M. Lo´pez-Pastor et al. / Journal of Molecular Structure 799 (2006) 146–152
147
studied by Raman spectroscopy. Malononitrile is a fre-
quently used reagent in organic synthesis because of its
exceptional reactivity due to its acid protons. Upon reac-
tion with bases malononitrile, as well as other CH acids
like malonate and cyanoacetic esters, form a carbanion act-
ing as reactive intermediate. Malononitrile is therefore
widely used in chemical synthesis especially in Knoevenagel
condensations and Michael additions [15–19]. The ionic
liquid emimBF4 has been chosen as solvent because this
ionic liquid already has found multiple use as reaction
medium in organic chemistry [20–22]. For analysis of the
in situ recorded Raman spectra and interpretation of the
course of reactions observed in the solvents under investi-
gation Two-Dimensional Correlation Spectroscopy as well
as Multivariate Curve Resolution have been used.
the basis of the freely available toolbox by Berry and Ozaki
[24]. The mean spectrum of the data set was used as refer-
ence spectrum to construct correlation maps in 2DCoS.
On the other hand, MCR decomposes the data set into
the product of two smaller matrices containing concentra-
tion profiles and spectra of the modeled components that
match best the recorded data set. The remaining spectral
information not explained by the modeled concentration
profiles and component spectra is captured in a residual
data matrix. When performing MCR on the experimental
data decision must be made as to how many components
are to be modeled. Furthermore, it is necessary to come
up with initial estimates for the spectra and/or concentra-
tion profiles prior to starting the iterative modeling process
of MCR. For this task chemometric tools like principle
component analysis (PCA), Evolving Factor Analysis
(EFA) among others are used [25]. During the iterative
modeling process constraints based on available physical
and chemical knowledge may be applied. These constraints
guide the iteration process to find chemically and physical-
ly meaningful results. Experimental data were treated by
MATLAB 5.3 software (The Maths Work Inc., Natick,
MA, USA) on the basis of the freely available toolbox by
Tauler and de Juan [26].
2. Experimental
2.1. Reagents and instrumentation
All the chemicals used were purchased from Fluka
(Steinheim, Germany), and they were of reagent grade.
The RTIL used, 1-ethyl-3-methylimidazolium tetrafluoro-
borate (emimBF4), was purchased from Solvent Innova-
tion GmbH (Ko¨ln, Germany) and used without any pre-
treatment. Reaction was carried out by mixing 10 ll of
malononitrile (1 M) in distilled water or pure RTIL and
5 ll of KOH (0.2 M) in distilled water. Solutions were host-
ed in a calcium fluoride plate where the reaction was mon-
itored with a Confocal Raman Microscope LabRam
HR800 (Jobin Ybon GmbH, Bernsheim, Germany). The
instrument was equipped with a CCD detector. The
632.817 nm He–Ne laser was used as excitation source
and the laser power was set to 14.5 mW. The resolution
of the Raman system was set at 0.9 cmÀ1. Each spectrum
was recorded with a duration of 14 s.
As the concepts of 2DCoS and MCR are fundamentally
different, the results they provide are complementary in
nature. It is therefore useful to combine both techniques
to obtain a complete picture of the process under study
[27].
3. Results and discussion
Malononitrile, 1, is a well-studied compound in organic
chemistry. It is easily deprotonated by strong bases such as
KOH [28,29]. This deprotonation yields a carbanion, that
exists in resonant structures, 2 and 3 (see Scheme 1).
Raman spectra recorded from the reaction between 1 and
KOH in emimBF4 are shown in Fig. 1. From the spectral
changes observed, it gets clear that the reaction proceeded
beyond simple deprotonation of 1. From the deprotona-
tion step itself, only two bands, one corresponding to both
nitriles of 2 and the other to the single nitrile of 3, should
be expected. However, as a more complicated spectral pat-
tern is observed it may be concluded that the reaction pro-
ceeds further with the carbanion as a plausible reactive
intermediate. At the beginning of the reaction a band
appeared at 2173 cmÀ1 with two further new bands arising
2.2. Data analysis
Two-Dimensional Correlation Spectroscopy (2DCoS)
and Multivariate Curve Resolution (MCR) are two com-
plementary data analysis techniques for elucidation of
spectral variations in evolving system. From a general
point of view, 2DCoS spreads the one-dimensional spectra
into a second spectral dimension with the same wavenum-
bers in both dimensions [23]. It provides both synchronous
and asynchronous correlation maps. While the former rep-
resents the simultaneous or coincidental spectral changes,
the latter one shows the correlation between bands chang-
ing with a different rate. Information on the temporal
sequence of the events causing these changes can also be
extracted when considering the signs of the asynchronous
peaks. Thus, the asynchronous map gives useful informa-
tion because it can help to identify overlapping bands in
case they change at a different rate during the experiment.
Raman spectra obtained were analyzed by MATLAB 5.3
software (The Maths Work Inc., Natick, MA, USA) on
at 2196 and 2158 cmÀ1
.
2DCoS was applied to the original data to identify
kinetically correlated bands, as well as, bands belonging
to different species. Synchronous and asynchronous maps
are shown in Fig. 2. The synchronous map reflects the
overall spectral changes. From the sign of the synchronous
peaks it can be said that the band at 2173 cmÀ1 decreases
when the bands at 2196 and 2158 cmÀ1 increase. A closer
look at the synchronous correlation plot revealed that the
synchronous peaks around 2273 cmÀ1 were slightly shifted.
López-Pastor, Mercedes
