ebook download Resolving Spectral Mixtures With Applications from Ultrafast Time-Resolved Spectroscopy to Super-Resolution Imaging 1st Edition Cyril Ruckebusch (Eds.) - eBook PDF all chapter

Resolving Spectral Mixtures With
Applications from Ultrafast

Time-Resolved Spectroscopy to
Super-Resolution Imaging 1st Edition
Cyril Ruckebusch (Eds.) - eBook PDF
Go to download the full and correct content document:
https://ebooksecure.com/download/resolving-spectral-mixtures-with-applications-from
-ultrafast-time-resolved-spectroscopy-to-super-resolution-imaging-ebook-pdf/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
(eBook PDF) Constructive Conflicts: From Escalation to

Resolution 5th Edition
http://ebooksecure.com/product/ebook-pdf-constructive-conflicts-
from-escalation-to-resolution-5th-edition/
Spectroscopy and Dynamics of Single Molecules: Methods

and Applications 1st edition - eBook PDF
https://ebooksecure.com/download/spectroscopy-and-dynamics-of-
single-molecules-methods-and-applications-ebook-pdf/
(eBook PDF) Introduction to Spectroscopy 5th Edition
http://ebooksecure.com/product/ebook-pdf-introduction-to-
spectroscopy-5th-edition/
Tissue Elasticity Imaging: Volume 2: Clinical

Applications 1st Edition - eBook PDF
https://ebooksecure.com/download/tissue-elasticity-imaging-
volume-2-clinical-applications-ebook-pdf/
Vibrational Spectroscopy Applications in Biomedical,
Pharmaceutical and Food Sciences 1st Edition Andrei A.
Bunaciu - eBook PDF
https://ebooksecure.com/download/vibrational-spectroscopy-
applications-in-biomedical-pharmaceutical-and-food-sciences-
ebook-pdf/
Molecular and Laser Spectroscopy: Advances and

Applications: Volume 2 1st Edition V.P. Gupta (Editor)
- eBook PDF
https://ebooksecure.com/download/molecular-and-laser-
spectroscopy-advances-and-applications-volume-2-ebook-pdf/
Progress in Heterocyclic Chemistry Volume 29 1st

Edition - eBook PDF
https://ebooksecure.com/download/progress-in-heterocyclic-
chemistry-ebook-pdf/
Innovation in Nano-polysaccharides for Eco-

sustainability: From Science to Industrial Applications
1st Edition - eBook PDF
https://ebooksecure.com/download/innovation-in-nano-
polysaccharides-for-eco-sustainability-from-science-to-
industrial-applications-ebook-pdf/
Theory and Modeling of Cylindrical Nanostructures for

High-Resolution Spectroscopy. A volume in Micro and
Nano Technologies 1st Edition Edition Stefano Bottacchi
And Francesca Bottacchi (Auth.) - eBook PDF
https://ebooksecure.com/download/theory-and-modeling-of-
cylindrical-nanostructures-for-high-resolution-spectroscopy-a-
volume-in-micro-and-nano-technologies-ebook-pdf/
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
First edition 2016
Copyright © 2016 Elsevier B.V. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopying, recording, or any
information storage and retrieval system, without permission in writing from the
publisher. Details on how to seek permission, further information about the Publisher’s
permissions policies and our arrangements with organizations such as the Copyright
Clearance Center and the Copyright Licensing Agency, can be found at our website:
www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under
copyright by the Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and
experience broaden our understanding, changes in research methods, professional
practices, or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge
in evaluating and using any information, methods, compounds, or experiments
described herein. In using such information or methods they should be mindful of their
own safety and the safety of others, including parties for whom they have a
professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or
editors, assume any liability for any injury and/or damage to persons or property as a
matter of products liability, negligence or otherwise, or from any use or operation of
any methods, products, instructions, or ideas contained in the material herein.
ISBN: 978-0-444-63638-6
ISSN: 0922-3487
For information on all Elsevier publications

visit our website at https://www.elsevier.com/
Publisher: John Fedor

Acquisition Editor: Kathryn Morrissey
Editorial Project Manager: Jill Cetel
Production Project Manager: Vijayaraj Purushothaman
Cover Designer: Victoria Pearson
Typeset by SPi Global, India
Contributors
Numbers in Parentheses indicate the pages on which the author’s contributions begin.
E. Aguado Sarrió (519), Multivariate Statistical Engineering Group, Universitat
Politècnica de València, Valencia, Spain
S. Aloı̈se (353), Universite de Lille, Sciences et Technologies, LASIR CNRS, LILLE,
France
Y. Altmann (185), School of Engineering and Physical Sciences, Heriot-Watt
University, Riccarton, Edinburgh, EH14 4AS, United Kingdom
J.M. Amigo (613), Faculty of Sciences, University of Copenhagen, Frederiksberg C,
Denmark
A. Bogomolov (53), Art Photonics GmbH, Berlin, Germany
D. Brie (279), Centre de Recherche en Automatique de Nancy, Universite de Lorraine,
CNRS, Vanduvre-lès-Nancy, France
N. Brun (185), Laboratoire de Physique des Solides, CNRS, Univ. Paris-Sud, Univ.
Paris-Saclay, 91405 Orsay Cedex, France
R. Calvini (613), University of Modena and Reggio Emilia, Reggio Emilia, Italy
C. Carteret (279), Laboratoire de Chimie Physique et Microbiologie pour
l’Environnement, Universite de Lorraine, CNRS, Villers-lès-Nancy, France
M. Cocchi (409), University of Modena and Reggio Emilia, Modena, Italy
A. de Juan (5), Universitat de Barcelona, Barcelona, Spain
O. Devos (353, 453), Universite de Lille, Sciences et Technologies, LASIR CNRS,
LILLE, France
N. Dobigeon (185), University of Toulouse, IRIT/INP-ENSEEIHT, 31071 Toulouse
Cedex 7, France
M. Dossot (279), Laboratoire de Chimie Physique et Microbiologie pour l’Environne-
ment, Universite de Lorraine, CNRS, Villers-lès-Nancy, France
L. Duponchel (477), LASIR CNRS UMR 8516, Universite Lille 1, Sciences et
Technologies, Villeneuve d’Ascq Cedex, France
A. Ferrer (519), Multivariate Statistical Engineering Group, Universitat Politècnica de
València, Valencia, Spain
C. Gomez (579), IRD, UMR LISAH (INRA-IRD-SupAgro), Montpellier, France
N. Gorretta (579), IRSTEA UMR ITAP, Montpellier Cedex 5, France
D.M. Haaland (381), Spectral Resolutions, Albuquerque, NM, United States
xv
xvi Contributors
P. de B. Harrington (311), Clippinger Laboratories, Center for Intelligent Chemical

Instrumentation, Ohio University, Athens, OH, United States
S. Hugelier (453), Universite de Lille, Sciences et Technologies, LASIR CNRS,
LILLE, France
H.D.T. Jones (381), HyperImage Solutions, Rio Rancho, NM, United States
D. Jouan-Rimbaud Bouveresse (225), UMR Ingenierie Procedes Aliments,
AgroParisTech, INRA, Universite Paris-Saclay, Massy, France
urß (135), Universit€at Rostock, Institut f€
A. J€ ur Mathematik, Rostock, Germany
S. Kucheryavskiy (53), Aalborg University, Aalborg, Denmark
M. Li Vigni (409), University of Modena and Reggio Emilia, Modena, Italy
A. Malik (101), Institute of Environmental Assessment and Water Research (IDÆA)
SPANISH COUNCIL OF SCIENTIFIC RESEARCH (CSIC), Barcelona, Spain
S. Miron (279), Centre de Recherche en Automatique de Nancy, Universite de
Lorraine, CNRS, Vanduvre-lès-Nancy, France
S. Moussaoui (185, 279), Ecole Centrale de Nantes, IRCCyN, UMR CNRS 6597,
Nantes Cedex 3, France
etivier (353), PPSM, ENS Cachan, CNRS, Cachan Cedex, France
R. M
K. Neymeyr (135), Universit€at Rostock, Institut f€
ur Mathematik; Leibniz-Institut f€
ur
Katalyse, Rostock, Germany
M. Offroy (477), LASIR CNRS UMR 8516, Universite Lille 1, Sciences et
Technologies, Villeneuve d’Ascq Cedex, France
K.M. Pierce (333), Seattle Pacific University, Seattle, WA, United States
D.K. Pinkerton (333), University of Washington, Seattle, WA, United States
J.-P. Placial (353), PPSM, ENS Cachan, CNRS, Cachan Cedex, France
J.M. Prats-Montalbán (519), Multivariate Statistical Engineering Group, Universitat
Politècnica de València, Valencia, Spain
D. Roberts (551), University of California, Santa Barbara, CA, United States
C. Ruckebusch (1, 353, 453), Universite de Lille, Sciences et Technologies, LASIR
CNRS, LILLE, France
D.N. Rutledge (225), UMR Ingenierie Procedes Aliments, AgroParisTech, INRA,
Universite Paris-Saclay, Massy, France
M. Sawall (135), Universit€at Rostock, Institut f€
ur Mathematik, Rostock, Germany
oder (135), Universit€at Rostock, Institut f€
H. Schr€ ur Mathematik, Rostock, Germany
M. Sliwa (353), Universite de Lille, Sciences et Technologies, LASIR CNRS, LILLE,
France
B. Somers (551), Division of Forest, Nature, and Landscape, KU Leuven, Leuven,
Belgium
R.E. Synovec (333), University of Washington, Seattle, WA, United States
Contributors xvii
R. Tauler (5, 101), Institute of Environmental Assessment and Water Research

(IDÆA) SPANISH COUNCIL OF SCIENTIFIC RESEARCH (CSIC), Barcelona,
Spain
J.A. Timlin (381), Sandia National Laboratories, Albuquerque, NM, United States
L. Tits (551), Flemish Institute for Technological Research (VITO), Remote Sensing
Unit, Boeretang 200, BE-2400 Mol, Belgium
A. Ulrici (613), University of Modena and Reggio Emilia, Reggio Emilia, Italy
E. Wetherley (551), University of California, Santa Barbara, CA, United States
W. Windig (53), Eigenvector Research Inc., Manson, WA, United States
Preface
This book deals with ways and means to resolve spectral mixtures, a long-
standing but topical issue that is tackled in many research fields, with an
unlimited number of possible applications. In chemistry, spectral mixture res-
olution investigates situations where the spectral signatures of a sample result
from the superposition of the various chemicals or materials composing it, as
observed in process spectroscopy or hyperspectral imaging. More broadly,
resolving a spectral mixture is an inverse problem that is addressed in differ-
ent ways in chemistry and in the different fields of science and technology.
This has led to the development of a wealth of methods.
Spectral mixture resolution remains a hot issue in chemometrics statistics
or signal and image processing. However, the diversity of methods (multivar-
iate curve resolution, blind source separation, linear unmixing, etc.) can be a
serious impediment to wider understanding and dissemination of the spectral
mixture problem. By taking a multi-angle and cross-disciplinary approach,
this book has the ambition to turn this limitation into an asset and to provide
a comprehensive and comprehensible description of the current state of
the art.
The book is multi-authored, written as a collection of independent chapters
that provide different perspectives and applications. Expressed altogether,
these perspectives translate into a real interplay between the chapters. Basic
concepts and main methods are presented in the first part of the book, whereas
the second part is more oriented toward applications in chemistry and remote
sensing. Some chapters are written as tutorials, others are reviews. Overall,
the book presents an extensive bibliography which is incomplete—
considering the amount of scientific literature on the subject—but good list
to start with.
Written by invited authors that are recognized experts in their field, the
book is addressed to graduate students, researchers, and practitioners in ana-
lytical science, emphasizing applications dealing with any type of spectral
data. Some background in data analysis (chemometrics, statistics, signal or
image processing, etc.) and knowledge of the basics of linear algebra will help
readability. However, each chapter was meant to contain enough information
to be, by itself, sufficient and to be read independently of the other chapters.
xix
xx Preface
Lastly, I would like to express my gratitude to all the invited authors that
have accepted to cooperate and contributed to this book, and to the numerous
people who, one way or another, were involved in this project.
Cyril Ruckebusch
Lille, January 2016
Foreword
The topic of this volume is, of course, very interesting and it is the right time
for its publication as a rapidly growing number of researchers rely on the
methods described here for the analysis of their data. Very likely this number
will increase significantly in the near future.
All work presented here deals with a series of measured spectra, and the
goal of the analyses is the resolution of this collection of spectra into the prod-
uct of contributions of the components in terms of their concentrations and
their spectral responses. The task is, in some ways, both surprisingly simple
and surprisingly difficult.
All analyses are based on the model that the overall signal at a particular
wavelength or channel is the sum over the contributions of all components
where each component contribution is the product of its concentration times
its signal for unit concentration. This is most clearly realized in solution
absorption spectroscopy where Beer–Lambert’s law reigns; at each wave-
length the absorption is the sum over the products of the component concen-
trations times their molar absorptivity at the particular wavelength. It is,
therefore, not surprising that most methods presented in this volume have
been developed for the resolution of series of absorption spectra, typically
an HPLC chromatogram equipped with a diode array detector. This volume
demonstrates that the methodologies have found much wider application.
However, the following explanations will be based on the language of absorp-
tion spectroscopy.
Eq. (1) describes the relationship for the absorption reading of spectrum i
measured at wavelength j, Di, j is the sum over the products of concentration
of species k, Ci,k, times its molar absorptivity at wavelength j, Ak, j:
X
Di, j ¼ Ci, k Ak, j (1)
k
The Equation (1) can most conveniently be written as a matrix equation

