Standardizing the Mask-Maker/Fab Gap
Alexander E. Braun, Senior Editor -- Semiconductor International, 8/1/2000
 |
Now, with subwavelength (Figure), features, data, the correlation between design and silicon, have become very non-linear. This means integration and couplings between designer, mask-maker and fab become even more crucial. Typically, from the paperwork, it does not appear mask-makers get much information about how the fab will use their product, just specs that sometimes must seem as if they were drawn out of a hat.
At the recently held Advanced Reticle Symposium, a question considered — and not for the first time — was how to bring mask makers and users together, particularly since users' main interest seems to lie in getting better reticles faster and cheaper.
Better communication and clearer specs require that a standard of some kind be adopted. More specific information is necessary to quantify a reticle's quality; just tightening mask specs is not the road to QA. Various parameters need controlling, such as CD errors — both global and localized — transmission loss from repair, butting errors, crystal formation, pellicle transmission loss, and feature fidelity errors such as line shortening, edge roughness and corner-rounding.
Once, it was enough to look at CD mean to target, CD uniformity, mask registration and defect size. Then came proximity and linearity effects. As device size shrinks continue, it becomes necessary to add an x and y component under mask registration, and look not only at butting errors, but feature integrity and defect type as well.
Specifications for many of these areas should be discussed and commonly standardized upon. A standard is needed to make the process more efficient, not just by tightening specs but by including more parameters. The mask house and fab must standardize on some production-worthy means of doing outgoing and incoming inspection. One way to do it at present is with an AIMS tool. Although this would ensure that what is supposed to print prints, it is a very time-consuming process. AIMS measurement is not yet sufficiently production-worthy to become the standard for evaluating outgoing and incoming masks. But it points to the technical viability of defect printability analysis as a means of improved communication.
For a mask analysis standard to be effective, it requires the selection of a technically acceptable, provable inspection technology that can incorporate many different individual standards in a quantification scheme that looks at the end result; i.e., will it print or not? This simplification can be attained through innovative combinations of hardware and software techniques. Otherwise, we will have to continue specifying things like corner rounding, the angle, the size, transmission, all of which could reside within the software. The standard would be whether the mask passes an agreed-to test technology, be it AIMS or any viable products already available.
Once the testing technology is agreed to, it might be validated by a standard format, such as a test reticle. Then the industry will have one validated standard instead of many.
Mask makers could ship and accept more masks with a quicker turnaround. In turn, fabs would benefit by minimizing delays caused by rejections, rework, repairs and problems associated with attending to defects that do not matter. Right now, there is too much miscommunication, because each applies a different standard and there is no common link for the activity that must take place for a reticle to be used in a timely way.
As CDs get smaller, masks are not going to get any easier to produce or lithography simpler to do. We no longer have the luxury of flying by the seat of our pants. •
