Mask technology challenges and 230-mm reticles
08/01/1998
Mask technology challenges and 230-mm reticles
Brian J. Grenon, Grenon Consulting Inc., Colchester, Vermont
Semiconductor lithographers have for years been pressuring mask fabricators to provide larger, 230-mm (9 in.) reticles for the purpose of increasing productivity. However, when faced with the choices of higher reticle quality, tighter mask specifications, optical proximity correction (OPC), and phase shift masks (PSM), the desire for 230-mm reticles becomes a low priority, according to a recent survey.
We conducted a 15-question survey during the 4Q97 and 1Q98 to ascertain the relative importance of larger reticles to the semiconductor industry. Participants in the survey represented captive mask shops, commercial mask shops, and large semiconductor manufacturers in the US, Japan, and Europe. The participants represented a good cross-section of the organizations that drive mask technology in the industry. The survey results provide insight into some of the key issues facing the mask industry.
Industry will not be able to move toward 230-mm reticles as it did during the transition from 5-in. to 152-mm reticles, which, for the most part, happened by default. At that time, most of the mask fabrication equipment was capable of handling both sizes of mask substrates. When users were ready for the next size reticles, no large capital expenditures were required. Additionally, the 152-mm mask was easy for manufacturing personnel to handle. On the other hand, the 230-mm reticle will require robotics, as indicated by the respondents to the survey. (All 13 respondents indicated robotic handling would be required.)
 |
The challenges and costs associated with the upgrade of a mask fabrication facility to 230-mm reticle capability might be in the range of $50-100 million. Critical level reticles will be at least three times more expensive than current leading edge reticles; noncritical level reticles will be about 1.8 times more expensive. For example, a 230-mm resist-coated blank alone will cost between $5000 and $10,000, an increase of more than $800-1000 for comparable 6-in. resist coated blanks. A full set of leading-edge masks can be about $1 million, a difficult cost to leverage, particularly when a company must account for writing errors and other mishaps. These costs would also be difficult to recover because of the low number of wafer runs/mask set, which is projected on the average of 2000.
Survey respondents were asked their top priorities relating to mask technology development. Surprisingly, only one respondent felt that production of 230-mm reticles was the highest priority. Table 1 shows the relative priority of the mask technology challenges facing the industry. Not all respondents provided more than one priority. (Some provided five priorities, in order of importance, others provided only one.)
For example, Table 1 shows that five respondents felt that OPC was their highest priority, and five respondents felt that PSM was the second highest priority. While one could look at these data in different ways, it indicates that OPC, PSM, mask lithography, and tighter mask specifications are the key challenges facing the participants in the survey. The responses indicate that 230-mm reticle development is currently a low priority.
Participants were also asked what areas would limit their ability to fully integrate 230-mm reticle production. Table 2 is a summary of their responses. The number one priority is what the participants considered the biggest technology limiter to fully integrating a 230-mm reticle facility.
While all of the technology items listed in Table 2 are required for successfully integrating 230-mm reticles into production, the preponderance of opinion is that mask lithography (pattern generation) will be the major technical challenge, followed by the ability to obtain quality 230-mm quartz substrates. A closer review of Table 2, however does indicate that all of the sectors of the mask fabrication process are of some concern.
 |
At some point, however, lithographers will have to accept the burden and go the way of 230 mm. Even with optical lithography options losing steam after 2006, the reticle need will remain, and 230 mm will ultimately have to happen. The only way to completely eliminate the need for reticles is to go to direct write.
Chip size will drive this movement. Survey participants were also asked what was the primary factor driving the need for 230-mm reticles. Table 3 shows the responses, and that there will be no one specific reason for lithographers to introduce the larger reticles. Historically, the argument has been that higher productivity could be achieved through use of multichip reticles; however, larger chip sizes will start to have an impact on the decision process.
Participants were also asked when they would be able to write 230-mm reticles and when their respective facilities would be fully integrated to produce larger reticles. Figures 1 and 2 on p. 50 show the responses. While one respondent would be fully integrated by 1999, the majority would be ready after 2001 or were undecided.
 |
Looking at several of the responses to the survey, one can conclude that the highest priorities are the introduction of reticles with resolution enhancement techniques, such as OPC and phase shifting. Higher quality reticles with tighter specifications are also a high priority. The earliest possible date that anyone will be producing larger reticles will be in 1999, and the full transition will take approximately four years. The SIA Roadmap indicates the need for the 230-mm reticles in 2001 for the 150-nm technology node.
One might ask why the deadline for the requirement of 230-mm reticles continues to be put off. Here are the reasons:
No single organization wants to pay all of the cost of the scale up. Development costs for new 230-mm mask systems are high, (lithography, metrology, process, and inspection).
The availability of high quality mask substrates has yet to materialize.
The cost of the larger substrates is too high to provide a cost-effective mask set ($1.0 million mask set).
Mask users need to accept the fact that they will have to pay the bill, that is, they need to help subsidize the scale up through higher mask prices.
There is no universally accepted method for handling the masks.
Pellicles need to be developed for the larger reticle area.
There is no robust repair technique that will allow masks to be repaired with a high assurance of quality. Laser damage/focused ion beam substrate damage will continue to affect mask yields.
Until these issues are resolved, the 230-mm reticle discussions will continue and their introduction will be stretched out; the mask industry has plenty of challenges to meet with a limited amount of resources. Perhaps the best indication of this is seen in the BACUS `98 Symposium on Photomask Technology and Management
Program. Twenty of the 70 submissions to the symposium deal with phase shift masks and optical proximity correction. None were submitted that address 230-mm reticle production.
Figure 1. The date when survey participants expect to have their first 230-mm capability.
Figure 2. The date when survey participants expect their facilities to be fully integrated to produce 230-mm reticles.
Nevertheless, demand for 230-mm reticles will start to take off. The industry will need about 7200, 230-mm reticles by 2000, and roughly 45,000 by 2004. Early demands are based on the need for test reticles for equipment and process debug and set-up. These early volume projections were partially derived from responses in the survey. Early demands for the larger reticles will be less than 2% of total reticle volume.
One or two Japanese companies will lead the way this year, supplying the industry with 200 production quality 230-mm reticles. But this picture can change as technology demands change. Price pressures could push the move to use 230-mm reticles for higher throughput for 300-mm wafers, the need for multichip reticles, or larger chip sizes.
Brian J. Grenon is the president of Grenon Consulting Inc., a mask technology consulting firm. He was formerly the lead mask engineer at IBM`s mask facility in Essex Junction, VT, where he worked for 19 years. He is the co-chair for the 1998 BACUS Symposium on Photomask Technology and Management. Grenon Consulting Inc., 3 Dunlop Way, Colchester, VT 05446; ph 802/862-4551, fax 802/658-8952, email [email protected].