CC BY
14
Научни трудове на Съюза на учените в България-Пловдив, серия Б. Естествени и хуманитарни науки, т.ХУЬ Научна сесия „Техника и технологии, естествени и хуманитарни науки", 30-31 Х 2013 Scientific researches of the Union of Scientists in Bulgaria-Plovdiv, series B. Natural Sciences and the Humanities, Vol. XVI.,ISSN 1311-9192, Technics, Technologies, Natural Sciences and Humanities Session, 30-31 October 2013
Spectral Libraries of Vibrational Spectra P. Penchev*, S. Tsoneva, Ts. Krusteva and S. Nachkova University of Plovdiv, Plovdiv, Bulgaria
ABSTRACT
Several spectral databases are created that are composed of 966 absorption FT-IR, 102 ATR FT-IR and 200 Raman spectra measured in our laboratory. There are compiled four other libraries of 13484 and 350 absorption FT-IR, 12 ATR FT-IR and 116 Raman spectra from other sources. The software IRSS previously developed in our lab is used to perform spectral search in the database. The library search routines programmed in IRSS are thoroughly tested. It is found that a successful identification is possible not only of unknown compounds but of the mixtures components if their vibrational spectra are present in the vibrational spectral collections.
INTRODUCTION
The infrared (IR) and Raman spectra reflect in a great extent the compound's structure, therefore both techniques are well suited for the process of structure elucidation [1,2]. Though Raman spectroscopy has only recently been introduced for analytical purposes, it was applied since the middle of the previous century as a vibrational technique that provides spectral information complementary to middle IR (MIR) spectroscopy. Similar to the MIR spectrum, the Raman spectrum is based on molecule fundamental vibrations and their overtones thus providing a detailed spectral "fingerprint" of the compound under study. As an analytical technique, Raman spectroscopy offers many advantages over MIR spectroscopy [2,3]; the most important of them are the following: (1) little or no sample preparation is required; (2) water as a liquid is a weak scatterer - no special accessories are needed for measuring aqueous solutions; (3) water vapors and carbon dioxide are very weak scatterers - sample compartment purging is unnecessary; (4) fiber optics (up to tens of meters in length) can be used for remote analyses; (5) Raman bands are narrower and both the overtones and combination bands are generally weak; (6) the spectral range reaches well below 400 cm-1, making the technique applicable to organic compounds containing heavier elements; (7) The symmetric molecular vibrations which appear as low-intensity bands in the IR spectrum exhibit very strong Raman bands.
Attenuated total reflectance (ATR) Fourier transform infrared (FT-IR) spectroscopy is one of the more popular sampling techniques used by FT-IR spectroscopists because the measurement is quick, non-destructive and requires virtually no sample preparation [2]. An ATR spectrum is slightly different from an MIR one of the same compound measured in KBr pellet or as a thin layer: ATR spectral bands are narrower and band relative intensities are also different from those of the absorption IR spectrum. That poses a challenge when IR spectra are searched in ATR library and vice-versa.
* Correspondence author; E-mail: plamen@uni-plovdiv.net.
Library search systems for MIR and Raman spectra and the corresponding spectral libraries are currently commercially available from both the instrument manufacturers and chemical software businesses together with search and retrieval software [4,5]. Also, vibrational spectral libraries consisting of several hundreds to thousands of spectra can easily be produced on a computerized instrument. Despite that, the number of commercially available and lab-produced spectra is way smaller than the number of currently known compounds (around 75 millions). That is why, the composition of new spectral collections is of a great importance for analytical chemistry.
This paper reports on a preparation of databases of FT-MIR, ATR FT-MIR and Raman spectra. Some of the spectra are measured in our lab but there are spectra taken from other sources. A library search in Raman spectral databases is thoroughly tested with spectra of individual organic compounds and their mixtures.
METHODS
The library search consists of comparison of a spectrum of the unknown with a collection of reference spectra of compounds with known structure [1,4-6]. The collection of spectra together with both chemical information and structure of the compounds is called a spectral library. The result obtained by a search, the so-called hitlist, is a list of spectra (hits) that are most similar to the query spectrum. The hits are sorted in the hitlist according to a real number reffered to as hit quality index (HQI) which reflects the similarity between the unknown and reference spectrum. If the unknown is among the library entries, then the correct answer often appears among the first several hits and a visual inspection of them makes it possible to identify the unknown: this is called identity search [6]. However, if the unknown compound is not among the library spectra, a more sophisticated interpretation of the hitlist is necessary [7]. Assuming that similar spectra indicate similar structures, the hitlist structures characterize the unknown structure: this is the essence of the similarity search [6].
