An identication approach as a prerequisite for quantitative electrochemical studies

  • Tom Breugelmans ((PhD) Student)
  • Els Tourwé (Promotor)
  • Steven Van Damme (Jury)

Student thesis: Doctoral Thesis


Electrochemical processes are widely studied because of their importance in a multitude
of industrial activities like electrowinning and electrorening of metals, electrosynthesis,
plating, electrochemical forming and machining, storage and conversion of energy, corrosion,.
. . . In order to remain competitive, product innovation and a continuous search for
new products are primordial. In the present economical climate, it is no longer possible
to use expensive trial and error methods. Instead, a profound insight in the electrochemical
phenomena is required. The latter kind of knowledge can solely be obtained if the
electrochemical process can be modelled in an accurate and reliable way. However, reliable
modelling results are only provided when correct experimental data are available and
when it can be evaluated whether the model is able to describe the experiments within the
experimental error. To tackle this problem, this work integrates those two elements in an
integrated identication methodology.
The main goal of this work is the development of an identication approach for electrochemical
processes. On the one hand, this requires the collection of reliable experimental
results for the investigated electrochemical processes by means of powerful electrochemical
techniques and combinations thereof. On the other hand, it demands the implementation
of data analysis to determine the quality of the results provided. Only by combining those
two aspects, the subsequent modelling step can be evaluated and reliable modelling results
can be obtained. Based on this strategy, the developed identication approach contains
four building blocks: (i) collecting experimental data, (ii) data analysis, (iii) data modelling
and (iv) a statistical evaluation of the results. The combination of the four steps
can be considered as an accurate, overall identication methodology. In this work, the
presented identication approach is implemented for the following three electrochemical
systems, each dominated by characteristic physical processes: the mass and charge transfer
phenomena for the ferri-/ferrocyanide redox couple, the corrosion of SiOx-layers and
BTSE-layers correspondingly on steel and aluminium surfaces and the adsorption kinetics
of 2-amino-5-mercapto-1,3,4-thiadiazole (AMTD) on a copper surface.
Collecting experimental data is a rst step. Moreover, the overall success of the identi-
cation approach depends on its ability to gather correct experimental data in step (i). A
key issue in this step, is the use of the most appropriate measuring technique. This importance
is illustrated by using the identication approach with both linear sweep voltammetry
(LSV) and electrochemical impedance spectroscopy with an odd random phase multisine
excitation signal (ORP-EIS) to study the ferri-/ferrocyanide redox system. LSV successfully
provides the kinetics parameters of the investigated system whereas ORP-EIS fails.
Nevertheless, the same ORP-EIS techniques shows powerful benets in studying the corrosion
and adsorption processes that are selected in this work. For the adsorbed AMTD-layer
on copper, the value of its thickness in the potential region of steady state adsorption is
calculated together with its error estimate by means of potentiodynamic ORP-EIS.
The data analysis step (step (ii)) involves two aspects. It it not only estimating the
measurement error but it also checking whether the basic requirements for reliable measurements
are fullled. The former aspect is dealt with in the study of the ferri-/ferrocyanide
redox system. Performing multiple experiments for the LSV measurements provides the
uncertainty on the mean voltammograms which is dened by two times the standard deviation
on the experiments. The latter aspect is especially worked out in the analysis of
the ORP-EIS data. A lot of eort is put in checking the necessary conditions of causality,
linearity and stationarity. That it is important to verify whether those requirements are
fullled, is shown for the studied corrosion processes. The data analysis of the ORP-EIS
measurements clearly indicates that the onset of both corrosion reactions is accompanied
by a non-stationary behaviour. This observation oers the possibility to obtain information
about the corrosion protection of the coated system without a time-consuming modelling
step for the experimental results.
If the data analysis indicates that the necessary conditions to obtain reliable experimental
data are not fullled, one can (1) redo the experiments in other experimental conditions,
(2) correct for them or (3) take their in
uences into account when modelling. Correcting
the experimental data for the non-stationary behaviour of the corrosion reaction is done
in this work by setting up a procedure to quantify the instantaneous impedance value for
the studied corrosion systems. By modelling skirt-like contributions that are induced by
the non-stationary behaviour in the corrosion systems, the time evolution of the collected
ORP-EIS data is estimated. Not using the instantaneous impedance value of the system,
can lead to wrong modelling results in the subsequent modelling step. But more important,
the provided time evolution is a crucial piece of information if one aims at unraveling the
mechanism of the studied corrosion reaction. By observing how the parameter values of
the model change over the dierent instantaneous impedance values, additional information
about the underlying corrosion mechanism is provided.
The last two steps of the identication approach cannot be considered separately. Only
when a proposed model succeeds its statistical evaluation, the model is accepted and an
output is provided by the identication approach. If not, the modelling step has to be
In step (iii), the dierences between the experiment and a selected model are always minimized
by means of the combination of the Gauss-Newton and Levenberg-Marquardt minimization
algorithms. An important feature of the tting procedure is its ability to use the
data analysis provided by step (ii) as a weighting factor. For ORP-EIS, not only the noise
level of the measurements, but also the contributions of non-stationary and non-linear behaviour
can be selected as weighting factors.
In step (iv), the proposed model is evaluated. For the LSV results, this is done by comparing
the dierence between experimental and modelled data with their calculated 95%
condence interval. For the ORP-EIS data, the statistical evaluation is done by comparing
the residual values between experiment and model with the noise levels. If they coincide,
the model is accepted. If it turns out that the model cannot completely describe the
experimental results, there is always a trade-o between the complexity of the selected
model and the minimization of the residual value. The tting procedure used in this work
oers a statistically founded judgement of the model. Deciding whether the model can be
accepted or whether the residual curve has to be further minimized, is in the hands of the
practitioner and depends on the nature of the studied system.
To end, it can be concluded that the dierent case studies have proven the importance
of each step and moreover, they have illustrated what can go wrong if he or she does not use
the identication approach. Only by coupling the two rst steps that deal with the collection
of reliable experimental data with the last two modelling steps, the true underlying
reaction mechanism can be detected. In addition, the approach provides new information
that can contribute to the insight into the dierent electrochemical phenomena.
Obviously, this identication approach is not limited to the case studies considered in this
work but can be used for other electrochemical processes. It is a prerequisite for every
quantitative electrochemical study. Moreover, it will provide accurate modelling results
and will contribute to the understanding of technologically relevant electrochemical
reactions that are indispensable in optimizing existing industrial processes.
Date of Award23 Sep 2010
Original languageEnglish
SupervisorSteven Van Damme (Jury) & Els Tourwé (Promotor)


  • modelling

Cite this