This content of chapter includes industrial background, motivation, research objectives and organization of this research for Automatic Material Handling Systems (AMHS) of 300mm Semiconductor Foundry.
1.1 Background and Motivation
The semiconductor technology is always continuously progressing forward larger wafer scale and smaller line width. The wafer scale has already moved from 200mm to 300mm since several years ago. The advance to 300mm semiconductor manufacturing is expected to reduce the manufacturing cost up to two-third, as depicted in Table 1 [1]. Comparing to the operations in 200mm semiconductor wafer manufacturing, a 300mm fab demands highly automated operations in both processing and material transfer for creating more cost effective affect. Seamless collaboration is crucially needed between automated processing and material transfer operations to optimize equipment utilization as well as product cycle time.
Table 1. Wafer Cost Comparison between 200mm vs. 300mm Wafer Fabs
200 mm 300mm
Processed Wafer Cost 1723.0 2247.7
Processed Wafer Cost / cm^2 1.4 0.8
1. Data from International SEMATECH Wafer Cost Comparison Calculator, Nov. 2003.
2. The process flow is an International SEMATECH 130nm logic Cu flow (7 metal layers, STI, 23 masks with five 193nm levels) for both 200mm and 300mm fabs.
3. Unit: US dollars.
Wafer foundry services demand production of short cycle time and on-time delivery in order to satisfy customers’ requirements. Due to occasional process changes and pilot or risk production, semiconductor manufacturing suffers from frequent process experiments or inspections. A lot will be granted as high priority, named as Hot Lot or Super Hot Lot, for process characterization, or design validation before releasing a new product for production, or customers’ special request. Hot lots are very important to both fab operations and services to customers. Operations of hot lots can be either preemptive against regular operations, or resource-reserved for no-wait services. Such an effect is usually expected to deteriorate in 300mm
semiconductor manufacturing due to highly automated material handling (AMH) operations involved.
Among the proposed AMHS solutions, overhead hoist transport (OHT) is one of the promising technologies in realizing fab-wide automatic tool-to-tool transportation.
We adopt OHT as our study vehicle to 300mm AMHS.
The increased size and weight of 300mm wafers, foundry manufacturing must face the following priority challenging problems when transiting to 300mm operations:
1. Pilot or risk productions are more frequent due to the increased variety of new products with small production volume. Pilot or risk production purposes are also usually given to higher priority than normal production wafers.
2. Process experiments or inspections are also more frequent needed due to occasional process changes. Wafers for process experiment or inspection purposes are usually given to higher priority for processing.
3. Manual operations are still needed for frequent process fine tunes and logistic support. Both automatic and manual operations would exist in a same time in 300mm foundry fabs.
4. Frequently lots blocking happens within the intraby. Since the OHT moving speed is fast, when OHT exercises a hoisting work, the other OHTs after the hoisting one will be blocked. This issue is a big impact for the delivery time of hot lots.
In addition to hardware limitation, OHT dispatching policy is also crucial to OHT delivery time, which is one of the important metrics for evaluation on intrabay transport efficiency as well as production delivery time. Otherwise, it is difficult to predict production cycle time and lot scheduling. Although many researchers and practitioners have paid lots of efforts to cycle time control and management [2], [3], [4], [5], it is still challenging in precisely determining the production cycle time.
Due to the complicated dynamics of a wafer fab, the estimation of cycle times usually relies on empirical experiences, historical data analysis and statistical projection [4], or computer simulation [5]. Human experiences are straightforward but difficult to explain their induction process, and heavily depend on the decision makers.
Statistical inferences based on historical data are more analytic but still questionable because of highly-coupled interactions among lots. Computer simulations are either too complex to model fab operations as well as the whole AMHS, or too much time-consuming to simulate with a full-scaled fab model.
Since 200mm semiconductor manufacturing era, automatic material handling
systems (AMHS) have played an important role in both the interbay lot delivery as well as the management of in-process inventory. Lot transportation time in the interbay AMHS becomes a non-neglectful factor to the production cycle time.
However, it is either unknown or difficult to predict the transport time in the complicated AMHS. In order to eliminate unnecessary transport delays in AMHS, hand-carrying is sometimes adopted to speed up the transportation of lots. In 300mm semiconductor manufacturing, the capability of automatic tool-to-tool delivery is considered as a must [6], [7]. Lot transport time between consecutive operations can be no longer neglected in such a fully automated operational environment. Seamless collaboration is expected between lot scheduling and material transfer to optimize equipment utilization and product cycle times. The less the OHT delivery time, the shorter the production cycle time. However, all of these researches focus on the wafer processing operations only and none on the wafer transport operations.
The 300mm automatic transport system should differentiate its services to different priorities of process requirements in order to cope with frequent process changes and fine tunes, and the small production volumes of products. The AMHS management should provide hot lots services to meet the operational requirements of lots of different priorities. Accurate forecast to production activities is crucial to streamline semiconductor fab operations. Moreover, an effective solution methodology is eager to determine lot transport time in 300mm AMHS. Operations for hot lots should be optimized by effectively integrating shop floor control functions, like lot scheduling and dispatching, with differentiated OHT services. However, the hot lots services and delivery time estimation of OHT are not realized in 300mm foundry fab, yet.
1.2 Research Objectives
There are two objectives of this paper. One purpose of this paper is to develop an effective OHT dispatching method, Preemptive Priority Policy (PPP), to provide the transport services for hot lots against the normal lots transport requirements under 300mm frequently blocking transportation situation. The objective of this rule is to minimize the transport delay of hot lots and to convey our idea for hot lots transport services; a simulation model is built based on realistic data from a Taiwan 300mm foundry fab.
The second objective of this paper is to propose a modular-like approach, Modularized Simulation Method (MSM), for OHT delivery time forecast to lots of various priorities in 300mm AMHS. Lot delivery time within an OHT loop is estimated by simulation and statistical techniques. A simulation model is built based
on realistic data from a Taiwan 300mm foundry fab to simulate the effects of MSM.
We then estimate the lop-to-loop delivery time by adding all the forecast delivery times of each OHT loop, along the transport path.
1.3 Research Process and Organization
The remaining of this paper is organized as follows. Chapter 2 is the literatures review, which introduce the 300mm OHT and survey the related papers. Chapter 3 proposes the OHT dispatching method for differentiated priorities, PPP. Chapter 4 formulates the OHT delivery problem and details the MSM method. Experiment designs and simulation studies based on realistic data from a local 300mm production fab are described in Chapter 5. Chapter 6 analyzes the experiment results. Final, in Chapter 7, conclusions are made with some future research directions.