Tool/Fab Support Strategies

Using measurement tools to assess the health of fab automation systems

Patrick Conarro, Peter Frost, and Gordon Beckhart, Brooks Automation

Today's 300-mm wafer fabs depend on automation. Throughput and cycle-time requirements demand the speed and precision of robots and lot movement systems, while contamination concerns dictate that the presence of people in the fab be minimized. Yet the integration of automation and process tools does not come about without challenges. Engineers must ensure that the reliability of material-handling systems does not affect the ability of the fab to do its essential job—processing wafers. The potential of automation to affect fab productivity is significant; one study revealed that if each wafer-handing robot were to fail just once per 1 million pick-or-place cycles, a fab could suffer a productivity loss of $10 million per year.¹

Other risks exist as well. If a robot drops a wafer, more is lost than just the product. The tool may be damaged, and silicon cleanup always requires extensive downtime. For transport and docking systems, out-of-adjustment loadports can fail to latch front-opening unified pod (FOUP) doors securely, which can cause a door to fall off of a pod. Beyond the obvious risk to production, such an occurrence poses a hazard to fab personnel. If a door drops when its FOUP is on an overhead transport rail, serious injury or even death can result.

Given these risks, it makes sense to keep a fab's automation systems "healthy"—that is, operating as close to ideal parameters as feasible. This means not only keeping the robotics and transport and docking systems within nominal specifications, but also implementing efficient maintenance procedures to minimize the time spent in setting up and adjusting equipment. The benefits of implementing a procedure to ensure healthy automation include increased mean time between failures, decreased mean time to repair, reduced particulation, improved fab predictability, and improved safety.

With the objective of achieving these goals in an operational 300-mm fab, an engineering team from Brooks Automation (Chelmsford, MA) performed a health check on the facility's robotics and FOUP loadports. Prior to the checkup, tool engineers had experienced significant downtime because of automation failures and extended recovery times of up to 48 hours following robot replacement as a result of difficulties encountered in setup and teaching. While many fabs devote considerable attention to the robots that handle wafers directly, all of the components of an automation suite should be considered together, since a misaligned loadport interface can also affect robot adjustment and operation. Accordingly, this study focused on the subsystem consisting of the loadports, shuttle robots, and pass-through elements.

Figure 1: The equipment used to perform the health check: (left) the loadport aligner, and (right) the electronic level.

The health check was performed using a suite of equipment developed by the Brooks MicroTool operation (Colorado Springs, CO). The equipment included a loadport aligner (shown in Figure 1, left), a precision tool designed to measure and capture multiple loadport parameters. The aligner form factor is that of an ideal FOUP. The area that underwent the health check contained six process tools; therefore, the system was used to measure alignment on 24 loadports. A stand-alone electronic level (shown in Figure 1, right) with an accuracy of ±0.03° was used to assess the coplanarity of critical surfaces and robot blades. The team also used this device to measure vibration in critical system components.

Loadport Alignment

The number of loadport interfaces, coupled with the importance of highly reliable automation, makes it imperative for equipment engineers to maintain critical loadport dimensions within SEMI tolerances. With 24 loadports in just one area, on the fabwide level this is clearly a task of significant scope. Variation in loadport designs adds complexity to the situation. A 300-mm production fab might purchase tools from 15 or more different loadport vendors. This means that maintenance engineers and technicians must learn a variety of troubleshooting and setup procedures. In addition, because some loadport manufacturers offer more than one design, maintaining consistent setups throughout the fab can be a complex task even when maintenance is performed by representatives from the vendor.

Prior to the health check, the subject fab had been experiencing problems and failure modes related to loadport performance. Significant problems included docking and undocking errors, and doors dropping off FOUPs that were on an overhead conveyor system. The latter problem had been reported for FOUPs from different manufacturers. In addition, fab engineers and the checkup team agreed that several other issues could be consequences of misaligned loadports, including scratched wafers, wafer movement during robot pickup routines, longer load and unload cycles resulting from multiple prealigner centering iterations, and particulation from wafers dragging on carriers. Based on these concerns, the health check examined nine loadport interface variables:

• Door distance—the distance between the vertical datum plane of the FOUP door opener and a parallel vertical plane at the center of the FOUP that passes through the centroid of the triangle formed by kinematic pins on the load port. If this distance is out of tolerance, problems occur with door removal and placement.

• Frame distance—a similar dimension measured from the surface around the door opener to the parallel plane through the FOUP. Maintaining this distance is important to proper FOUP docking and to effective operation of the FOUP interface mechanical system (FIMS).

• Door pitch—the rotational variance of the FOUP door opener from vertical (rotation around the horizontal axis). Excessive pitch can result in docking interference as the door opener could contact the FOUP door at either top or bottom. It could also result in interference between opener keys and sockets.