(Eq. 2)
D ¼ CA (2)
where D contains row-wise the spectra taken typically during a process, it is
known; the matrix C contains the concentrations of the components and
A contains their spectra or responses. These matrices are unknown and the
goal is to determine them. It is useful to represent the situation graphically:
xxi
xxii Foreword
D = C
Each element of D represents an equation of type (1), and the elements of

C and A are the unknowns. A prerequisite for all analyses presented here is
that there must be more elements in D (equations) than unknowns in C and
A; in fact, in most instances the area of D is much larger than the areas of
C and A combined.
The first question one has to ask is: as there are more equations than
unknowns, is there a solution at all? The answer is yes, and there must be a
solution, the true concentrations, and molar absorptivities. Second, is the solu-
tion unique? As all elements of C and A must be positive, there is hope that
there is a unique solution. However, in most case the solution is not unique,
and there are a range of feasible solutions.
In the following chapters, several issues related to the task of decomposing
a measured data set into two meaningful matrices are discussed:
How can the solution be computed?
(Chapters 2, 3, 7, 8, and 10)
This is clearly a central issue as the computation of the unique solution or of
one of a range of feasible solutions is not a trivial task. In the chemometrics
literature the alternating least squares, ALS, algorithm is most commonly
used; it is based on an elegant and simple algorithm.
What if the structure of the data is more complex than the matrix equation (2)?
(Chapters 2, 10, and all image chapters)
Eq. (2) is very simple, if the structure of the data is more complex; e.g., in all
hyperspectral images, the data can be rearranged in such a way that they fit
into the structure of Eq. (2), and this is also the case for higher order data sets
where individual matrices are concatenated into the basic structure of Eq. (2).
Is there a unique solution?
(Chapters 4, 5, and 8)
Even if there are many more equations than unknowns, for most data sets the
solution is not unique; there are a range of feasible solutions for the matrices
C and A which resolve the data matrix D with exactly the same accuracy. We
call this rotational ambiguity. There are readily available methods that either
approximate this range (Chapter 4) or perform a comprehensive analysis
(Chapter 5).
Foreword xxiii
Can additional constraints be applied?

(Chapters 3, 14, 17, and 18)
Nonnegativity of the elements of the matrices C and A is the universally
applied constraint. Rotational ambiguity can be reduced or even eliminated
altogether by application of other constraints or assumptions, such as maxi-
mum smoothness of or dissimilarity between the results.
What if the signal is not strictly following Eqs. (1) and (2)?
(Chapter 6, all image chapters)
Eqs. (1) and (2) describe the ideal case of linear signal contribution; quite
often, this is not strictly the case, and it certainly is only an approximation
in hyperspectral images. This does not render the analysis irrelevant; the result
is an approximation and often a very useful one. Chapter 6 specifically
explores the possibility of a nonlinear relationship between component
concentration and signal.
What if the data set is huge?
(Chapters 9, 12, and 13)
Wavelet transformation is an option for the reduction of the size of the origi-
nal data matrix D. Other corrections like baseline subtraction, etc., can also be
useful.
Can the spatial resolution be improved?
(Chapter 15)
Hyperspectral images have a certain given spacial resolution; surprisingly, it
can be improved by clever algorithms.
How widely can the methods be applied?
(Chapters 11 and 16)
There are two chapters that report on the application of the methods presented
here in very different fields of chemical research.
Are the results qualitative or quantitative?
(Chapter 2)
The typical result of most analyses presented here is a set of concentration
profiles/distributions and response vectors/spectra. They can be represented
in graphs. Primarily, the results do not allow quantitative interpretation. The
extension of the methods toward quantitative analyses is most important.
To conclude this foreword I would like to add an important comment:
Every scientist knows that a reported value for a parameter should be further
characterized by an error estimate, e.g., K ¼ 350 25. Similarly, a spectral
mixture analysis without analysis of rotational ambiguity should be rejected.
In my experience, too many users take the analysis result as a fact, rather than
xxiv Foreword
one of a range of possible solution. Implementing further constraints may

result in unique solutions, but if the additional constraint is not based on
actual facts but rather on wishful thinking, the unique outcome may well
not be the true one. The authors in this volume are the leaders in the field
and should educate the novice or casual user about this problem and lead with
the proverbial good example. (I will not list the offenders.).
As we have seen, there are many difficulties, should the present methods
not be used? The answer is a resounding no! The methods are very powerful
and while they might not give ultimate and detailed answers they nonetheless
deliver a lot of very useful information which would be difficult if not impos-
sible to gather otherwise. More work will be done in the field, there is
no doubt.
Marcel Maeder
University of Newcastle, Australia
Chapter 1
Introduction
C. Ruckebusch1
Universite´ de Lille, Sciences et Technologies, LASIR CNRS, Lille, France
1
Corresponding author: e-mail: cyril.ruckebusch@univ-lille1.fr
Chapter Outline
1 Introduction 1 3 Book Content and
2 The Spectral Mixture Problem 1 Organization 3
1 INTRODUCTION
This chapter introduces first very basic information about the topic of the
book and sets the overall context. It provides broad definitions and clarifies
some points regarding the terminology. The second part provides information
about the organization of the book. A first insight into the content of the
19 chapters composing the book, and their interplay, is given. The intention
of these few words of introduction is mainly the presentation of the issues that
will be tackled more comprehensively along the chapters of the book. These
questions can be roughly put as follows:
– What is a spectral mixture?
– What does resolving a spectral mixture mean?
– What are the different ways to tackle the spectral mixture issues?
– What difficulties remain?
– And what are the perspectives?
2 THE SPECTRAL MIXTURE PROBLEM

A spectral mixture is a data that results from the observation of a chemical
system composed of (mixed) individual components and submitted to some
variation. This variation is related to the change of an external factor, which
is usually a physical or chemical variable. It can be for example sampling
time, position, or pH. The spectral data thus consist of a superposition, or mix-
ture, of the pure spectra of the individual components and their associated
proportions. When dealing with evolving systems such as chemical reactions
or processes, these proportions correspond to concentration profiles.
Data Handling in Science and Technology, Vol. 30. http://dx.doi.org/10.1016/B978-0-444-63638-6.00001-2

Copyright © 2016 Elsevier B.V. All rights reserved. 1
2 Data Handling in Science and Technology
Spectral mixture data are usually arranged in a matrix with columns as

spectral variables (wavelength, wavenumbers, etc.) and objects (time, posi-
tion, etc.) as rows. Objects can be of different nature, but should always be
clearly related to the state of the before mentioned physical or chemical vari-
able. Ideally, the variations contained in the spectral data translate what is
supposed to be relevant information for the problem at hand. Spectral mixture
resolution aims to decompose the variations of the spectral data into a model
of the contributions from the individual unknown components. These compo-
nents are composed of source proportions and spectral signatures. It is impor-
tant to realize that, more than often, this decomposition is aimed at situations
for which little a priori information is available. It should also be noted
that, in practice, some physical perturbations or chemical interferences may
complicate the ideal situation.
In chemistry, spectral mixture resolution corresponds to the resolution of
complex mixture spectra into pure contributions, consisting of concentration
distributions and spectra of the different chemical components. The basic
model underlying this decomposition, usually termed multivariate curve resolu-
tion (MCR) in chemometrics, corresponds to the Lambert-Beer law written in a
matrix form. This factorial model states a bilinear relation between the matrix
of observations and the two matrices of contributions containing concentration
profiles and spectra, respectively. It should be noted that this extends to the
analysis of spectral and hyperspectral images when investigating a specimen
(in microscopy) or a scene (in remote sensing). Also, the bilinear model can
be extended for the analysis of multiple data sets that are meant to connect dif-
ferent experiments together. Overall, MCR can be applied in situations where a
reasonable approximation of the bilinear model, or any other fundamental basic
equation that has the same mathematical structure, holds.
Application of MCR methods is broad, quite straightforward, and provides
results which are readily chemically/physically interpretable. These assets
explain why MCR has spread in the chemical literature and in many other sci-
entific fields. However, considering the mathematical conditions for exact res-
olution of the MCR problem, some theoretical issues remain and are currently
the subject of intensive research. The most puzzling of these issues is the
so-called rotational ambiguity of the resolution. In more common words, this
translates into the fact that a unique solution cannot be obtained in general.
Then, particular attention should be paid to the initial condition, or to the con-
straints applied during resolution, and it is important to assess the extent of
rotational ambiguity before any definitive conclusion to be drawn. Consider-
ing these aspects, one may notice a certain antagonism in MCR between wide
applicability and high interpretability on the one hand and mathematical com-
plexity of the resolution on the other hand. This explains to a large extend the
continuous development of this topic into a proper research field, still very
much in progress.
Introduction Chapter 1 3
Taking a broader perspective, spectral unmixing enters the more general

category of inverse problems, important, and ubiquitous problems in analyti-
cal science and data analysis. From a set of (spectral) observations, one aims
to extract the unknown sources that produced the data but could not be
observed directly. Mixture analysis, MCR, blind source separation, linear
unmixing, etc. are methods that share this objective but were developed in dif-
ferent scientific fields, chemistry, statistics, or signal and image processing.
3 BOOK CONTENT AND ORGANIZATION

The book starts with Chapter 2 that introduces the key concepts and provides
an overview of the progress in MCR with an emphasis on applications to
spectroscopic data. Focus is on constraints, multiset analysis, and quantitative
aspects in multivariate curve resolution alternating least squares (MCR-ALS).
Next, Chapter 3 revisits the concept of variable purity, with purity defined as
the observation of a nonzero contribution from one and only one of the mix-
ture components. Issues and solutions relative to rotational ambiguity of the
MCR solutions, currently a very active research topic, are then discussed in
Chapters 4 and 5. Chapter 4 sets the basis of the problem and focuses on a
nonlinear constrained optimization approach for the direct calculation of max-
imum and minimum band boundaries of feasible solutions. In contrast,
Chapter 5 aims to provide a systematic introduction to the concept of area
of feasible solutions, from which feasible solutions can be derived. With
Chapter 6, spectral unmixing and spectral mixture analysis are introduced.
These methods aim at extracting the spectral characteristics and quantifying
the spatial distribution over a spectral image. This chapter goes beyond the
state of art by introducing nonlinear approaches to SU which allows to take
into consideration more complex mixing process or spectral variability of
the sources. Chapter 7 covers the basic of independent component analysis,
a source separation method initially developed in the field of telecommunica-
tions and now applied in different domains including chemometrics and spec-
troscopy. Chapter 8 deals with a Bayesian positive source separation approach
of the MCR problem which is motivated by the search of unique solutions.
The second part of the book, oriented more towards applications, starts with
Chapter 9. It introduces a wavelet compression strategy that facilitates the
application of MCR to large data sets. Chapter 10 deals with chromatography
coupled with spectral detection, the type of data which originally motivated
development of MCR, and extends to the application of trilinear approaches.
With Chapter 11, the focus is on the application of MCR-ALS for ultrafast
time-resolved absorption spectroscopy data. Chapter 12 tackles the analysis
of hyperspectral images of biological samples with the use of automated data
preprocessing and improved MCR methods, increasing the sensitivity and
accuracy of the chemical images obtained. In Chapter 13, the integration of
wavelet transform with multivariate image analysis in a multiresolution anal-

ysis approach opens the possibility of simultaneously accomplishing denois-
ing and feature selection. With Chapter 14, a new constraint that allows
forcing some information related to the low-frequency character of the com-
ponents profiles and distribution maps in MCR-ALS is introduced.
Chapter 15 discusses the potential of super-resolution in vibrational spectros-
copy imaging, merging instrumental and algorithmic developments.
Chapter 16 deals with the current topic of biomarker imaging for early cancer
detection applying MCR to magnetic resonance images. Chapters 17 and 18
provide ways and means to deal with remotely sensed data. Chapter 17
focuses on the use of spectral libraries for spectral mixtures analysis. How
to compose, handle, and optimize endmember libraries are the issues dis-
cussed in detail. Chapter 18 reviews the recent developments of spectral
unmixing algorithms that incorporate spatial information, termed spatial-
spectral unmixing. To close, Chapter 19 presents a sparse approach for spec-
tral unmixing of hyperspectral images, which provides better interpretability
of the results obtained.
Chapter 2
Multivariate Curve
Resolution-Alternating Least
Squares for Spectroscopic Data
A. de Juan*,1 and R. Tauler†
*
Universitat de Barcelona, Barcelona, Spain
†
Institute of Environmental Assessment and Water Research (IDÆA) SPANISH COUNCIL
OF SCIENTIFIC RESEARCH (CSIC), Barcelona, Spain
1
Corresponding author: e-mail: anna.dejuan@ub.edu
Chapter Outline
1 MCR: The Concept and the Link 4 MCR-ALS Applied to HSI
with Spectroscopic Data 5 Analysis 28
2 MCR-ALS: Algorithm and Data 4.1 Encoding Image Information:
Set Configuration 7 The Spatial Dimension 28
2.1 MCR-ALS Algorithm: Steps 11 4.2 Image Multiset Analysis 31
2.2 Constraints 12 4.3 MCR Postprocessing 35
3 MCR-ALS Applied to Process 5 MCR-ALS and Quantitative
Analysis 17 Analysis 37
3.1 Encoding Process 5.1 Second-order Calibration 37
Information: Sequentiality 5.2 First-order Calibration:
and Physicochemical Models 17 Correlation Constraint 40
3.2 Multiset Analysis: 6 MCR-ALS and Other Bilinear
Multiexperiment Analysis Decomposition Methods 42
and Data Fusion 21 References 44
1 MCR: THE CONCEPT AND THE LINK WITH

