Spectral searching

One great advantage with digitized spectra is the capabihty to perform comparisons between the spectra of unknowns and reference spectra in a library. Infrared spectra are largely, but not completely, immune to collection conditions, that is, most spectral collection techniques (e.g., transmission, attenuated total reflection, diffuse reflection, photoacoustic spectrometry) will all produce approximately equivalent (but not identical) spectra if appropriate care is taken. Of course, some methods will be more successful than others for some samples or sample matrices, and those effects must be taken into account. Sample conditions are also important Spectra can change with temperature, solvent, or crystallinity, for example, and as the samples change their physical states, the infrared spectra will reflect those changes. 

AU search algorithms have a goodness-of-fit or hit quality index metric to indicate which spectrum is the best match. In a true Euchdean distance metric, where the spectra are normalized to unity, the metric wiU report a value between zero and 1, where the value of zero represents a perfect match. Commercial software will often scale or adjust this number. For example, if the Euclidean distance is subtracted from 1 and the result multiplied by 1000, a perfect match gives rise to a hit quality index of 1000, and a complete mismatch gives a score of zero. Regardless, the algorithm to compare the spectra is virtually identical only the way the results are reported varies. 

In earlier versions of Euclidean distance searches, the spectra were compressed by normalizing to a constant maximum absorbance and setting the absorbance data to an 8-bit precision that is, the absorbance was known to 1 part in 256. The entire spectrum was reduced in resolution to about 16cm . Finally, the region of the spectrum from about 2800 to 1800 cm was also eliminated, as very few fundamental absorption bands occur in that region. All these modifications to the spectra meant that the spectra were greatiy smoothed and less susceptible to noise than full-resolution spectra. 

When the baseline is not at zero, it may be better to use the first derivative spectrum rather than the absorbance spectrum as measured. An equation similar to Eq. 10.15 or 10.16 may again be used as the search metric, but in this case, N = S(y, — s,) and D = S(y,), where y, = v, x,+i and whae s, = r, r,+i. The derivative algorithm essentially functions to correct baseline offset or tilt An offset is reduced to zero, and a tilt is reduced to a constant when a first derivative of a spectrum is calculated. If the baseUne is curved, the derivative algorithm tends not to perform as well as the standard EucUdean distance algorithm. In the case of mixtures, however, the derivative system is better than a direct Euclidean distance as a major component of the mixture can often be identified more frequently. If it is necessary to distinguish between very similar spectra, the derivative algorithm does not produce good results. 

The vectors used in the correlation algorithm are defined as follows  

---

With the availability of computerized data acquisition and storage it is possible to build database libraries of standard reference spectra. When a spectrum of an unknown compound is obtained, its identity can often be determined by searching through a library of reference spectra. This process is known as spectral searching. Comparisons are made by an algorithm that calculates the cumulative difference between the absorbances of the sample and reference spectra. For example, one simple algorithm uses the following equation... 

Spectral searching and stripping in the analysis of a mixture of mannitol and cocaine hydrochloride, (a) IR spectrum for the mixture (b) Library IR spectrum of mannitol (c) Result of subtracting mannitol s IR spectrum from that of the mixture ... 

Table 74-1 A listing of Classic Spectral Search Algorithms and Terminology...

Sections on matrix algebra, analytic geometry, experimental design, instrument and system calibration, noise, derivatives and their use in data analysis, linearity and nonlinearity are described. Collaborative laboratory studies, using ANOVA, testing for systematic error, ranking tests for collaborative studies, and efficient comparison of two analytical methods are included. Discussion on topics such as the limitations in analytical accuracy and brief introductions to the statistics of spectral searches and the chemometrics of imaging spectroscopy are included. 

C. Carbon-13 Nuclear Magnetic Resonance (CNMR) Spectral Search System. 

Heller, S. R., and Milne, G. W. EPA/NIH Mass Spectral Search System (MSSS), A Division of CIS. Washington, DC U.S. Government Printing Office. An interactive computer searching system containing the spectra of over 32,000 compounds. These can be searched on the basis of peak intensities as well as by Biemann and probability matching techniques. 

NIH/EPA Chemical Information System (1983) Carbon-13 NMR Spectral Search System, Mass Spectral Search System, and Infrared Spectral Search System, Arlington, VA, Information Consultants... 

Spectral searches using a library of reference spectra can be a useful tool in identification. Search algorithms have improved over the years and now use the concept of artificial intelligence. Several software packages can be used to conduct searches in spectral libraries in which the main peaks of known compounds are encoded. The compounds offering the best matches are retained as potential candidates. Library searches involve three stages ... 

Figure 5.9. Spectral search at Spectral Database Systems (SDBS). The infrared (IR), nuclear magnetic resonance H-NMR and 13C-NMR), electron spin resonance (ESR), and mass (MS) spectra of organic compounds and common biochemical compounds can be viewed/retrieved from SDBS.

Spectral subtraction and spectral search aid the identification of evolved gases, which are often a mixture of products. Nevertheless, for unambiguous identification of unknown volatiles more powerful methods are required. Jansen and co-workers [86] have incorporated a parallel mass spectrometer onto the FTIR stage of a thermogravimetry-FTIR (TG-FTIR). The sample is thermally decomposed by TGA and the products collected in a Tenax (absorbent charcoal) trap. After desorption, the products are separated by a GC and the sample split, with 99% going to the IR spectrometer and 1% to the mass spectrometer. 

M.T. Soderstrom, R.A. Ketola and O. Kostiainen, Identification of some nerve agents homologues and dialkyl methylphosphonates by gas chro-matography/Fourier transform infrared spectrometry, Part II Spectral search with the help of retention indices, Fresenius J. Anal. Chem., 352, 550-556 (1995). 

S.R. Lowry and D.A. Huppler, Infrared spectral search systems for gas chromatography/Fourier transform infrared spectrometry, Anal. Chem., 53, 889-893 (1981). 

Although individual laboratories find it useful to compile their own reference library files, access to very large collections of mass spectra and to published data [55] is essential. A compilation of many thousands of spectra by the Aldermaston Mass Spectrometry Data Centre and the Division of Computer Research and Technology at the National Institutes of Health [56-58] has been made available commercially. The file can be searched in a number of ways using an interactive conversational mass spectral search system via a teletype and acoustic link over telephone lines. 

---