Currently, the software on the FTIRsearch.com web site does not have the ability to perform spectral subtraction for mixture samples. This was considered during the original development of the web site; however, Internet technology is not yet capable of performing this operation with the level of real-time interaction required to make it useful. It is possible to implement an "Auto-Subtract" function, but this will not solve the problem 100%. Hopefully, the web will become more interactive as new standards and web browsers evolve, and this feature can be incorporated in the future.

There are actually two ways to execute a "region-only" search. First, if your spectrometer software allows you to manipulate the spectrum, truncate the spectrum down to the region you want to search, then save it as a new file. When you run the spectrum search, upload the data file of the truncated version of the spectrum.

The Spectrum display does not support region clicking at this time. This has been discussed as a possible future enhancement and if there is enough interest on the part of FTIRsearch.com members, this feature can be added in the future. If you are interested, please let us know by filling out a request in the Feedback form or by sending an email to [email protected].

There are a variety of different methods used in commercial spectral searching software to match unknown samples to large databases. These range from correlating peak positions and/or intensities to full spectral pattern matching.

FTIRsearch.com uses a series of well-established methods which employ mathematical algorithms that perform point-by-point comparisons of your submitted spectrum with the spectra in the database. Historically these algorithms have been shown to give the most accurate and reliable matches possible.

When submitting a spectrum for searching, care should be taken to avoid using spectra that exhibit deviations from Beer's Law. In most commercial spectrometer systems, the direct relationship between the absorption at a given frequency and the pathlength and concentration of the sample is only maintained over a controlled range (0.0001 to 2.5 Absorbance units).

Spectra that are measured on samples that are too thick or with air bubbles in the cell or between the plates will often have absorption bands that exceed the capabilities of the spectrometer's detector. This will typically appear as a flattening along the tops of the largest peaks in the spectrum. There are also cases where the actual flattening is not observed, but peak maxima occur in the range of 4 to 5 Absorbance units.

When performing spectral library searching, usually the entire spectrum is compared to the database. Most of the spectra in the FTIRsearch.com database are of pure components or well-characterized industrial materials. For this reason, using spectral searching techniques with mixtures or contaminated samples generally leads to poor match results.

Baseline deviations are one of the most significant contributing factors to the quality of the matches resulting from a spectral search. All of the spectra in the FTIRsearch.com database have been processed to have baselines as near zero as possible. When comparing a submitted spectrum to the database, typically small constant offsets are easily compensated by the spectral search comparison algorithms. However, sloping, curved or irregular baselines will have a noticeable negative effect on the quality and accuracy of the matching process.

The FTIRsearch.com search engine automatically applies a baseline correction method during the search process to try to get the best matches possible. While this algorithm is excellent at removing offset and slope effects, it cannot compensate for curved or irregular baselines.

If you use the Advanced Spectrum Search option, you can also employ one of the algorithms that use a first derivative comparison (such as the First Derivative Correlation method). These are very useful for searching spectra with irregular baselines, but will also tend to enhance the noise. This will have the effect of placing more emphasis on the individual peak positions than the overall shape of the spectrum during the hit quality calculation.

Certain searching algorithms such as the Correlation method can compensate for noisy spectra and produce reasonable matches. However, there is no substitute for well-measured sample spectra when searching spectral databases.

This is especially true if you plan to use algorithms that employ first derivatives to remove sloping baselines. These algorithms will amplify the noise in a spectrum relative to the signal, which can adversely effect the quality of the spectral matches.

The signal-to-noise ratio (SNR) in a spectrum can generally be improved by averaging more scans when measuring your spectrum. Note that the SNR increases with the square root of the number of scans, so you must increase the number of scans by a factor of four to improve the SNR by a factor of two.