SPECTROSCOPIC DATA
Multivariate curve resolution (MCR) is the generic denomination of a family
of methods meant to solve the mixture analysis problem, i.e., able to provide a
chemically (scientifically) meaningful additive bilinear model of pure contri-
butions from the sole information of an original data matrix including a mixed
measurement [1–7]. Thus, MCR analysis describes the total signal of a multi-
component data set (D) as the sum of the signal contributions coming from
Data Handling in Science and Technology, Vol. 30. http://dx.doi.org/10.1016/B978-0-444-63638-6.00002-4

Copyright © 2016 Elsevier B.V. All rights reserved. 5
each of the pure components, Di (Eq. 1). Each term of this additive model can
be expressed by the product of a dyad of profiles, cisTi , where sTi is the pure
response of the component weighted by the related concentration profile, ci
(Eq. 2). Finally, the classical compact expression of the bilinear additive
model is shown in Eq. (3):
X
D¼ Di + E (1)
i
X
D¼ ci sTi + E (2)
i
D ¼ CST + E (3)
T
where C contains the concentration profiles of all components and S the
related pure responses. E is the matrix expressing the error or variance unex-
plained by the bilinear model in all earlier equations.
The expressions shown in Eqs. (1)–(3) remind clearly the natural bilinear
model of spectroscopic measurements, the Beer–Lambert law, where the var-
iation of a data table containing the spectra of a multicomponent system (D)
can be expressed through the product of the pure spectra (absorptivities)
(ST) of the components by their related concentration profiles (C). Then, it
is not surprising that the first applications of MCR were devoted to analyze
spectroscopic data because of the identical underlying models of the measure-
ment and the method [1,5]. In difference to other bilinear decomposition
methods, MCR provides meaningful profiles because chemical properties
related to the concentration profiles and spectra are actively incorporated in
the optimization of the bilinear model.
There are many multicomponent systems described by spectroscopic data.
MCR adapts to this diversity by tuning the way of application of the method
according to the characteristics of the spectroscopic technique and the concen-
tration profiles. In a very general way, we can distinguish between process and
mixture data sets. Process data sets show very structured concentration profiles,
displaying a smooth variation as a function of a process variable. Typical exam-
ples may be an HPLC–DAD chromatographic data set, where the concentration
(elution) profiles have a peak-shaped signal or reactions monitored by spectro-
scopic techniques (see Fig. 1A). Instead, the concentration profiles of a mixture
data set can vary in a nonpatterned way. This less structured variation can come
from the nature of the data, e.g., a set of independent samples, or from the data
set configuration needed for MCR application, e.g., hyperspectral images
(HSIs) are treated by using a data table of pixel spectra, which does not pre-
serve the spatial organization of the original measurement (see Section 4 and
chapters 6,12,13,15,17 and 19 in this book for more detail on this kind of mea-
surements). MCR applies in all the scenarios described, but specificities of the
different data set typologies should be taken into account in the data set config-
uration and in the different steps of application of the algorithm.
Before describing the details and application of MCR algorithms, a neces-
sary comment has to be made regarding one of the fundamental assumptions
MCR-ALS for Spectroscopic Data Chapter 2 7
FIG. 1 (A) Bilinear model of an HPLC–DAD chromatogram and (B) bilinear model of a
thermal-dependent process monitored by circular dichroism.
of MCR, i.e., the bilinearity of spectroscopic data. Indeed, the natural fulfill-
ment of the Beer–Lambert law may be affected by signal artifacts, such as
scattering in near infrared spectroscopy [8,9], or fluorescence contributions
in Raman spectra [10]. Most of the times, the bilinear behavior is easily
recovered by suitable preprocessing, e.g., scatter or baseline correction,
adapted to the spectroscopic measurement of interest [11–14]. Because of
the nature of the spectroscopy used, some techniques need more dedicated
and intensive preprocessing, like ultrafast spectroscopy measurements (see
Chapter 11) [15,16] and, in extreme cases, nonlinear unmixing methods can
be applied (see Chapter 6).
2 MCR-ALS: ALGORITHM AND DATA SET CONFIGURATION

Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS) is an
iterative algorithm that solves the bilinear MCR model by optimizing
concentration profiles and pure spectra in an alternating least squares way

under constraints [5,6,17–20]. To trigger the optimization, a set of initial esti-
mates of spectral or concentration profiles are needed. The general minimiza-
tion step in the iterative optimization can be expressed as follows:

min D CST (4)
taking into account that C and ST are calculated in each iterative cycle.
The main assets of this algorithm are the variety of information that can be
included in the optimization process and the capability to work with single
data tables and complete and incomplete multisets [3,6,20–22].
Multisets are defined as data structures formed by more than one data table.
Multiset structures can be extremely flexible and can be formed by data matri-
ces of different size and meaning. All multiset structures present a bilinear
model and the obvious requirement is that the different subsets share compo-
nents in common. Different multiset typologies can be obtained depending on
how the data subsets are connected. Multisets can be structured as: (a) row-wise
augmented matrices, when data tables are appended besides each other, e.g.,
multitechnique process monitoring (Eq. 5), (b) column-wise augmented matri-
ces, when data tables are one on top of each other, e.g., different batches moni-
tored in a production process (Eq. 6), and (c) row- and column-wise augmented
matrices, when the multiset extends in the row and column direction (Eq. 7).

½D1 D2 D3 … DL ¼ C ST1 ST2 ST3 … STL + ½E1 E2 E3 …EL ¼ CSTaug (5)
0 1 0 1 0 1
D1 C1 E1
B D2 C B C2 C B E2 C
B C B C T B C
B D3 C ¼ B C3 CS + B E3 C ¼ Caug ST + Eaug (6)
B C B C B C
@…A @…A @…A
DK CK EK
2 3 0 1
D11 D12 D13 … D1L C1
6 D21 D22 D23 … D2L 7 B C2 C
6 7 B C
6 D31 D32 D33 … D3L 7 ¼ B C3 C ST ST ST … ST
6 7 B C 1 2 3 L
4 … … … … … 5 @…A
DK1 DK2 DK3 … DKL CK
0 1 (7)
E11 E12 E13 … E1L
B E21 E22 E23 … E2L C
B C
+B C
B E31 E32 E33 … E3L C ¼ Caug Saug + Eaug
T
@… … … … … A
EK1 EK2 EK3 … EKL
The minimization function expressed in Eq. (4) also holds, just replacing
the C and/or ST matrices by their augmented versions when needed.
Looking at the different expressions in Eqs. (5)–(7), it can be understood

that the data tables appended in a multiset should have, at least, a common
mode (the concentration or the spectral mode). This means that the size
(dimensionality) and profile behavior must be the same. For instance, when
a process is monitored simultaneously with different analytical techniques,
i.e., giving a row-wise augmented matrix, the measurements (rows) collected
should refer to the same process stages in all data matrices because the single
concentration matrix will define the process behavior for all data tables trea-
ted together. Likewise, when different experiments are carried out using the
same spectroscopic technique, all of them should be performed using the same
spectral range and spectral resolution, because a single matrix of pure spectral
signatures has to be valid for all experiments.
A recent particular situation is the one presented by incomplete multi-
sets [22]. In this case, a regular multiset structure cannot be built and missing
blocks are present (see Fig. 2B).
To cope with this kind of data structure, the minimization problem is
slightly modified. The incomplete structure can be generally defined by sev-
eral complete multisets with their related bilinear models (DA and DB in
D1 D2 C1 ST1 ST2
ST
D3 D4 C2
D C
D1 D2 C1 ST1 ST2
DA
ST
D3 C2
D C
DB
FIG. 2 Multiset examples and related bilinear models. (A) Complete row- and column-wise aug-
mented matrix and (B) incomplete row- and column-wise augmented matrix.
Fig. 2B). The regular multisets are optimized in each iterative cycle and the
minimization function optimizes all residuals from all calculated bilinear
models in a simultaneous way (see Eq. 8).

min DA CA STA + DB CB STB (8)
Averaged profiles or selected profiles from the different regular multisets

optimized are provided as the resolved profiles in each iterative cycle or as the
final MCR solution [22,23].
Although the equations and minimizations presented take as a basis an
ordinary least squares optimization, the same algorithm can be used when
information about the noise level and structure is available. In this case,
weighted MCR-ALS variants are also available [24–28].
The flexibility of data configurations and the possibility to introduce gen-
eral information in MCR-ALS models is outstanding, but all MCR algorithms
are affected by the so-called ambiguity phenomenon. This means that differ-
ent combinations of concentration profiles and spectra can describe equally
well, in terms of model fit, the original data set [19,29]. Indeed, three main
modalities of this phenomenon can be encountered
(a) Permutation ambiguity. There is no sorting order on the MCR compo-
nents. Therefore, they can be shuffled in the concentration and spectra
matrix (always keeping the right correspondence of the dyads) and pro-
vide identical result.
(b) Intensity ambiguity. Dyads of profiles having the same shape but differ-
ent relative scales between concentration profile and spectrum reproduce
equally well the original data set. Indeed, concentration values and pure
spectra intensities in ci and sTi profiles are always in arbitrary units unless
reference information on real intensities is available and actively used in
the resolution process. This can be seen taking Eq. (2) as initial expres-
sion of the bilinear model skipping the error term. Eq. (2) can be easily
transformed into Eq. (9) by introducing an unknown and different scaling
factor for each i-th component, ki, to define the related dyad cisTi .
X
D¼ ci sTi (2)
i
X
1
D¼ ðc i k i Þ sTi (9)
i
ki
(c) Rotational ambiguity. Sets of concentration profiles and spectra with dif-
ferent shapes can reproduce the original data set with the same fit quality.
This is the most relevant kind of ambiguity and can be expressed taking
as a basis the bilinear model in Eq. (3) and modifying it into Eq. (10),
by introducing a T transformation matrix, as follows:
D ¼ CST + E (3)

D ¼ ðCTÞ T1 ST + E (10)
Different ways to assess the extent and location of ambiguity in an MCR
solution will be explained in Chapters 4 and 5 [30–33]. The way to decrease
or suppress this phenomenon is by introducing chemical or mathematical con-
straints to model the resolved profiles or by working with more information
related to the same chemical system, i.e., with multiset structures.
2.1 MCR-ALS Algorithm: Steps

