Hands-on Electrochemical Impedance Spectroscopy

June 28, 2022 by Dino Klotz

This blog post accompanies the webinar "Hands-on Electrochemical Impedance Spectroscopy", which introduced basics of electrochemical impedance spectroscopy (EIS), plus practical tips and tricks in order to measure EIS consistently.

After the webinar, there was a vivid Q&A session where most questions of the attendees could not be answered due to time limitations. This blog post provides you with the questions and answers that compliment some of the topics that were introduced in the webinar. Before moving on to the Q&A part, we first want to clarify one aspect about the difference of capacity and capacitance.

Capacity/Capacitance

During the live session, the terms capacity and capacitance were not always consistently used according to their physical definitions. That is why we want to state clearly:

  • The capacitance C is the specification for a capacitor, which is given in Farad. Capacitance is the ratio of the amount of electric charge stored on a capacitor (more generally, a conductor) to a difference in electric potential. 
  • Capacity refers to a charge capacity that specifies the electrical charge that can be stored on a battery, for example. Capacity thereby only specifies the number of charges and not the related energy.

Watch the Webinar Recording

Hands-on Electrochemical Impedance Spectroscopy | Zurich Instruments Webinar

The full set of slides is available here.

Questions and Answers

Here is a summary of the main questions asked during the webinar. Some of them have already been answered live.

1. How to calculate capacity of a battery from EIS Nyquist plot?

This question was partially answered in the webinar, but here we want to give a more precise answer. The capacitance that is seen in the Nyquist plot represents the “differential capacity”. It means the portion of the charge that is charged or discharged at the operating point of the measurement (linearized). For determining the complete capacity of the battery, one would need to measure at many operating points (states of charge), and integrated over the local differential capacity. Some more information about capacities and capacitances in batteries can be found here:

  • M. D. Levi and D. Aurbach, J. Phys. Chem. B 101, 4630-4640, 1997 (DOI: 10.1021/jp9701909)
  • J. P. Schmidt, P. Berg, M. Schönleber, A. Weber, E. Ivers-Tiffée, Journal of Power Sources 221, 2013, 70-77 (DOI: 10.1016/j.jpowsour.2012.07.100)
  • M.Schönleber, E.Ivers-Tiffée, Electrochemistry Communications 61, 45-48, 2015 (DOI: 10.1016/j.elecom.2015.09.024)

2. Could you give some more details on the user compensation methods possible to increase the impedance range?

The MFIA Impedance Analyzer and the LabOne® software offer user compensation (short and load and combinations thereof). For a "short" compensation, you can use a "short" on a holder that comes with the instrument. Then, the instrument saves the values measured in that setup ("short" plus potential parasitics from cables or setup) as zero Ohms. Any deviation from this zero will then be the "user compensated" value for the impedance. Similarly, you can do "load" compensation. In that case, you need to use a high-precision resistor and enter its value into the software before running the compensation procedure. It is then calibrated for this value and will measure any resistor close to this value with very good precision. It is difficult to specify the precision, but if you measure the same resistor again, the one with which you calibrated, the error will be almost zero, only noise can be added to the impedance. That means you can bring down the measurement error significantly for area in the accuracy chart where it does not show the 0.05% base accuracy. It thereby expands the range where you can measure with high precision.

It is difficult to extent the range significantly beyond the specified range between 1 mΩ and 1 TΩ though because below and above those, the measurement values for voltage and current input become too small, respectively.

3. Can you comment about the points per decade for the Nyquist plot? What would be a good rule of thumb?

That is a very good question and this was not mentioned in the webinar. Actually, the more points per decade, the better for fitting and evaluation. But also the measurement time increases. Another consideration is the complexity of the impedance, with which I mean, how many processes are contributing to the impedance response. If there is only one or two processes, 5 points per decade should be enough. If you are measuring a difficult impedance, such as the impedance of a perovskite solar cell with potential negative hooks and a large number of processes, it might be advisable to increase the points per decade to more than 10. The measurement points should always describe nice curves without corners or edges when connecting the points. If there are corners or edges, it is an indication that you should increase the number of points per decade in order to resolve the features better.

4. Is it possible to run DC measurement with this instrument?

Yes, it is possible. There is a function for DC measurements in LabOne. However, we still recommend impedance measurements because of the higher accuracy.

Or did you mean impedance measurements under DC bias? In that case, the answer is also yes. A bias can be set up to 10 V, but in 4 Terminal mode, the maximum for the measured AC voltage is 3 V.

