new approaches to generalized two-dimensional correlation...
TRANSCRIPT
New Approaches to GeneralizedTwo-Dimensional Correlation Spectroscopy
and Its Applications
Young Mee Jung
Department of Chemistry, Kangwon National University, Chunchon,
Korea
Isao Noda
The Procter & Gamble Company, West Chester, Ohio, USA
Abstract: This review is focused on the recent approaches to generalized 2D
correlation spectroscopy, a technique widely used for the analysis of spectral data. A
brief introduction of generalized 2D correlation spectroscopy is described first. Then
the powerful combination of generalized 2D correlation spectroscopy and multivariate
chemometircs techniques, such as the data reconstruction by principal component
analysis (PCA), eigenvalue manipulation transformation (EMT), and self-modeling
curve resolution (SMCR) analysis are explored. Examples of successful applications
of new approaches to generalized 2D correlation spectroscopy are highlighted.
Keywords: Two-dimensional (2D) correlation spectroscopy, principal component
analysis, eigenvalue manipulating transformation, self-modeling curve resolution
INTRODUCTION
Generalized two-dimensional (2D) correlation spectroscopy has attracted a
high level of interest of analytical science community, as it provides considera-
ble utility and benefit in many fields of spectroscopic studies (1–8). In gene-
ralized 2D correlation spectroscopy, a sample is subjected to an external
Received 16 April 2006, Accepted 30 June 2006
Address correspondence to Young Mee Jung, Department of Chemistry, Kangwon
National University, Hyoja 2-dong, Chunchon 200-701, Korea. E-mail: ymjung@
kangwon.ac.kr
Applied Spectroscopy Reviews, 41: 515–547, 2006
Copyright # Taylor & Francis Group, LLC
ISSN 0570-4928 print/1520-569X online
DOI: 10.1080/05704920600845868
515
perturbation (e.g., temperature (9–21), pressure (22–24), concentration
(25–37), of pH (88) while being monitored with an appropriate spectroscopic
probe, such as IR (17, 18, 20, 25, 27, 39–68), Raman (21, 32–36, 69–74), NIR
(11–15, 28–32, 75–84), fluorescence (85–87), UV-Vis (88–90), and X-ray
(91, 92). The systematic variation in the spectral signal intensities induced
by the perturbation is then treated as a function of the physical variable repre-
senting the perturbation. 2D correlation is finding applications even outside of
conventional spectroscopic studies. For example, an interesting application of
generalized 2D correlation spectroscopy to gel permeation chromatography
was reported by Izawa et al. (93–95).
Some of the most notable features of generalized 2D correlation spectra
are simplification of complex spectra consisting of many overlapped peaks;
enhancement of spectral resolution by spreading peaks along the second
dimension; establishment of unambiguous assignments through the corre-
lation of bands of selectively coupled by various interaction mechanisms;
and determination of the sequence of the spectral peak emergence.
A very intriguing possibility in 2D correlation spectroscopy is the idea of
2D hetero-spectral correlation analysis, where two completely different types
of spectra obtained for a system using multiple spectroscopic probes under a
similar external perturbation are compared (35, 73, 74, 84, 91, 96–98). If there
is any commonality between the response patterns of system constituents
monitored by two different probes under the same perturbation, one should
be able to detect the correlation even between different classes of spectral
signals. There are two types of applications of 2D hetero-spectral correlation
analysis. One is the investigation of the correlation between closely related
spectroscopic measurement (35, 73, 74, 84, 97) as IR and Raman spectra,
which is very attractive from the point of better understanding of its comp-
lementary vibration spectra. It provides especially rich insight and clarifica-
tion into the in-depth study of vibrational spectra. Another is the hetero-
correlation between completely different types of spectroscopic or physical
techniques, which is useful for investigating the structural and physical prop-
erties of materials under a particular external perturbation (91, 96, 98). It has,
for example, been applied to X-ray absorption and Raman spectroscopy (91).
Hetero-spectral correlation has become one of the most active areas of
research in 2D correlation spectroscopy.
Recently, many interesting developments in generalized 2D correlation
spectroscopy have been proposed (82, 99–115). For example, sample-sample
2D correlation spectroscopy was promoted by Sasic et al. (82, 102–104). In
the sample-sample 2D correlation spectroscopy, which has two independent
sample axes defining the 2D correlation coordinates instead of traditional
variable axes (e.g., wavenumber or wavelength), the relationship between
different samples observed under different states of perturbation (time, tem-
perature, concentration, etc.) is studied by examining the similarity or difference
of their spectral trace patterns along the spectral variable. Information obtained
by conventional (variable-variable) and sample-sample 2D correlation
Y. M. Jung and I. Noda516
spectroscopy is often complementary, and general features of variable-variable
correlation maps are expected to be equally applicable to the sample-sample cor-
relation maps. Recently, to further extend the concept of sample-sample 2D cor-
relation spectroscopy, Wu et al. proposed hybrid 2D correlation spectroscopy for
two spectral data sets obtained separately, which can potentially explore the
latent correlation between different perturbation variable (108, 109). Classical
statistical analysis (99–101) and global phase 2D maps (105–107) were also
explored by Morita et al., which is much more straightforward to extract pure
correlation information. A very powerful recent development in generalized
2D correlation spectroscopy is the incorporation of multivariate chemometrics
techniques proposed by Jung et al., (110–115).
In this review, the principle of generalized 2D correlation spectroscopy
and recent developments in the field are reviewed. The fruitful combination
of multivariate chemometircs techniques and generalized 2D correlation spec-
troscopy is discussed in detail, and its applications are explored.
BACKGROUND
Generalized 2D Correlation Spectroscopy
In a generalized 2D correlation spectroscopy experiment, a series of a
perturbation-induced dynamic spectra are collected first in a systematic
manner, e.g., in sequential order during a process. Such a set of spectra
y(v, t) observed as a function of the perturbation variable t (e.g., time, temp-
erature, or concentration) during the interval between Tmin and Tmax, is then
transformed into a set of 2D correlation spectra by a form of cross correlation
analysis. The dynamic spectrum y(v, t) of a system induced by the application
of an external perturbation is formally defined as
~yðv; tÞ ¼yðv; tÞ � �yðvÞ for Tmin � t � Tmax
0 otherwise
�ð1Þ
where y(v) is the reference spectrum of the system. Reference spectrum is
customary to set y(v) to be the stationary or averaged spectrum given by
�yðvÞ ¼1
Tmax � Tmin
ðTmax
Tmin
yðv; tÞ dt ð2Þ
The intensity of 2D correlation spectrum X(v1, v2) represents the quantitative
measure of a comparative similarity or dissimilarity of spectral intensity varia-
tions y(v, t) measured at two different spectral variables, v1 and v2, during a
fixed interval. In order to simplify the mathematical manipulation, X(v1, v2)
is treated as a complex number function
Xðv1; v2Þ ¼ Fðv1; v2Þ þ iCðv1; v2Þ ð3Þ
Generalized Two-Dimensional Correlation Spectroscopy 517
comprising two orthogonal (i.e., real and imaginary) components, known
respectively as the synchronous and asynchronous 2D correlation intensities.
The generalized 2D correlation function given below
Fðv1; v2Þ þ iC ðv1; v2Þ ¼1
p ðTmax � TminÞ
ð1
0
~Y1ðvÞ � ~Y�
2ðvÞdv ð4Þ
formally defines the synchronous and asynchronous correlation intensity
introduced in Eq. (3). The term Y1(v) is the forward Fourier transform of
the spectral intensity variations y(v1, t) observed at a given spectral variable
v1 with respect to the external variable t. It is given by
~Y1ðvÞ ¼
ð1
�1
~yðv1; tÞe�ivtdt
¼ ~YReðvÞ þ i ~Y
Im
1 ðvÞ ð5Þ
where Y1Re (v) and Y1
Im (v) are, respectively, the real and imaginary com-
ponents of the Fourier transform. It is useful to remember that the real
component Y1Re(v) is an even function of v, while Y1
Im (v) is an odd
function. The Fourier frequency v represents the individual frequency
component of the variation of y(v1, t) traced along the external variable t.
According to Eq. (1), the above Fourier integration of the dynamic
spectrum is actually bound by the finite interval between Tmin and Tmax.
The conjugate of the Fourier transform Y2
�
(v) of the spectral intensity
variation y(v2, t) observed at another spectral variable v2 is given by
~Y�
2ðvÞ ¼
ð1
�1
~yðv2; tÞeþivtdt
¼ ~YRe
2 ðvÞ þ i ~YIm
2 ðvÞ ð6Þ
Once the appropriate Fourier transformation of the dynamic spectrum y(v1, t)
defined in the form of Eq. (1) is carried out with respect to the variable t,
Eq. (4) will directly yield the synchronous and asynchronous correlation
spectrum, F(v1, v2) and C(v1, v2).
