Optimizing an operation for world-class maskmaking
02/01/2007
The semiconductor photomask industry business model has been changing since the advent of merchant shops in the 1980s, and the past 10 years have brought changes of epic proportions. In the 1980s, there were many captive mask operations within major independent device manufacturers (IDMs), as well as many small “mom and pop” shops to service customers that did not (or could not) make their own masks. In the 1990s, rapidly rising technology-development costs forced a consolidation in the merchant sector into several large mask companies whose scale allowed development costs to be amortized over larger mask volumes than were available to most captives (Fig. 1).
 |
Faced with increased development costs and the rise of “fab-lite” wafer manufacturing in the 21st century, photomask companies must redefine what it means to be world class.
Forces of change
Here are the major economic, technical, and logistical challenges affecting the mask industry:
- The mask user base is declining (Fig. 2). The number of companies building wafer fabs at the leading edge has declined, primarily due to the increased cost of building a fab. At 130nm, about 50 companies built wafer fabs. Between 10 and 20 companies will be able to afford 65nm fabs, and the number projected to build fabs at 32nm is between five and 10.
- The user base is relocating, following the semiconductor manufacturing base. In the mid-1990s, half the mask market was in Japan, one-quarter was in North America, and the other quarter was equally split between Europe and Asia (ex-Japan). Almost half the market was served by captive mask companies, which were especially strong in Japan and Europe. Today, around one-quarter of the market is still in North America, Japan’s share has declined to about one-quarter, Europe’s share is around 10%, and over 40% is now in Asia ex-Japan (especially Taiwan). Captive mask companies have almost disappeared in Europe and Japan, but they are a significant factor in Asia ex-Japan, serving about 40% of the market.
- Semiconductor yields now depend on more than just defects. With today’s technology, design-related issues and process variability contribute significantly to yield loss, as well. Design complexity drives mask complexity.
- Thanks to deep subwavelength imaging, today’s state-of-the-art scanner delivers roughly 1000× more pixels per exposure than could be delivered in 1985. In response, mask designers introduced OPC and phase shifting, and mask complexity increased significantly. Scanners also deliver more exposures per hour than ever before, fueling the dramatic reduction in per-transistor manufacturing costs.
- Mask manufacturing costs, driven by high capital expenditures, have increased, but throughputs have decreased. Mask inspection tool costs have increased more than 50% faster than overall manufacturing capital costs.
- Thanks to OPC and phase shifting, the mask is now engineered as an optical element for the wafer-exposure process. On the other hand, the mask is the first physical embodiment of the circuit design and can therefore be considered the last step in design. It is the place where design data meets manufacturing processes (Fig. 3).
Responses to change
Considering their complexity and manufacturing cost, photomasks deliver more value than ever before. In 1985, mask-industry revenues equaled about 3% of semiconductor-industry sales, but the ratio declined to 1.5% by 1995. This trend has continued. Last year, maskmakers’ revenues were about 1.3% of semiconductor-industry sales. A key reason for this change is the consolidation of the mask industry into a few large merchants, resulting in a more cost-effective industry business model.
To examine this evolution, let’s consider the technology imperatives of the captive and merchant mask business models. A captive mask operation is a strategic asset, used to guarantee capacity for critical products or technology development. In a wafer foundry, a captive operation is also a profit center. Sized properly, a captive mask operation can run in a fab-lite mode, running at or near peak capacity, while using merchant suppliers for surge capacity. This mirrors the wafer-foundry manufacturing model with one significant difference: The wafer foundry has historically received manufacturing technology from the IDMs that outsource to it, while the top merchant photomask companies must develop the technologies necessary to build the most advanced masks internally.
Another difference between wafer foundries and photomask merchants concerns geography: A fabless semiconductor company can design in the US and manufacture wafers using a wafer foundry in Asia. The long design-cycle time greatly exceeds the time incurred transferring the design from the fabless company to the wafer foundry. On the other hand, since mask manufacturing is the first step of the manufacturing process, response time counts!
 Figure 3. A schematic diagram of the photomask’s role in design for manufacturing (DFM). |
The mask shop should be close to the customer; the mask set is delivered to the manufacturing process as a sequence of discrete parts with short cycle times and directly impacts wafer-manufacturing cycle time. Thus, world-class mask merchants manufacture in every major region (and most countries) in which semiconductors are produced. Forced to address its own geographical issues, the merchant manufacturer must be able to bridge distances instantaneously via a seamless, invisible manufacturing-execution system that links its global manufacturing network together. World-class merchant maskmakers must deliver the latest technology via a global manufacturing network.
Success factors
The key factors for success in photomask manufacturing include: providing advanced technology, deploying the technology close to global customers, and exploiting economies of scale. Let’s look at these success factors in more detail.
Advanced technology. As noted above, the IDMs generally have not supplied photomask technology to the mask merchants. In the past, the merchant supplier could develop technology in its own shop and qualify the resulting masks at the customer’s site. Recently, however, mask technology has become so complex, development costs have become so high, and time to market has become so critical that shared development throughout the mask supply chain is now necessary. To ensure the mask performs properly as a complex optical element, the mask supplier and mask customer collaborate on mask design and even on lithography strategy.
