Introduction
The application of measuring surface texture with a white light optical profiler has been well-known for many years. As the capabilities of optical manufacturing and precision machining increase, the production of ‘super smooth’ or ‘sub-angstrom’ surfaces has become more common, and quantification of these surfaces is critical for effective process control. The NewView™ 6000 series of optical profilers using Scanning White Light Interferometry (SWLI) with MetroPro™ software and patented FDA analysis enable rather straightforward quantification of surfaces with texture measured on the order of fractions of a nanometer. With good control over the measurement environment, proper selection of measurement parameters, and effective instrument calibration, quantification of surfaces with roughness measured in tens of picometers (1x10-12 m) is possible. With the NewView 6300’s best-in-class acquisition speed and resolution, areal measurement of supersmooth surface texture has never been so easy.
Understanding System Noise
The first step required in making quantitative measurements of super smooth surfaces is to understand that every measurement system has an inherent baseline system noise. This noise results from a number of factors including electronic noise, sensor noise, small irregularities in the reference surface, and small vibrations caused by changes in the measurement environment. For most samples, the measurement noise in the NewView can be essentially ignored, as the measurement value is much larger than the noise floor. However for very smooth samples, this is not the case. For these samples it is important to understand the noise sources and to control them as tightly as possible. Many sources of noise can be virtually eliminated or at least significantly reduced by both tightly controlling the measurement environment (acoustics, air currents, temperature, etc.) and also by performing a number of measurements and averaging them together into a single data file.
Environmental Controls
The first task in setting up measurements for a super smoothpart is establishing control over the measurement environment. The ideal environment would be one which:
- is mechanically and acoustically quiet to minimize part vibrations;
- has tight temperature control to minimize sample and objective changes during the measurement period;
- has well controlled airflow to minimize air currents between the microscope and the part.
In order of importance, vibration, noise, and temperature rank at the top. When the objective working distance is small, airflow control may essentially be a non-issue after mechanics, acoustics, and temperature have been addressed. With long working distance objectives, however, air currents will be more critical to control.
Measuring the System Merit Function
ZYGO has developed a process for quantifying the expected system noise as a function of measurement averages which we will call the System Merit Function (refer to the last page of this document for an illustration of this measurement method). The measurement process involves acquiring a number of measurements—typically 10 or more—with a given number of averages. Each of these data files are saved as D1, D2,…Di. These data are averaged together into one single file, Dse which represents the total system error during the measurement period. This Dse file is then subtracted from each of the component Di files to create an error map indicative of the expected system noise for a single measurement. The rms of the individual difference maps are recorded, and the mean and standard deviation are calculated for the series of differences. By adding twice the standard deviation of the series to the average rms of the series, the expected system noise for a given number of measurement averages can be estimated with good confidence. This process can be automated by using standard MetroPro and some very simple MetroScripting. An example application is available upon request from ZYGO.
Predicting System Noise
 | Once the value of the System Merit Function for measurements with no averaging is known, the predicted system noise for a specific number of averages can be predicted using the formula. |
where SN1 is the system merit value measured with no measurement averages and AVG is the desired number of averages. For larger numbers of averages taking a longer time period to measure (typically greater than 32 measurements) this prediction can only hold true in very well controlled environments. For critical applications, it is recommended to test the measured noise floor against the predicted value and ensure that the environment is controlled well enough – morewell-controlled environments will generally require fewer averages. In the event that the environment is not satisfactory, the line for the measured values in Figure 1 will typically turn upward again and diverge from the predicted values. If a noise floor associated with averages beyond the upturn point were desired, it would be necessary to further improve the environment. |  Figure 1 –Excellent correlation is observed between predicted and actual system noise on the NewView 6300 |
Phase Res - Which Level?
Depending upon the smoothness of the surface to be measured, it may be necessary to increase the internal precision of the calculations made by the MetroPro software. Starting with version 8.1.1, MetroPro allows for three levels of precision with the Phase Res Measurement Control—Normal, High, and Super.Normal is the lowest resolution—useful primarily for large steps and rough surfaces. High is the standard setting for most typical measurement situations using scans up to 150µm and texture down to approximately 0.050 nm. The newest and highest precision setting, Super, enables measurement of very smooth surfaces using a large number of averages. Only very smooth surfaces will require the use of Super. This should be taken into consideration when determining baseline system noise.
What Noise Floor do I Need?
There is no one right way of selecting the number of averages (and by extension, the noise floor) for a particular application. For rougher surfaces, where the surface is measured in tens of nanometers or more, striving for a system noise on the order of 10x lower is often recommended. However for a surface which is on the order of 0.05 nm (0.5 Å) achieving a noise floor 10 times lower would theoretically require approximately 3000 averages! Rather than a hard and fast rule, a more empirical rule of thumb employed by ZYGO is that the lowest practical noise floor for an application is recommended, but that system noise should be at least 2 to 4 times smaller than the desired measurement surface features.
System Error Characterization
After determining the phase resolution and the number of averages required for the desired noise floor, it is recommended that the user perform a system error characterization. This process entails measuring a number of physical locations on an optical grade flat using the desired number of averages per site.
Typically, at least 8 distinct sites with no overlapping regions are recommended for generation of a system error file. For specific information and procedures for creating a system error file, please refer to the NewView MetroPro Microscope Application Booklet, OMP-0360 or Section 8, MetroPro Reference Guide, OMP-0347. The error map created will then be subtracted from each of the surface measurements made on the actual sample.
Figure 2 - A graphical representation for the process of measuring the System Merit Function
Conclusion
Using the methods and procedures described here, ZYGO has demonstrated the capability of measuring surfaces smoother than 0.05 nm. Tightly controlling the measurement environment, selecting an appropriate internal precision, and choosing the number of phase averages based on the System Merit Value all combine to provide the highest quality surface texture measurements available from an optical profiler.