The intensity of a synchronous 2D correlation spectrum represents the
simultaneous or coincidental changes of two separate spectral intensity varia-
tions measured at v1 and v2 during the interval between Tmin and Tmax of the
externally defined variable t. The intensity of peaks located at diagonal
positions mathematically corresponds to the autocorrelation function of
spectral intensity variations observed during an interval between Tmin and
Tmax. The diagonal peaks are therefore referred to as autopeaks, and the
slice trace of a synchronous 2D spectrum along the diagonal is called the
autopower spectrum. The magnitude of an autopeak intensity, which is
always positive, represents the overall extent of spectral intensity variation
observed at the specific spectral variable v during the observation interval
Y. M. Jung and I. Noda518
between Tmin and Tmax. Thus, an autopeak represents the overall susceptibility
of the corresponding spectral region to change in spectral intensity as an
external perturbation is applied to the system. Cross-peaks located at the
off-diagonal positions of a synchronous 2D spectrum represent simultaneous
or coincidental changes of spectral intensities observed at two different
spectral variables v1 and v2. Such a synchronized change, in turn, suggests
the possible existence of a coupled or related origin of the spectral intensity
variations. The sign of synchronous cross-peaks becomes positive if the
spectral intensities at the two spectral variables corresponding to the coordi-
nates of the cross-peak are either increasing or decreasing together as
functions of the external variable t during the observation interval.
However, the negative sign of cross peaks indicates that one to the spectral
intensities is increasing while the other is decreasing.
The intensity of an asynchronous 2D correlation spectrum represents
sequential or successive, but not coincidental, changes of spectral intensities
measured separately at v1 and v2. Cross-peaks develop only if the intensity
varies out of phase with each other for some Fourier frequency components
of signal fluctuations. The sign of asynchronous cross-peak provides useful
information on sequential order of events observed by the spectroscopic
technique along the external variable. If the signs of synchronous and asyn-
chronous cross-peaks are the same, the intensity change at v1 occurs before
v2. If the signs of synchronous and asynchronous cross-peaks are different,
the intensity change at v1 occurs after v2.
Principal Component Analysis–Based 2D (PCA 2D) CorrelationSpectroscopy
It is natural to consider the combined use of various multivariate chemo-
metrics tools and 2D correlation spectroscopy. However, most workers have
focused their attention to the parallel or sequential use of 2D correlation
analysis (1–8) and chemometrics (116–118) as complementary but essen-
tially independent data analysis techniques (59, 77, 79, 81, 119, 120). Jung
et al. have recently reported the promising possibility of the combination of
2D correlation spectroscopy and principal component analysis (PCA) (110–
115). In this approach, PCA treatment of spectral data set now becomes an
essential and integral part of the subsequent 2D correlation analysis.
Let us set the raw data matrix A, comprising the original set of pertur-
bation-dependent spectra, to be an n � m matrix with n spectra and m wave-
number points. The loading matrix V is an m � r matrix, where each column
is the loading (i.e., the eigenvector of the dispersion matrix ATA) obtained by
PCA. Here, AT stands for the transpose of A. The total number of loadings r
selected for the analysis must be less than or equal to n. It is customary to
normalize each column of V (i.e., each loading), such that the product VT
V is an identity matrix. Associated with the PCA loading factors are
Generalized Two-Dimensional Correlation Spectroscopy 519
the scores (sometime called latent variables). Score matrix W is a relatively
small n � r matrix.
The fundamental idea of factor analysis based on PCA is that the signifi-
cant part of the data matrix can be expressed as the product of score and
loading matricies
A ¼W VT þ E ¼ A�� þ E ð7Þ
where E is the residual matrix often associated with pure noise. Thus, the
matrix product A�� can be regarded as the noise-free reconstructed data
matrix of the original data A.
A�� ¼W VT ð8Þ
In generalized PCA-based two-dimensional (PCA 2D) correlation
analysis, the reconstructed data matrix A�� is utilized for the calculation of
2D correlation spectra instead of the original data matrix. The analysis of
this new data matrix, reconstructed strictly from a few selected significant
scores and loading vectors of PCA, has been proposed as a useful method
to substantially improve the data quality for 2D correlation analysis by
focusing the attention more efficiently to the dominant feature of spectral
intensity variations. Furthermore, the 2D correlation analysis of such a recon-
structed data matrix accentuates only the most important features of synchro-
nicity and asynchronicity without being hampered by the noise, or even by
minor signal components, if the number of principal components is restricted.
Eigenvalue Manipulating Transformation (EMT)
Jung et al. have also proposed a new concept of eigenvalue manipulating
transformation (EMT) for the generalized PCA-2D correlation analysis
(111–113). The PCA-reconstructed data matrix A�� is expressed in the
familiar form known as the singular value decomposition (SVD) (116–118).
A�� ¼ USVT ð9Þ
and
S ¼ L1=2 ð10Þ
where U, S, and V are, respectively, the orthonormal score matrix, diagonal
matrix containing the singular values, and loading matrix. Here L ¼W0 Wis a diagonal matrix where each diagonal element corresponds to the eigen-
value of principal component. The score matrix W is expressed in the form
W ¼ U S and can be obtained directly from W ¼ A V.
Y. M. Jung and I. Noda520
The new transformed data matrix A���� will be obtained by manipulating
and replacing eigenvalues of A�� as
A���� ¼ US����VT ð11Þ
where S�� is given by varying the corresponding eigenvalues in S by raising
them to the power of m.
S���� ¼ Sm ð12Þ
The concept of EMT of spectral data involves the systematic substitution of
individual eigenvalues associated with the original data set. This process
will generate a new set of transformed data with a very different emphasis
place on specific information content. This new EMT-reconstructed data
matrix A���� will be used instead of A�� for the calculation of enhanced 2D
correlation spectra.
By uniformly raising the power of a set of original eigenvalues, the
influence of factors associated with major eigenvalues becomes more
prominent, while the minor eigenvectors primarily arising from the noise
component are no longer strongly represented (111). Thus, this transformation
of the data matrix becomes a gradual noise reduction scheme with attractive
flexibility of continuously fine-tuning the balance between the desired noise
suppression and retention of pertinent spectral information.
However, by uniformly lowering the power of a set of eigenvalues associ-
ated with the original data, the smaller eigenvalues becomes more prominent
and the contributions of minor components but potentially interesting factors
become amplified (112, 113). Thus, much more subtle difference of spectral
behavior for each component is now highlighted.
When the power parameter m is reduced to near zero, the singular value
matrix S is essentially replaced with the identity matrix I. The final trans-
formed data matrix approaches A���� ¼ U VT. The synchronous 2D correlation
analysis of such reconstructed data generates the so-called projection
spectrum, F ¼ V VT (associated with the subspace spanned by the set of
object eigenvectors) instead of more familiar form of the synchronous 2D cor-
relation spectrum based on the covariance or dispersion matrix F ¼ AT A.
The intensity at the diagonal of a synchronous 2D projection spectrum corre-
sponds to the squared Mahalanobis distance of the spectral intensity variations
of the data set from the reference (mean) spectrum.
In more advanced case of the spectral selectivity enhancement, where the
power parameter m is now taken to be a negative number, EMT operation will
further exaggerate the effect of minor PC factors such that asynchronicity is
very strongly accentuated. When the power parameter m becomes 21, the
diagonal matrix S is replaced with S21, which by definition is the inverse
matrix of S. It is interesting to point out that the corresponding synchronous
spectrum V S22 VT is nothing but the Moore-Penrose pseudo inverse of the
conventional 2D synchronous spectrum V S2 VT. In such an inverted 2D
Generalized Two-Dimensional Correlation Spectroscopy 521
spectrum, the information content associated with small eigenvalues of the
original data matrix is preferentially represented, while the contribution
from the dominant variance is mostly attenuated.
The elimination of the contribution of the first PC, which is closely
aligned to the autopower spectrum, results in the PC1-free data matrix A����
given by
A���� ¼ A�� � s1U1VT1 ð13Þ
where s1 is the first (i.e., most dominant) singular value of the data matrix,
and U1 and V1 are the first eigenvectors of matrices A AT and AT A.
The motivation behind the elimination of PC1 is similar to that of the
negative power EMT discussed earlier. The beneficial outcome of such
selective manipulation of data is the efficient detection and discrimination
of individual features of spectral intensity variations. Thus, highly overlapped
signals with only slight differences in the variational pattern can be readily
identified. It is noted that, the obtained PC1-free data matrix A�� is the
same as that reconstructed by using only PC2 and PC3 in regular PCA 2D
correlation analysis.
2D Correlation–Based Self-Modeling Curve Resolution (SMCR)
Analysis
Furthermore, Jung et al. have reported the use of 2D correlation spectroscopy
in conjunction with alternating least-squares (ALS)-based self-modeling
curve resolution (SMCR) analysis of spectral data see (114). A classical
problem of SMCR is to extract a set of concentration profiles and spectra of
pure components from a set of unknown mixture spectra A without any
prior knowledge about the system.