5. If I measurement starts from high frequency then running to low frequency, is there different result with measuring from low to high frequency?

Generally, it should not make a difference for the measured values. Sweeping direction is only relevant for practical reasons.

As frequency increases, the current increases because you can think of it as more and more capacitive elements acting as short circuits. Then, it might be necessary to switch to the next higher current range. It is easier for the algorithm that determines the best suitable current range in auto-ranging mode, to just switch to higher values instead of finding a suitable smaller one. Larger ranges are usually accompanied by higher noise level, which means there might be a better suitable smaller range but the algorithm still sees too much noise to switch.

But that is only true for auto-ranging. When you know the expected impedance, and you can specify one current range, then there should be no difference whether you sweep from low to high or from high to low frequencies. The only benefit for sweeping from high to low frequencies is that you see the major part of the spectrum faster because measurement at high frequencies is faster than for low frequencies. This can be useful to judge whether the measurement will be successful.

Please also see this blog post, where our new feature current zone ranging is described.

6. How to measure impedance of liquid samples?

Liquid electrochemistry is actually where it all started. Two metal electrodes in a solution. Here are some examples for liquid measurements covering several interesting applications of impedance measurements on liquid samples. First, second, third example are linked here.

7. What is the frequency range for MFIA?

MFIA 500 kHz has a frequency range from 1 mHz to 500 kHz.

MFIA 5 MHz has a frequency range from 1 mHz to 5 MHz.

Please visit the product page.

8. What is about the polarization resistance?

If I understand the question correctly, you are asking why the resistance Rpol = RDC - R0 (for ZDC having only a real component) is called polarization resistance. Is this correct?

First of all, polarization can have different meanings and occurs for many physical phenomena (dielectric/magnetic/ionic/... polarization). If a physical body is polarized, one or more of its inherent properties (fields or concentrations) point to one direction or show a gradient. In the case of EIS, it means, electrical charges are moved, but not transferred.

Polarization can only happen when a DUT can be charged or discharged, which, as was discussed in the Webinar, can happen at different orders of magnitude (depending on capacitance). The polarization behavior is frequency-dependent and that provides us information. Processes occurring at different timescales can be distinguished through the EIS measurement and quantified through subsequent analysis and modeling.

9. How can I design/ select the electrodes size/ shape if I would like to measure the moisture of the concrete block? Electrodes should be put on the surface or inside the block?

We are surely not experts in measuring moisture in concrete but it sounds like a very interesting application. I can share some thoughts on it.

If you want to measure the moisture in a concrete block, I would say as a first guess, you will have to use a 4-Terminal setup. The setup would be similar to this application, but your sample dimensions could differ. At the two ends, LCUR and HCUR should be connected to apply the current, and more to the center, two more electrodes to measure the voltage drop over that length. Reason is that those electrodes themselves will cause quite large contact and/or charge transfer resistances/impedances. Measuring the voltage via dedicated voltage lines does not lead to current flow, as explained in the Webinar, so the two center electrodes will be able to properly measure the voltage drop inside the concrete. If you expect significant surface effects, you might be required to drill channels into the block to put the voltage electrodes inside the concrete block as you suggest, and seal them.

10. How can we know true value of the DUT?

When you look at the accuracy chart of the MFIA, you see that in the white box, you will measure the "true value" of the DUT with an accuracy of 0.05%, or better.

11. What does it mean when the Nyquist plot is not semicircle rather something different?

A non-ideal semicircle (flattened, depressed,...) is a sign for a non-ideal process. Deviation from an ideal process can have many reasons. One example for such inhomogeneity is that your sample might not be 100% at the exact same temperature over its whole area. Then you will have a temperature distribution and correspondingly a distribution of related resistances, rather than one specific resistance. Another example would be a porous electrode with a certain grain size distribution. Then, every grain/electrolyte interface can have a slightly different capacitance value. In summary, there can be internal (such as grain size distribution) or external (such as local variation of ambient conditions such as temperature or gas/pressure) reasons to make a process non-ideal and thus lead to the flattened semicircle in the Nyquist plot.

12. Why do we measure impedance and not admittance? Admittance is the opposite of impedance and it gives the ionic conductivity of a material at a certain temperature.

We can measure the admittance with the MFIA as well - and we can display it in the Plotter and Sweeper modules. So, we could also call the MFIA an "Admittance Analyzer". Impedance spectroscopy is popular in many areas because it takes advantage of the fact that resistances add up to the total resistance (R1 + R2 = Rtot). In many cases, you can distinguish R1 and R2 very easily. That does not work for the admittance or the conductivity, respectively.

