I've performed some EIS experiments of a gold alloy deposited on a non-noble metal in a solution containing Cl- ions. The electrode should be stable but I also hypothesized a dissolution or corrosion behavior.
I've tried to fit the resulting data (reported in figure: Nyquist plot, zoom of Nyquist at high frequencies and Bode plot) but I can't fin any good equivalent circuit.
I found problems at both high freq. (Z and phase tend to grow up, and in the Nyquist plot there is a weird trend) and low freq. (the phase seems to follow an inclined line).
I tried simple and double Vogit element using CPE, Randles circuit with Warburg element and even something with an inductance...
I've obtained the lowest X^2 with the simplest circuit but sometimes the resistance in parallel is fitted with extremely high (Tohm) or negative values.
Could anyone suggest me a good model to explain this setup and data?
Hello,
It is not very easy to fit an equivalent circuit. What I can see from your graph is a problem in high frequencies. It is probably related to an artefect with the reference electrode. What is system : WE, CE and REF? and what is the electrolyte? Measure at OCP or under polarization?
First, try to remove the points at high frequencies and test a Randles circuit with Warburg element with a good choice of the elements : Rs, Rp, Cdl...
Best Regards
Thomas
Thank you Thomas for your answer. Yes it was a 3 electrodes system, NaCl water solution measured at OCP. But actually the OCP gradually changes over very long period.
Unfortunately even removing the high frequencies the fitting is not good at low frequencies. I'm wondering if could I admit a negative resistance if I suppose the presence of a galvanic couple between the coating and the substrate and a slow dissolution of a metal.
Hello, I think that data after 20-30 kHz are not reliable. I mean there might be a strong effect of wires and connections. Try to cut this part and fit it again. So I think the same as Thomas Perrier . It is not easy to fit this generally due to many parameters ... You need to find some model which is in theoretically in agreement and then try to estimate some of parameters and lock them and try fit ... Using this way you may find some convergence and in the end unlock all parameters. Galvanic couple is also possible but it depends how powerful it may be ... perhaps not much.
Ok, but Gold should be stable in NaCl solution, but of course it depends of the thickness and porosity of the layer.
You tell that the OCP is not stable. Maybe the problem is here. For EIS measurement your system must be stable, in steady state. If you are not in this state, the EIS data could be invalidate. May try a Kronig-Kramers test, to validate your EIS data.
What you can try is to use multisines in EIS to reduce the time of the EIS measurement, let the OCP stabilise during few minutes of hours if necessary. Do OCP measurement, EIS, OCP measurement to see if the second OCP is egal to the first. It can give an idea on how stable your system is. You can also repeat the EIS measurmeent over immersion time to see if there is an evolution or not.
EIS is not very simple. Each system is different and the this method is very sensitive. Good luck...
Thank you both.
I did the Kronig-Kramers test, I am not familiar with it, but it gave me an error of less than + -3.3% on the whole frequency range, so I suppose the data is correct.
In any case, repeating the fitting on the same simple model (R (CR)) several times with different starting values and running fit for only one parameter at a time, I got something decent.
Dear Dr. Walter Giurlani,
let me just mention in this context that the conditions which an equivalent electric circuit should comply with are the coherence with the experimental data and the conformity with a physical meaning, since very often to only one impedance spectrum various electric models can be associated.
Good work and best regards, Pierluigi Traverso
First, this spectra is not completed yet, you should extend your measurements down to 0.01 Hz, not only to 10 Hz as shown in the Bode diagram. Second, you have to remove the first 6 points at the high frequency range. After these two important points you can try to fit the results using Randles circuit with Warburg element and I am sure that you will get good fitting with the proposed model. Finally, if this simple model does not work you can try the two parallel time constants model in series with the solution resistance. Good luck..
Just a two final questions about the data elaboration:
Generally do you prefer to keep low the estimated error on the single values or the total X^2? Sometimes when one goes down the other goes up.
In the NOVA software it is possible to activate or disable a weight factor (values with greater |Z| worth more in the fitting), do you think it should be on, off, or I should decide based n the estimated error and X^2?
Thank you
Hello,
Best Regards
Thomas