• Door roll—the rotational variance of the door side to side (around the vertical axis). Too much roll can create interference between the opener and the sides of a door, or result in problems with key operation.

• Frame pitch and roll—rotational variance related to the door frame. Misalignment here can result in off-plane docking of the FOUP. Excessive pitch or roll can be a significant concern even if the door opener and door frame remain coplanar, since problems may occur with wafer loading and unloading or with alignment of the wafer-handling robot. At the extremes, out-of-tolerance conditions with respect to pitch or roll can cause severe docking, door-removal, door-placement, and door-latching problems.

• Upper and lower pin positions—x- and z-axis placement of the registration pins on the door opener. Out-of-position pins can cause docking interference, an inability to pull vacuum on the door for removal, and problems with key operation.

• Registration pin diameter—pins that are too large can cause registration problems with FOUPs. On the other hand, if pins are too small, a door opener might successfully engage an out-of-position door, causing removal, replacement, or latching problems. In addition, pin diameters must be known in order to calculate relative pin positions.

• Level pitch—the angular rotation of the FOUP platform about its horizontal axis, with the door considered to be at the front. Correct pitch is required for proper docking and door-opener operation.

• Level roll—the rotation of the FOUP platform around the vertical axis. Roll misalignment can cause most or all of the interface problems noted above.

The procedure used for measuring loadport alignment involved placing the aligner on the kinematic locator pins of the FOUP platform, moving the device to its docked position, and taking the desired measurements. Once the device was docked, it took from 2 to 4 seconds to assess 14 dimensions on each loadport. Results were displayed on the device controller.

The checkup team found significant variations in alignment across all the loadports tested. Although the incidence of gross out-of-tolerance conditions was minimal, numerous dimensional readings were beyond the SEMI tolerance limits. None of the six process tools in the area under study was served by a single loadport having all 14 dimensions within specifications.

Table I shows the measurements from the process tool loadport set that exhibited the highest rate of compliance with SEMI tolerances, at 77%. Table II lists comparable data for the four loadports serving another process tool. In that case, only 64% of the measured dimensions were within tolerance. In both tables, out-of-tolerance dimensions are displayed in red.

Figure 2: Percentage of dimensions within tolerance, by loadport.

Figure 2 summarizes results for all of the loadports measured. Overall, the percentage of dimensions in tolerance ranged from 50 to 86%. One loadport set had a very wide variation in tolerances, with rates of 50, 64, 71, and 86%. Fabs must identify such a condition because a wide degree of variation in adjustment across mechanical interfaces is likely to cause problems with the FIMS and wafer-handling subsystems associated with the tool. The availability of alignment data can enable tool owners and maintenance technicians to prioritize their efforts to maintain equipment in calibration.

Figure 3: Percentage of dimensions within tolerance, by dimension.

Figure 3 shows in-tolerance rates from a dimensional perspective. Overall, 67% of all dimensions checked were within tolerance limits. The dimension with the highest in-tolerance rate, at 94%, was level pitch on loadport FOUP platforms, with only one failure seen among the 24 loadports tested. At the opposite end of the spectrum was frame distance, a dimension critical to FOUPs being docked at the optimal plane both for FIMS door-opening and -closing operations and wafer pickup. This dimension was within tolerance limits on only 3 of 24 loadports.

Level Measurement

The purpose of the level measurement phase of the health check was to assess the coplanarity, referenced to earth gravity level, among key wafer support surfaces: the loadports, shuttle robot arms, and pass-through cassettes. Effects of wafer-handling surfaces being significantly off level can include particulation from end effectors scraping wafer backsides, wafers scraping against (or even colliding with) carrier supports, and dropped wafers. A coplanarity tolerance of ±0.2° would maintain a height difference between opposite points on a 300-mm wafer edge of about a millimeter or less. The accuracy of the electronic level used was 0.03°.

Figure 4: Percentage of level measurements within a ±0.2° tolerance limit, by dimension.

Figure 4 presents level results by dimension measured. Overall, 75% of the measurements were within the specified ±0.2° tolerance. As the figure shows, however, roll (rotation about the y-axis) on the single-blade shuttle robot end effectors was a problem, with only 33% of measurements within tolerance limits. Notably, one robot was measured at greater than 1.5° from level, which would result in opposite points on a wafer edge differing in height by nearly 8 mm. Based on these results, this robot in particular, and the shuttle robots in general, would be a high priority for level adjustment by maintenance technicians.

Vibration Analysis

It has been reported that vibration in robot end effectors can contribute to backside wafer damage, particulation, and wafer "walking," and that process tool efficiency can suffer from excessive system vibration.^1,2 In fact, vibration had been an issue for some time in the fab under investigation.