The earlier section described the general concept behind the MCR-ALS algo-
rithm, i.e., the iterative alternating least squares optimization of concentration
profiles and pure spectra under constraints. Below, the general steps in the
MCR-ALS algorithm are detailed
1. Determination of the number of components.
2. Generation of initial estimates of C or ST.
3. Iterative alternating least squares optimization of C and ST under con-
straints until convergence is achieved.
The number of components in a data set can be known beforehand or deter-
mined by auxiliary means, such as principal component analysis (PCA). Both
methods aim at describing the maximum variance of the data set, and the
assumption is that a data set can be described with a number of MCR contribu-
tions equal to the number of significant principal components used for the same
purpose. In this context, examination of the magnitude of singular values is typ-
ically used, i.e., singular values related to chemical contributions are large,
whereas singular values related to noise are small and similar among them. This
diagnostic can be complemented by looking at the emergence of noisy patterns
in scores and loadings profiles, typical in noise-related components. When in
doubt, several MCR models with different number of components can be calcu-
lated. The final model will be selected as the smallest one providing an optimal
model fit and chemically meaningful resolved profiles.
Initial estimates in MCR-ALS can be concentration profiles or spectra.
A good hint in the generation of estimates is starting with sensible initial
guesses that present the same general properties sought in the profiles to be
recovered. This means that initial random values are generally of no help
because the profiles expected do not have this numerical structure [4–7,34].
Initial estimates can be based on previous knowledge, e.g., spectra of compo-
nents known to be present in the data set, spectra at maximum of chro-
matographic peaks, or can be obtained by auxiliary methods. The selection
of methods to obtain initial estimates will depend essentially on the structure
of the data set, particularly on the presence or absence of systematic behavior
in the concentration direction (process data vs. mixture or image data).
A general family of methods apt to generate initial estimates for any kind
of data sets are the so-called purest variable selection methods, such as
SIMPLISMA [35], orthogonal projection approach [36], and key set factor
analysis [37,38]. All these methods have the common goal of selecting the
most dissimilar rows or columns in the original data set D and provide, as a
consequence, initial estimates of spectra or concentration profiles, respec-
tively. They will be explained in more detail in Chapter 3. A good property
of these methods is that, when selective rows or columns for a particular com-
ponent exist, the related pure spectrum or concentration profile is readily
recovered. However, it is generally incorrect associating the profiles extracted
by these methods with pure component profiles because selectivity cannot
be always ensured. Methods related to the generation of initial estimates for
process data will be explained in Section 3.
Once initial estimates are obtained, optimization takes place until conver-
gence is achieved. To control convergence, figures of merit related to the
model fit are used, such as the lack of fit (% LOF) or variance explained
(in Eqs. 11 and 12, respectively).
sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P 2
i, j eij
% LOF ¼ 100 P 2 (11)
i, j dij
P 2!
i, j eij
% var:expl: ¼ 100 1 P 2 (12)
i, j dij
where eij is the residual associated with the MCR reproduction of the related
element in the original data set, dij.
The optimization is usually finished when the difference of model fit
between consecutive iterations does not improve significantly (e.g., a differ-
ence of less than 0.1% among the lack of fit between consecutive iterations).
Other possibilities are using a maximum number of iterations or criteria based
on the comparison of shapes of the optimized profiles among iterations.
2.2 Constraints
Constraints are the essential part of the MCR-ALS algorithm and they have
two main objectives: (a) introducing chemical and mathematical knowledge
to provide chemical meaning to the concentration profiles and pure spectra
and (b) reducing or suppressing the ambiguity related to the MCR solutions
[4–7,19].
A constraint can be defined as any chemical or mathematical property
systematically present in the component profiles of the data set. Within
each iterative cycle, the computed profiles by least squares are modified
so that they obey the conditions defined by the preselected constraints
(see Fig. 3).
A Nonnegativity B Unimodality
0.35 0.35
0.3 0.35
0.3 0.3
0.25 0.3
0.2 0.25 0.25
0.25
0.15 0.2 0.2
0.2
0.1 0.15 0.15
0.05 0.15
0.1 0.1
0 0.1
0.05 0.05
–0.05 0.05
–0.1 0 0
0 0 5 10 15 20 25 30 35 40 45 50
0 10 20 30 40 50 0 10 20 30 40 50 0 5 10 15 20 25 30 35 40 45 50
Elution times Elution times Elution times Elution times
C Closure
S = ctotal
Mass balance
0.35
0.3 ctotal 0.3 ctotal
0.25 0.25
0.2 0.2
0.15 0.15
0.1 0.1
0.05 0.05
0
5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
FIG. 3 Examples of constraints. (A) Nonnegativity, (B) unimodality, and (C) closure. Unconstrained profile is black bold and constrained profile is red bold
(light gray bold in print version).
There are many approaches to implement constraints: smooth least squares

[39,40] or penalty-based [41,42] procedures and other methodologies that
work directly replacing erroneous profiles or elements in a profile by different
values. The degree of tolerance in the fulfillment of constraints can also be
tuned according to the properties of the data set to be treated, e.g., depending
on the noise level of the data.
In the MCR-ALS framework, flexibility in the application of constraints is
a key element. There is no default list of constraints, because default chemical
data sets do not exist. Selection of constraints is an active step that MCR users
have to be responsible for since they know the properties of their data sets and
of the profiles to be resolved. Hence, application of all constraints is optional
and they can be used differently (a) mode-wise, i.e., the concentration and the
spectral profiles may have different properties, (b) compound-wise, i.e., the
different compounds (profiles) within the C or ST matrix may behave in a dif-
ferent way, and (c) subset-wise, i.e., the properties of different submatrices in
C or ST in a multiset structure may be very diverse.
The most common constraints, applicable to any kind of data set
structure are
(a) Nonnegativity. Forces the profiles to be formed by positive values. Can
be implemented replacing negative values by zeros or with softer algo-
rithms, such as nonnegative least squares or fast nonnegative least
squares [39]. It applies to all concentration profiles and to many spectro-
scopic responses. It should be avoided in certain kinds of spectroscopic
profiles, such as those coming from circular dichroism or electronic para-
magnetic resonance, which provide natural negative measurements, or
when working with derivative spectra.
(b) Unimodality. Allows the presence of a single maximum per profile. It
refers to peak-shaped profiles, such as elution profiles in chromatogra-
phy, and to monotonic profiles, i.e., steadily increasing or decreasing,
like some reaction profiles. Least squares implementations of this con-
straint exist, usually linked simultaneously to nonnegativity [40]. In con-
straints working by replacement of nonunimodal values, variable degrees
of tolerance on the fulfillment of the constraint can also be set [43].
When working with multisets, unimodality is applied separately to the
profiles of each of the subsets of the augmented C or ST matrix.
(c) Closure. It is the expression of the mass balance condition. Applies to
some concentration profiles of reaction systems and rescales the concen-
tration profiles so that the total sum of the concentration of compounds in
the closure amounts to a constant value during all the reaction. It can be
applied as a strict equality or using a least squares implementation [44].
When working with multisets, closure is applied separately to the profiles
of each of the subsets of the augmented C matrix.
(d) Selectivity and local rank. This constraint reflects that the local rank in
some concentration windows or spectral ranges can be lower than the
total rank of the data set, i.e., not all compounds are present everywhere
in the data set. When selective concentration or spectral regions for a par-
ticular compound exist, the rest of compounds are set to zero and the pure
spectrum or concentration profile for that compound is readily recovered.
Local rank constraints apply when only some compounds are absent in a
certain concentration or spectral range; however, these constraints also
play a very relevant role in the data set to decrease and, in appropriate
conditions, suppress ambiguity. In next sections, it will be described
how local rank and selective regions can be gathered from a process or
an image data set.
(e) Equality constraints. They apply when some elements in a profile or a
full profile are known, e.g., a pure spectrum of a compound is known.
They set the profile or the known elements in it to be equal to the known
information. Local rank and selectivity can be considered equality con-
straints when absent compounds are set equal to zero. Likewise, nonnega-
tivity can be applied as an equality constraint when negative values are
set to be equal to zero.
Some additional constraints are associated only with multiset structures:
(a) Correspondence of species. This constraint extends the idea of local rank
and selectivity to a multiset context and has the same beneficial effect. In
this case, for column-wise augmented multisets, this constraint expresses
the presence/absence of compounds in the original experiments. Therefore,
full concentration profiles in certain C submatrices are forced to have null
profiles and a correct correspondence between the compounds present and
absent in the different experiments is explicitly defined [5,6,20,21].
(b) Model constraints. Classical MCR works under the assumption of a bilin-
ear model to describe the data. However, there may be multisets that
behave according to multiway models, such as trilinear or multilinear
models [20,21,45,46], or factor interaction (Tucker) models [21,47].
These more complex models can also be implemented under the form
of constraints in profiles from augmented C or ST matrices, e.g., trilinear-
ity can be imposed in profiles of augmented matrices by forcing them to
have the same shape and preserving their natural scale. To do so, as can
be seen in Fig. 4, the profiles of a certain compound in the different C
submatrices, cni , are arranged one besides the other in a single data
matrix. Singular value decomposition is then applied to this matrix. The
first score profile is adopted as the basic shape of this compound for all
C submatrices. To recover the properly scaled constrained profile for
each C submatrix, the common score profile is multiplied by the related
loading, which includes the scale information.
The main difference with this constraint and the approach used by
natural multiway multilinear methods, such as parallel factor analysis
(PARAFAC) [48] or direct trilinear decomposition (DTD) [49], is the
Caug ST
D1 cn1
D2 cn2
D3 cn3
Daug
Selection of profile n Constrained profile n

cn1
Folding SVD zn
cn2 cn1cn2cn3 cn ⊗ cn Rebuilding
augmented
zn
Cn scores
cn3 1st score
FIG. 4 Implementation of trilinearity constraint.
compound-wise application of the model constraint. Therefore, hybrid mod-

els, which have some compounds behaving in a bilinear way and others in a
trilinear or multilinear way, can coexist [50,51]. When trilinearity or multili-
nearity is applied as constraints in MCR, the related compounds are resolved
in a unique way, without ambiguities [19,29].
Finally, there are some constraints linked to particular tasks performed
often by other data analysis methods, such as hard-modeling methods or mul-
tivariate calibration approaches. These are
(a) Hard-modeling constraint. The mechanism or the physicochemical law
driving a reaction can be included as an additional constraint. Process
profiles can be fitted with a hard model and the related model parameters,
i.e., reaction constants, can also be retrieved. More detail on this con-
straint is given in Section 3, devoted to process analysis [21,52].
(b) Correlation constraint. This constraint includes an inner calibration
model within MCR to provide quantitative information, in real concentra-
tion units, to the resolved concentration profiles of the components of
interest. More detail on this constraint will be given in Section 5, dedi-
cated to MCR for quantitative analysis [34,53].
In principle, any systematic property of a concentration or spectral profile can
be potentially encoded as a constraint. The possibilities grow to give response
to new kinds of problems and measurements. Clear examples of new con-

straints are related to image analysis, where the spatial pattern of the image
starts to be taken into account (see Chapters 14 and 18), or to particular kinds
of response or concentration profiles that present properties closely defined by
mathematical conditions, such as sparseness (see Chapter 19).
3 MCR-ALS APPLIED TO PROCESS ANALYSIS

MCR was born to provide a solution to describe the evolution of multicompo-
nent processes. The term process was applied to define a wide family of
problems, all of them sharing the fact of having a continuous nonrandom evo-
lution in the concentration direction. Therefore, a process could be from a
classical reaction-driven system to an evolving system with no chemical trans-
formation involved, like a chromatographic elution, a blending process, or a
conformational or polymorphic transition. In all these systems, the concentra-
tion profiles are never a set of random numbers; instead, they reflect the
smooth evolution of a particular compound as a function of a certain process
variable (time, temperature, etc.). The structured concentration direction in
processes enabled the use of particular kinds of information under the form
of constraints to improve the description of these systems by MCR.
3.1 Encoding Process Information: Sequentiality

and Physicochemical Models
3.1.1 Sequentiality
A very common fact in processes is the sequentiality in the evolution of com-
ponents, i.e., the fact that a system can be defined by a set of components that
emerge and decay in an ordered way. This property explains the frequent use
of certain constraints, such as unimodality, in a process context. Indeed, a sin-
gle emergence-decay curve for a component is linked to peak-shaped signals,
such as elution profiles in chromatography, or to monotonic and peak-shaped
concentration profiles in processes, i.e., the initial compound in a process can
only decay, the intermediate ones emerge and decay and the last product
keeps increasing until stabilizing at the end of the process.
The sequential structure of many processes is also the core assumption of
auxiliary local rank analysis methods, such as evolving factor analysis (EFA),
that provide information on the sequential emergence-decay concentration win-
dows of the different components in a process [54,55]. Taking advantage of this
knowledge, initial estimates of concentration profiles can be generated and
selectivity and local rank constraints, i.e., concentration windows of presence
and absence of certain compounds along the process direction, can be set [4,19].
Indeed, at the very beginning of MCR methods, many algorithms were
based on the proper setting of local rank subspaces to solve in a correct and
unique way the concentration profiles and related pure spectra of the
compounds involved in a process [3,56–59]. In 1995, Manne proposed two

theorems that could define when a data set could be uniquely and correctly
resolved based only on the patterned local rank structure in the concentration
direction [44]. These theorems stated
(a) The concentration profile of a compound can be correctly resolved if all
components inside its concentration window are also present outside,
(b) The pure spectrum of a compound can be correctly recovered if its con-
centration window is not fully embedded into the window of another
component.
These pioneering theorems refute the false and too common statement saying
that selectivity for all compounds is needed for unique resolution in MCR. They
also warn about the importance of the active use of local rank information for a
proper recovery of concentration profiles and pure spectra in process analysis.
The retrieval of local rank information from a data set can be done with
local rank analysis methods, such as EFA [55,59]. Modifications of this parent
algorithm are fixed size moving window-EFA (FSMW-EFA) [60], more pow-
erful to detect the presence of minor compounds, or exhaustive EFA [61],
devoted to check the validity of the assumption of sequentiality in a process.
EFA can be easily defined as a chemometric process monitoring. Thus, EFA
performs many local rank analyses including every time a new process observa-
tion (spectrum) in the data subset analyzed. This is done from the beginning to
the end of the process (forward EFA) and in the opposite direction (backward
EFA; see Fig. 5A). The information of all these analyses is displayed in two
plots, related to forward and backward EFA. In these plots, the eigenvalues of
each local rank analysis are plotted in a vertical line as a function of the process
variable designing the last row included in the subset analyzed (being an elution
time, temperature, etc.). All 1st, 2nd, …, n-th eigenvalues are connected with a
line. When the complete data set is considered, at one extreme of the plot, the
number of lines above the noise level indicates the total rank of the system.
Inside the plot, the point (time, temperature, etc.) where an eigenvalue rises
from the noise level indicates the emergence (forward EFA plot) or the decay
(backward EFA plot) of a new process contribution.
Joining the forward and backward EFA plot in a single picture and connect-
ing the first eigenvalue emerging (in the forward EFA plot) with the first decay-
ing (in the backward EFA plot), the second emerging with the second decaying,
and so forth, initial estimates of concentration profiles for a sequential process
can be obtained. Besides, the definition of the concentration window for each
compound allows for setting related local rank constraints, i.e., zones of absence
for each particular compound in the concentration direction (see Fig. 5B).
3.1.2 Physicochemical Models

