Known as the Profile of Sequential Adjustment by Regression (SPAR), this new statistical analysis technique identifies and eliminates the bias, noise and artifacts related equipment. It could then lead to experimental measurements on nanomaterials and nanostructures more accurate and reliable and, subsequently, allow for industrial application.

Indeed, a precise understanding of the properties at the nanoscale is essential for manufacturers of nanomaterials and nanostructures. The development of future industrial applications with a large-scale production requires consistent performance. According CF Jeff Wu, a professor at Georgia Institute of Technology, “our statistical model will be useful when the nanomaterials industry to produce more upscale products developed in laboratories, for industrial users can not allow a detailed study on each production line. The significant errors related to experimental conditions can be filtered directly, which means that this model could be used in a manufacturing process.

Measurements of properties at the nanoscale, and separation of signals, noise and artifacts are still problems for many years for research in this area. Indeed, there are many sources of error measures, such as the shift of the studied structure, surface irregularities or imprecise placement of the tip of the microscope on the samples. In addition, the measured effects are sometimes only 2 or 3 times the noise level, then they are very difficult to differentiate properly, as well as noise may possibly mask other interesting effects.

To develop the SPAR technique, researchers have relied on a sample of data measuring the deformation of a material called “Nanobelts, made of zinc oxide, this to determine how elastic the material. Theoretically, by applying a force on the material through an atomic force microscope, a linear deformation occurs. However, experimentally, this is rarely measured. Additional forces arise and create a greater distortion, sometimes even non-symmetrical. To rectify these errors, the team from the Georgia Institute of Technology has developed a model of correction applied to data collected, corrected step measurements, using a regression technique.

It is true that ideally in Physics, the researchers want to understand and correct directly the real causes of experimental errors, but as noted by V. Roshan Joseph, a professor at Georgia Tech, for such scales this is very difficult to achieve. He added: “The physical models are based on some assumptions that can sometimes be distorted into reality. We could try to identify all sources of error can be corrected, but this would take considerable time. The statistical methods have the advantage of correcting these errors more easily, so this process seems better for industrial applications.

Besides correcting errors, the greater accuracy of this new statistical technique SPAR would reduce the effort required to produce reliable experimental data on the properties of nanostructures. As Professor Wu said: “with only half the effort typically required to conduct an experiment, it is possible to obtain the same deflection with methods without correction. This saves significant time.”. In addition, the team plans to use this correction on past experiences to ensure that signals of interest were not obscured by noise.

Funded by the National Science Foundation, the research was published in the issue of June 25, 2009 journal Proceedings of the National Academy of Sciences .
[ad#468x60text-only]