By using a continuous-logging function on the electronic level system, the checkup team was able to characterize vibration (angular deflection over time) on a number of critical points. The team entered the logged data into a spreadsheet containing time and amplitude (extremes of angular position) for both wafer pitch and roll. The data were then normalized to correct for starting position and to center the average amplitude at zero (earth level). Examples of the results are presented in Figures 5 and 6.

Figure 5: Comparative vibration data for two cassettes serving a process tool: (a) the left pass-through cassette, and (b) the right pass-through cassette.

The figures reveal that there was substantial variation in vibration amplitude and frequency among the points measured, with significant peak-to-peak excursions on some surfaces. Figures 5a and 5b depict the difference in amplitude, particularly with respect to pitch, between the left and right pass-through cassettes serving one process tool. The ability to observe and measure such differences gives maintenance engineers the means to assess their significance, then troubleshoot and solve problems where necessary.

Figure 6: Comparative vibration data for two components serving a process tool: (a) a batch-transfer apparatus, and (b) a single-blade shuttle robot end effector.

On a different process tool, an even greater difference in vibration was seen between the batch-transfer apparatus and the single-blade shuttle robot and effector. Figure 6a depicts the relatively quiet vibration signature of the batch-transfer device, with pitch varying only about 0.02° (0.1 mm) peak to peak. Roll varied somewhat more, but not to an alarming degree. In contrast, the single-blade configuration (Figure 6b) vibrated to a significant extent. Both pitch and roll consistently varied between ±0.045°, with a maximum peak-to-peak excursion of more than 0.14° (0.73 mm). Although data such as these may not establish that a problem exists, being able to see differences between components gives engineers direction on where to look if a problem does occur. For preventive maintenance, technicians may be able to identify places where a simple servo adjustment or extra bit of isolation can prevent chronic or hard-to-troubleshoot particulation problems.

Conclusion

In the 300-mm process area that was studied, the health check marked the first time that an integrated set of tools had been used to quantify multiple parameters on loadports and automation. Previously, adjustments had been performed using alignment jigs, introducing operator-to-operator variation. The checkup procedure took about 10 hours, and the results showed that the equipment was in a suboptimal state of alignment. While the checkup offered a one-time snapshot of these conditions, the methodology can provide the basis for long-term studies correlating alignment results with reported problems involving handling errors, scratched wafers, particulation, or excessive prealigner iterations.

When new loadports and pass-through robotics are being installed in a fab, using this methodology during the initial setup can reduce subjectivity and personnel-dependent variation. Once a piece of equipment, such as a loadport, is tuned up to optimum SEMI dimensions, engineers can quickly and easily track drift over time and take timely preventive action before the automation equipment can begin to adversely affect either productivity or product.

References

1. K Janac, "The Expanding Role of Robotics in Process Tool Productivity," Solid State Technology 42, no. 1 (1999): 40–44. Available from Internet: http://sst.pennet.com/home.cfm.

2. P Attaway and Z Kemeny, "White Vibration Filtering for Equipment and Facilities," Solid State Technology 41, no. 6 (1998): 149–159. Available from Internet: http://sst.
pennet.com/ home.cfm.

Patrick Conarro is director of engineering at the MicroTool Alignment Services Division of Brooks Automation (Colorado Springs, CO). A cofounder of MicroTool, he has nine years of experience in the semiconductor industry. Previously, he was a tooling manager, overseeing the development of FOUPs, SMIF pods, and cassette injection modeling design at Empak. A member of SEMI since 1997, Conarro holds four patents in the area of robotics. He received an associates degree in tool design from Williamsport Community College in Pennsylvania. (Conarro can be reached at 719/471-9888 or pat.conarro@brooks.com.)

Peter Frost is the Eastern region customer service manager for Brooks Automation (Chelmsford, MA), where he is responsible for assisting customers in the setup, integration, training, and troubleshooting of robotics systems. Previously, he was an ion implant mechanical installation technician at Varian Semionductor. He received a BS in organizational management from Daniel Webster College in Nashua, NH. (Frost can be reached at 978/265-5405 or peter.frost@brooks.com.)

Gordon Beckhart is general manager at the MicroTool Alignment Services Division of Brooks Automation. A cofounder of MicroTool, he has 15 years of experience in the IC industry. Previously, he was business unit manager and manufacturing manager at Empak, and an R&D engineer for GaAs crystal growth operations at Crystal Specialties. He holds four patents in the area of robotics. He received BS and MS degrees in materials science and engineering from the Massachusetts Institute of Technology in Cambridge. (Beckhart can be reached at 719/471-9888 or gordon.beckhart@brooks.com.