For both types of searches important aspects are the representation of the spectra, the spectral similarity measures used and the search algorithm applied [1]; for the similarity search another key characteristic is the method of analysis of the hitlist structures. Other characteristics of a library search system include the speed and versatility of the implemented search algorithms, the size, contents and reliability of the database, the options for an update of spectral library, the availability of modules for analysis of the hitlist entries and the possibility to derive spectrum-structure correlations.
EXPERIMENTAL
The method for library search is implemented into a Windows-based user-friendly program, called IRSS [7,8]. Seven different algorithms for the comparison of IR spectra are implemented: three methods for peak matching [8,9] and four methods for comparing full spectral curves [7]. Furthermore, IRSS contains software tools for an import of IR spectra in JCAMP-DX format, peak picking, and an interactive analysis of IR spectra of mixtures based on multiple linear regression techniques.
Some of the MIR spectra (see Table 1) are registered on a Perkin-Elmer 1750 FT-IR Spectrometer from 4000 cm-1 to 450 cm-1 at resolution 4 cm-1 with 25 scans. All these spectra are subjected to curvilinear baseline correction and off-line converted into JCAMP-DX files. The latter are transferred to an IBM compatible computer with the standard protocol for data exchange KERMIT. The new MIR spectra (see Table 1) are measured on a VERTEX 70 FT-IR spectrometer (Bruker Optics) from 4000 cm-1 to 400 cm-1 at resolution 2 cm-1 with 25 scans. For all MIR spectra the solid samples are prepared as KBr pellets and liquids are measured as a capillary film between KBr plates.
The ATR FT-IR spectra are recorded with a VERTEX 70 FT-IR spectrometer; the spectrum is measured from 4500 cm-1 to 600 cm-1 at resolution 2 cm-1 with 16 scans. The used ATR accessory is MIRacle™ with a one-reflection ZnSe element (Pike). For solid samples the stirred crystals of
compound are pressed by an anvil to the reflection element; for liquid samples a drop is placed directly upon the crystal plate of the accessory.
The FT-Raman spectra are measured on Bruker Optics RAM II module coupled to a Vertex 70 FT-IR instrument with a focused laser beam of Nd:YAG laser (1064 nm) from 4000 cm-1 to 51 cm-1 at resolution 2 cm-1 with 25 scans: the laser power varies from sample to sample from 10 mW to 1000 mW. The solids are placed as stirred crystals in aluminium disc and liquid samples are recorded in standard NMR tubes.
A very large-size spectral library of 13 484 spectra, CC (Chemical Concepts), has been created earlier from the MIR spectra of SpecInfo database [7]; its composition is described there in detail.
Two libraries, an ATR and a Raman one, are composed from spectra measured in Julius KuhnInstitute, Beriln; these spectra are used with the kind permission of Dr. Gennady Gudy. The ATR FT-MIR spectra have been recorded in the wavenumber range from 4000 to 375 cm-1 with a portable ATR diamond crystal infrared spectrometer (Alpha, Bruker Optics GmbH, Ettlingen, Germany). The corresponding FT-Raman spectra are recorded on an RFS-100 Bruker FT-spectrometer.
The Bruker Demo FT-IR library of 350 entries (which is delivered with the OPUS software) have been converted into our spectral format.
For test purposes additional Raman spectra of (a) 20 individual compounds (see Table 2a), (b) two types of binary mixtures prepared with varying volume fraction (see Table 3), and (c) three individual compounds available as small crystals are measured (see Table 2b). The Raman spectra in (a) and (b) lists are measured on RAM II module as described above but the three tiny crystal samples are measured on RamanScope II microscope (with Nikon objective x10) coupled to RAM II module of the Vertex 70 FT-IR instrument.
All spectra except those registered on the old Perkin-Elmer instrument have been available on PC computer as ASCII files in JCAMP-DX format, v. 4.24. The spectra in JCAMP files are converted into our spectral format.
The structures are prepared with ISISDraw, v. 2.4 (MDL, Inc.): MOL files are exported by this software and read by IRSS to create the chemical structure files. The IRSS software and all spectral libraries created with our spectra are available from the corresponding author.