For a given spectral data matrix A, where each row corresponds to a
spectrum of a mixture, one have the following bilinear form
A ¼ TPT þ E ð14Þ
Here, T and P are the score and loading vector matricies comprising the
information on concentration profile of pure components and their spectra,
respectively; PT stands transpose of P, and E is the residuals comprising
mainly noise. The total number of pure variables should be the same as the
total number of components to be modeled. It can be estimated by the
simple PCA or SVD, where the number of “significant” eigenvectors will
be equated to the number of components.
Asynchronous 2D correlation spectrum will provide the excellent can-
didates for the most distinct variation points of spectral intensities (corre-
sponding closely to the concept of pure variables). Although they may not
Y. M. Jung and I. Noda522
be the best estimates in terms of the lack of signal overlaps, the correlation
peak base band positions will give the clearest idea of how the most
dominant individual component intensities are changing with samples.
Let us start the ALS iteration as
P ¼ ATðTT TÞ�1ð15Þ
P is the estimated loading matrix calculated from the initial guess of T. One
then replace all the negative elements of P with zero according to the
so-called non-negativity constraint for spectral intensities. One then nor-
malizes each column of P by dividing the Euclidian norm (i.e., square
root of pTp).
One can then estimate new T from the constrained and normalized P as
T ¼ AT PðPT PÞ�1ð16Þ
and then replace all negative elements with zero according to the non-negative
constraint for component concentrations. The process is repeated until T and P
are stabilized. If the above process converges successfully, one should have T
and P consisting strictly with non-negative elements. The ALS-estimated
scores and loading matricies come very closely to the true concentrations
and pure spectra of individual chemical components. Thus, the reconstructed
data matrix can be
A� ¼ TPT ð17Þ
Applications of PCA 2D Correlation Spectroscopy
In this review, we give an overview of applications of new approaches to
generalized 2D correlation spectroscopy very briefly, instead of reviewing
extensive amount of its applications. Two examples of recent applications
of PCA 2D correlation are discussed. One is the applications of PCA 2D cor-
relation analysis to experimental spectra with a finite but unknown level of
noise contribution (115, 121). This work demonstrates the potential of PCA
2D correlation analysis for noise filtering effect. Another is the application
of PCA 2D correlation spectroscopy with EMT operations for spectral selec-
tivity enhancement (122, 123). EMT technique which uniformly lowers the
power of a set of eigenvalues associated with the original data highlights
the subtle contributions from minor eigenvectors. Thus, subtle differences in
the thermal responses of polystyrene-block-poly(n-pentyl methacrylate)
(PS-PnPMA) copolymer, which are difficult to observe by conventional 2D
correlation analysis, are accentuated much more strongly than the original
data.
Generalized Two-Dimensional Correlation Spectroscopy 523
Application of PCA 2D Correlation Spectroscopy for Noise Filtering
Effect
The PCA 2D correlation spectroscopy was applied to the time-dependent
FTIR spectra of a mixture of methyl ethyl ketone (MEK), deuterated
toluene, and polystyrene (PS) during the solvent evaporation, which were
combined with a substantial amount of artificial noise to emphasize the
practical benefit of this technique (121).
Figure 1 shows the original data set examined in this study, where sub-
stantial amount of artificial noise is injected to the raw transient IR spectra
of a PS/MEK/toluene solution mixture during the solvent evaporation
process. A detailed information of these spectra including bands assignment
was already given elsewhere (4) This original spectral data set was decom-
posed into the scores and loading vectors by standard PCA analysis. PCA
factor 1 (PC1), factor 2 (PC2), and factor 3 (PC3) accounts, respectively,
for 83.0, 16.1, and 0.2% of the total variance of spectral intensities along
the time axis.
Figure 1. Synthetic noisy spectra in which substantial amount of artificial noise is
injected to the time-dependent FTIR spectra of a mixture of a PS/MEK/toluene
solution mixture during solvent evaporation, measured every minute over a time
range of 0 to 12 min, in the region of 1320–1520 cm21. Reproduced with permission
from Bulletin of the Korean Chemical Copyright Society, 24 (9): 1345. Copyright
(2003) Korean Chemical Society.
Y. M. Jung and I. Noda524
Figure 2 displays the conventional asynchronous 2D correlation spectrum
from these original spectra in Figure 1. The basic properties and interpreta-
tional procedure of asynchronous 2D correlation spectrum have already
been described in more detail previously (4), so no further in-depth discussion
will be given here. In this demonstration, the asynchronous 2D correlation
spectrum is focused, because the asynchronous 2D correlation spectrum is
often contaminated heavily by artifactual peaks attributed to the fortuitous
correlation of noise, while the synchronous 2D correlation spectrum is
relatively insensitive to noise contribution.
The reconstructed data matrix A�� obtained by Eq. 8 from the only two
principal components was used instead of the original spectral data matrix
A for the subsequent 2D correlation analysis. Figure 3(a) depicts the PCA-
reconstructed spectra represented in the matrix A�� from loading vectors and
scores of PC1 and PC2 with average spectrum added back. Figure 3(b)
shows the raw transient IR spectra of a PS/MEK/toluene solution mixture
during the solvent evaporation process for comparison. It shows clearly that
the reconstructed spectra (Figure 3(a)) are sufficiently close enough to the
original spectra (Figure 3(b)), suggesting that most of the pertinent infor-
mation is retained in the PCA-reconstructed data.
Figure 2. Conventional asynchronous 2D correlation spectrum obtained from the
spectra in Figure 1. Solid and dashed lines represent positive and negative cross-
peaks, respectively. Reproduced with permission from Bulletin of the Korean Chemical
Society, 24 (9): 1345 Copyright (2003) Korean Chemical Society.
Generalized Two-Dimensional Correlation Spectroscopy 525
Figure 4 depicts the asynchronous 2D correlation spectrum generated
from the PCA 2D reconstructed spectral data matrix A�� with PC1 and PC2.
It is similar to the conventional 2D correlation spectrum in Figure 2, but is
remarkably less noisy and somewhat simplified. The noise component E
can be successfully truncated from A to calculate A�� by applying classical
PCA treatment to the original data without using a mathematical smoothing
filter. This result confirms again that the key features of 2D correlation
spectrum can be faithfully reconstructed from PC loading vectors and scores.
In addition, finer features appearing in the asynchronous 2D correlation
spectrum now can be taken as real manifestations with a much higher level
of confidence, since the possibility of noise-induced correlation intensity
effect is now substantially eliminated.
Application of PCA 2D Correlation Spectroscopy with EMT
Operation for Spectral Selectivity Enhancement
Polystyrene-block-poly(n-pentyl methacrylate) [PS-PnPMA] exhibits a
closed-loop phase behavior bounded by a lower disorder-to-order transition
(LDOT) at lower temperature and an upper order-to-disorder transition
(UODT) at higher temperature. The analysis of conventional 2D IR
Figure 3. (a) PCA-reconstructed spectra from loading vectors and scores of PC1 and
PC2 with average spectrum added back. Reproduced with permission from Bulletin of
the Korean Chemical Society, 24 (9): 1345. Copyright (2003) Korean Chemical Society
(b) The raw time-dependent IR spectra of a PS/MEK/toluene solution mixture during
the solvent evaporation process. Reproduced with permission from Applied
Spectroscopy, 57 (5): 564. Copyright (2003) Society for Applied Spectroscopy.
Y. M. Jung and I. Noda526
correlation spectra revealed that the conformation of C-C-O group of PS-
PnPMA is changed near the transitions and two disordered states occurring
at lower and higher temperatures are different(20). Jung et al. then demon-
strated that the phase behavior of PS-PnPMA is more clearly understood by
using PCA 2D correlation spectroscopy through EMT technique (123). To
better understand the different characteristics of the two disordered states
found in PS-PnPMA more clearly, PCA 2D correlation spectroscopy
through EMT of spectral data set was applied to temperature-dependent IR
spectra of PS-PnPMA copolymer.
Figure 5 shows the temperature-dependent IR spectra of PS-PnPMA
measured during heating from 100 to 2608C at an interval of 58C. The
original spectral data set in Figure 5 was decomposed into the scores and
loading vectors by standard PCA analysis. PCA factor 1 (PCl), factor 2
(PC2), and factor 3 (PC3) account for 98.4, 1.3, and 0.2%, respectively, of
the total variance of spectral intensities along the time axis. The reconstructed
data matrix A�� obtained by Eq. 8 from the three principal components was
used instead of the original raw spectral data matrix A for the subsequent
2D correlation analysis.
Figure 4. Asynchronous 2D correlation spectrum obtained using reconstructed data
from the reconstructed spectral data matrix A� with PC1 and PC2. Solid and dashed
lines represent positive and negative cross-peaks, respectively. Reproduced with per-
mission from Bulletin of the Korean Chemical Society, 24 (9): 1345. Copyright
(2003) Korean Chemical Society.