Chemical process modeling was classically done fitting physicochemical
models to experimental data [62,63]. Actually, physicochemical models based
A Forward EFA Backward EFA
(Beginning → end of process) (Beginning ← end of process)
PCA PCA PCA PCA PCA PCA PCA

PCA
11
11
10.5
log(eigen values)
10.5
log(eigen values)
10
10
9.5
9.5
9
9
8.5
8.5
8 8
7.5 7.5
0 5 10 15 20 25 20 35 40 45 50 0 5 10 15 20 25 20 35 40 45 50
Retention times Retention times
Emergence of compounds Noise level Decay of compounds
B A B C D
EFA plot
11
0 0 0 0
log(eigen values)
10.5
0 0 0 NaN
10
9.5
0 0 0 NaN
9
0 0 NaN NaN
8.5 0 0 NaN 0
8 0 0 NaN 0
7.5
5 10 15 20 25 30 35 40 45 50 0 NaN 0 0
Retention times NaN NaN 0 0
NaN NaN 0 0
NaN 0 0 0
NaN 0 0 0
FIG. 5 Modus operandi of evolving factor analysis (EFA). (A) Forward and backward EFA plots, (B) joint forward (solid black lines) and backward EFA plot
(dashed black lines), and derived concentration profiles. At the bottom, concentration windows of components (in bold colored (different bold gray shades in the
print version) lines) and (C) translation of concentration window information into a mask matrix of local rank information.
on first principles are the best way to describe a process evolution in an accu-
rate way as long as a reliable model can be found to describe the phenomenon
under study.
Classical hard-modeling works considering that concentration profiles can
be described by mathematical functions based on the use of suitable process
parameters, e.g., rate constants [63]. Tuning these parameters appropriately
with iterative optimization algorithms, the process evolution can be properly
modeled. Classical hard-modeling approaches also take into account implic-
itly the bilinear model of spectroscopic data. Thus, the function to be mini-
mized is the same as in Eq. (4), conveniently transformed to the form
presented in Eq. (14):

min D CST (4)
1
ST ¼ CT C CT D ¼ C + D (13)
min kD CC + Dk ¼ min kDðI CC + Þk, where C ¼ f ðki Þ (14)

The form of Eq. (14) implies that only the parameters linked to the func-
tion describing the concentration profiles need to be fitted. Pure spectra of the
process compounds can be afterward calculated using Eq. (13). In classical
hard-modeling approaches, there is a single set of parameters, ki, providing
concentration profiles that optimally fit the variation of the experimental data
and, therefore, a unique answer can be generally obtained. However, the nec-
essary condition to ensure the successful application of hard-modeling
approaches is that all contributions to the signal measured must be described
by the physicochemical model. This condition is mandatory because the
model fits the raw spectroscopic data, D.
Since physicochemical models are very accurate expressions to describe
the shapes of process concentration profiles, they can also be used in MCR
under the form of a hard-modeling constraint [52,64,65]. In this case, as any
other constraint, the physicochemical model will act on the concentration pro-
files, C, and the function to be minimized will be
min kCALS CHM k (15)
CALS meaning the concentration profiles calculated by least squares within
the MCR optimization and CHM the profiles shaped according to a physico-
chemical model. In this constraint, the related parameters of the model will
be iteratively optimized until CALS is optimally fitted by CHM. The con-
strained profiles will be equal to CHM and, as with any other hard-modeling
fitting, the optimized parameters will be obtained as an additional result.
The main advantage of using hard-modeling as a constraint within MCR
can be understood by comparing the two minimization functions in Eqs. (14)
and (15). In classical hard-modeling, the raw measurement D is fitted using
the model, whereas the model fits the concentration profiles, C, in MCR.
Another random document with
no related content on Scribd:
“The poor Prince de la Paz is covered with wounds and
contusions, and is cast into prison, where he constantly invokes the
terrible moment of his death. He thinks of nobody but his friend the
Grand Duke of Berg, and says he is the only person to whom he
looks for his salvation.
“My father, mother, and I have talked with your respected
commander. He will tell you all. I trust in your friendship, and that by
that you will save us all three and the poor prisoner.
“I have not time to say more, but I trust in you. My father will add
two lines to this letter.
“I am, from my heart,
“Your most affectionate Sister and Friend,
“Maria Luisa.”
To this letter Carlos IV. added a postscript:
“Sir and very dear Brother,

“Having talked to your worthy commander, and informed
him of all that has happened, I beg you to tell the Emperor that I
intreat him to set free the poor Prince de la Paz, who only suffers
from having been a friend of France, and at the same time beg of
him to let us go to a place which will suit us, and take with us the
same Prince. We are going now to Badajoz. I beg your reply before
then, in case we are absolutely left without means of seeing each
other, for my life is only in you and in the Emperor. In the meanwhile
I am,
“Your very affectionate Brother and Friend,
“Carlos.”
The General was also given a letter from the Queen to Murat,
which ran thus:
“Sir, my dear Brother,

“I have no friend but Your Highness. The King, my beloved
husband, writes to you imploring your friendship, for in that lies our
only hope. We both beg of you to prove you are our friend by
informing the Emperor of our sincere friendship, and of the affection
we have always professed for him, you, and all the French.
“Poor Prince de la Paz, who is wounded and imprisoned for being
our friend, is passionately attached to all France, and he is suffering
now for having desired the arrival of your troops, and for having been
our only permanent friend. He would have gone to see you had he
been free, and now he does not cease to speak of you and express
his desire to see the Emperor. Help us to end our days quietly in a
place suitable to the health of the King, which, like mine, is delicate,
and let it be in company with our friend, who is also that of Your
Highness.
“My daughter will be my interpreter if I do not have the
satisfaction of knowing Your Highness personally and talking to you.
Could you make an effort to see us, if only for a minute, by night or
when you like? Your worthy officer will tell you all we have said.
“I hope you will be able to manage what we want, and that you
will pardon all the slips and omissions in the matter, for I do not know
where I am, and you must believe that this has been from no slight to
you nor lack.
“May you live many years!
“Your most affectionate
“Luisa.”
The Queen became quite desperate as the days went by,
bringing no definite help from the Grand Duke of Berg, and in one of
her letters to her daughter she writes:
“If the Grand Duke does not see that the Emperor gives orders
for the stoppage of the intrigues against his friend the Prince de la
Paz, against me and my daughter, none of us will be safe. All the
malevolent people get round my son, and he believes them like
oracles, and on his own part he is not very inclined to magnanimity
and clemency. He must expect sad results from all this. I and my
husband think that, if my son sees the Emperor before he has given
his orders, he and those with him will tell him so many lies that he
will doubt the truth. For this reason we would beg the Grand Duke to
let the Emperor know that we are absolutely in his hands, hoping he
will give tranquillity to the King, my husband, me, and the Prince de
la Paz, whom we desire to have with us, and end our days
peacefully in a place suited to our health without giving the least
trouble to anybody. We urgently beg the Grand Duke to let us have
daily news of our mutual friend, the Prince de la Paz, because we
know absolutely nothing.”
The King added the following words in his own handwriting:
“I asked the Queen to write this, as my pains prevent my writing
much.”
The next letter from the Queen of Spain to her daughter for the
Grand Duke of Berg is without a date:
“The King, my husband, and I do not wish to be importunate nor

troublesome to the Grand Duke, who has so much to do; but we
have no other friend but him and the Emperor, and in him rest the
hopes of the King, those of the Prince de la Paz, the friend of the
Grand Duke and our own intimate one, and those of my daughter
and myself. My daughter wrote me yesterday afternoon what the
Grand Duke had said, and our hearts are filled with gratitude and
comfort, hoping for all that is good from the sacred and incomparable
personages of the Emperor and Grand Duke. But we do not want
him to be ignorant of what we know in spite of nobody telling us
anything or answering our questions, important as it was for us to
have a reply. However, we regard it all with indifference, and the only
thing which interests us is the welfare of our only and innocent
friend, the Prince de la Paz, who is also the Grand Duke’s, as he
exclaimed in his prison in the midst of the horrible treatment to which
he was exposed; for he always called the Grand Duke his friend, as
he did before the conspiracy, and he says constantly: ‘If I could only
have the good fortune for the Grand Duke to come here, I should
have nothing to fear.’
“He wanted you to come to the Court, and he was flattered by the
pleasure the Grand Duke showed in accepting his house as a
dwelling. He had some presents ready to give you, and he thought of
nothing but the moment when he could present himself to the
Emperor and the Grand Duke with all imaginable ardour. But now we
are in continual fear that he will take his life, or that he will be more
closely imprisoned if his enemies know that there is a question of his
being saved. Would it not be possible to take some precautionary
measures before the definitive resolution? The Grand Duke could
send some troops without saying why. Could they not come to the
prison and disperse the guard over him, without giving it time to fire a
shot or do anything against the Prince? For there is reason to fear
that it would do so, as they all know his wish to die, and they would
glory in killing him. So the guard could be absolutely under the
command of the Grand Duke; and if not, the Grand Duke can be
sure that the Prince de la Paz will die if he continue in the power of
the worthless traitors and in the hands of my son. Hence we repeat
the plea that he should be removed from the power of the
bloodthirsty gardes de corps, my son, and his evil companions; for
we are in continual fear of his life, although the Emperor and the
Grand Duke wish to save him. We repeat, therefore, the entreaty
that the Grand Duke should take every measure for this object,
because if time be lost his life is not safe, as it would certainly be
easier to protect the Prince in the midst of carnivorous lions and
tigers.
“After dinner yesterday, my son was with Infantado, Escoiquiz,
who is a malignant cleric, and San Carlos, who is worse than all; and
this makes us tremble, as the secret conference lasted from half-
past one till half-past three. The gentil hombre who is with my son
Charles is a cousin of San Carlos; he has talent and some learning,
but he is a malignant American and a great enemy of ours, like his
cousin San Carlos, in spite of all they have received from the King,
my husband, at the request of the Prince de la Paz, to whom they
say they are related. All those who are with my son Charles are
mixed up in the same intrigue, and inclined to do all possible harm,
and what is reported as true is the greatest untruth.
“I hope the Grand Duke will pardon all my blunders and mistakes
when I write French, as it is forty-two years since I came to Spain at
thirteen and a half years of age, so, although I speak French, I do
not speak it well.
“The Grand Duke will know what helps me, and will pardon all my
faults of the language.
“Luisa.”
Ferdinand, in his blind belief in Escoiquiz, disregarded the

