility
Normalized exposure tool cost 3628751 3628751 4535939 4535939 4535939 2721563
Normalized track cost 558269 558269 488486 209351 488486 488486
Raw throughput 180 100 100 20 100 100
Normalized exposure cost/layer 1.00 1.80 2.17 10.37 2.17 1.37
Normalized consumable cost/layer 0.91 2.27 1.43 1.43 1.05 1.05
Normalized expsure + consumable cost/layer 1.91 4.07 3.61 11.80 3.22 2.42

 

Update #1: EUV throughput and tool cost

In Lin’s analysis, EUV throughput was assumed to be 100wph. However, the projected throughput of ASML’s most recent EUV system, being installed at imec in May 2011, is 5-6wph [2]. While this is a substantial improvement from previous versions, thanks to higher EUV source power, it is almost 20X slower than the 100wph used in Lin’s calculation.        

Lin also assumed that an EUV tool would cost 25% more than an 193i tool. Given 193i’s current price tag of $50 million, this formula puts EUV tool price at $62.5 million. However, since Lin’s publication, new information has come to light. G. Dan Hutcheson of VLSI Research projected in November 2010 EUV tool cost to be $125 million [3]. This is 2X Lin’s 2009 projection and 2.5X today’s cost of a 193i tool.

Low throughput and high tool cost are challenges for EUV, but EUV is making progress. A 5X boost in EUV source power, for example, could potentially increase throughput to 25-30wph for the $125 million tool price.

Update #2: EUV masks

Mask infrastructure for EUV is still in development. SEMATECH in 2007 estimated that the average critical layer mask for EUV would cost $300,000 and pattern 5000 wafers [4]. This adds $60 per layer. There remain unresolved issues in EUV infrastructure including mask blanks, mask making, mask defect inspection, and mask defect compensation -- a requirement unique to EUV.

Update #3: Complementary e-beam lithography (CEBL)

Complementary lithography, first proposed by Yan Borodovsky of Intel in 2010 [5], uses two lithography technologies to complement each other for better patterning at lower cost.

Multibeam Corporation is developing CEBL for patterning critical layers to complement optical lithography for advanced logic and system-on-a-chip (SoC) manufacturing. In other words, CEBL focuses on cutting poly and metal lines as well as contact and via holes, while optical patterns all other layers. Our version of CEBL is optimized for cutting applications and requires no masks.

Lin [6] and Borodovsky [7] have each stated publicly that electron-beam lithography (EBL) tool price is tied to throughput. They both want 100wph for about €50 million (roughly $71.4 million). This puts pressure on EBL developers to increase throughput and reduce tool cost. Because of application focus, Multibeam's CEBL systems are expected to have a lower cost than generic EBL systems.

It's well known that EBL is too slow for high-volume manufacturing. This is still true if EBL is used to pattern all layers. CEBL plays a limited yet crucial role: patterning only critical layers, which typically have low feature density of ~5%. Our technology also enables e-beam columns to be small in size, fast in beam deflection and arrayable in a multi-column module. The scalable architecture delivers high throughout to complement and extend optical lithography. In addition, support infrastructure for CEBL is available today, thus eliminating major investments in new infrastructure.  

References:
1. B.J. Lin, SPIE Proceedings vol. 7379, pp. 737902-1~11, 2009.
2. A. Steegan, imec, interview with Debra Vogler of Solid State Technology at the Confab, Las Vegas, May 15-18, 2011.
3. D. Hutcheson, “EUVL: and the price is? The answer is...$125 million” VLSI (2010).
4. P. Seidel, “EUV Lithography Cost of Ownership Considerations,” The 2007 International UEVL Symposium (2007).
5. Y. Borodovsky, Lithovision 2010, Feb. 21 2010, San Jose, CA.
6. M. LaPedus, “TSMC tips litho roadmap, backs maskless,” EE Times, (2/27/2009).
7. M. Levenson, “E-beam clusters, curves, and characters,” BetaBlog, (3/8/2010).

David K. Lam, Chairman of Multibeam Corp. at 4008 Burton Drive, Santa Clara, CA USA 95054; email contact (Lynn Barringer): barringerl@multibeamcorp.com; www.MultibeamCorp.com

Subscribe to Solid State Technology/Advanced Packaging.

Follow Solid State Technology on Twitter.com via editors Pete Singer, twitter.com/PetesTweetsPW and Debra Vogler, twitter.com/dvogler_PV_semi.

Or join our Facebook group

MOST READ



SUBSCRIBE

LATEST ISSUE

05/03/2012
Volume 55, Issue 4

Article Archive for Solid State Technology.

© 2012. PennWell Corporation. All Rights Reserved. PRIVACY POLICY | TERMS AND CONDITIONS