[fsonnichsen]

I have been working with a 4-pole conductivity sensor that is to operate for long periods of time (weeks).  I note that during the course of a week the sensor reading drifts by around 3%. The sample flask is sealed from evaporation and the temperature is held constant in a bath and further corrected with an RTD temp sensor. Concentrations run from 0 -> 60,000uS. I am using KCl dissolved in DI as a test reagent.
   I can see no reason for this drift.
> The 4-pole sensors are either graphite, platinum, or platinum black.
> Voltages are held below 0.3 and even 0.1Vpp which should be well below any electrochemical levels
> I have impressed 100, 1000 and 10000 Hz square waves on the excitation (outer) electrodes to break down any polarization layers.
The probe has a "memory"--if I disconnect it for a while and reconnect it the reading resumes where it left off.

So I am convinced that there is some type of physical interaction between the electrode and solution that permanently damages the electrode. If the voltages were higher I would attribute this to electrochemical plating/dissolution but this is not the case. So I am seeking information on physical absorption by the electrodes with K, Cl or perhaps OH/H. I know that Pt has some interesting surface properties but a search of the literature reveals nothing here.

Has anyone experienced this type of phenomenon in performing sensitive electrochemical analysis?  Anything in the literature or texts delving more deeply into the physical nature/interaction of the electrodes? (I have around 10 electrochemistry texts here but nothing is discussed on this important issue).

Thanks
Fritz

[Enthalpy]

Hi Fritz, 3% drift isn't that much for electrochemistry. In my microelectronics lab we used to read the exponent of the conductivity, not the mantissa.

Rather than an explanation, here are suggestions for trials:
- After a week of run, rub the probe with DI water, check where the measure restarts. The hard back of a sponge is great to clean copper, should work on platinum too.
- Measure the initial conductivity with two or more probes. Run with one for a week, then check if the other probe's reading has drifted too.
- Has the solution's pH changed? Any odour after using the higher concentration?
- Take a bigger flask. Is the drift smaller? After a week, move the electrolyte gently, check if the reading changes. Or isolate somehow a smaller portion of the electrolyte around the probe, and mix again after a week.

Depending on the flask's size, it is numerically conceivable that one week worth of small current changes the electrolyte's composition, more so if the measurement frequency is low.

[fsonnichsen]

Ha--I agree on the 3%. I work a lot with ocean chemists and they demand drift rates around 0.01%/day in recent times! Apparently it is doable, at least according to the claims of some commercial salinity probe manufactures. Putting it in perspective, evaporation of 0.1cc in 1 day will lead to this drift and I have seen it in a few poorly sealed flasks.
  Cleaning the probe does not work here-the slightest perturbation to the electrodes changes the whole value and it needs calibration.
  Regarding comparing two probes--yes-this works so apparently excitation of the probe is causing the drift. -leaving it in the flask with the power off shows no change upon repeating the measurement.
  No odor but I have not checked pH. Are you suspicious of oxidation here leading to H+ ions?
  The use of different sized flasks to evaluate possible build-up of reactants was done earlier with spurious results due to my original circuit design. I would like to try this again some time. Regardless I think this would imply some type of reaction at the electrodes and I do not see why this might happen at such a low voltage. Wouldn't we have to exceed the electropotental of at least one of the ions present?

   I have found an apparent correlation of drift with frequency of measurement--taking measurements every minute for 1 week leads to the drift---taking a measurement every hour does not.

thanks
fritz

[Enthalpy]

My proposals try to distinguish between a drift at the electrodes and at the solution. Hence the suggested tests, with a spare probe inserted after a week, rubbing the probe, bigger flask, renewed electrolyte...

It's not very intuitive, but the small current might alter the electrolyte's composition over time.

• Imagine that a 60mS/m solution gives 1mS conductance between the outer electrodes.
• At 0.1V peak, the current is 100µA peak.
• Over 1 week = 6*10⁵s, the moved charge is 60C or 600µmol.
• 1S/m needs 0.5%wt KCL so 60mS/m would be a 300ppm wt solution or 4mmol/L.
• If the flask contains 50cm³ (does it?) it's 0.2µmol KCl only, which hopefully diffuse enough in a week.
• With 0.3V, little is expected, and AC minimizes the effect, but I suspect some secondary effect of the 600µmol current on the 0.2µmol ions.

As I'm unable to put more exact figures on this suspicion, I suggest experiments to decide it.

[Borek]

That's why we use AC for conductance measurements.

[fsonnichsen]

Thanks again for the responses. I understand the ploy here--to attribute the drift to either the sensor or to changes in the solution. My initial impression is that the latter would not be the case since the electrical potential is too low to induce chemical changes. I did try an experiment with a 4L vessel (normally I test in a 500mL erlenmyer sealed at the top).  The large vessel was difficult to seal and the result inconclusive. (Tiny amounts of evaporation can upset the whole operation). Each test takes a week and I have limited space and circuit boards etc so one can understand the practical problems here. At any rate I intend to repeat the test with some other flasks. Cleaning the sensor is not an option since it is so sensitive.

I was remiss in reporting units for the conductivity range --it is expressed in uS/cm which is the unit used by my Thermo-Fisher meter and also generally used for ocean sciences. Using 60,000uS/cm we get 6S/m. If we assume the cell has a constant of 1.0 (as is the case here) it can be modeled as two 1cm sq plates separated by 1cm.  This leads to 16.6ohms----leading to a current of 6mA or 3600 coulombs over the week. This translates to 3.7E-2 moles.
Key here is that we are not impressing a DC voltage here-we are using a square wave centered on a virtual ground. I do suspect problems with voltage offsets however. The op-amps have very small bias currents and there is also expectable leakage in the bypass capacitors so I would expect a small DC bias voltage which nonetheless would be quite immeasurable with conventional tools but over weeks could lead to a charge build up.

More important however I don't see how this leads to the final drift. Since we are not dealing with electorlysis here and consumption of ions at the electrodes, we must look elsewhere. A charge layer lingering at the biased electrode would diffuse off upon power cycling (How long does it take to diffuse?). So I was inclined to think about some physical amalgamation of ions with the electrodes but I cannot imagine what the chemistry would be in this case.

Thanks
Fritz

[Enthalpy]

AC vs DC: you do see gases evolve at the electrodes with 50Hz AC. Less than with DC, but here the parasitic effects shall be less than 3% × 3×10^-4 times smaller than what DC would achieve, so considering only the simplest effects can't suffice. Since other effects are too complicated and unusual, I recommend to experiment.

I strongly suppose that the solution drifts rather than the electrodes.

Several probes: put two probes in the flask, measure once with each probe, then use one probe for a week while the other idles, finally check if the other probe has drifted too. Use a smaller flask to observe the drift after a shorter time.

If you post the electronic circuit I'll look at it. It's probably not the culprit, because measuring AC under low impedance is easy, and creating a drift over time isn't really what such circuits can do.