Table 1. The compiled spectral libraries. a) with spectra measured in our laboratory
b) with spectra from other sources
Lib. Name Spec. Type Num. of Spectra1 Intrument
IR01 MIR 105 / 105 PE
IR02 MIR 181 / 161 PE, Br
IR03 MIR 197 / 144 PE
IR04 MIR 179 / 179 PE
IR05 MIR 52 PE, Br
IR06 MIR 197 PE
IRSub MIR 55 / 37 PE, Br
ATR MIR 102 Br
Raman Raman 200 RamlI
Lib.
Name
Spec.
Type
Num. of Spectra
Intrument
CC3 MIR
Bruker MIR ATRp MIR
RamP
Raman
13 484 350 12 116
unknown PE Alpha RFS
1 The second entry is the number of old spectra.
RESULTS AND DISCUSSION
PE = Perkin-Elmer 1750 FT-IR Spectrometer; Br = Bruker FT-IR Vertex 70 Spectrometer; RamII = Bruker Optics RAM II module; Alpha = portable ATR diamond crystal infrared spectrometer; RFS = RFS-100 Bruker FT-spectrometer.
Chemical Concepts, SpecInfo MIR database.
In years 1992-98 three FT-IR libraries of 608 spectra have been created [10]. In 2011-12 with the support of the Bulgarian National Science Fund, Contract DDWU02/37, these databases have
2
been expanded with other nearly 500 FT-IR spectra and an FT-Raman library of 200 entries is also created. For the project NI13ChF006 (University of Plovdiv) an ATR FT-IR library of 102 spectra is composed [11]. All three types of spectral databases are maintained by the IRSS software that has been created for absorption FT-IR spectra. That is why, it is necessary to test the library search routines with Raman and ATR spectra; the IRSS software is thoroughly tested for absorption FT-IR spectra [9-10]. The comprehensive test of the performance for search of ATR FT-IR spectra (as unknowns) in absorption FT-IR spectral libraries and vise versa has been reported earlier [11]; here only the performance of the library search of Raman spectra in an FT-Raman database is described.
Table 2. The test spectra. The laser power of the corresponding reference spectra (Lib) and test spectrum (Unk) is given.
a) measured on Bruker Optics RAM II module.
# Compound Laser power / mW
Lib Unk
1. 1,4-Dioxane 700 500
2. 2,2,4-Trimethylpentane 500 100
3. Dichloromethane 500 100
4. 3-Methyl-1-butanol 500 100
5. n-Butyl acetate 500 50
6. 1 -Decanol 700 100
7. Dodecane 500 100
8. Phenyl-acetonitrile 200 20
9. Acetic acid linalylester 200 50
10. Squalene 500 100
11. 1 -Hydroxy-3 -nitrobenzene 100 20
12. 1 -Hydroxy-4-nitrobenzene 200 10
13. 3 -Nitro-benzamide 500 100
14. Diethyl ether 700 100
15. Carbon disulfide 400 100
16. Carbon tetrachloride 1000 100
17. Acetonitrile 500 100
18. Dimethyl sulfoxide 1000 100
19. 1 -Nitro-3 -trifluoromethyl-benzene 500 100
20. Tetracyanoethylene 100 50
b) measured on RamanScope (Bruker), x10.
# Compound Laser power / mW
Lib Unk
1. Oxalic acid 200 700
2. 2-(2,2,2-Trichloro-acetylamino)-benzamide 200 500
3. L-Histidin-methylester-dihydrochloride 500 500
As the absolute band intensity in the FT-Raman spectrum is dependent on the laser power, the Raman spectra vary nearly linear with the change of the latter [3]. As the higher laser power means higher signal to noise (S/N) ratio, the Raman spectra are preferably measured at highest laser power that does not decompose the sample (the so-called burning is observed only for solid samples). When we measured the library spectra we could allow ourselves to increase the power up to the burning level and use the last good spectrum as a library one. That is why, the library spectra are measured at very high laser power and are with the best-possible S/N ratio. But when the spectroscopist measures a sample of an unknown organic compound he/she try to preserve it
intact; usually the sample is recovered back from the disk - otherwise the advantage of Raman spectroscopy as non-destructive technique is lost. As a result, the Raman spectrum of an unknown is usually measured at a laser power lower than that of the library spectrum.
Figure 1. Raman spectra of oxalic acid: (1) measured on Ram II in aluminium disk, (2) registered on Raman microscope, x10.