13. Regarding AC stimulation you have mentioned an optional DC bias. What is the advantage?/Do you have an example of situation in which it is useful?

For some devices, it is required. For example, a lithium ion battery. Its voltage ranges from 2.8 V to 4.2 V (depending on the electrode materials) and it cannot be measured at 0 V. It would break. Similarly, an LED shows blocking behavior at 0 V, while the processes of charge transfer and photon emission occur at voltages > 1.5 V (depending on color and chemistry). Therefore, you will need a certain DC bias to characterize the junction in the LED. Another aspect is that in equilibrium, one process might be dominating while under a bias, other processes could be analyzed with better accuracy.

14. What is double layer effect in battery?

Electrochemical double-layers are formed at electrochemical interfaces. Such interfaces are in batteries for example electrode/current collector or electrode/electrolyte.

15. Are there more ways to validate EIS other than DUT and KK test: like computational, of conventional simulation methods?

There are more theoretical ways to validate EIS. Measure twice is the simplest in my view (but not systematic and can only make a statement of the stability). The KK test is a more systematic test that catches several categories of systematic errors, but as stated, it is not a sufficient criterion for the validity. Other tests are based on similar versions of the KK relations, such as the Hilbert transform. Further, you can compare your measurement with a suitable model that is supposed to describe the behavior of your sample. You can further validate linearity if you vary the measurement amplitude - in the linear regime, this should not change the impedance you measure. Computationally, you can predict an impedance, but that is not best suited to check the validity of the measured spectrum.

16. In general, how does performance differ between using this MFIA module and using your HF2LI along with a HF2TA for impedance measurements?

The MFIA includes instrument calibration, built-in circuit modeling, and auto-ranging, which are designed for accurately measuring impedance. That is how the 0.05% basic accuracy is achieved in our reactance chart. With the HF2LI Lock-in Amplifier and HF2TA, you can also measure impedance by Ohm's law calculation, but the accuracy is not well defined.

17. Can we use this setup for impedance gas sensing? I asked this because just blowing on the stage changed the temperature.

Interesting question. The blowing onto the heater stage was a pure temperature effect though, no gas could enter the heater stage directly.

Gas sensing is indeed an important impedance application. Actually, there is a number of techniques for gas sensing. Please have a look at this blog post.

18. Do you suggest multi frequency impedance measurement? And, how do you measure the operating voltage graph?

Multi-frequency measurements can have several purposes. One purpose would be to measure the high-frequency capacity (high frequency required) and at the same time monitor the device's low frequency resistance, basically the overall losses (low frequency required). You could do that with the MFIA and the MD option, where you can measure simultaneously at 2 different frequencies.

Another purpose is to save measurement time by measuring at several frequencies simultaneously. There was quite some interest on this topic a couple of decades ago, especially for battery research because they require low frequencies which produce very long measurement times. However, the concept has never really picked up. One reason is that the signal energy has to be shared among several frequencies and cannot be concentrated on one frequency, as it is the case for classical impedance spectroscopy with the corresponding good signal to noise ratio. There are concepts to decrease the maximum amplitude by using phase angles and frequencies intelligently, but the signal to noise ratio will always be worse for multi-frequency measurements and the setup and detection is more complex. That is why we do not recommend it.

The operating voltage graph, or iv curve for current-voltage characteristics can be done by the MFIA (voltage sweep in the sweeper module) or a different device such as a controllable source meter.

19. What does the semi circle indicate inside the battery?

I think you are referring to the semicircles in the diagram that were shown left of the capacitive branch. Those are polarization processes that occur inside the battery. For example, charge transfer at the electrodes. That is a lossy processes and it adds an RC circuit in series. Since this process usually has a smaller time constant than the intercalation (meaning the actual charging itself), it shows up as additional semicircle left of the capacitive branch. The series resistance, mostly attributable to the electrolyte, also shows up left, as it is "faster" and adds to the impedance additively.

20. What is the main difference between MFLI and MFIA?

The MFLI Lock-in Amplifier is an instrument with a reference signal output and a measurement input. Those can be correlated. So, in principle, one could measure an impedance also with the MFLI, but it does not have a calibrated impedance calculation, auto-ranging and automated user compensation. Only the MFIA, or the MFLI plus the IA-option, can calculate an impedance within LabOne and deliver an impedance result with a base accuracy of 0.05%.

21. Can we use the device to measure impedance spectrum of multilayer solar cell?

The MFIA can measure voltages up to 3 V in 4 Terminal mode. For a triple-junction solar cell, this is sufficient if it is not operated under concentrated light. For the current record holder in multi-junction solar cells under 1 Sun (Vmax = 4.7 V), this would not be possible.