Generalized Two-Dimensional Correlation Spectroscopy 527
Since synchronous 2D correlation spectra constructed from these raw
spectra are already reported (20), the interpretation of conventional synchro-
nous 2D correlation spectra is not further discussed. For comparison, conven-
tional synchronous 2D correlation spectra for two disordered states are shown
in Figure 6. Figure 7 shows the spectrum of the new reconstructed data
obtained by varying the value of the power parameter m, as m ¼ 1, 1/2,
1/3, etc. Such a data matrix will emphasize the subtle contributions from
minor eigenvectors much more strongly than the original data. It looks like
the EMT effect is already apparent at m ¼ 1/2, well before reaching the
m ¼ 0 condition.
Synchronous 2D correlation spectra generated from the EMT-recon-
structed spectral data matrix A���� obtained by replacing the original eigen-
values with m ¼ 1/2 for disordered state at lower temperature, ordered
state, and disordered state at higher temperature are shown in Figures 8(a),
(b) and (c), respectively. By lowering the power of a set of eigenvalues associa-
ted with the original data, the contribution of the minor but potentially interes-
ting factors such as hidden property of phase transition is greatly accentuated
compared with conventional 2D correlation spectra. As shown in Figure 8, the
EMT effect for spectral selectivity enhancement is very much apparent. The
Figure 5. The temperature-dependent IR spectra of polystyrene-block-poly(n-pentyl
methacrylate) (PS-PnPMA) measured during heating from 100 to 2608C at an interval
of 58C Reprinted from Journal of Molecular Structure, published on the Web, April 24,
2006. Copyright (2006), with permission from Elsevier.
Y. M. Jung and I. Noda528
synchronous spectrum of the ordered state is completely different from those
in the two disordered states. When the EMT effect is included, the synchro-
nous spectra for two disordered states are also clearly different. The
intensity of peaks located at the diagonal positions in the synchronous 2D
spectrum represents the overall susceptibility of the corresponding spectral
region to change in spectral intensity as an external perturbation is applied
to the system. The power spectrum, extracted along the diagonal line of the
synchronous 2D correlation spectrum, given in the top of Figures 8(a) and
Figure 6. Conventional synchronous 2D correlation spectra obtained from the raw
spectra in Figure 5 for disordered states in lower (a) and higher (b) temperatures.
Solid and dashed lines represent positive and negative cross peaks, respectively.
Reprinted from Journal of Molecular Structure, published on the Web; April 24,
2006. Copyright (2006), with permission from Elsevier.
(continued )
Generalized Two-Dimensional Correlation Spectroscopy 529
8(c), reveals that the intensity changes with temperature for bands at 1185,
1198, 1386, 1596, 1721, and 1750 cm21 at lower temperatures are larger
than those at higher temperatures. On the other hand, the intensity change
with temperature for bands at 1496 and 1456 cm21, assigned to phenyl ring
stretching of PS, at lower temperatures is smaller than that at higher tempera-
tures. Intensities of bands from C–CiO stretching, C–H deformation, and
C55O stretching of PnPMA and those from phenyl ring of PS change
greatly at lower and higher temperature, respectively. The distinct differences
in two disordered states in the cross correlations of the bands from phenyl
group in PS with that fiom C–C–O group in PnPMA are clearly observed.
These results again confirm that the conformation of PS-PnPMA in the two
disordered states is different, and that the weak directional interaction
between phenyl group of PS and the side chain of PnPMA in the two
Figure 6. Continued.
Y. M. Jung and I. Noda530
Figure 7. Spectra of the EMT-reconstructed data obtained by replacing eigenvalues from loadings and
scores of PC1, PC2, and PC3. Each spectrum is obtained by varying the value of power parameter m, as
m¼1 (a), 1/2 (b), 1/3 (c), and 0 (d).
Gen
eralized
Tw
o-D
imen
sion
al
Co
rrelatio
nS
pectro
scop
y5
31
disordered states is different. This kind of difference was not clearly observed
in conventional 2D correlation spectra, as shown in Figure 6. Therefore, the
EMT technique with even m ¼ 1/2 could distinguish the very subtle differ-
ences of spectra which are not detected by conventional 2D correlation
spectra. Consequently, PCA 2D correlation spectroscopy with EMT
technique provides a very promising way for studying various systems.
Figure 8. Synchronous PCA2D correlation spectra obtained from the EMT recon-
structed data with m ¼ 1/2 for disordered state in lower temperature (a), ordered
state (b), and disordered state in higher temperature (c), respectively. Solid and dashed
lines represent positive and negative cross peaks, respectively. Reprinted from Journal
of Molecular Structure, published on the Web, April 24, 2006. Copyright (2006), with
permission from Elsevier.
(continued )
Y. M. Jung and I. Noda532
Application of 2D Correlation–Based Self-Modeling Curve
Resolution (SMCR) Analysis for Interpreting
Highly Convoluted Spectra
Jung et al. have demonstrated a great potential of the application of 2D
correlation-based self-modeling curve resolution (SMCR) analysis for inter-
preting highly convoluted spectra (124–126).
The system used in this article was electrochemical reduction of p-benzo-
quinone (p-BQ) in acetonitrile, which produces anion radicals and dianions at
its first and second reduction potentials (125). The dianions undergo a fast
comproporationation reaction with neutral p-BQ molecules to produce
anion radicals back, complicating the spectral analysis. For this reason, the
Figure 8. Continued.
Generalized Two-Dimensional Correlation Spectroscopy 533
absorption band of a large amount of Q2† might have buried those resulting
kom Q22 or other species, which makes the spectroelectrochemical data unin-
terpretable by a simple examination of the spectra when the data are obtained
in a bulk cell, where reaction (QþQ22 O 2Q2†) is allowed to occur. In
general, the recording of Q22 spectrum was possible in the optically transpar-
ent thin layer electrochemical (OTTLE) cell because exhaustive electrolysis
was achieved in it and no neutral p-BQ molecules were available for the
following reaction. However, Kim et al. reported a complete interpretation
of the highly convoluted in situ spectroelectrochemical data obtained during
p-BQ reduction by 2D correlation analysis technique in conjunction with
the SMCR analysis.
Figure 8. Continued.
Y. M. Jung and I. Noda534
Figure 9 shows a series of spectra recorded in the potential region
between 21.1 and 21.8 V, where Q2† is reduced to Q22. A dianion peak,
which can be clearly observed at 370 nm in OTTLE cell, was not identified
in Figure 9. To interpret the spectra shown in Figure 9, 2D correlation
analysis was performed on them. Figure 10 shows the 2D correlation
spectra for those shown in Figure 9.
The analysis of 2D correlation spectra reveals that the broad autopeak at
342 nm in the synchronous 2D correlation spectrum is clearly resolved into
two peaks, one at 330 and the other at 351 nm in the asynchronous 2D corre-
lation spectrum. The band at 351 nm is influenced by the band at 330 nm, and
it appears to have originated from Q22, which is not readily noticeable in the
1 D spectra. We also determine the order of peak emergence of (416 and
442)! 330! 351 nm from the analysis of the asynchronous 2D correlation
spectrum.
To confirm that the band at 351 nm observed by 2D correlation analysis
indeed arises from Q22, 2D correlation based SMCR analysis was employed
to extract individual spectra from the spectral data obtained between 21.1 and
21.8 V. For the successful fitting of the spectral data by ALS iteration, of par-
ticular importance is the initial selection of the pure variables. As an initial
pick of the pure variables, the dominant cross peaks in the asynchronous 2D
correlation spectrum (Figure 10(b)) representing the pure variable bands;
i.e., (330, 351), (351, 416), and (351, 442) nm, were selected.
Figure 9. A series of spectra recorded in the potential region between 21.1 and 21.8 V,
where Q2† is reduced to Q22.
Generalized Two-Dimensional Correlation Spectroscopy 535
Figure 10. (a) Synchronous and (b) asynchronous 2D correlation spectra obtained from the spectra the
potential scan between 21.1 and 21.8 V (Figure 9). Solid and dotted lines indicate regions of positive and
negative correlation respectively. Reproduced with permission from Analytical Chemistry, 76: 5236.
Copyright (2004) American Chemical Society.
Y.
M.
Ju
ng
an
dI.
No
da
53
6
Figure 11. (a) The spectra of individual chemical components obtained from 2D correlation spectroscopy based
SMCR analysis of the spectra of BQ shown in Figure 10 between 21.1 and 21.8 V and (b) their concentration profiles.
Solid, dashed, and dotted lines represent BQ2†, BQ22, and BQH†, respectively. Reproduced with permission from
Analytical Chemistry, 76: 5236. Copyright (2004) American Chemical Society.
Gen
eralized
Tw
o-D
imen
sion
al
Co
rrelatio
nS
pectro
scop
y5
37
The spectra of individual chemical components obtained from the SMCR
analysis and their concentration profiles are shown in Figures 11(a) and (b),
respectively. The SMCR-estimated spectra thus obtained show well
resolved spectra of three species from what appeared to be a single
spectrum. The 351 nm band assigned to BQ22 in the 2D correlation
analysis actually corresponds to the broad band at around 356 nm in the
SMCR-extracted spectrum. A reasonable set of concentration profiles
during the potential sweep between 21.1 and 21.8 V was also obtained. It
is seen from this figure that the intensity of the Q2† band at 320 nm (solid)
stays nearly constant, while both that for Q22 at 356 nm (dashed) and the
absorption at 330 nm (dotted), which is very similar to that of Q2†,
increases until both of them reach equilibrium concentrations at the increas-
ingly negative potential. Note that the 330 nm band has already a fairly
high concentration at 21.1 V.
