Browse Prior Art Database

Laser Energy vs. Voltage Error Limiter (LEVEL) Disclosure Number: IPCOM000034138D
Publication Date: 2005-Jan-18

Publishing Venue

The Prior Art Database

Related People

Other Related People:


TITLE: Laser Energy vs Voltage Error Limiter (LEVEL) Purpose and advantage of invention: Reduction of energy transients and duty cycle effects in a mode similar to external voltage will greatly improve laser energy performance in a manner which is transparent to the customer Short description of invention (Include or attach appropriate sketches, and point out novel features): Problem description Some of Cymer’s customers use the laser in “External Voltage” mode. In this mode, the customer directly commands the voltage applied by the pulse power system. The voltage command is computed by the scanner usually based upon energy readings taken from somewhere in the beam path. Figure 1 illustrates this architecture. Figure 1. External voltage control mode. This architecture suffers from a major liabilities. Since all voltage control is handled by the scanner, the scanner controller is responsible for handling burst transients, duty cycle effects, drift, and energy vs. voltage sensitivity (dE/dV) changes. This is a problem for our customers since they want the same controller to work not just with our lasers, but with the competition’s lasers as well. The control required to optimally compensate for the above mentioned affects will depend upon the details of the laser design. It is unlikely that an optimal controller for our laser will be the same as that for someone else’s, hence the customer has to custom design controllers depending on laser. In practice, this does not happen. Instead, the scanner manufacturer dictates requirements about the maximum size of burst transients, drift, and dE/dV changes. This often forces us to invent solutions to problems which could be easily solved using active control (e.g. energy transients) using more expensive methods, such as chamber redesign. A different approach is to let the laser handle energy control. In this case, the laser receives energy data either from an on-board sensor or from a sensor within the scanner. The laser then computes the correct voltage to apply in order to achieve an energy time history which meets certain performance metrics such as moving average (dose), or moving standard deviation (sigma). Figure 2 shows this configuration. Figure 2. Energy control mode. This approach also has limitations. First, the scanner must provide an energy target instead of a voltage command. This is a significantly different style of operation and would force the customer to have different controllers for Cymer lasers. Second, in addition to correcting for laser errors (e.g. drift, energy transients, dE/dV changes, duty cycle effects), the laser controller must also tailor the energy time history to maximize performance metrics such as dose or sigma. The problem is that if the customer decides that they have different performance goals for the controller, they must get Cymer to modify the controller to accommodate them. Further, the customer must share more information about what is important to the exposure process than they currently do. They are reluctant to do this because it skirts dangerously close to details about their product that they would rather hide from their competition. Invention Figure 3 shows the proposed invention. From the scanner side of the system, the laser looks exactly like it does in “External Voltage” mode. The scanner commands a voltage to the laser and receives a beam of light. The difference is on the laser side of the system. Instead of going directly to the pulse power system, the voltage command goes to the laser controller. This controller also receives data from an energy sensor which measure the energy in the beam at the output of the laser. The laser controller adjusts the voltage commanded by the scanner to: • Remove energy burst transients • Correct for changes in dE/dV • Correct for slow drift in output energy • Compensate for changes in effienciency due to changes in duty cycle. To the scanner, the laser appears to have constant dE/dV, no energy drift and no energy transients. The shot to shot variability will look very much like white noise with tonal components at frequencies corresponding to blower blade passage frequency. Figure 3. Laser voltage correction. Fundamentally, the controller is not doing anything very different from what it does in energy control mode. The major difference is the control objective. The controller is not making high bandwidth corrections to the output energy to optimize energy dose or sigma. Instead it is making corrections which will ensure that the energy output by the for a given voltage command is as close to an ideal E vs. V curve as possible. Figure 4 shows an embodiment of the voltage correcectinon algorithm. There are three major components: Adaptive feed forward, dV/dE estimator, and Energy Servo. Adaptive feed forward makes corrections to the voltage signal on a burst to burst basis. It monitors the ouput energy for burst correlated transients. For the first n shots in a burst (typically 20), a fixed value is added to the voltage (dependent on shot number) to remove the transient. An adaptation scheme is used to adjust the correcting voltages. The adaptive feed forward block also receives duty cycle as an input. It adds an additional adjustment to the voltage signal to compensate for efficiency losses which occur at low duty cycles. The dV/dE estimator (inverse of dE/dV) adds a small, high frequency dither signal to the voltage. By correlating this signal with the adjusted voltage and the output energy, the estimator can determine a reasonably good value of the laser’s actual energy to voltage sensitivity. The voltage is adjusted to correct for the deviation of the actual dE/dV from the desired dE/dV. Finally, the energy servo monitors the error between the laser energy and the energy determined by a desired E vs. V characteristic. The servo adjusts the voltage to compensate for energy drift during a burst. It should be emphasized that the loop gain of the energy servo is kept very low. This is to ensure that the servo corrects for slow drift errors and there is no chance of this controller destabilizing the scanner dose controller. The scanner dose controller still has the primary role of shaping the energy time history to optimize exposure based metrics. Figure 4. Energy controller architecture