counsel of other men, and, as Escoiquiz only thought of conciliating
the Corsican so as to advance his plan of Ferdinand’s union with a
member of the House of Bonaparte, the power of the French
increased daily.
It was believed that all the intrigues of Beauharnais were only to
keep the sceptre in the hand which held it, and the silly credulity on
the part of Escoiquiz was the chief cause of the consequent
misfortunes.
To a genius like Napoleon the situation of Spain was an easy
prey to his ambition, and its state of submission to the French was
seen in the fact of Caballero conforming to Murat’s desire to become
the possessor of the sword which was surrendered to Charles V. by
Francis I. of France after the Battle of Pavia.
The function in which Spain lost this heirloom is described in the
Gaceta de Madrid of April 5, 1808. The sword was borne in state to
the Grand Duke’s house. It was placed on a silver tray covered with
a puce-coloured silk cloth trimmed with a wide bright fringe, and Don
Carlos Montarges, the honorary Chief Armourer, and his attendant,
Don Manuel Trotier, went in the gala carriage with the trophy. The
carriage was drawn by mules in gala attire, and three royal lackeys
in full livery walked by the side of each. In the other carriage, also
drawn by four mules and accompanied by lackeys, came the Duke
del Parque. The sword was borne into Murat’s presence by the two
armourers, and, after giving him the King’s letter, they solemnly
presented him with the historic weapon, which was received with
many expressions of thanks.
Murat now set no bounds to his ambitious aims, especially as he
knew that his brother-in-law had decided on the dethronement of the
Bourbons in Spain. So, dazzled by the brilliance of his position, he
precipitated matters by his intrigues. He suggested the advisability of
the Infante Don Carlos going to meet Napoleon as far as Burgos, so
this journey of the Spanish Prince was arranged, Pedro Macanáz
and Don Pascual Vallejo being in attendance.
As Napoleon did not trust entirely to the perceptions of Murat, he
sent the astute Savary to reconnoitre the state of affairs in Madrid.
The clever Frenchman was as successful in Spain as he had been in
Russia, and it was soon arranged for Ferdinand to take the
undignified course of going to meet Bonaparte at Burgos, for
Escoiquiz thought that it would gain the favour of the great
Frenchman.
Before starting, Ferdinand wrote to his father begging for a letter
in which he would assure Napoleon that he (Ferdinand) professed
the same sentiments of friendship with the French as his father. The
reply to this request came from the Queen, and she said that the
pains in the King’s hand prevented his writing himself, but she had
written to the Grand Duke of Berg saying that the desired letter had
not been sent because they knew that Ferdinand had no love for
France.
CHAPTER III
HOW NAPOLEON I. CHECKMATED THE SPANISH ROYAL FAMILY
1808–1814
As Napoleon was not quite satisfied with Murat’s reports, he

determined to go himself to Spain, and Ferdinand was advised by
Escoiquiz to go to Bayonne to meet the Emperor. After holding a
council on the subject at Vittoria in the bedroom of Escoiquiz, who
was ill, Ferdinand wrote a humble letter to the Emperor, promising to
go and meet him, in spite of Savary’s objections to the want of
dignity in the suggested proceeding. In his letter to Napoleon,
Ferdinand declared that he had been raised to the throne by the free
and spontaneous abdication of his father, and to this epistle the
Emperor replied:[7]
[7] Published in the Moniteur in 1808.
“In Bayonne,
“April 16, 1808.
“My Brother,
“I have received the letter of Your Royal Highness. You will
have seen by your father’s papers what an interest I have always
shown in him, so you will allow me now to speak to you with
frankness and loyalty.
“I had hoped to come to Madrid and persuade my august friend
to make certain necessary reforms in his dominions which would
give public satisfaction. The separation of His Majesty from the
Prince of the Peace seemed to me absolutely necessary for his
happiness and that of his vassals. Events in the North retarded my
journey, and the occurrences of Aranjuez have intervened.
“I do not constitute myself a judge of what happened, or of the
conduct of the Prince of the Peace; but I know very well that it is very
dangerous to Kings for the people to become accustomed to
shedding blood in their own attempts to obtain justice. God grant that
Your Highness may not find it so yourself! It would not be for the
interest of Spain to persecute a Prince who has married a Princess
of the Royal Family, and who has so long governed the kingdom. He
has no friends already, and Your Highness will have none, either, if
you come to be disgraced one day, for people like to avenge
themselves for the respect they have had to show us.
“Moreover, how could a Cause be framed against the Prince of
the Peace without framing it also against the King and Queen, your
parents? This Cause would foment hate and seditious passions, and
the result would be fatal to the crown. To this crown Your Royal
Highness has no rights beyond those transmitted by your mother. If
the Cause soils her honour, Your Highness destroys your own rights.
Do not listen to weak, perfidious counsels. Your Highness has no
right to judge the Prince of the Peace; the sins which are imputed to
him disappear in the rights of the throne.
“I have often expressed my wish for the Prince of the Peace to be
removed from affairs. If I have not been more insistent, it has been
because my friendship for King Charles overlooked the weakness of
his affection. Oh, miserable humanity! Weakness and error are our
lot. But all this can be made right if the Prince of the Peace is exiled
from Spain, and I offer him an asylum in France.
“As the abdication of Charles IV. took place at the moment when
my armies were occupying Spain, it will seem in the eyes of all
Europe and of posterity that I sent these troops with the sole object
of dethroning my ally and friend. As a Sovereign and a neighbour, I
must therefore hear all about the event before recognizing the
abdication.
“I tell Your Royal Highness that if the abdication of Charles was
spontaneous, and he was not forced to it by the insurrection and
consequent meeting in Aranjuez, I have no objection to admitting it,
and acknowledging Your Royal Highness as King of Spain. I
therefore desire to confer with Your Royal Highness on this matter.
“The circumspection I have observed for the past month in the
matter ought to convince Your Highness that you will always have
my support if factions of any kind disturb you on the throne.
“When King Charles told me of the recent events in October, I
flattered myself that I had contributed by my entreaties to the
peaceful conclusion of the Escorial matter.
“Your Highness is not free from faults; the letter you have written
me is sufficient to show that, and I have always wished to forget it.
Being a King, you know how sacred are the rights of the throne; any
step of an hereditary Prince towards a foreign Sovereign is criminal. I
consider the marriage of a French Princess with Your Royal
Highness would be conformable to the interests of my people, and,
above all, as a circumstance which will unite me by fresh bonds to a
house which I have had every wish to honour ever since I ascended
the throne.
“Your Royal Highness ought to beware of the consequences of
popular insurrections; you might be able to make an assault on my
scattered soldiers, but it would only lead to the ruin of Spain.
“I have seen with regret some letters from the Captain-General of
Catalonia which tried to rouse the people.
“Your Royal Highness knows all the depth of my heart; you will
observe that I am full of many ideas which require consideration; but
you can be sure that in any case I shall behave to you as I have to
the King your father.
“Your Royal Highness must be assured of my desire to conciliate
matters, and to find occasions of giving you proofs of my affection
and perfect esteem.
“May God have you in His holy and worthy keeping!
“Napoleon.”[8]
[8] From the “Memorias” of Nellerto and the “Manifestation” of

Don Pedro Ceballos.
The King, oblivious of the veiled insult of the Emperor, that he

had no right to the throne beyond that transmitted by his mother, still
cringed to the Frenchman, and wrote:
“Vittoria,
“April 18, 1808.
“Señor, my Brother,
“I have received with great satisfaction your letter of the
16th, sent by General Savary. The confidence with which Your
Majesty inspires me, and my desire to show you that my father’s
abdication was the consequence of his own impulse, have decided
me to go immediately to Bayonne. I hope therefore to leave to-
morrow for Irun, proceeding thence to the country-seat of Marrae,
where Your Majesty is.
“I am, my good Brother, with the highest esteem and sincerest
affection,
“Ferdinand.”
When Ferdinand arrived at Bayonne, the Emperor went to see

him at once, and Ferdinand went down to the door to meet him. The
interview was short, but the Spanish King was invited to dinner that
night. It was noticeable that, although Napoleon was very friendly, he
never addressed his guest as “Majesty”; and hardly was Ferdinand
back in his rooms, when a message was conveyed to him by Savary
to the effect that the Emperor had determined that the Bourbons
should not return to the throne of Spain, and that, as he had
determined to put a French Prince upon the throne, he required the
son of Charles IV. to renounce the diadem of both worlds in his own
name and in that of all his family.
Pedro Ceballos was loud in his indignation at such usurpation,
when Napoleon, who had heard his remarks from the next room,
entered the apartment, upbraided him for his treachery to Charles,
and declined to enter further into the matter until Ferdinand’s father
was there to speak for himself.
After Charles had sent Napoleon a protest against his abdication,
he concentrated all his efforts on gaining the liberty of the Prince of
the Peace. Indeed, the old man seemed more upset at the risks run
by his ex-Minister than he was at the treatment he had himself
received.
Pursuant to Murat’s advice, Charles and his wife repaired to the
Escorial, and there, in this imposing but gloomy abode, they brooded
over the turn in their affairs until despair filled their hearts.
Murat, faithful to the promise made to Charles in the presence of
the Queen of Etruria on the eve of his departure for the Escorial, did
his best for the dethroned Sovereigns, and persuaded the Union to
depute him to accompany them to Bayonne to take part in the
conference with Napoleon. The fact of Godoy being in Bayonne was
another reason for the royal couple to wish to go there, as they had
not seen him since his release from captivity. The following letter,
which the King wrote to Napoleon announcing his departure for
Bayonne, shows the esteem in which they held Bonaparte:
“Aranda,
“April 25, 1808.
“Sir and Brother,

“A prey to rheumatic pains in my hands and knees, I should
be completely miserable were not my troubles alleviated by the hope
of seeing you in a few days. I cannot hold a pen, so I beg of Your
Majesty to pardon my not writing with my own hand to express the
great pleasure I have in going to enjoy your generous kindness, for I
am obliged to use a secretary.
“The Queen also writes to Your Imperial Majesty, and we beg you
to accept our united sentiments of love and confidence.
“Your protection is balm to the wounds of my heart, and I feel that
the moment in which I shall find myself in your arms will be one of
the happiest of my life, and the first, after all that has happened, on
which I shall feel sure of my existence.
“May my wishes be fulfilled!
“My sir and Brother,
“I am, Your Imperial Majesty’s faithful Ally and Friend,
“Charles.”[9]
[9] Published in the Moniteur, 1810.

The Queen’s letter to Napoleon ran thus:
“Sir and Brother,

“I should have written before to Your Imperial Majesty if the
trying situation in which we undertook the journey had not presented
so many obstacles. We have now just arrived at Aranda of Duero.
The King is in a terrible state. He is troubled with rheumatic pains in
his hands and knees, but, in spite of all, we are longing for the happy
moment of throwing ourselves into the arms of Your Imperial
Majesty, whose great generosity is beyond all expressions of our
gratitude.
“We ought to have arrived at Bayonne before now, but,
unfortunately, circumstances do not correspond with our ardent
desires, because my son’s journey has left us without horses,
money, and all other necessaries. Heaven grant that the moment of
our interview will be as interesting to Your Imperial Majesty as it will
be to us, your faithful, worthy friends! We are quite sure of the
protection of Your Majesty, and nothing in the world can compare
with the complete and sweet confidence which leads us to place our
fate under the most powerful protection of Your Majesty, whose
immutable equity is so great, as the critic of the situation of his
faithful friend and ally, since the unhappy epoch of the unheard-of
events at Aranjuez.
“If Your Majesty’s troops had arrived then, they would have
protected our legitimate rights as their great captain deigns to do, but
Heaven sent us calamities which came like thunderbolts because we
had no help, nor had we anyone to support us.
“I do not know what day we shall arrive at Bayonne, because, if
the King’s indisposition permit it, we hope to take double journeys
every day. Your Imperial Majesty may be sure that we shall fly to
your arms, so great is our desire to strengthen the sweet ties of
alliance and friendship.
“May God have you in His safe keeping!
“Sir and Brother,
“I am, Your Imperial Majesty’s most affectionate Sister,
“Luisa.”
The affectionate tone of these royal letters shows that the royal
couple thought that Napoleon was about to restore to them the
sceptre which had been torn from their hands.
When the King and Queen arrived at Villareal, they asked what
reports were circulated about affairs, and the Duke of Mahon replied:
“It is said that the Emperor of the French is calling the Royal Family
of Spain together at Bayonne in order to deprive them of the throne.”
The Queen looked surprised, but she thought for a moment, and
then said:
“Napoleon has always been a great enemy of our family.
Nevertheless, he has made Charles repeated promises to protect
him, and I cannot believe he is now acting with such scandalous
perfidy.”
The royal arrival at Bayonne was announced by a salute of 101
guns, the garrison lined the streets, and Charles, on dismounting
from his carriage, showed his pleasure at the reception vouchsafed
to him by talking even to those he did not know.
A shadow came over the King’s genial countenance when he
saw Ferdinand standing with his brother at the foot of the staircase,
and it was only the younger Prince who was given a cordial “Good-
day” by the King, and who was embraced fondly by his mother.
Although Ferdinand saw that he was ignored, he made a step
forward to greet his parents. But Charles stopped, made a
movement of indignation, and began mounting the stairs with a
severe face. The Queen, however, who was behind, could not forget
that she was a mother, and folded her treacherous son to her
bosom.
Then the Princes repaired to their apartments, and their parents
hastened to greet the exile Godoy with tears of joy.
The Emperor of the French lost no time in paying his respects to
the royal travellers, but he did not ask them to dinner until the
following day.
As Charles’s rheumatism gave him some difficulty in mounting
the stairs of the imperial abode, he gladly accepted Napoleon’s arm,
saying: “I have not the strength that I had. It has been all knocked
out of me.”
“We will soon see about that,” returned the Emperor. “Lean on
me, and I will find strength for both.”
Thereupon the King stopped, and said emphatically: “So I
believe, and I base all my hopes upon you.”
On taking their seats at the table, Charles noticed the absence of
Godoy, and he exclaimed with tender concern: “And Manuel? Where
is Manuel?”
So Napoleon, anxious to please his ally, sent for the Prince of the
Peace, and the party was complete.
At the meeting at which it was hoped Napoleon would bring the
Royal Family to a satisfactory understanding there were very violent
scenes. It was natural that the sight of their renegade son should
revive all the bitterness of the King and Queen’s recent trials, but it
was a pity that they did not restrain the passions which made them
lose their royal dignity.
The Emperor announced that Ferdinand would restore on the
morrow to His Majesty the crown he had snatched from his father’s
brow. This Ferdinand stoutly declared he would not do, and Maria
Luisa, who had destroyed the proofs of her son’s guilt in the
conspiracy of the Escorial, was now so mad with rage that,
according to the report of Caballero, she cried to the Emperor to
punish the crimes of her son by committing him to prison.
Ferdinand was silent during the interview, but a few hours later
he wrote to his father, maintaining that the abdication had been a fait
accompli and declaring that he would only give up the crown at the
request of the Cortes and all the tribunals.
To this letter the King replied:
“My Son,
“The perfidious counsels of the people about you have
brought Spain into a very critical condition, and only the Emperor can
save it.... You have been too easily led away by the hatred which
your late wife had for France, and you have thoughtlessly shared her
unjust feelings against my Ministers, your mother, and myself.
“I was obliged, in support of my rights as a King and a father, to
have you arrested, for your papers contained proof of your crime.
But as I am approaching the end of my life, and I was miserable at
the idea of my son dying in a dungeon, I let myself be softened by
your mother’s tears. And yet my subjects have been upset by the
deceitful courses of the faction you formed, and from that time I have
had no peace in my life....
“You introduced disorder into my palace, you summoned the
Royal Guard against my own person. Your father has been your
prisoner; my Prime Minister, whom I created and received into my
family, was covered with blood, and taken from one prison to
another.... I am King by the right of my fathers. My abdication was
due to force and violence. I have nothing to accept from you, nor can
I consent to any meeting or to any new and base suggestion on the
part of the people about you.”
However, Ferdinand was obstinate, and there seemed no chance