It is suggested that the effective decay of Q2† via the protonation reaction
is responsible for the steady state concentration of Q2†, because the concen-
tration of the species absorbing at 330 nm is already fairly high at 21.1 V
where the dianion is not even produced. Also, this species is already
generated well before Q22 is generated as pointed out above. It is therefore
concluded that the species represented by the dotted spectrum is protonated
Q2† or QH† produced from the reaction of Q2† with a trace amount of Hþ
in acetonitrile. It is the first time that the QH† is identified spectroscopically
during the reduction of p-BQ, which explains many observations described
in the literature. Results obtained in the buk cell using the 2D and SMCR
analysis techniques is in excellent agreement with that obtained in the
OTTLE cell in that the dianion spectrum was observed and extracted.
However, the spectrum from a third species (QH†) was also extracted by
the SMCR technique, which had not been possible from a simple examination
of the spectra obtained in the OTTLE cell. In conclusion, the 2D correlation
analysis used along with the SMCR analysis offers a powerful technique for
complete analysis of complex spectroelectrochemical data due to following
chemical reactions.
ACKNOWLEDGMENTS
We thank all of our colleagues who have provided us valuable contributions to
the completion of this review paper.
REFERENCES
1. Noda, I. (1993) Generalized two-dimensional correlation method applicable toinfrared, Raman, and other types of spectroscopy. Appl. Spectros., 47: 1329.
Y. M. Jung and I. Noda538
2. Noda, I. (2000) Determination of two-dimensional correlation spectra using the
Hibert transform. Appl. Spectros., 54: 994.
3. Ozaki, Y. and Noda, I. (eds.) Two Dimensional Correlation Spectroscopy;
American Institute of Physics: Melville, NY.
4. Noda, I., Dowrey, A.E., Marcott, C., Story, G.M., and Ozaki, Y. (2000) General-
ized two-dimensional correlation spectroscopy. Appl. Spectrosc., 54: 236A.
5. Appl. Spectros., 2000, 54 (7) (Special Issue on Two-Dimensional Correlation
Spectroscopy).
6. Noda, I. (2002) General theory of two-dimensional (2D) Analysis. In Handbook
of Vibrational Spectroscopy; Chalmers, J.M. and Griffith, P.R. (eds.), John Wiley
& Sons, Ltd: Chichester; Vol. 3, 2123–2134.
7. Ozaki, Y. 2D Correlation spectroscopy in vibrational spectroscopy. In Handbook
of Vibrational Spectroscopy; John Wiley & Sons, Ltd: Chichester; Vol. 3,
2135–2172.
8. Noda, I. and Ozaki, Y. (2004) Two-Dimensional Correlation Spectroscopy:
Applications in Vibrational Spectroscopy; John Wiley & Sons, Inc: New York.
9. Muler, M., Buchet, R., and Fringeli, U.P. (1996) 2D-FTIR ATR spectroscopy of
thermo-induced periodic secondary structural changes of poly-(L)-lysine: A
cross-correlation analysis of phase-resolved temperature modulation spectra.
J. Phys. Chem., 100: 10810.
10. Noda, I., Liu, Y., and Ozaki, Y. (1996) Two dimensional correlation spectroscopy
study of temperature-dependent spectral variations of N-methylacetamide
in the pure liquid state. 1. Two dimensional infrared analysis. J. Phys Chem.,
100: 8665.
11. Noda, I., Liu, Y., Ozaki, Y., and Czarnecki, M.A. (1995) Two dimensional
Fourier transform near-infrared correlation spectroscopy studies of tempera-
ture-dependent spectral variations of oleoyl alcohol. J. Phys. Chem., 99: 3068.
12. Liu, Y., Ozaki, Y., and Noda, I. (1996) Two dimensional Fourier transform near-
infrared correlation spectroscopy study of dissociation of hydrogen-bonded
n-methylacetamide in the pure liquid state. J. Phys. Chem., 100: 7326.
13. Ozaki, Y., Liu, Y., and Noda, I. (1997) Two dimensional near-infrared correlation
spectroscopy study of premelting behaviour of nylon 12. Macromolecules, 30:
2391.
14. Czarnecki, M.A., Maeda, H., Ozaki, Y., Suzuki, M., and Iwahashi, M. (1998)
Resolution enhancement and band assignments for the first overtone of OH
stretching mode of butanols by two-dimensional near-infrared correlation
spectroscopy. Part I: sec-butanol. Appl. Spectros., 52: 994.
15. Czarnecki, M.A., Maeda, H., Ozaki, Y., Suzuki, M., and Iwahashi, M. (1998)
Resolution enhancement and band assignments for the first overtone of OH
stretching mode of butanols by two-dimensional near-infrared correlation
spectroscopy. 2. Thermal dynamics of hydrogen bonding in n- and tert-butyl
alcohol in the pure liquid states. J. Phys. Chem., 46: 9117.
16. Noda, I., Liu, Y., and Ozaki, Y. (1996) Two-dimensional correlation spec-
troscopy study of temperature-dependent spectral variations of n-methylaceta-
mide in the pure liquid state. 2. Two-dimensional Raman and infrared-Raman
heterospectral analysis. J. Phys. Chem., 100: 8674.
17. Shin, H.S., Jung, Y.M., Lee, J., Chang, T., Ozaki, Y., and Kim, S.B. (2002) Struc-
tural comparison of Langmuir-Blodgett and spin-coated films of poly(tert-butyl
methacrylate) by external reflection FTIR spectroscopy and two-dimensional
correlation analysis. Langmuir, 18: 5523.
Generalized Two-Dimensional Correlation Spectroscopy 539
18. Jung, Y.M., Shin, H.S., Czarnik-Matusewicz, B., Noda, I., and Kim, S.B. (2002)
Characterization of transition temperatures of a Langmuir-Blodgett film of
poly(tert-butyl methacrylate) by two-dimensional correlation spectroscopy and
principal component analysis. Appl. Spectros., 56: 1568–1574.
19. Zhang, J., Tsuji, H., Noda, I., and Ozaki, Y. (2004) Weak intermolecular inter-
action during the melt crystallization of poly(L-lactide) investigated by two-dimensional infrared correlation spectroscopy. J. Phys. Chem. B, 108: 11514.
20. Yang, I.-S., Jung, Y.M., Kim, S.B., and Klein, M.V. (2005) Two dimensional cor-
relation analysis of superconducting YNi2B2C Raman spectra. Int. J. Mod. Phys.
B, 19: 281.
21. Kim, H., Kim, S.B., Kim, J.K., Jung, Y.M., Ryu, D.Y., Lavery, K.A., and
Russell, T.P. (2006) Phase behavior of a weakly interacting block copolymer
by temperature-dependent FTIR spectroscopy. Macromolecules, 39: 408.
22. Smeller, M. and Heremans, K. (1999) 2D FT-IR spectroscopy analysis of the
pressure-induced changes in proteins. Vib. Spectros., 19: 375.23. Noda, I., Story, G.M., and Marcott, C. (1999) Pressure-induced transitions of
polyethylene studied by two-dimensional infrared correlation spectroscopy.
Vib. Spectros., 19: 461.
24. Dzwolak, W., Kato, M., Shimizu, A., and Taniguchi, Y. (2000) Comparative
two-dimensional fourier transform infrared correlation spectroscopic study on
the spontaneous, pressure-, and temperature-enhanced H/D exchange in
a-Lactalbumin. Appl. Spectros., 54: 963–967.
25. Sefara, N.L., Magtoto, N.P., and Richardson, H.H. (1997) Structural characteri-
zation of b-lactoglobulin in solution using two-dimensional FT mid-infraredand F near-infrared correlation spectroscopy. Appl. Spectros., 51: 536.
26. Nakashima, K., Ren, Y., Nishioka, T., Tsubahara, N., Noda, I., and Ozaki, Y.
(1999) Two-dimensional infrared correlation spectroscopy studies of polymer
blends: conformational changes and specific interactions in blends of atactic
polystyrene and poly(2,6-dimethyl-1,4-phenylene ether). J. Phys. Chem. B,
103: 6704.
27. Czanik-Matusewicz, B., Murayama, K., Wu, Y., and Ozaki, Y. (2000) Two-
dimensional attenuated total reflection/infrared correlation spectroscopy of
adsorption-induced and concentration-dependent spectral variations ofb-lactoglobulin in aqueous solutions. J. Phys. Chem. B, 104: 7803.
28. Wang, Y., Murayama, K., Myojo, Y., Tsenkova, R., Hayashi, N., and Ozaki, Y.
(1998) Two-dimensional Fourier transform near-infrared spectroscopy study of
heat denaturation of ovalbumin in aqueous solutions. J. Phys. Chem. B, 34: 6655.
