Industry Insights: Lithography's need for partnerships
09/01/1999
 Moshe Preil, Sam Harrell |
Lithography continues to be the largest capital investment and the largest operating cost component of leading-edge fabs. The increasing acceleration of the industry roadmap is driving lithography to ever smaller feature sizes, while next-generation lithography (NGL) technologies continue to remain out of reach for production. As a result, optical lithography remains the technology of choice for the foreseeable future, and production fabs have been forced to lower k1 factors beyond levels previously considered possible or practical (resolution equals k1 times the wavelength of exposing light divided by the numerical aperture of the exposure tool).
Unfortunately, there is no k1 knob that an engineer can turn to get smaller features. Low k1 lithography simply means doing more with less: trying to print sharp features from fuzzier aerial images or with more exotic techniques.
 Wafer CD error vs. reticle CD error (MEF). |
Where once conventional wisdom dictated that k1 must be at least 0.7-0.8 for production and 0.5-0.6 for R&D, some advanced production fabs today are running at k1 values of 0.4, and values at 0.3 and lower are being seriously considered for coming generations. Fabs are being forced well beyond their comfort zone, with dramatic changes in engineering complexity and economic risk.
The lower k1 does not come from one single source, but from a range of contributors. The error budget for each part of the process has been tightened, and the advantages of wavefront engineering are being exploited heavily. Both of these factors lead to an incredible increase in complexity and cost, but they have bought a lower tolerable value of k1 .
The latest SPIE International Symposium on Microlithography provided many indications of just how high the cost of low k1 lithography will be. The dreaded MEF or mask error factor was the subject of numerous presentations. The figure shows the MEF problem and how, as k1 shrinks, a small percentage error in the critical dimension (CD) of the reticle becomes a sharply higher percentage of wafer CD error. CD control is the first casualty of low k1, both technically and economically.
Features drawn on the reticle are distorted by the MEF, by optical proximity correction (OPC) and phase shift structures, by the individual lens distortion signatures, and by the resist and etch processes. Predicting the final results at the wafer becomes a complex systems-engineering problem. Each of the "levers" that control the CD becomes ever more sensitive. Small errors in focus, reticle dimensions, or process parameters can have a large impact on device yield and performance, requiring more precise control over all of the levers than ever before.
In addition, more attention must be paid to evaluating the economic risks and error budgets associated with the entire process. From an economic perspective, the industry has been forced to build its prime direction around DUV technology and drive more of the technical difficulty into the mask shop. This will require increased control of total reticle CD performance and intelligent monitoring of lithography and etch processes. It is no longer enough for fab engineers to deliver a specified CD measured at a few test points on a small number of wafers. Systems must be implemented to ensure that the final electrical performance of the entire device meets the desired targets for profitable operation.
The above requires more attention to data integrity: ensuring that there is acceptable correlation between CD measurements of the reticle, of the wafer features after lithography and etch, and of the final parametric performance. Simulation of the interactions of small reticle errors and design tolerances on the process window will be critical to understand the full impact of reticle quality on device yield.
The economic and technical trade-offs taking place in lithography today make the reticle a limiting factor in both cost of ownership and technology development. The industry has traditionally pursued a one-size-fits-all approach, assuming that each new technology will be the correct solution for DRAM, MPU, and ASIC fabs. With the skyrocketing cost of reticles, this approach may no longer be valid. Reticle costs that are manageable for MPU and DRAM fabs may well be too expensive for ASIC fabs, especially for low-volume parts.
Even for MPUs, the costs of prototyping complex devices are becoming prohibitively expensive. New approaches to questions of lens reduction ratio, field size, field stitching in the prototype phase, and even direct writing of prototypes must all be considered to achieve an economically viable model for future development.
These same economic and technical trade-offs also must be considered in selecting the technology that will succeed optical lithography. It is not enough to have a few parts of the puzzle; all of the parts must be in place and coordinated or the system will not work. The reticle has clearly emerged as a key gating item in selecting the best candidates for NGL.
Despite years of intensive research and investment, 1X x-ray has waned as a primary choice on the roadmap because of the lack of a viable mask technology and infrastructure. The winning NGL technology will be the one for which durable, repairable, full-field reticles with adequate CD, overlay, and defect capability are available reliably, in quantity, and at affordable costs.
As the demands of low k1 lithography push economic and technical limits, enormous investments are being made, and will continue to be required, to develop the tools needed to continue the relentless pursuit of smaller feature sizes. KLA-Tencor, like many other leading suppliers, is investing heavily in many aspects of lithography control.
Successful development of complete system solutions will require partnerships between suppliers and leading adopters of low k1 lithography to deliver acceptable reticle quality and lithography module control for sustainable circuit performance. The true challenge will be to develop solutions that are feasible not only technically, but economically as well.
Authors
Sam Harrell is senior VP of strategic business development; Moshe Preil is director of advanced lithography applications at KLA-Tencor Corp., 160 Rio Robles, San Jose, CA 95134; ph 408/875-3000, fax 408/875-2815.