Specific Process Knowledge/Characterization/Stylus Profiler Measurement Uncertainty: Difference between revisions

← Older edit

VisualWikitext

@@ Line 1: / Line 1: @@
 =Stylus Profiler Measurement Accuracy=
-The accuracy of a height measurement with the profiler depends on the measurement settings, the sample, the instrument calibration and the resolution.
+The accuracy of a height measurement with the profiler depends on the measurement settings, the sample, the step or trench shape and height, the instrument calibration, and the resolution.
 ==Optimal Measurement Settings==
 Both the force setting and the scan speed are important: Too high force may compress a soft material like Al, Au or some polymers, while too low force may lead to the stylus "jumping" over features, especially if the scan speed is high. Too low scan speed may result in drift of the measurement and noise from vibrations while too high scan speed may mean that the stylus tip does not have time to reach the bottom of the features you are measuring and can also give rise to increased noise levels (see the [http://labmanager.dtu.dk/d4Show.php?id=2346&mach=304 Dektak XTA manual] on LabManager, Figure 3 for details).
-A sharp vertical step is easiest to measure. If the step is gradual or the surface is rough, it can be difficult to determine where to measure and how the scan should be leveled. Underlying sample curvature can also make it hard to level the scan. See further below under '''Total Uncertainty for steps < 1 µm'''.
+A sharp vertical step is easiest to measure. If the step is gradual or the surface is rough, it can be difficult to determine where to measure and how the scan should be leveled. For very small steps, underlying scan noise can also make it hard to level properly and will convolute with the shape of the step. See further below under '''Total Uncertainty for steps < 1 µm'''.
 <br>
 <br>
@@ Line 36: / Line 36: @@
 ==Total Uncertainty for steps < 1 µm==
-The underlying sample curvature can influence the measurement accuracy for steps < 1 µm. We have in June 2025 found both random and repeatable scan bowing or scan noise in the range of 10-60 nm across a 2 cm (2000 µm) scan. This is enough to influence the measurement accuracy of steps < 1 µm, especially steps up to 500 nm.
+Apart from the uncertainty described above, the underlying "shape" of the scan due to the instrument itself will influence the measurement accuracy for steps <1 µm. Although the stylus profilers use an optically flat surface as a reference, the basic underlying scan is never truly completely flat. We have in June 2025 found both random and repeatable scan bowing / noise in the range of 10-60 nm across a 2 cm (2000 µm) scan. This is enough to influence the measurement accuracy of steps <1 µm and especially <500 nm.
-We therefore cannot recommend using the stylus profilers for measuring steps <100 nm. One must expect relatively large uncertainty on small steps < 1 µm.
+We therefore cannot recommend using the stylus profilers for measuring steps <100 nm and one must expect relatively large uncertainty on steps in the range of hundreds of nm. See further discussion below.
+[[image:Si 6 in left_50umprsec_aveof2.png||right|thumb|Sample average of 2 scans of a blank, new 6" Si wafer made with the P17 Stylus Profiler.]]
+[[image:BlankSi 6in center-2-altlevel2.PNG||right|thumb|Sample scan of a blank, new 6" Si wafer made with the DektakXT Stylus Profiler.]]
 If measuring steps <100 nm it is important to check a measurement of a "flat" area nearby and to play with positioning the measured step in different locations along the scan length (to show variation in the underlying scan noise right under the step).
-Note that the step determination will usually be better than the noise of the underlying scan. In the DektakXT we regularly measure shadow masked metal films made by e-beam evaporation that are around 100 nm thick. We have found good agreement between our measurements with the DektakXT of a 500 µm wide step compared to XRR measurements of films of the same thickness. Probably this is because we level the scan close to the step, so only the curvature in an 800 µm range in x is important, and since we average the thickness across approx. 400 µm across the top of the step compared to 50 µm intervals nearby where the average height is set to zero, the variation comes out much smaller than what is seen across a full 2000 µm.
+The figures on the right show representative "flat" scans made with the P17 and DektakXT. The shape is influenced a lot by the position of the leveling cursors and a bow-like shape as the one seen in the P17 scan can easily be found in DektakXT scans of flat surfaces as well. More "flat" scans for the Dektak and P17 can be seen here: [[:File:Dektak XT and P17 scan flatness comparison summer 2025.pptx]].
-In practice we have found the step heights around 100 nm can be measured with higher accuracy in the DektakXT than the P17: We tend to get a measurement with the DektakXT within a few nm of the XRR measurement while the measurement with the P17 is off by around 10 We don't yet have a good explanation for this - nor do we have enough data to prove it's a general trend.