of a peaceful settlement of the disgraceful family feud.
The above letter was dated May 2, 1808, and it was on that day
that the historic blow was struck in Madrid for Spain’s emancipation
from the French. It was the sight of the young Infante Francisco’s
tears at leaving the Palace of Madrid at the call of Napoleon which
acted like a match to gunpowder. The valiant Velarde, Daoiz, and
Ruiz were martyrs on this occasion, and the dramatic way in which
the Spaniards always keep this anniversary shows that those who
struck that blow are not forgotten in the land.
When Charles IV. heard the news of the riot, he at once thought
that it had been instigated by his sons.
“Manuel, send for Charles and Ferdinand,” he said, in a firm tone.
Napoleon remained in the room restless and gloomy; Charles
and Maria Luisa looked worried and anxious. They were all seated
when Ferdinand appeared and silently stood alone before them, for
his brother was ill in bed.
The King then asked his son if he had heard the news from the
capital. When Ferdinand replied in the negative, Charles returned
vehemently, “Very well, I will tell you,” and rapidly related what had
happened. “Judge, then,” he added, “if it be possible to persuade me
that you had no part in this? And did you hasten your miserable
associates to dethrone me in order to massacre my subjects? Who
advised you to this carnage? Do you only aspire to the glory of a
tyrant?”
The Duke of Rovigo, who gives us this scene in his “Mémoires,”
says that he and the other people who were listening in the adjoining
salon could not catch Ferdinand’s reply, but they heard the Queen
exclaim: “Didn’t I always presage your perdition? See into what
abysses you throw yourself and us! Ah, you would have killed us if
we had not left Spain! What! you have made up your mind not to
answer? You do not forget your old ways. You never know anything
when you do something bad.”
During this dialogue Charles IV. angrily moved about the cane
which he used when walking, and he so far forgot his dignity as to
raise it in a threatening way to his son, in his anger at his
impenetrable countenance. When Maria Luisa finished her diatribe,
she lifted her hand as if to strike the Prince, but she checked herself
in time.
The final touch to Ferdinand’s humiliating position was given
when the Emperor said in cold, clear, chilling tones:
“Prince, I had formed my resolution from the events which
brought you to France, and now the blood spilt in Madrid confirms
my decision. This carnage can only be the work of the band which
calls you chief, and I will never recognize as King of Spain one who
breaks the old alliance of two nations and orders the assassination
of the French soldiers, whilst asking me to sanction the impious act
of dethroning your father. Such is the result of bad counsels. You are
brought to the precipice. It is to your father alone that I am in any
way bound, and if he wish it I will restore him to his throne and
accompany him to his capital.”
But Charles IV. exclaimed vehemently: “But I don’t wish it. What
could I do in a country where they have worked up such passions
against me? And I, who have always rejoiced at seeing my country
peaceful in the midst of the upset of Europe—I should dishonour my
old age if I made war in the provinces and condemned my subjects
to prison. No, no; I don’t wish it. My son will undertake it with more
pleasure than I.” Then, looking at Ferdinand with majesty mingled
with pity, he said: “Do you think it costs nothing to reign? You have
followed these perfidious counsels. I neither aspire to command nor
can I do anything. Now you must avoid the precipice as best you
can.”
As Napoleon told Ferdinand that resistance about his resignation
was useless, and would only make his fate worse, it was agreed that
the crown should be handed over to France.
So the Treaty of Bayonne was formally signed on May 6 by the
Prince of the Peace for Charles IV., and by Marshal Duroc for
Napoleon, and this step, disastrous to the nation, can thus be
distinctly traced to the family feuds induced by the Queen’s unbridled
passion for the Prince of the Peace.
Charles had passed the twenty years of his reign in a self-
indulgent, simple life, and although he did nothing to show great
devotion to his kingdom, he certainly of his own accord would have
done nothing to disturb its peace. The Count of Toreno repeats the
account which Charles gave of his daily routine to the Emperor:
“Every day, winter and summer, I hunt till twelve o’clock, when I
dine. Directly afterwards I hunt again till evening. Manuel tells me
how things are going on, and I go to bed, to begin the same life next
day, unless there is some important ceremony.”
With a Sovereign so inert, Godoy did not demur on signing the
deed of renunciation of the throne, and as Escoiquiz sanctioned the
deed, it shows that he also felt that Ferdinand was but a broken
reed.
After the humiliating events of Bayonne, the poor Queen of
Etruria sought to return to Etruria, but was detained at Nice.
Miserable at having been obliged to leave her young son ill at
Compiègne, she tried to escape to England, but, the plot being
discovered, one of her two agents was shot, the other died in prison,
and she herself was condemned to confinement in a convent at
Rome; so she did not recover her liberty nor see her child again until
the fall of Napoleon. The Queen’s claims on Etruria were
subsequently nullified by the Congress of Vienna, and she had to be
contented with the nomination of her sons to the dukedom of Lucca.
Although after the Treaty of Bayonne the city of Madrid was in the
hands of Napoleon Bonaparte, the palace could not count the
Emperor as one of the residents in the palace, for during his stay in
the Spanish capital he was installed in the mansion of the Duque del
Infantado at Chamartin, and it was from this house that he made his
entry into Madrid. “Je la tiens en fin cette Espagne si désirée,” said
the French conqueror as he passed up the magnificent staircase of
the royal palace, and placed his hand upon one of the lions on the
balustrade; then, as his eyes travelled up the matchless marbles and
fine panels and pictures of the staircase, he turned to his brother
Joseph and said: “Mon frère, vous serez mieux logé que moi.”
When passing through the magnificent apartments, he stopped
before a portrait of Philip II., and after gazing at it for some minutes
in silence turned away. Who knows what recollections may have
passed through the conqueror’s mind, of stories of this Sovereign
read in boyhood, and how little he had then thought that the throne
of this King would ever be at his disposal!
Thus ended the rapid and only visit of Napoleon to the Spanish
capital, for he went back to Chamartin, and from thence set out for
Galicia.
King Joseph soon found he had a difficult part to play at the royal
palace as ruler of a foreign nation, but, although the Spaniards could
not be supposed to be fond of him, tribute was paid to the kindness
of his heart. After a meeting held at the palace to concert steps for
dealing with the fearful famine which was devastating Madrid, the
father of Mesoneros Romanos said to his son: “Joseph has certainly
not lost his head at his elevation, neither is he unduly set up by his
rank. He seemed profoundly moved at the misery of the people, and
proclaimed his intention to do all in his power to assist them.
Certainly,” concluded the speaker, “the man is good. It is only a pity
he is called Bonaparte!”
The preference entertained by Joseph Bonaparte for a beautiful
lady, the Countess Jaruco, widow of the Governor of Havana, is well
known. The lady died, and on the night of her burial her body was
exhumed (one can imagine by whose orders), and was interred
under a shady tree in her own garden. Joseph subsequently married
the Countess’s daughter by General Merlin. The hatred of the people
got on the poor Frenchman’s nerves, and for the last four years of
his enforced reign in Madrid he kept quite in retirement, spending a
good deal of time in the Casa del Campo, to which he passed by a
tunnel entrance.
But it was not very long, as we know, before the day came for
Joseph to leave Spain.
“The excitement in our house,” writes Mesoneros Romanos, “at
the news of the evacuation of the royal palace by the French was
extreme, and it was the same in every Spanish home. The hatred of
the foreigners who had taken possession of us was very deep-
rooted, and those who had joined the Gallic banner were not safe
from actual persecution.
“The shades of a dreadful nightmare were passed, and men
talked excitedly, and women and children laughed for joy. The
Virgins del Carmen and of the Paloma were promised new robes,
and the children ran to light up the altar, backed with a valuable
picture of the Immaculate Conception—a relic of the sacking of
Godoy’s house; and after a Paternoster and a Salve my father said:
‘Now we must go to bed, for we must be up early to-morrow to see
the entry of our friends.’
“By this was meant the Anglo-Spanish army, with its chief, Lord
Wellington, and the Generals Alava, España, and Conde de
Amarante. It was indeed a fine sight; the streets were decorated, and
after a repast served in the Town Hall the English Commander-in-
Chief appeared at the windows in response to the vociferous cheers
of the crowds, and his speech, which was as cordial as was
compatible with the stiff English manner, was received with the
enthusiasm of our Southern nature.”
Then Wellington repaired to the royal palace, which the
municipality had put at his disposal. The English General’s official
proclamation, placed at the corners of the streets, struck cold on the
hearts of the Spaniards, for it savoured more of a fierce Murat than
of the General of a liberating force. The following copy is taken from
the only remaining one in the archives of the city:
“The inhabitants of Madrid must remember that their primary duty