29. Ren, Y., Shimoyama, M., Ninomya, T., Matsukawa, K., Inoue, H., Noda, I., and
Ozaki, Y. (1999) Two-dimensional near-infrared correlation spectroscopy
studies on composition-dependent spectral variations in ethylene/vinyl acetate
copolymers: assignments of banks due to ethylene units in amorphous,
disordered, and orthorhombic crystalline phases. Appl. Spectros., 53: 919.
30. Ren, Y., Murakami, T., Nishioka, T., Nakashima, K., Noda, I., and Ozaki, Y.(2000) Two-dimensional near-infrared correlation spectroscopy studies of
compatible polymer blends: composition-dependent spectral variations of
blends of atactic polystyrene and poly(2,6-dimethyl-1,4-phenylene ether).
J. Phys. Chem. B, 104: 679.
31. Czamik-Matusewicz, B., Murayama, K., Tsenkova, R., and Ozaki, Y. (1999)
Analysis of near-infrared spectra of complicated biological fluids by two-
dimensional correlation spectroscopy: Protein and fat concentration-dependent
spectral changes of milk. Appl. Spectros., 53: 1582.
Y. M. Jung and I. Noda540
32. Wu, Y., Czarnik-Matusewicz, B., Murayama, K., and Ozaki, Y. (2000) Two-
dimensional near-infrared spectroscopy study of human serum albumin in
aqueous solutions: Using overtones and combination modes to monitor
temperature-dependent changes in the secondary structure. Phys. Chem. B,
104: 5840.
33. Ren, Y., Shimoyama, M., Ninomiya, T., Mtsukawa, K., Inoue, H., Noda, I., and
Ozaki, Y. (1999) Two-dimensional Fourier transform Raman correlation
spectroscopy study of composition-induced structural changes in a series of
ethylene/vinyl acetate copolymers. J. Phys. Chem. B, 103: 6475.
34. Abderrazak, H., Dachroui, M., Canada, M.J.A., and Lendl, B. (2000) Quantitative
phosphate analysis is industrial raw phosphoric acid based on evaluation of band-
shifts in FT-Raman spectroscopy. Appl. Spectros., 54: 1610.
35. Jung, Y.M., Czamik-Matusewicz, B., and Ozaki, Y. (2000) Two-dimensional
infrared, two-dimensional raman, and two-dimensional infrared and Raman
heterospectral correlation studies of secondary structure of b-lactoglobulin in
buffer solutions. J. Phys. Chem. B, 104: 7812.
36. Chae, B., Jung, Y.M., Wu, X., and Kim, S.B. (2003) Characterization of a series
of sodium molybdate structures by two-dimensional Raman correlation analysis.
J. Raman Spectros., 34: 451.
37. Jung, Y.M., Czarnik-Matusewicz, B., and Kim, S.B. (2004) Characterization of
concentration-dependent infrared spectral variations of urea aqueous solutions
by principal component analysis and two-dimensional correlation spectroscopy.
J. Phys. Chem. B., 108: 13008.
38. Murayama, K., Wu, Y., Czarnik-Matusewicz, B., and Ozaki, Y. (2001) Two-
dimensional/attenuated total reflection infrared correlation spectroscopy
studies on secondary structural changes in human serum albumin in aqueous
solutions: pH-dependent structural changes in the secondary structures and in
the hydrogen bondings of side chains. J. Phys. Chem. B, 105: 4763.
39. Marcott, C., Noda, I., and Dowery, A. (1991) Enhancing the information content
of vibrational spectra through sample perturbation. J. Anal. Chim. Acta, 250: 131.
40. Nakano, T., Shimada, S., Saitoh, R., and Noda, I. (1993) Transient 2D IR corre-
lation spectroscopy of the photopolymerization of acrylic and epoxy monomers.
Appl. Spectros., 47: 1337.
41. Noda, I., Dowrey, A.E., and Marcott, C. (1993) Recent developments in two-
dimensional infrared (2D IR) correlation spectroscopy. Appl. Spectros., 47: 1317.
42. Marcott, C., Dowrey, A.E., and Noda, I. (1993) Instrumental aspects of dynamic
two-dimensional infrared spectroscopy. Appl. Spectros, 47: 1324.
43. Shilov, S.V., Okretic, S., Czarnecki, M.A., and Siesler, H.W. (1996) Segmental
mobility of liquid crystals and liquid crystalline polymers in an electric field:
A study by time-resolved rapid-scan and step-scan FTIR-spectroscopy. Appl.
Spectros., Rev., 31: 125.
44. Marcott, C., Story, G.M., Dowrey, A.E., Reeder, R.C., and Noda, I. (1997)
Photoacoustic depth profiling, dynamic rheo-optics, and spectroscopic imaging
microscopy of polymers by step-scanning FT-IR spectroscopy. Michrochim.
Acta, (Suppl.), 14: 157.
45. Noda, I., Story, G.M., Dowrey, A.E., Reeder, R.C., and Marcott, C. (1997)
Applications of two-dimensional correlation spectroscopy in depth-profiling
photoacoustic spectroscopy. Near-infrared dynamic rheo-optics and spectro-
scopic imaging microscopy. Macromol. Symp., 119: 1.
Generalized Two-Dimensional Correlation Spectroscopy 541
46. Jiang, E., McCarthy, W.J., Drapcho, D.L., and Crocombe, A. (1997) Generalized
two-dimensional Fourier transform infrared photoacoustic spectral depth-
profiling analysis. Appl. Spectros., 51: 1736.
47. Nabet, A. and Pezolet, M. (1997) Two-dimensional FT-IR spectroscopy: A
powerful method to study the secondary structure of proteins using H-D
exchange. Appl. Spectros., 51: 466.
48. Czarnecki, M.A., Jordanov, B., Okretic, S., and Siesler, H.W. (1997) Two-dimen-
sional correlation analysis of time-resolved step-scan FT-IR spectra of a liquid
crystallime guest-host system in an electric field. Appl. Spectros., 51: 1698.
49. Graff, D.G., Pastrana-Rios, B., Veniaminov, S., and Prendergast, F.G. (1997) The
effects of chain length and thermal denaturation on helix-forming peptides: A
mode-specific analysis using 2D FT-IR. J. Am. Chem. SoC., 119: 11282.
50. Ataka, K. and Osawa, M. (1998) In situ infrared study of water-sulfate coadsorp-
tion on gold (111) in sulfuric acid solutions. Langmuir, 14: 951.
51. Buffeteau, T. and Pezolet, M. (1998) Photoinduced orientation in azopolymers
studies by infrared spectroscopy: Cooperative and biaxial orientation in semicrys-
talline polymers. Macromolecules, 31: 2631.
52. Pancoska, P., Kubelka, J., and Keiderling, T.A. (1999) Novel use of a static
modification of two-dimensional correlation analysis. Part I: Comparison of the
secondary structure sensitivity of electronic circular dichroism, FT-IR, and
Raman spectra of proteins. Appl. Spectros., 53: 655.
53. Kubelka, J., Pancoska, P., and Keiderling, T.A. (1999) Novel use of a static
modification of two-dimensional correlation analysis. Part II: Hetero-spectral
correlation of protein Raman, FT-IR, and circular dichroism spectra. Appl.
Spectros., 53: 666.
54. Magtoto, N.P., Sefera, N.L., and Richardson, H. (1999) Two-dimensional FT-IR
correlation analysis of the chemisorption of nitric oxide on Pt(100). Appl.
Spectros., 53: 178.
55. Halttunen, M., Tenhunen, J., Saarinen, T., and Steius, P. (1999) Applicability of
FTIR/PAS depth profiling for the study of coated papers. Vib. Spectros., 19: 261.
56. Fabian, H., Mantsch, H.H., and Schultz, C.P. (1999) Two-dimensional IR
correlation spectroscopy: Sequential events in the unfolding process of the l
Cro-V55C repressor protein. Proc. Natl Acad. Sci. USA, 96: 13153.
57. Nagasaki, Y., Yoshihara, T., and ozaki, Y. (2000) Polarized infrared spectro-
scopic study on hindered rotation around the molecular axis in the smectic-C�
phase of a ferroelectric liquid crystal with a naphthalene ring. Application of
two-dimensional correlation spectroscopy to polarization angle-dependent
spectral variations. J. Phys. Chem. B, 104: 2846.
58. Tian, G., Wu, Q., Sun, S., Noda, I., and Chen, G.Q. (2001) Study of thermal
melting behavior of microbial polyhydroxyalkanoate using two-dimensional
Fourier transform infrared correlation spectroscopy. Appl. Spectros., 55: 888.
59. Wu, Y., Murayama, K., and Ozaki, Y. (2001) Two-dimensional infrared
spectroscopy and principle component analysis studies of the secondary
structure and kinetics of hydrogen-deuterium exchange of human serum
albumin. J. Phys. Chem. B, 105: 6251.