Shared costs, facilities, and personnel are becoming common in the most advanced mask development environments; one example is the Advanced Mask Technology Center in Dresden, Germany. The most advanced mask companies perform basic research, which supports collaboration with their suppliers. Overall, collaborative development helps reduce time to market and provides a better fit between the mask and the wafer lithography process.
Global reach. In the past 10 years, the Asian photomask market has expanded to comprise more than 40% of the global market. Compared with other locales, Eastern Asia (China, Taiwan, Korea, and Singapore) spans a vast area and consists of multiple cultures, nations, legal systems, and local customs. Coupled with increased mask complexity, the global photomask landscape makes building and staffing mask-manufacturing facilities an exciting, and sometimes daunting, proposition. It demands multinational engineering and management structures and a robust technology-transfer process. If this difficult integration is successful, it will result in a rich mixture of ideas and people driving innovation and customer awareness beyond that achievable with a geographically limited company.
Scale. The wafer-manufacturing toolkit for a given technology node covers a wide range of technologies. For example, a 65nm wafer fab has ArF, KrF, and i-line exposure tools. Roughly 20% of the tools in a wafer fab are unique at each node. The photomask-manufacturing toolkit turns over quickly as well, perhaps more quickly than the wafer-manufacturing toolkit.
A 65nm mask-manufacturing line might have two classes of electron-beam exposure tools and one class of laser pattern generator. While similar generational spans are found in mask fabs for other process modules, some modules demand the most advanced technology for every layer. For example, the same distance-metrology tool may be required on every mask in a set to guarantee registration and overlay. Further, the mask manufacturer must provide redundancy across its equipment fleet to guarantee delivery of masks with short lead times. The capital investment required by this combination of rapid technology turnover and guaranteed delivery is only affordable if the mask company has the scale to spread the cost over a large revenue base.
Design meets manufacturing
Collaboration, sharing, and networking are common threads throughout much of what has been presented. Indeed, collaboration-between companies in the supply chain, across engineering cultures, and national boundaries, for example-is actually the fourth success factor in a world-class mask company, and it deserves separate treatment because it is manifest nowadays as design for manufacturing (DFM).
Nowhere is collaboration more discussed, and perhaps more misunderstood, than in DFM. On a simple level, DFM means adherence to manufacturing restrictions placed on design; if these restrictions aren’t followed, the design may not be manufacturable. A more useful definition might be “the mutual concurrence of both design and manufacturing requirements.” In a global, increasingly distributed (some might say “dis-integrated”) semiconductor business, DFM might become the new framework for integration across disciplines and companies.
The photomask was introduced earlier as both a piece of precision optical tooling and as the transformation of design intent from an abstraction to a physical state. The phtomask is the filter through which all design intent must pass to be realized on the wafer. Figure 3 shows several concrete examples of what DFM interactions mean to the mask manufacturer. Most of the interactions are one step up- or downstream in the design or manufacturing chain. For example:
- Mask capability-precision and repeatability-contributes to litho yield.
- Mask specifications are set by litho requirements.
- From the mask perspective, RET/OPC equals “design”:
- A mask cost model helps RET/OPC determine how much the OPC solution will cost, or if it’s even possible.
- The mask litho transform, the litho model for masks, becomes part of the OPC model and defines the minimum OPC feature size. - Internally, the mask manufacturer has process flows that can be considered part of DFM, including mask manufacturing rules checks, mask and wafer simulation, and OPC application.
- The mask manufacturer’s sole link upstream to design at this time is the possibility of using “design intent” to help define defect criticality. Currently, every polygon on the mask has the same degree of importance, i.e., gate or CMP fill appear the same from a defect standpoint. Proper information transfer can allow the mask manufacturer to tune inspection sensitivity to reflect design intent.
World-class operation
So how does one run a world-class mask operation in the 21st century? It requires four key and tangible actions:
- Develop cutting-edge technology with customers and suppliers so that suppliers can provide the materials and toolkit best suited to the mask company’s needs, and the mask technology plugs into the customer’s lithography process.
- On a global scale, deploy manufacturing technology consistent with the needs of every important local or regional customer base, and bridge large distances with a manufacturing-execution system that is seamless and invisible to the customer.
- Operate on a scale large enough to make rapid technology turnover and manufacturing redundancy affordable.
- Develop and integrate DFM into the manufacturing flow.
On paper, this may seem obvious, even simple. But one of the most challenging aspects to attaining and maintaining world-class status is operational excellence throughout the enterprise. That includes establishing and nurturing a culture in which employees are dedicated to helping customers master their challenges. It also requires a tight integration of global resources, as well as collegial cooperation among employees representing a diversity of cultures. These requirements also are among the changes that have re-shaped the maskmaking industry.
Acknowledgments
The Advanced Mask Technology Center is a joint venture of Advanced Micro Devices, Infineon, and Toppan Photomasks. Peter Buck of Toppan Photomasks articulated the section on DFM and generated Fig. 3.
Franklin Kalk received his PhD in physics from the U. of Rochester. He is CTO of Toppan Photomasks, 131 Old Settlers Blvd., Round Rock, TX 78664; ph 512/310-6521, fax 512/310-6544, e-mail [email protected].