is to maintain order, and to render the Allied Armies every assistance
in their power to continue their operations.
“The Constitution established by the Cortes in the name of H.M.
Ferdinand VII. will be proclaimed to-morrow, after which will follow
the immediate formation of the Government of the City according to
the form it prescribes.
“In the meanwhile the existing Authorities will continue in the
exercise of their functions.
“Lord Wellington,
“Duke of Ciudad Rodrigo.”
It was soon seen that Wellington did not intend to rest upon his
laurels, for he scoured the rural park of the Retiro, where a French
detachment still lingered, and took 2,000 prisoners and 200 pieces of
artillery. This act completely confirmed the confidence of the
Spaniards in the English commander, and the heads of families
eagerly repaired to the churches to take the oath of the Constitution,
which, however, conveyed little to them beyond their emancipation
from the French and the approaching return of King Ferdinand VII.
It cannot be said that Wellington made himself very popular whilst
he was at the royal palace. He received the attentions showered
upon him in his cold and stiffly courteous way, and took little pains to
be cordial with the people of importance who called upon him.
Wellington’s fancy to have his portrait painted by Goya nearly led
to the future hero of Waterloo losing his life.
For, be it known, the illustrious Spanish painter was irascible to a
degree, the more so that he was completely deaf. So when the great
General made his appearance in the studio on the banks of the river
Manzanares, the painter’s son interpreted the Englishman’s wishes
in deaf and dumb language to his father.
The sittings took place, and the artist worked assiduously; and
when he thought the portrait was far enough advanced to be seen by
the General, he placed it before him. But, unfortunately, the picture
did not please the commander, who shrugged his shoulders
contemptuously, and said in English to his friend that he would not
accept such a caricature as a gift. General Alava declined to
translate this depreciatory remark; but the artist had noticed the
scornful gestures of the Englishman, and the son in alarm saw his
father turn his eyes to the loaded pistols which always lay ready to
hand on the table. The young man’s fear was increased when
Wellington rose from his seat in a discourteous way, and put on his
hat preparatory to departure. Then Goya, enraged at the officer’s
contemptuous manner, seized the pistols, and the General clapped
his hand to his sword.
The scene would have ended in a tragedy had not Lord Alava
assured the irate General that the artist was suffering from sudden
mental aberration, and young Goya restrained his father by force
from using the deadly weapons.
Wellington gave a great ball at the Town Hall the night before he
left Madrid, and with this return for the bull-fights, serenades, and
fêtes, which had been given in his honour, he took his departure
from the Spanish capital.
The English camp in the Retiro was raised a month later by
General Hill, and it is a matter of regret that the step was
accompanied by the blowing up of the royal manufactory of
porcelain, for the fabrication is now extinct. The magnificent walls
and ceilings of one of the salons of the royal palace, decorated with
cherubs, fruit, and flowers, in this beautiful ware, show that Spain
boasted an industry which rivalled that of Sèvres, Dresden, or
Worcestershire.
The reason given for this act of vandalism was that the French
might have used the building as a barrack; but it did not satisfy the
Spanish, who could not contain their indignation at the deed, which
was made worse by the English withdrawing to Portugal and leaving
the capital.
Ferdinand, with his usual duplicity, wrote to Berthémy from
Valençay, where he was practically a prisoner. In this letter he
pleaded in a cringing way for the protection of Napoleon, who had
robbed him of his crown.
“My greatest desire,” he writes, “is to be the adopted son of His

Majesty the Emperor, our Sovereign. I believe I am worthy of this
adoption, which would make the happiness of my life, by reason of
my love and affection to the sacred person of His Majesty, and by my
submission and entire obedience to his intentions and desires.
“Moreover, I am anxious to leave Valençay, for this place is in
every way disagreeable to us and in no way suits us.
“I am glad to trust in the magnanimity of conduct and the
generous beneficence which always distinguish Your Imperial
Majesty, and to hope that my ardent desire will be soon fulfilled.
“Receive, etc.,
“Ferdinand.”
When Napoleon decided to publish this correspondence with

Ferdinand, he wrote and asked him to send a letter to show that he
had his authorization for doing so.
So, before the appearance of the letters in Le Moniteur,
Ferdinand, in obedience to the imperial request, wrote to Napoleon:
“Valençay,
“May 3, 1810.
“Señor,
“The letters now published in Le Moniteur show the whole
world the sentiments of perfect love which I entertain for Your
Imperial Majesty, and the deep desire I cherish of becoming your
adopted son. The publicity which Your Imperial Majesty has deigned
to give my letters makes me hope that you do not disapprove of my
sentiments nor of the desire I have formed, and this hope fills me
with joy.
“Permit me, sire, to confide to you the thoughts of a heart which I
do not hesitate to say is worthy of your adoption. If Your Imperial
Majesty would unite me to a French Princess, you would fulfil my
most ardent wish. By this union, apart from my personal happiness,
all Europe would be convinced of my unalterable respect for the will
of Your Majesty, and it would see that you deign to make some
return for such sincere feelings.
“I will venture to add that this union and the sight of my happiness
will exercise a beneficial effect on the destiny of all Spain, and will
rob a blind and furious people of the pretext of covering a country
with blood in the name of a Prince, the eldest son of an ancient
dynasty, who has, by a solemn treaty by his own choice and by the
most glorious of all adoptions, made himself a French Prince and a
son of Your Imperial Majesty.
“I venture to hope that such ardent wishes, and an affection so
absolute, will touch the magnanimous heart of Your Majesty, and that
you will deign to make me share the fate of the many Your Majesty
has made happy.
“Señor, I am, etc.,
“(Signed) Ferdinand.”
Charles Leopold, Baron de Colly, an astute and intriguing youth,

proposed to the Duke of Kent a plan for releasing Ferdinand from his
ignoble position at Valençay by taking him on board an English man-
of-war to a port of Spain.
The Duke of Kent referred the matter to his father, who sent
Ferdinand two letters by the Baron. Provided with a set of passports
and all papers necessary for the undertaking, besides supplies, in
the form of diamonds and an open draft on the house of Maensoff
and Clanoy, and a ship loaded with provisions for five months, Colly
commenced operations. He reached Paris in safety, sold part of the
diamonds, and began his preparations; but the police got wind of the
plot through Colly’s secretary Albert, and he was promptly shut up in
the Castle of Vincennes.
Fouché tried to persuade Colly to continue his work, so that
Ferdinand might be caught in the act of escaping; but the
Englishman preferred his prison to such treachery, and in this prison
he remained until the fall of Napoleon.
In the meanwhile Fouché sent to Ferdinand a man called
Richard, personating Colly. But the Prince was not caught in the trap,
for, in his rooted desire to conciliate the Emperor of the French, he
sent at once for Berthémy, the Governor, and said to him:
“The English have done great harm to the Spanish nation by
using my name, and they are now the cause of the blood which is
being spilt. The English Ministry, in their mistaken idea that I am kept
here by force, have sent an emissary to me who, under the pretext of
selling me curios, has given me a letter from His Majesty the King of
England.”
The letter from George III. to Ferdinand, which was subsequently
published in Le Moniteur, ran thus:
“Sir, my Brother,
“I have for a long time wished for an opportunity to send
Your Majesty a letter signed by my hand, to express the deep
interest and the profound feeling which I have entertained for you
since you were taken from your kingdom and your faithful subjects.
Whatever the violence and cruelty with which the usurper of the
throne of Spain oppresses that nation, it ought to be of great
consolation to Your Majesty to know that your people retains its
loyalty and love for its legitimate Sovereign, and Spain makes
continual efforts to maintain the rights of Your Majesty and to re-
establish those of the monarchy. The resources of my kingdom, my
squadrons, and my armies, will be employed in aiding the vassals of
Your Majesty in this great cause, and my ally the Prince Regent of
Portugal has also contributed with all the zeal and perseverance of
his faithful friend.
“The only thing which is wanting to your faithful subjects and your
allies is the presence of Your Majesty in Spain, where it would give
fresh energy. Therefore I ask Your Majesty, with all the frankness of
alliance and friendship which bind me to your interests, to think of
the most prudent and efficacious way of escaping from the
indignities which you suffer, and to present yourself in the midst of a
people unanimous in its desire for the glory and happiness of Your
Majesty.
“I beg Your Majesty to be sure of my sincere friendship, and of
the true affection with which I am—in the palace of the Queen,
Monday, January 31, 1810—sir, my Brother,
“Your worthy Brother,
“George R.”
“By command of the King,
“Wellesley.”[10]
[10] “Monitor de Paris, traducido por Don Juan Maria Blanco en

el ‘Español’ publicado en Londres,” tomo i., p. 136.
But Ferdinand’s cross-grained nature was unable to follow any

straightforward advice or adopt any clear course. However, we all
know how the people’s desire to have a Spaniard on the throne,
aided by the troops of England, was finally successful, and
Ferdinand the Desired entered his capital on May 13, amid cries of
delight from his people, who were wild with joy.
CHAPTER IV
KING FERDINAND VII. AND HIS HOME LIFE
1814–1829
So Spaniards once more had a King of their own blood. The pity of
the matter was that the man himself was so unworthy of the people’s
trust. Brought up in a Court honeycombed with intrigue, truth and
sincerity seemed unknown to Ferdinand, and although he constantly
said, “I hate and abhor despotism,” there never was a Sovereign
more despotic than this son of Charles IV.
Being untrustworthy himself, he thought everybody was
unreliable, and so he set spies on his entourage, and stooped to
listen to stories from his servants.
Thus, no Minister or officer was safe from being sent off to prison,
and with the duplicity which had been perfected by constant practice
in his youth sentence of condemnation would be given by Ferdinand
with an air of friendliness, with a wave of his cigar or the offer of his
caramels, followed by thrumming on the table, or the pulling of his
ear, or the slapping of his forehead, with which his courtiers were
familiar as signs of bad temper.
The Duke of Alagon was the King’s most constant attendant in
any gallant adventure, and, indeed, his departures in that respect
were those of a man who seemed to atone for his want of personal
attractions by a surplus of gallantry to the fair sex. It was whilst
pursuing one of these intrigues with a charming widow at the royal

ebook download Resolving Spectral Mixtures With Applications from Ultrafast Time-Resolved Spectroscopy to Super-Resolution Imaging 1st Edition Cyril Ruckebusch (Eds.) - eBook PDF all chapter

Uploaded by

Copyright:

Available Formats

You might also like

ebook download Resolving Spectral Mixtures With Applications from Ultrafast Time-Resolved Spectroscopy to Super-Resolution Imaging 1st Edition Cyril Ruckebusch (Eds.) - eBook PDF all chapter

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ebook download Resolving Spectral Mixtures With Applications from Ultrafast Time-Resolved Spectroscopy to Super-Resolution Imaging 1st Edition Cyril Ruckebusch (Eds.) - eBook PDF all chapter

Uploaded by

Copyright:

Available Formats

Resolving Spectral Mixtures With

Applications from Ultrafast

(eBook PDF) Constructive Conflicts: From Escalation to

Spectroscopy and Dynamics of Single Molecules: Methods

(eBook PDF) Introduction to Spectroscopy 5th Edition

Tissue Elasticity Imaging: Volume 2: Clinical

Molecular and Laser Spectroscopy: Advances and

Progress in Heterocyclic Chemistry Volume 29 1st

Innovation in Nano-polysaccharides for Eco-

Theory and Modeling of Cylindrical Nanostructures for

For information on all Elsevier publications

Publisher: John Fedor

P. de B. Harrington (311), Clippinger Laboratories, Center for Intelligent Chemical

R. Tauler (5, 101), Institute of Environmental Assessment and Water Research

The Equation (1) can most conveniently be written as a matrix equation

Each element of D represents an equation of type (1), and the elements of

Can additional constraints be applied?

one of a range of possible solution. Implementing further constraints may

2 THE SPECTRAL MIXTURE PROBLEM

Data Handling in Science and Technology, Vol. 30. http://dx.doi.org/10.1016/B978-0-444-63638-6.00001-2

Spectral mixture data are usually arranged in a matrix with columns as

Taking a broader perspective, spectral unmixing enters the more general

3 BOOK CONTENT AND ORGANIZATION

wavelet transform with multivariate image analysis in a multiresolution anal-

1 MCR: THE CONCEPT AND THE LINK WITH

Data Handling in Science and Technology, Vol. 30. http://dx.doi.org/10.1016/B978-0-444-63638-6.00002-4

2 MCR-ALS: ALGORITHM AND DATA SET CONFIGURATION

concentration profiles and pure spectra in an alternating least squares way

Looking at the different expressions in Eqs. (5)–(7), it can be understood

Averaged profiles or selected profiles from the different regular multisets

2.1 MCR-ALS Algorithm: Steps

There are many approaches to implement constraints: smooth least squares

Selection of profile n Constrained profile n

FIG. 4 Implementation of trilinearity constraint.

compound-wise application of the model constraint. Therefore, hybrid mod-

to new kinds of problems and measurements. Clear examples of new con-

3 MCR-ALS APPLIED TO PROCESS ANALYSIS

3.1 Encoding Process Information: Sequentiality

compounds involved in a process [3,56–59]. In 1995, Manne proposed two

3.1.2 Physicochemical Models

PCA PCA PCA PCA PCA PCA PCA

Emergence of compounds Noise level Decay of compounds

min kD CC + Dk ¼ min kDðI CC + Þk, where C ¼ f ðki Þ (14)

To this letter Carlos IV. added a postscript:

“Sir and very dear Brother,

“Sir, my dear Brother,

“The King, my husband, and I do not wish to be importunate nor

Ferdinand, in his blind belief in Escoiquiz, disregarded the

As Napoleon was not quite satisfied with Murat’s reports, he

[8] From the “Memorias” of Nellerto and the “Manifestation” of

The King, oblivious of the veiled insult of the Emperor, that he

When Ferdinand arrived at Bayonne, the Emperor went to see

“Sir and Brother,

[9] Published in the Moniteur, 1810.

“Sir and Brother,

However, Ferdinand was obstinate, and there seemed no chance

“The inhabitants of Madrid must remember that their primary duty

“My greatest desire,” he writes, “is to be the adopted son of His

When Napoleon decided to publish this correspondence with