Of course, library and unknown spectra are scaled in 0.0 - 1.0 range in ordinate but a different band intensity means also a different band width and sometimes two bands with peaks, close in wave number, merge at higher laser power into one. That is why, it is necessary to test the search algorithms with spectra of the library compounds that are measured at different laser power. This is done with the Raman spectra of 20 library compounds from Table 2a. All of them appear as first hit when searched in the library of 200 Raman spectra. The three spectra from Table 2b, measured on Raman microscope at laser power higher than that of the corresponding reference, are searched in the same library. Two of them are ranked as first hit, the other, oxalic acid, as second hit. (Here has to be mentioned that the microscope Raman spectra are usually measured at higher power because of their low S/N ratio; the latter could be seen from spectra in Figure 1.)
Table 3. Backward peak search results: hit position of the mixture components. Search tolerances are AA = 11 cm-1 and Av = 1.0 a.u.
a) mixtures of butyrophenone and valerophenone.
# Mixture composition v/v 1) Butyrophenone Valerophenone
1. 4:1 1 2
2. 3:2 1 2
3. 1:1 2 1
4. 2:3 2 1
5. 1:4 2 1
b) mixtures of cyclopentanone and benzylacetone.
# Mixture composition v/v 1) Cyclopentanone Benzylacetone
1. 4:1 1 10
2. 3:2 1 3
3. 1:1 1 2
4. 2:3 1 2
5. 1:4 9 1
i)
volume ratio
The Raman spectra of two series of binary mixtures are measured; their composition and the hit position of both components are given in Table 3. The components in the first series can be successfully identified but the 4:1 v/v and 1:4 v/v mixtures from the second series give worse results - the low-concentration component appear in 10 and 9 hit position, respectively. As the high-concentration component appear always as first hit, the problem could be solved with spectrum subtraction as it has been described in [9].
CONCLUSIONS
The created spectral libraries are of significant help in compound identification if the query compound has a library record. A search in Raman spectral library is thoroughly tested with spectra of individual organic compounds and their mixtures. Similarly to MIR spectra, the Raman spectra provides a fingerprint by which the compound or mixture component can be identified.
ACKNOWLEDGMENT
This work has been supported by the Bulgarian National Science Fund, Contract DDWU02/37 (IR & Raman Spectral Libraries) and Plovdiv University Project NI13ChF006 (ATR FT-IR Library). We are grateful to Prof. Kurt Varmuza (Technical University, Vienna) for providing the SpecInfo IR database.
REFERENCES
1. H. Luinge; Automated Interpretation of Vibrational Spectra. Vib. Spectrosc., 1990, 1, 3-18.
2. P. Larkin; Infrared and Raman Spectroscopy. Principles and Spectral Interpretation, Elsevier, 2011.
3. E. Smith and G. Dent; Modern Raman Spectroscopy - A Practical Approach. John Wiley & Sons, 2005.
4. W. Warr. Computer-Assisted Structure Elucidation. Part 1: Library Search and Spectral Data Colections. Anal. Chem, 1993, 65, A1045-A1050.
5. В. Вершинин, Б. Дерендяев, К. Лебедев; Компьютерная идентификация органических соединений. (монография, 182 стр.) Изд. «Наука», Москва, 2002.
6. J. Clerc; Automated spectra interpretation and library search systems. In: Meuzelaar, H.L.C., Isenhour, T.L. (Eds.), Computer-Enhanced Analytical Spectroscopy. Plenum, New York, 1987, 145-162.
7. K. Varmuza, P. Penchev, H. Scsibrany; Maximum Common Substructures of Organic compounds Exhibiting Similar Infrared Spectra. J. Chem. Inf. Comp. Sci., 1998, 38, 420-427.
8. P. Penchev, N. Kochev and G. Andreev; IRSS: A Programme System for Infrared Library Search. Compt. Rend. Acad. Bulg. Sci., 1998, 51, 67-70.
9. P. Penchev, V. Miteva, A. Sohou, N. Kochev, G. Andreev; Implementation and Testing of Routine Procedure for Mixture Analysis by Search in Infrared Spectral Library. Bulg. Chem. Commun., 2008, 40, 556-560.
10. P. Penchev; Application of Chemometric Methods for Identification of Organic Compounds from their Infrared Spectra. Ph. D. Thesis. Plovdiv, Bulgaria, 1998.
11. S. Tsoneva, S. Nachkova and P. Penchev; ATR Spectra Database of Organic Compounds, accepted for publication in НАУЧНИ ТРУДОВЕ НА РУСЕНСКИЯ УНИВЕРСИТЕТ, 2013.
12. P. Penchev, A. Sohou and G. Andreev; Description and Performance Analyses of an Infrared Library Search System. Spectrosc. Lett, 1996, 29, 1513-1522.