+===How much flatness can we expect of a "flat" scan?===
+KLA, the manufacturer of the P17 profiler, provides some information on the flatness one can expect from the scans, some of which is freely available in [https://www.kla.com/wp-content/uploads/KLA_AppNote_Stylus_2D_Stress.pdf this document]. Here we see that we can expect a variation of up to 40 nm across a flat scan of 3 cm. Thus the variation of up to 40 nm across 2 cm that we have seen in practice is not far beyond the presumably best-case scenario stated by the manufacturer (our tool could use some lubrication of the scan axes at the time of measurement).
-===How flat is a "flat" scan?===
+For the P17 the underlying scan noise or bow is extremely reproducible for many scans in the same position whether or not the vacuum hold is turned on. However the underlying noise varies when measuring different locations on a blank wafer. Therefore it seems that the non-flatness of the scans derive from some underlying structure on the optically flat surface or the scan rails.
-KLA, the manufacturer of the P17 profiler, provides some information on the flatness one can expect from the scans, some available in [https://www.kla.com/wp-content/uploads/KLA_AppNote_Stylus_2D_Stress.pdf this document]. From this we can expect a variation of up to 40 nm across a flat scan of 3 cm, which means the variation we have measured is not far beyond the presumably best-case scenario painted by the manufacturer (our tool could use some lubrication of the scan axes at the time of measurement).
+For the Dektaks, we have in some cases seen that the scan noise is reproducible not only for scans in the same location, but even for scans in different locations on the stage, which must be due to dust or bumps on the "feet" of the stage as they move across the optical flat. However in other cases, as in all scans shown in the document above, the bowing/noise of the scan varies even though the scan coordinates and sample are the same. The latter might be because the stage positioning is a little less accurate than for the P17 or because there is more environmental noise influencing the measurement.
-The variation depends partly on the location of the leveling cursors during data analysis. We have generally found the worst bowing far from the center of the chuck.
+===How does the underlying scan noise affect the measurement accuracy?===
+The step determination will usually be better than the noise of the underlying scan - we don't expect an error in the measured step sizes as big as 30-60 nm. This comes down to removing some of the noise due to averaging parts of the scan during leveling and measuring. Also, usually it is not necessary (or recommended) to measure such steps across more than 1 cm. The closer together the measurement points, the "flatter" the underlying scan will be.
-For the P17 the underlying scan noise or bow is consistent for many scans in the same position whether or not the vacuum hold is turned on. Therefore it seems that the non-flatness of the scan derives from some underlying structure on the optical flat or the scan rails.
+For example, in the DektakXT we regularly measure shadow masked metal films made by e-beam evaporation that are around 100 nm thick. We have found good agreement between the DektakXT measurements of a 500 µm wide step and XRR measurements of films of the same thickness, so we see no reason not to trust these measurements to within around +/- 5 nm. We level the scan with the leveling cursors close to the step before and after it, so only the scan noise in an ~700 µm range is important. Additionally, since we average the thickness across approx. 400 µm at the top of the step compared to the ~40 µm leveling intervals nearby (where the height is also averaged), the variation comes out much smaller than what is seen across a full 2000 µm scan.
-For the Dektaks, the scan noise is also in some cases reproducible for scans in the same location, but in other cases the bowing of the scan varies even though the scan coordinates and sample are the same. The latter might be because the stage positioning is a little less accurate than for the P17 or because there is more environmental noise influencing the measurement.
+In practice we have found the step heights around 100 nm can be measured with higher accuracy in the DektakXT than the P17: We tend to get a measurement with the DektakXT within a few nm of the XRR measurement while the measurement with the P17 has been off by around 10-15 nm. We don't yet have a good explanation for this - nor do we have enough data to prove it's a general trend.
-A lot of "flat" scans for the Dektak and P17 can be seen here: [[:File:Dektak XT and P17 scan flatness comparison summer 2025.pptx]].
+It may be even better to just level with cursors near each other on what one knows is a flat surface and then measure just before and after the step very near the leveling cursors to get the best height estimate.