Improving accuracy between tools using ToolMatch software

Introduction

In the semiconductor industry, Wavelength Dispersive X-ray Fluorescence (WD-XRF) is an important metrology technique, as it has proven to meet strict criteria when it comes to reproducible results on an individual spectrometer. In addition to excellent measurement reproducibility, results matching of different tools is often essential. With Malvern Panalytical’s ToolMatch software it is possible to match different tools to get highly similar results and thus improve the accuracy of the results across multiple tools.

In this application note, ToolMatch is demonstrated for an application containing Al, Cu and Ti, which was measured on four different Malvern Panalytical 2830 ZT Wafer analyzers.

Instrumentation

Four Malvern Panalytical 2830 ZT spectrometers equipped with fixed channels for Al Kα, Cu Kα and Ti Kα are used. All spectrometers have the same hardware configuration, see Table 1.

Procedure

The first step in improving accuracy between different spectrometers using ToolMatch is creating a recipe (called an “application”) in SuperQ, the analytical software. In this report, for the AlCuTi application, calibration lines for compounds Al (nm), Cu (nm) and Ti (nm) were created. An example of such a calibration line is shown below.

Based on these calibration lines, a set of reference wafers is measured, to obtain references results on tool 1 (now called Mother tool). 

Once the application is created, and the reference measurements are obtained on the Mother tool, the complete recipe, which includes application, monitor, calibration standards, calibration, calibration update, wafer patters and ToolMatch correction set, can be transferred to the secondary (now called Daughter) tool via the Application Export/Import functionality which is included in the ToolMatch software package.

On the Daughter tool, after all calibration wafers are measured under identical conditions as on the Mother tool, the calibration is completed with a simple push of a button. Then the measurement of the ToolMatch reference wafers is executed.

Note: The ToolMatch software even checks for the location on the wafer that is measured. The ToolMatch correction set is only updated in case both the wafer identification and the location of the measurement are identical. This prevents automatic update of TMC when selecting a different wafer or measurement spot, to overcome differences introduced by wafer-inhomogeneity.

As we can see, the results obtained on the daughter tool are similar to the ones obtained on the mother tool, but still there is a small difference that will be corrected using ToolMatch in the following steps.

Based on the reference from the Mother tool and the measured Daughter tool values, correction lines can be calculated. These are stored in a ToolMatch Correction set (TMC-set), which is used to correct future unknown measurements, to achieve the same results on the Daughter tool as on the Mother tool.

For every compound, correction lines can be calculated, based on a selection of different models: No correction or correction using a Constant, Linear, Quadratic or Exponential model. It is even possible to create multiple correction lines for one single compound, using boundaries. The TMC-set decides which correction line needs to be applied for the unknown measurement, for each parameter (element concentration / layer thickness) individually.

In the example the Al #1 line will be used from 0-100 nm, where the Al #2 line will be applied when the layer thickness is 100-600 nm thick.

Due to very strict process demands in today’s production control, which are often well below 1% or even 0.1% relative, we may see differences between spectrometers, even though the spectrometers are maintained within the specifications. At signals that go from low to very high count rates, this could e.g. be caused by small deviations in the detector linearity (specified to be ±1% relative) or crystal properties. With ToolMatch these differences can be automatically corrected for. During routine operation, drift correction in combination with an automatic update of the TMC-set will provide an easy to operate procedure to maintain the ToolMatching. 

Results

With the TMC sets applied, wafers were measured over a period of 3 months on 4 different spectrometers (> 1500 measurements). This is compared to the same period in previous year (with manual corrections for matching) on 3 different spectrometers (>1200 measurements). The critical parameter which is compared is the Process capability (C_p):

As can be seen in Table 5, a huge improvement in C_p has been achieved switching from manual correction to automatic correction using ToolMatch software. A C_p value > 1.33 is typically accepted stating that the process is well within control.

This improvement in Cp is also visualized in the SPC chart for Cu:

Using the automatic TM correction, the results on all tools are closer matched to each other. With time consuming manual corrections, tools were matched within ±1.5% relative difference. With automatic TM correction, this improved to ±1% relative difference. Hence, automatic TM correction helps in improving the quality control process across multiple tools.

Daily operation

Once ToolMatch is set up, the procedure is referred to in the application, just like e.g. the calibration. Once selected, updating of the correction is done fully automatically by simply measuring the wafers from the ToolMatch set. Where manual Matching usually is a procedure that is very time-consuming and can be carried out by experts only, this can now be activated and completed by a routine operator. As the measurement itself activates the updating, the procedure can be integrated seamlessly in process monitoring, where quality check measurements are carried out on a regular basis. In case where, despite instrument monitor correction, the quality check measurement fails, the ToolMatch measurements will bring the results back in-line with the expected results. Because these measurements are activated by the routine operator, the need for expert intervention is significantly reduced. This improves redundancy, prevents delays, and increases efficiency.

Conclusion

ToolMatch software offers a straightforward way to match results on different tools. The combination of ToolMatch and application import/export software makes it easy to create identical recipes on spectrometers which, combined with Tool Matching corrections, provide nearly identical results. In practice this leads to a greatly improved process capability (Cp), without the need of manual correction by an expert, therefore improving the process efficiency significantly.