FIAlab incorporates a multivariate analysis feature in FIAsoft and SIAsoft. This tool allows our analyzers to reach lower quantitation limits of detection than ever before. This advanced data processing tool reduces noise 3-5-fold uncovering peaks with concentrations in single digit parts per billion. This result yields a striking distinction between pre and post-processed data shown below:
Figure 1: Effect of multivariate data processing. The left picture is preprocessed and the right is post-processed.
One can easily see the advantage of using this tool from Figure 1. The processing algorithm was developed by FIAlab’s dedicated scientists and engineers, who utilized both our new software platform, FIAsoft, and an array-type photodetector, a UV-vis spectrometer. Traditionally, FIA instrument manufacturers have relied on conventional photosensors that can only detect one wavelength at a time and thus preclude any multivariate approach to data handling. The use of an array spectrometer offers an entirely new degree of freedom in allowing data collection and analysis across the entire spectral range (400-900 nm). As a result, FIAlab is proud to be the first among FIA suppliers to offer a multivariate data processing feature to benefit our customers. The following is a brief description of the underlying principles of this technology and how it works.
The Principles of Multivariate Analysis
A colorimetric assay uses reagents that, when mixed with the sample, react to develop a colored product. When light passes through this colored product, it gets absorbed in different amounts at different wavelengths, which the spectrometer detects. See an example of a blue dye spectrum below:
The figure above has blue dye absorbance on the y-axis and the UV-vis spectrum in wavelength on the x-axis.
Although the absorbance at a given wavelength will differ based on the color intensity, the spectrum will have the same shape. This relationship allows FIAsoft to fit model spectra with little to no noise to actual data. The software minimizes the error between the data and the model by scaling the model. The software then automatically extracts the wavelength of interest from the scaled model spectrum and sends the corresponding absorbance to the output the user sees. This output is time-dependent, with the y component being absorbance as shown below:
Figure 3: Time-Dependent Absorbance before and after multivariate treatment. The red trace shows preprocessed data, the blue trace shows post-processed data.
FIAsoft repeats the process described above at specified time intervals dictated by the data acquisition time in FIAsoft. With each time interval, the fitted absorbance is added to the data analysis tab and a smooth, discernable peak appears on the user’s screen.
There are two main takeaway points from this paper. One is that the multivariate data processing tool yields lower quantitative limits of detection in the ppb range. The second is that the spectrometer makes it all possible because it generates multivariate spectral data not attainable by conventional photosensors. This is an interesting feature to keep in mind when considering capabilities of new flow injection equipment.