Product In Action

Control System Speeds Start-up, Stabilizes Process Air in DUV Litho Tool Clusters

As chipmakers grapple with the stringent requirements of deep ultraviolet (DUV) photolithography, they confront a wide range of organic contaminants--from amines, halogens, and radicals to chlorine, bromide, and arsine. To remove such contaminants, most fabs have added chemical filters to their process air control systems. But as chipmakers have discovered, coupling chemical filters with conventional environmental control units (ECUs) can pose significant challenges.

For DUV, the environment within the stepper or track enclosure should be temperature-controlled within ±0.1°C, and humidity should be maintained within ±0.5% RH. Typically, an ECU can reach these set points and deliver process air to the tool within a few minutes, unless the fab has installed chemical filters between the ECU and the process enclosure. Introducing chemical filtration to a standard system retards the ECU's response time by a factor of 10 to 20. Fab operators have found that it can take an entire shift or more to reach the specified temperature and humidity.

Many fabs have addressed this problem with labor-intensive, costly manual workarounds. When San Jose­based Silicon Valley Group, a major photoresist coater/developer track supplier, sought an alternative, Semifab (Hollister, CA) developed a solution. Semifab's design automatically adjusts for the output of chemical filters and adapts to temperature and humidity fluctuations at the point of use. The company recently designed this control architecture into its RAM-E and CD-E product lines.

Conventional process air systems rely on a single control loop, with a sensor at the point of use relaying measurements to the ECU. As air temperature and relative humidity fluctuate at the process point, the ECU senses these changes, modifies the airstream, and the tool environment returns to the set point. Whenever parameters shift, the ECU responds and quickly stabilizes conditions at the point of use.

When a fab installs a chemical filter between the ECU and the tool enclosure, the system begins to lag. Because they contain charcoal and other materials, such filters have a large thermal mass. This thermal capacitance provides stability in heating and humidifying air but also retards the process of reaching a steady state. With chemical filtration, it takes much longer for air modified in the ECU to have any measurable impact at the point of use. Because of this lag, parameters at the process point often keep overshooting the mark.

For example, as temperature- and humidity-controlled air from the ECU passes through the filters, the charcoal or other filtration material absorbs moisture. Thus, air that has been correctly adjusted for humidity by the ECU becomes dehumidified as it passes through the filter, a situation that persists until the filter medium becomes water saturated. Meanwhile, the sensor at the stepper/track tool cluster discerns a prolonged deficiency in humidity. The ECU responds by increasing humidification of the airflow until the filter finally becomes saturated. Then the humidity at the point of use quickly spikes, and the ECU swings to the opposite extreme and begins dehumidifying its output.

Temperature undergoes analogous swings. As charcoal filters absorb moisture, an exothermic reaction heats the airstream. In this case, the heat rise is detected at the process area, and the sensor tells the ECU to reduce the heat, until the system inevitably undershoots the air temperature set point. Positioning the ECU between the chemical filter and a stepper doesn't work well because this creates a negative pressure downstream of the filter. Negative pressure makes it possible to suck contaminants into the system.

Customers told Semifab's engineers that they were experiencing a continuous problem with ECUs and chemical filters. After integrating chemical filters configured with the ECU in front of the filter, it was taking hours or even days--instead of minutes--to stabilize humidity and air temperature. The standard single-control-loop architectures were not addressing the increasing demand for chemical filter process performance.

For fab technicians, the short-term solution was to mount the ECUs sensor at the output of the air handler and before the filter. The technicians would enter the chamber and measure conditions with a handheld probe. Based on the delta between this reading in the chamber and the sensor's measurement at the ECU output, technicians would offset the air handler to achieve the desired set point. This allowed them to circumvent the filter-induced response lag and subsequent overshoot. The problem with this configuration was that the air handler remained blind to changes within the enclosure and could not adapt to inevitable fluctuations, keeping the process labor intensive and time consuming.

Semifab presented a straightforward solution: configure two controllers in a cascade system and make them interact to reach the process air set point quickly with minimal overshoot. Dubbing this technique dynamic dual-point control (DDC), engineers combined lessons from both conventional automation and the manual workarounds. As Figure 1 illustrates, DDC uses two specially designed sensors, one at the traditional location within the tool enclosure, and a second sensor just outside the ECU, before the filter. Leveraging two control loops, DDC correlates input from both sensors to quickly reach the set point at start-up and then gradually adjusts the air handler in response to changing conditions.

 

Figure 1: The dynamic dual - point control system uses two specially designed sensors, one located in the tool enclosure and a second one just outside the ECU before the filter.

This system receives input from the sensor in loop no. 1 (at the tool) and the sensor in loop no. 2 (air output from the ECU) determines the delta and, in essence, adjusts the set point for loop no. 2. This adjustment in the loop no. 2 set point is advanced or moderated to achieve rapid responses without overshooting the target. The amount of adjustment and the rate of adjustment are both scaled according to standard proportional, integral, and derivative (PID) settings. In addition, this technique greatly reduces surges in air-handler output, so that the system reacts to sustained changes, rather than spikes. This is particularly important at start-up.

With this architecture, the operator defines the set point at the tool, and the controller uses data gathered at two locations to achieve that goal as quickly as possible. So, for example, at start-up the system may instruct the ECU to reach a higher set point than the operator has chosen, then proportionately lower the setting, based on the PID algorithms. By adjusting the set point and heat-up rate, the system achieves and maintains the desired set point equilibrium much faster.

SVG has tested this twin-control-loop approach for more than a year and found the technique highly effective. Operators consistently have reported fast start-up and general stability of process air values over time, using this cascade control technique.

 

Figure 2: Results of system speed in attaining target process air values at start-up.

 

Figure 3: Steady-state RH and temperature performance over time of dynamic dual-point control system.

In addition, Semifab conducted its own testing, using a Semifab RAM 2400 and both Donaldson and Extraction Systems filters. In these trials, the DDC system achieved temperature control within RAM specs of ±0.1°C controlling from 20° to 23°C and ±0.5% RH controlling from 35 to 45% RH. The performance for temperature control was ±0.021°C. The performance for humidity control was ±0.098% RH. Figure 2 shows that the system quickly reached the target process air values at start-up. As Figure 3 illustrates, DDC steadily maintained relative humidity and temperature control of process air over time.

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit