Applying chemometrics to study battery materials: Towards the comprehensive analysis of complex operando datasets
Introduction
In situ or operando analyses, which provide unique insights on both structural and electronic properties of battery materials under electrochemical cycling conditions, are becoming increasingly popular in battery research [1], [2], [3], [4], [5], [6], [7]. In fact, while on the one hand the number of available techniques applied on a working battery increases steadily, on the other hand, time, energy and spatial resolution are continuously improving thanks to the technological development. As a consequence, these continuous improvements result in ever growing data sets, making the thorough data analysis more complicated and more time consuming.
As an example, a typical operando X-ray absorption spectroscopy (XAS) study of a working battery material at one or several transition metal edges (which is usually necessary if the material contains different possible active sites) easily produces hundred of spectra for a single discharge/charge process [8]. Such datasets were traditionally treated individually, in a slow spectra-by-spectra approach. Besides being laborious, this approach makes it very challenging to extract reliable information regarding general trends and intermediate process occurring during electrochemical cycling.
In order to handle such complex datasets without missing important information and in a reasonable time, a fruitful approach can rely on chemometrics. This interdisciplinary area between analytical chemistry and statistics, which owes its origin to the rise of use of computer and automated data acquisition in the 1970s, has led to the development of a wide range of multivariate tools, among which, a wealth of methods that aim at the decomposition of a data matrix into a linear model of dyads (the bilinear model) [9]. Principal Component Analysis (PCA), Independent Component Analysis (ICA) and Multivariate Curve Resolution (MCR) are the most common methods sharing the same global objective: the resolution of complex mixtures into pure-components contributions based on no or little available information. The first of these methods, PCA, provides a totally blind abstract decomposition of a matrix of experimental data by computing orthogonal components aligned to the directions that explain the maximum of the variance of the data matrix [10]. Similarly, also ICA can be proposed to solve the blind-source separation problem by finding a mathematical transformation of the data into a linear combination of components that are statistically independent, or as independent as possible [11]. Both PCA and ICA, however, produce abstract matrices without direct chemical or physical meaning. MCR, conversely, aims at decomposing the data matrix in a product of two matrices containing objects characterised by a physical or a chemical meaning, i.e., real spectra and concentrations, and does this decomposition by imposing specific constraints [12], [13]. Even though this strategy makes the solution of MCR dependent upon the chosen constraints, and thus non-unique, MCR has the important advantage of producing “pure” spectral components and concentration profiles (if Beer's law is valid for the investigated spectroscopic system) which can be analysed as real spectra/patterns and compared with real data. This advantage designates MCR as the method of choice for the analysis of large sets of data, and since its first application in 1971, many different algorithms have been developed in different research fields, all sharing the common assumption that the data should be bilinear or multi-linear [12]. Nowadays, the most popular and flexible MCR algorithm is undoubtedly Multivariate Curve Resolution by Alternating Least Squares (MCR-ALS), first proposed by Tauler in 1995, [14] which can be easily implemented as a freely available Matlab® Graphical User Interface [15].
In the last few years, MCR-ALS has been applied in many fields ranging from catalysis [16], [17], [18], [19], polymer chemistry [20], environmental science [21] and pharmaceutics or biomedical analysis [22] to battery materials involving many different characterization techniques such as electrochemistry, infrared spectroscopy, chromatography, nuclear magnetic resonance, X-ray fluorescence, etc. (see refs. [23], [12], [24] and references therein.) In the field of batteries and battery materials, after the first applications by Rodriguez et al. in 2007, [25] Muto et al. in 2009, [26] and Conti et al. in 2010, [27] a large number of studies based on the analysis by MCR-ALS of datasets deriving from different analytical methods such as X-ray absorption spectroscopy, X-ray microscopy, Mössbauer spectroscopy, etc. have emerged recently [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45].
In this article, we aim at providing a short review about this chemometric tool and its rapidly growing application in the field of batteries. After a short summary of theoretical grounds MCR-ALS, a few specific cases studies are resumed highlighting the advantages of the use of such methods for the analysis of ex situ and operando data providing relevant information about the electrochemical mechanism in battery materials. In this regard, the considerable shortening of analysis time, the reduction of ambiguity in data interpretation, as well as the unique asset to unveil transient species stand out as main advantages of the MCR-ALS approach.
Section snippets
MCR-ALS: theoretical grounds
A detailed description of MCR-ALS from a theoretical point of view is given by Tauler, [14], [13] who also proposed this method for the analysis of in situ spectroscopic data, while the intrinsic limitations of this method and of its application are further discussed by Ruckebusch et al. [12].
Shortly, the underlying principle of MCR-ALS is to decompose a complex two-way data matrix into two simpler matrices and containing concentration profiles and pure spectra of z
First applications of MCR-ALS in the field of batteries
The introduction in the field of battery materials has only been comparatively recent. To the best of our knowledge, the first application of MCR to a study of battery materials dealing with operando experiments concerns the investigation by XAS of the evolution during charge of a positive electrode based on a xerogel [27].
This study, after an introduction of the analytical methodology and on the spectra data treatment, allowed obtaining relevant information on the cell charging
Identification of elusive intermediates and transient phases: the reversible sodiation of SnSb
Elucidating chemical species and reaction paths in complex mixtures during the electrochemical lithiation or sodiation process are the basis for understanding the electrochemical mechanism of electrode materials. The identification of intermediates formed through such complex multi-phase and multi-step reactions is however not always straightforward, and one of the outstanding assets of the chemometric approach is its ability to unveil them. A particular intriguing example of this is the recent
Conclusion
It is now several years that MCR-ALS is increasingly applied for the analysis of operando datasets collected using various spectroscopic techniques. By carefully choosing the examples depicted here, we tried to expose some of the advantages of its application. Indeed, in the example of the sodiation of SnSb, it was shown as MCR-ALS allows spotting elusive intermediates, while with the example of the ionic redox in LRSO it was demonstrated how this data analysis method can be used to separate
Acknowledgements
Alistore-European Research Institute is gratefully acknowledged for financial support through the postdoc grant to M. Fehse. Valérie Briois (ROCK beamline, Synchrotron Soleil, Gif-sur-Yvette, France) and the research group “Catalyse, Réactivité de Surface et Rayonnement Synchrotron” (GDR CNRS 3590) are gratefully acknowledged for their information and supporting activity for the application of chemometrics to the analysis of spectroscopic data.
References (81)
- et al.
J. Power Sources
(2015) - et al.
Curr. Opin. Chem. Eng.
(2016) - et al.
Talanta
(2014) - et al.
Neural Netw.
(2000) - et al.
Anal. Chim. Acta
(2013) Chemom. Intell. Lab.
(1995)- et al.
Chemom. Intell. Lab.
(2015) - et al.
Anal. Chim. Acta
(2014) - et al.
J. Pharm. Biomed. Anal.
(2008) - et al.
TrAC Trends Anal. Chem.
(2017)
Nano Energy
J. Electroanal. Chem.
Anal. Chim. Acta
J. Electroanal. Chem.
Microelectron. Reliab.
J. Power Sources
Solid State Ion.
J. Solid State Chem.
J. Power Sources
J. Power Sources
Adv. Energy Mater.
Nat. Energy
Chem. Soc. Rev.
Chem. Mater.
Small Methods
Handbook of Chemometrics and Qualimetrics: part A
Anal. Methods
RSC Adv.
J. Phys. Chem. C
ACS Catal.
Anal. Bioanal. Chem.
J. Appl. Cryst.
Mater. Trans.
Anal. Chem.
Acta Cryst.
J. Phys. Conf. Ser.
J. Phys. Chem. C
Cited by (48)
Research progress in anode materials based on multiple potassium storage mechanisms
2022, Sustainable Materials and TechnologiesCitation Excerpt :Notably, ranges of in situ electrochemical analytical techniques (Fig. 3) have also been explored to characterize the energy storage processes in the anode of PIBs. In situ XRD [59,60] has been developed to detect phase alteration during reactions; In situ Raman spectroscopy [61,62] to obtain structural and chemical changes; In situ scanning/transmission electron microscopy (SEM/TEM) [63] to detect morphological and structural evolution; In situ atomic force microscopy (AFM) [64] to dynamically observe interfacial behavior and mechanical properties; And in situ X-ray absorption spectroscopy (XAS) [65] to explore geometrical and electronic structural transitions. However, the current application of in situ AFM and in situ XAS in PIBs is insufficient.
An innovative spectroscopic approach for qualitative and quantitative evaluation of Mb-CO from myoglobin carbonylation reaction through chemometrics methods
2022, Spectrochimica Acta - Part A: Molecular and Biomolecular SpectroscopyCitation Excerpt :MCR-ALS is a method that needs proper initialization (concentration or spectral profile) and definition of constraints, so this parameter definition is an important phase for its operation [68,69]. The strategy of imposing constraints makes the MCR-ALS solution dependent on the defined parameters and, therefore non-unique [70]. Therefore, MCR-ALS solutions must be critically evaluated [53].
Monitoring and quantifying morphological and structural changes in electrode materials under operando conditions
2020, Journal of Power SourcesCitation Excerpt :It probes element-specific the electronic and next-neighbour structure, thus providing insights on the evolution of oxidation state and local structural transformations. XAS is especially useful for battery materials that undergo strong nanostructuring upon electrochemical reaction, which impedes the use of x-ray diffraction (XRD) for their analysis [2]. Furthermore, it has been shown that operando XAS can be applied to explore the multistep reaction mechanism and reveal reaction intermediates [3,4].