# Conductivity measurements of demineralized water - Unexpected results. Looking for explanation

I'm trying to measure the conductivity change of demineralized water over a few days. But I get results I can't explain.

## Setup

This is the circuit and cell I'm using:

Where $R_x$ is the conductivity cell, R is $1797 \Omega$.
The right picture shows the cell ($R_x$). The metal plates shown (barely visible) are made of stainless steel.

## Evaluation

On ch 1 and ch 2 I measure the voltage by consecutively taking a sample of $10^4$ points over 4 ms on each channel then I calculate the RMS. Using ohms law I calculate the resistance of the cell:

$R_x = R \frac{Ch_1-Ch_2}{Ch_2}$

The applied current is a sine curve of 0.5V with varying frequency between 1, 5, 10, 50 and 100 kHz. It is turned on for about 2-3 seconds during the reading of the two channels.

For each frequency I make 10 measurements and take the mean of those.
These measurements are then repeated every 30 min. Shown in the graphs below.

(Sample 1-3: I changed the water and repeated the measurement to see if it's caused by some residue in the cell from previous measurements, although I cleaned it thoroughly)

Why do the curves look like this? If it's caused by polarization of the ions in the solution wouldn't the resistance increase with time?
I would be happy if someone could point me in the right direction finding a theory explaining this.

• CO2 dissolve in water. This lead to increased ionic concentration so increased conductivity. Water evaporate too. This also lead to increased ionic concentration and thus conductivity. – Alchimista Oct 8 '17 at 11:34
• @Alchimista Would it remove this effect or reduce it below the sensitivity of my measurement if I put the cell in a plastic bag then flood it with $N_2$ and afterwards fill it with water? – Alex Oct 8 '17 at 12:38
• If you are looking for a boring almost flat curve in your graph then do it. Jokes a part, yes. More you hinder the contact between water and atmosphere more constant will be the conductivity you measure. Better you do it more constant conductivity. You can bubble the water too but prior of measuring. Then stop else you improve evaporation. Also avoid a dry stream of N2. Ideally the N2 should function as a static layer. But also seriously, what you have seen is the only interesting outcome... Else, ideally, you have a conductivity stable for ever. In case of constant conc only T effect is left – Alchimista Oct 8 '17 at 12:45
• After everything is in place keep little inlet outlet for N2 and cabling, and just go with a minimal overflow – Alchimista Oct 8 '17 at 12:51