22. What do you mean by the EIS during degradation? is it mean during discharge of the/charge of the DUT? MFIA can measure EIS online?

Degradation measurement could be a measurement of a test cell over 1000 hours, for example. It could be advisable to measure one spectrum ever hour for the first 100 hours, and then measure a spectrum every 10 hours. This way, you could see if the sample is stable over time, and if not, analyze what part of the sample is degrading. Such kind of measurements can possible help to predict (by extrapolating) the lifetime of your sample or device.

I do not fully understand what you mean by "online". As has been discussed in the Webinar, EIS can be measured in situ or even in operando, which means the respective device is operated in a sensible enduring operating point. The MFIA can do that. But as it is the principle of impedance spectroscopy, any impedance analyzer should be able to do that as well.

23. Sometimes when in low frequencies (eg. below 1 Hz) the end of the semicircle becomes quite noise for my fuel cell experiments. Any idea what that might be? Degradation? Somehow not really in a steady-state?

There can be several reasons for this behavior. One reason could be that the process responsible for the impedance response at 1 Hz and below, depends on the applied voltage. If the amplitude is chosen too large, you might polarize the sample beyond the linear regime and it may react in a way that seems unsystematic (erroneous, faulty). If this might be the case, try decreasing the amplitude.

By the way, this is the case for the oxygen exchange electrode on the thin film sample that I showed in the example in the webinar. That is one reason why I did not go to really the low frequency regime during the example measurement.

Another reason could be slow drift in the measurement, such as temperature fluctuations (as also shown in the demo in the Webinar). In that case, high frequencies can be measured easily and accurately because a frequency point can be acquired faster than the fluctuation would change the measurement value. Below 1 Hz, measuring one point takes several seconds and fluctuations on that timescale will disturb the measurement.

24. Could a version of this instrumentation offer more than 1 signal input/output channel in order to allow multiple simultaneous impedance measurements?

The MFIA has one calibrated impedance channel. There are Auxiliary inputs and outputs (2 and 4, respectively), but they do not come with different measurement ranges. As there is no other version of our MFIA impedance analyzer, you would have to use an additional instrument in this case. By the way, it is possible to synchronize several MFIAs.

25. Can a battery module that contains multiple cell in series be characterized by EIS? In terms of getting average values for all cells?

Theoretically, it could be done. However, the voltage of several cells in series increases with the number and the maximum voltage is a major limitation for all impedance analyzers.

What is done frequently though, is to apply a current signal across the series connection of cells and measure only the voltage drop over one cell with a 4-probe measurement.

Then, you only get the impedance of one cell. But that is mostly more relevant than an average impedance because the average impedance will average out any misbehavior of a single cell.

26. Do you have a suggestion in the procedure for creating the model from the impedance measurements?

Yes, I do. But this is a complex topic. It will also be very different depending on the application, and there was not enough time to go into details in the Webinar. Please get in touch and I will be glad to assist and give a few hints.

27. What is the difference between dielectric and electrochemical impedance spectroscopy?

The only difference is the display of the results, because other parameters are relevant for the respective fields. The measurement principle is the same and the calculation of the impedance/admittance/dielectric properties is the same from the raw signal, just the calculated and relevant parameters are different. It is therefore easily possible to perform dielectric spectroscopy with the MFIA, and you can choose between many representations how to display the measured impedance in LabOne.

28. Low frequency impedance measurement takes very long for eg from 100mHz to 500kHz it takes 20min. Is it normal?

One period at 100 mHz takes 10 seconds. This is the absolute minimum of the measurement time for this particular frequency. Depending on the number of points per decade that you are using, you can calculate this minimum time. With the "1period averaging", you should be able to get quite close to this value, if the noise level of your setup allows for it. Which analyzer and software are you using? It should be able to measure 100 mHz to 500 kHz faster than in 20min, but that depends also on the sample and the noise. I know systems where you will need to average over at least 5 periods to get a low-noise measurement, and then, measurement times of 20min are quite normal.

29. How accurate would be if I measure the impedance at room temperature in open environment? Even when the parallel electrodes are on the surface of the dielectric block!

That is a difficult question. It surely depends on the nature of the dielectric block. In an open environment, there might be issues with moist on the surface of your sample. If it is not sensitive to that, and also not sensitive to small temperature variations (that are imminent in open environments), then there should be no problem measuring the sample in an open environment as you describe it.