Long-term Matching Of Optical Metrology Tools
The Therma-Wave tools involved each contained these optical metrology technologies:
* Beam Profile Reflectometry (BPR). This patented tool compares measured reflectivity vs. angle profiles to theoretical models for determining the characteristics of film stacks.
* Beam Profile Ellipsometry (BPE). This patented single-wavelength ellipsometer uses a small spot about 1Å in diameter. Changes in signal strength are proportional to the thickness of the film.
* Visible and Broadband Spectrometer (VS and BB). A visible spectrometer and combination visible-UV spectrometer compare measured reflectivity vs. wavelength profiles for determining the characteristics of film stacks.
Before this calibration procedure, the tools involved in the study had not been rigorously calibrated for several years. The tools had in large part been tuned for measurement of specific structures, and required frequent adjustment to maintain this condition. Metrology of structures other than those targeted by the tuning was achieved only with difficulty. A fifth and sixth technology, the spectroscopic ellipsometer and absolute ellipsometer, were present on these tools, but were not included in the matching procedure.
Matching of the tools for long-term reliability involved three steps:
* Optimising the configuration and calibration of hardware for each tool. The calibration must be to identical standards and performance requirements for each tool.
* Optimising the measurement recipes for each tool.
* Defining procedures that will maintain the tools in their matched condition.
Optimising hardware
The metrology targets of the tools fall into two groups. One group consists of thin oxides, which are measured with the BPE. Thicker films and film stacks are measured with BPR and VS. One of the reasons for carrying out this matching program was to enable conversion from visible to broadband spectrometry, because of the wider range of frequencies (190nm to 800nm) available in broadband.
As a first step, the adjustment of the hardware for each tool was carefully inspected and optimised, even if optimisation involved very small adjustments. None of the four tools was found to be out of specification, but some optimisation adjustments were made anyway.
Thick film calibration centred first on the very stable BPR technology. The reference targets used were thick silicon dioxide of known thickness and a wedged oxide wafer that provided a range of oxide thicknesses. Together these two targets were used to calibrate the BPR tool at over 100 points across the entire width of the wafer, thus ensuring maximum coverage of points where the oxide had different thicknesses. After calibration of the BPR, the three remaining technologies (BPE, VS and BB) were calibrated to the BPR and to the known optical dispersion properties of the silicon dioxide film and silicon substrate. First, a spot of known thickness on the wedged oxide wafer that had been measured by the BPR was measured using one of the other technologies. If the measurement differed from that of the BPR, the calibration was adjusted to bring the technology into a good match with the BPR. Calibration of the BPE also included the initial calibration of thin oxides at the thin edge of the wedge. Most of these measurements were less than 200Å.
The calibration of the BPE for thin oxides was also required to maintain matching with previous history - a requirement that was not needed for measurements of thicker films. It was possible to carry out the planned matching of the BPE and still maintain matching to as many as three known standards. It was anticipated that the calibration for thicker films and for film stacks would cause the measured values to shift slightly, requiring changes in process charts, but this shift was considered acceptable in the light of the long-term reliability benefits.
Recipe optimisation
Before the matching procedure was carried out, a separate recipe was often required for individual wafer or process steps, even when the only difference from one process to the next might be the thickness of a single film layer. One of the anticipated outcomes of tool matching was the writing of single recipes that could be used in measuring a given type of film stack, even if the thickness of the critical layer in that film stack varied. These recipes would also provide optimum matching under standard calibration conditions.
For example, prior to the matching process, seven separate recipes were required to measure all the different poly thicknesses and get matched results among all the Opti-Probe tools. After the matching process, all poly measurements use a single recipe. In total, 15 old recipes have been replaced by four new recipes (one for each type of film stack), greatly simplifying measurement operations.
Keeping the match
Three procedures were developed that can be implemented on a long-term basis to maintain the matching of these tools, and to preserve the benefits derived from having a closely matched tool set. These procedures are:
1. Whenever the tool or the environment causes monitor wafers to move out of control, the last step in corrective action will be to recalibrate the technology involved and to reverify matching of that technology.
2. At every preventative maintenance (PM) the tool will be recalibrated and the matching reverified.
3. The calibration wafers will be recertified once a year.
Results
The results of matching across the tools are shown in Figure 1 and Figure 2. It is anticipated that long-term effects of tool matching will include more effective process control and overall lower production costs.
Authors: Curry Scheirer, Agere Systems. Xiaoping Liu, Phil Graves, Paul Hilton, Kevin Peterlinz, Xiaohong Chen, Therma-Wave