60. Elmore, D. and Dluhy, R.A. (2001) bn-correlation analysis: A modified two-
dimensional infrared correlation method for determining relative rates of
intensity change. J. Phys. Chem. A, 105: 11377.
61. Amari, T. and Ozaki, Y. (2002) Generalized two-dimensional attenuated
total reflection/infrared and near-infrared correlation spectroscopy studies of
Y. M. Jung and I. Noda542
real-time monitoring of the initial oilgomerization of bis(hydroxyethyl tere-
phthalate). Macromolecules, 35: 8020.
62. Zuo, L., Sun, S.-Q., Zhou, Q., Tao, J.-X., and Noda, I. (2003) 2D-IR correlation
analysis of deteriorative process of traditional Chinese medicine Qing Kai Ling
injection. J. Pharm. Biomed. Anal., 30: 1491.
63. Jhao, J.G., Tatani, K., Yoshihara, T., and ozaki, Y. (2003) Electric-field-inducedreorientation of a ferroelectric liquid crystal molecule without a carboxylate
group near the stereocenter studied by time-resolved infrared spectroscopy
combined with normalized sample-sample two-dimensional correlation
spectroscopy. J. Phys. Chem. B, 107: 4227.
64. Huang, H., Malkov, S., Coleman, M.M., and Painter, P.C. (2003) Application of
two-dimensional correlation infrared spectroscopy to the study of miscible
polymer blends. Macromolecules, 36: 8156.
65. Yu, Z.W., Liu, J., and Noda, I. (2003) Effect of noise on the evaluation of corre-
lation coefficients in two-dimensional correlation spectroscopy. Appl. Spectros.,57: 1605.
66. Zhang, J., Tsuji, H., Noda, I., and Ozaki, Y. (2004) Structural changes and crys-
tallization dynamics of poly(L-lactide) during the cold-crystallization process
investigated by infrared and two-dimensional infrared correlation spectroscopy.
Macromolecules, 37: 6433.
67. Zhou, Q., Sun, S-Q., and Zuo, L. (2004) Study on traditional Chinese medicine
‘Qing Kai Ling’ injections from different manufactures by 2D IR correction
spectroscopy. Vib. Spectros., 36: 207.
68. Nishikawa, Y., Nakano, T., and Noda, I. (2006) Two-dimensional correlationstudy of uniaxially drawn poly(ethylene terephthalate) films by using attenuated
total reflection based dynamic compression modulation step-scan Fourier
transform infrared in combination with spectral simulation analysis by density
functional theory. Appl. Spectros., 60: 145.
69. Ebihara, K., Takahashi, H., and Noda, I. (1993) Nanosecond two-dimensional
resonance Raman correlation spectroscopy of benzil radical anion. Appl.
Spectros., 47: 1343.
70. Gustafson, T.L., Morris, D.L., Huston, L.A., Bulter, R.M., and Noda, I. (1994)
Solvent/solute interactions probed by picosecond transient Raman spectroscopy.Springer Proc. Phys., 74: 131.
71. Matsushita, A., Ren, Y., Matsukawa, K., Inoue, H., Minami, Y., Noda, I., and
Ozaki, Y. (2000) Two-dimensional Fourier-transform Raman and near-infrared
correlation spectroscopy studies of poly(methyl methacrylate) blends: 1.
Immiscible blends of poly(methyl methacrylate) and atactic polystyrene. Vib.
Spestros., 24: 171.
72. Ren, Y., Matsushita, A., Matsukawa, K., Inoue, H., Minami, Y., Noda, I., and
Ozaki, Y. (2000) Two-dimensional fourier-transform-Raman and near-infrared
correlation spectroscopy studies of poly(methyl methacrylate) blends: 2.
Partially miscible blends of poly(methyl methacrylate) and poly(4-vinylphenol).Vib. Spestros., 23: 207.
73. Schultz, G., Jirasek, A., Blades, M.W., and Turner, R.F.B. (2003) Identification
and interpretation of generalized two-dimensional correlation spectroscopy
features through decomposition of the perturbation domain. Appl. Spectros.,
57: 1561.
74. Choi, H.C., Jung, Y.M., and Kim, S.B. (2004) Characterization of Raman spectra
of size-selected TiO2 nanoparticles by two-dimensional correlation spectroscopy.
Bull. Kor. Chem. Soc., 25: 426.
Generalized Two-Dimensional Correlation Spectroscopy 543
75. Czarnecki, M.A. and Ozaki, Y. (1999) The temperature-induced changes in
hydrogen bonding of 1-decanol in the pure liquid phase studied by two-
dimensional Fourier transform near-infrared correlation spectroscopy. Phys.
Chem. Chem. Phys., 1: 797.
76. Ozaki, Y., Liu, Y., and Noda, I. (1997) Two-dimensional infrared and near-
infrared correlation spectroscopy: applications to studies of temperature-
dependent spectral variations of self-associated molecules. Appl. Spectros., 51:
526.
77. Murayama, K., Czamik-Matusewicz, B., Wu, Y., Tsenkova, R., and Ozaki, Y.
(2000) Comparison between conventional spectral analysis methods,
chemometrics, and two-dimensional correlation spectroscopy in the analysis of
near-infrared spectra of protein. Appl. spectros., 54: 978.
78. Czarnecki, M.A., Czarnik-Matusewicz, B., Ozaki, Y., and Iwahahi, M. (2000)
Resolution enhancement and band assignments for the first overtone of OH(D)
stretching modes of butanols by two-dimensional near-infrared correlation
spectroscopy. 3. Thermol dynamics of hydrogen bonding in butan-1-(ol-d) and
2-methylpropan 2-(ol-d) in the pure liquid states. J. Phys. Chem. A, 104: 4906.
79. Sasic, S. and Ozaki, Y. (2001) Comparison of principal component analysis and
generalized two-dimensional correlation spectroscopy: Spectral analysis of
synthetic model system and near-infrared spectra of milk. Appl. Spectros., 55: 29.
80. Amari, T. and Ozaki, Y. (2001) Sample-sample and wavenumber-wavenumber
two-dimensional correlation analyses of attenuated total reflection infrared
spectra of polycondensation reaction of bis(hydroxyethyl terephthalate). Anal.
Chem., 73: 5184.
81. Segtnan, V.H., Sasic, S., Isaksson, T., and Ozaki, Y. (2001) Studies on the
structure of water using two-dimensional near-infrared correlation spectroscopy
and principal component analysis. Anal. Chem., 73: 3153.
82. Sasic, S. and Ozaki, Y. (2001) Wavelength-wavelength and sample-sample two-
dimensional correlation analyses of short-wave near-infrared spectra of raw milk.
Appl. Spectros, 55: 163.
83. Murayama, K. and Ozaki, Y. (2002) Two-dimensional near-IR correlation
spectroscopy study of molten globule-like state of ovalbumin in acidic pH
region: Simultaneous changes in hydration and secondary structure. Biospectro-
scopy, 67: 394.
84. Awich, A., Tee, E.M., Srikanthan, G., and Zhao, W. (2002) Identification of
overlapped near-infrared bands of glucose anomers using two-dimensional near-
infrared and middle infrared correlation spectroscopy. Appl. Spectros., 56: 897.
85. Roselli, C., Burie, J.R., Mattioli, T., and Boussac, A. (1995) Two dimensional
correlation method applied to Yb3þ bibronic sideband spectroscopy: Discrimination
of fluorescence spectral features arising from different Yb3þ binding sites. Biospec-
troscopy, 1: 329.
86. He, Y., Wang, G., Cox, J., and Geng, L. (2001) Two-dimensional fluorescence
correlation spectroscopy with modulated excitation. Anal. Chem., 73: 2302.
87. Wang, G. and Geng, L. (2005) Statistical and generalized two-dimensional
correlation spectroscopy of multiple ionization states fluorescence of neurotrans-
mitter serotonin. Anal. Chem., 77: 20.
88. Liu, Y., Chen, Y.R., and Ozaki, Y. (2000) Two-dimensional visible/near-infrared
correlation spectroscopy study of thermal treatment of chicken meats. J. Agric.
Food Chem., 48: 901.
Y. M. Jung and I. Noda544
89. Fukutake, N. and Kabayashi, T. (2002) Size distribution of pseudoisocyanine
(PIC) J-aggregates studied by near-field absorption spectroscopy. Chem. Phys.
Lett., 356: 368.
90. Zhao, W., Song, C., Zheng, B., Liu, J., and Viswanathan, T. (2002) Thermal
recovery behavior of fluorinated single-walled carbon nanotubes. J. Phys.
Chem. B, 106: 293.
91. Choi, H.C., Jung, Y.M., Noda, I., and Kim, S.B. (2003) A study of the mechanism
of the electrochemical reaction of lithium with CoO by two-dimensional soft
X-ray absorption spectroscopy (2D XAS), 2D Raman, and 2D heterospectral
XAS-Raman correlation analysis. J. Phys. Chem. B, 107: 5806.
