For many years we have been aware of numerous problems in obtaining good calibration of the PZT scanners under various conditions of operation of the SPM. Many of these problems are due to creep in the PZT material. A creep compensation circuit has been implemented is now standard on all systems. The benefits are varied, and in some cases surprising.

        Creep causes the response of the PZT to vary with time. In the X and Y axes, we have always observed about 20% difference in the maximum scan size between X and Y, with Y being the slow direction where creep has a longer time to act and increase the scan size for a given applied voltage. With the creep compensator, this difference is less than 1%. It now becomes possible to change the direction of scan without causing gross distortion of the image.

        The nonlinearity in scan was always a mystery. Physical principles predict that the nonlinearity should be an S shaped or antisymmetric curve in contrast to the concave curve observed. Any nonlinearity can be represented by a series expansion of sinusoidal terms or by a power series. We chose to use a series of sinusoids. Because the creep function does not yield to a rapidly converging expansion, and the terms vary with the speed of scan, it was difficult to obtain accurate compensation. The creep compensator removes the concave hysteresis, represented by the first nonlinear correction coefficient, leaving only the next term representing cubic antisymmetric nonlinearity. Because the terms are smaller, the correction is more exact, and corrected scans show only 0.2% nonlinearity. This also apparently reduces the changes in linearity with scan size. It is to be expected that linearity should improve as the scan size is reduced, but since the creep is independent of scan size, this was not observed.

        It was always difficult to do a hard zoom except by moving in small steps over time. The Y calibration number was used in the X direction so that the final zoom point would be close. Due to the continuous creep during zoomed scans, the image was distorted by movement in X and Y. The creep circuit improves the response significantly. The left image is a filtered scan of the calibration grating showing the linearity, and the small correction needed.

        In the Z axis, the artifacts are related to features on the sample, and the identical circuit is used here. Overshoot over four orders of time scale has been the obvious defect which is visible even in competitors' published images. The defect makes it impossible to make accurate measurements of step heights unless there is a step height standard almost identical in height and width to the sample. The creep compensation makes a marked improvement here as well. Because of creep, the Z calibration using the slide as a ramp gave too large a value, and showed convex curvature at the start. The ramp below shows the result with the creep compensation.

        It is always important to develop proprietary technology. This is the only thing except price that can convince customers to buy our system instead of the competitors'. I believe we are the first company in the industry to solve the problem of creep compensation.