92. Choi, H.C., Jung, Y.M., and Kim, S.B. (2003) Characterization of the electro-
chemical reactions in the Li1þxV3O8/Li cell by soft X-ray absorption spectros-
copy and two-dimensional correlation analysis. Appl. Spectros., 57: 984.
93. Izawa, K., Ogasawa, T., Masuda, H., Okayabashi, H., O’Connor, C.J., and
Noda, I. (2002) Two-dimensional correlation gel permeation chromatography
(2D GPc) study of 1H,1H,2H,2H,-perfluorooctyltriethoxysilane sol-gel polymeri-
zation process. J. Phys. Chem. B, 106: 2867.
94. Izawa, K., Ogasawa, T., Masuda, H., Okayabashi, H., and Noda, I. (2002) Two-
dimensional correlation gel permeation chromatography study of octyltriethoxy-
silane sol-gel polymerization process. Macromolecules, 35: 92.
95. Izawa, K., Ogasawa, T., Masuda, H., Okayabashi, H., O’Connor, C.J., and
Noda, I. (2002) 2D Gel permeation chromatography (2D GPC) correlation
studies of the growth process for perfluoro-octyltriethoxysilane polymer aggre-
gates. Phys. Chem. Chem. Phys., 4: 1053.
96. Noda, I. (1990) Two-dimensional correlation approach to the dynamic rheo-
optical characterization of polymers. Chemtract-Macromol. Chem., 1: 89.
97. Czarnecki, M.A., Wu, P., and Siesler, H.W. (1998) 2D FT-NIR and FT-IR corre-
lation analysis of temperature-induced changes of nylon 12. Chem. Phys. Lett.,
283: 326.
98. Nagai, N., Yamaguchi, Y., Saito, R., Hayashi, S., and Kudo, M. (2001) IR and
EPR study of the Na ion-limplanted SiO2/Si system. Appl. Spectros., 55: 1207.
99. Barton II, F.B., Himmelsbach, D.S., Duckworth, J.H., and Smith, M.J. (1992)
Two-dimensional vibration spectroscopy: Correlation of mid- and near-infrared
regions. Appl. Spectros., 46: 420.
100. Barton II, F.B. and Himmelsbach, D.S. (1993) Two-dimensional vibration
spectroscopy II: Correlation of the absorptions of lignins in the mid- and near-
infrared. Appl. Spectros., 47: 1920.
101. Windigm, W., Margevich, D.E., and McKenna, W. (1995) A novel tool for
two-dimensional correlation spectroscopy. Chemom. Intel. Lab. Syst., 28: 109.
102. Sasic, S., Muszynski, A., and Ozaki, Y. (2000) A new possibility of the general-
ized two-dimensional correlation spectroscopy. 1. Sample-sample correlation
spectroscopy. J. Phys. Chem. A, 104: 6380.
103. Sasic, S., Muszynski, A., and Ozaki, Y. (2000) A new possibility of the
generalized two-dimensional correlation spectroscopy. 2. Sample-sample and
wavenumber-wavenumber correlation of temperature-department near-infrared
spectra of oleic acid in the pure liquid state. J. Phys. Chem. A, 104: 6388.
104. Sasic, S. and Ozaki, Y. (2001) Statistical two-dimensional correlation
spectroscopy: Its theory and applications to sets of vibrational spectra. Anal.
Chem., 73: 2294.
Generalized Two-Dimensional Correlation Spectroscopy 545
105. Morita, S., Ozaki, Y., and Noda, I. (2001) Global phase angle description of
generalized two-dimensional correlation spectroscopy: 1. Theory and its
simulation for practical use. Appl. Spectros., 55: 1622.
106. Morita, S., Ozaki, Y., and Noda, I. (2001) Global phase angle description of
generalized two-dimensional correlation spectroscopy: 2. Its application to temp-
erature-dependent infrared spectra of a Langmuir-Blodgett film of 2-dodecyl-
7,7,8,8-tetracyanoquinodimethane. Appl. Spectros., 55: 1622.
107. Morita, S. and Ozaki, Y. (2001) Pattern recognitions of band shifting, overlap-
ping, and broadening using global phase description derived from generalized
two-dimensional correlation spectroscopy. Appl. Spectros., 56: 502.
108. Wu, Y., Yuan, B., Zhao, J.G., and Ozaki, Y. (2003) Hybrid two-dimensional
correlation and parallel factor studies on the switching dynamics of a surface-
stabilized ferroelectric liquid crystal. J. Phys. Chem. B, 107: 7706.
109. Wu, Y., Meersman, F., and Ozaki, Y. (2006) A novel application of hybrid two-
dimensional correlation infrared spectroscopy: Exploration of the reversibility of
the pressure- and temperature-induced phase separation of poly(N-isopropylacry-
lamide) and poly(N-isopropylmethacrylamide) in aqueous solution. Macromol-
ecules, 39: 1182.
110. Jung, Y.M., Shin, H.S., Kim, S.B., and Noda, I. (2002) New approach to
generalized two-dimensional correlation spectroscopy. I: Combination of
principal component analysis and two-dimensional correlation spectroscopy.
Appl. Spectros., 56: 1562.
111. Jung, Y.M., Kim, S.B., and Noda, I. (2003) New approach to generalized two-
dimensional correlation spectroscopy. III: Eigenvalue manipulation transform-
ation (EMT) for spectral selectivity enhancement. Appl. Spectros., 57: 564.
112. Jung, Y.M., Kim, S.B., and Noda, I. (2003) New approach to generalized
two-dimensional correlation spectroscopy. II: eigenvalue manipulation trans-
formation (EMT) for noise suppression. Appl. Spectros., 57: 557.
113. Jung, Y.M., Kim, S.B., and Noda, I. (2003) Application of two-dimensional
correlation spectroscopy. IV: Eigenvalue manipulation transformation (EMT)
for partial attenuation of dominant factors. Appl. Spectros., 57: 850.
114. Jung, Y.M., Kim, S.B., and Noda, I. (2003) New approach to generalized two-
dimensional correlation spectroscopy to chemometrics: Self-modeling curve
resolution analysis of spectral data sets. Appl. Spectros., 57: 1376.
115. Jung, Y.M. (2004) Application of 2D correlation spectroscopy and outer product
analysis to infrared spectra of sugar beets. Vib. Spectros., 36: 279.
116. Malinowski, E.R. (1991) Factor Analysis in Chemistry, 2nd ed.; Wiley:
New York.
117. Martens, H. and Næs, T. (1991) Multivariate Calibration; John Wiley & Sons:
New York.
118. Vandeginste, B.G.M., Massart, D.L., Buydens, L.M.C., De Jong, S., Lewi, P.J.,
and Smeyers-Verbeke, J. (1998) Handbook of Chemometrics and Qualimetrics:
Part B; Elsevier Science B. V: Amsterdam, The Netherlands, 88–104.
119. Robert, P., Mangavel, C., and Renard, D. (2001) Infrared spectroscopy as applied
to glycinin film and gel formation kinetic. Appl. Spectros., 55: 781.
120. Jung, Y.M., Shin, H.S., Czarnik-Matusewicz, B., Noda, I., and Kim, S.B. (2002)
Characterization of transition temperatures of a Langmuir-Blodgett film of
poly(tert-butyl methacrylate by two-dimensional correlation spectroscopy and
principal component analysis. Appl. Spectros., 56: 1568.
Y. M. Jung and I. Noda546
121. Jung, Y.M. (2003) Principal component analysis based two-dimensional(PCA-2D) correlation spectroscopy: PCA denoising for 2D correlation spec-troscopy. Bull. Kor. Chem. Soc., 24: 1345.
122. Jung, Y.M. and Kim, S.B. (2005) Thermal behavior of Langmuir-Blodgett film ofpoly(tert-butyl methacrylate) by principal component analysis based two-dimensional correlation spectroscopy. Bull. Kor Chem. Soc., 26: 2027.
123. Jung, Y.M., Kim, H.J., Ryu, D.Y., Kim, S.B., and Kim, J.K. (2006) Published onthe web, application of principal component analysis-based two-dimensionalcorrelation spectroscopy to characterization of order-disorder transition ofpolystyrene-block-poly(n-pentyl methacrylate) copolymer. J. Mol. Struct.
124. Jung, Y.M. (2003) Characterization of pH-dependent IR spectra of oxalic acid:Comparison of self-modeling curve resolution analysis with calculation of IRfrequencies. Bull. Kor. Chem. Soc., 24: 1410.
125. Kim, Y.-O., Jung, Y.M., Kim, S.B., and Park, S.-M. (2004) Two-dimensionalcorrelation analysis of spectroelectrochemical data for p-benzoquinone reductionin acetonitrile. Anal. Chem., 76: 5236.
126. Hong, S.Y., Jung, Y.M., Kim, S.B., and Park, S.-M. (2005) Electrochemistry ofconductive polymers. 34. Two-dimensional correlation analysis of real-timespectroelectrochemical data for aniline polymerization. J. Phys. Chem. B., 109:3844.
Generalized Two-Dimensional Correlation Spectroscopy 547