Overview

As part of my design to implement MAT in the laboratory, I am encapsulating the anode in a peristaltic pump to evacuate the anodic reactants, particularly chlorine gas. The goal of this anode pump testing is to calculate the efficiency rates of the pump and estimate how much chlorine gas may enter the system.

As a reminder, here is the system:

Electrochemistry Half Reactions

1. @ the steel cathode, oxygen/water reduction

• $O_{2(g)} + 2H_20{(aq)} + 4e^- \rightarrow 4OH^-{(aq)}$
• $2H_2O_{(aq)} + 2e^- \rightarrow H_{2(g)} + 2(OH)^- _{(aq)}$
• $OH^-{(aq)} + HCO_3^{2-} \iff CO_3^{2-} + H_2O{(aq)}$

1. @ the platinum anode, oxidation of chloride

• $2Cl^-{(aq)} \rightarrow Cl{2(g)} + 2e^-$

For steel, the potential needs to be < -0.356V/SSC (potential relative to silver-silver chloride electrode) to be below the corrosion domain. From -0.406 to -1.006V/SSC, the cathode is in the oxygen reduction domain; above 1.006V/SSC, the steel is in the water reduction domain. On a steel cathode, potentials more cathodic than −1.044 V/SCC, the pH value at the steel–seawater interface will exceed 9.0, which may favor the formation of brucite Mg(OH)2, which requires a minimum value of 9.3.

Further, at these higher potentials, the rate of chloride oxidation increases. Thus, there is a two-fold advantage of operating at lower cathodic potentials between -0.4 and -1.0V/SSC to minimize chloride evolution and maintain within the oxygen reduction domain to precipiate calcium carbonate and not brucite.

Dye Testing

The end-goal of this test is to generate a table that looks like the following:

Table 1. Idealized results for conceptual purposes only, not actual data

This table shows that the initial volume of the aquarium is 75L and has no dye. 30mL of dye were added to the aquarium while the pump was running at 305mL/min for 8 minutes and 12 seconds, which decanted 2.5L into a separate container. Within that 2.5L, 29.9mL of the dye was collected and removed from the aquarium. The aquarium now has 72.53L of solution, which contains only 0.1mL of the dye that was not collected by the pump. Thus this idealized situation has a pump collection rat eof 99.7%

First Test

The first test I ran on a brushed motor DC pump that was pumping at a rate of 30mL/min. This matched the 30mL/min rate of dye I planned to inject into the system.

Immediately, I noticed problems that prevented further analysis. The dye was much denser than seawater so the majority of the dye was falling to the bottom of the tank. The chlorine gas will be buoyant bubbles that should follow the flow of water it is in, and so the dye is not a good proxy to test the effiency of the pump. Further, the pump rate was far too low to be of any use.

Second Test

I used one of the new ERL2 peristaltic pumps which was getting a maximum rate of 305 $\pm$ 3 mL/min (mean $\pm$ 1SD of 5 tests). I then took the stock dye solution and measured its density to be approximately 1.16 $g \text{ } cm^-3$. I created 100mL of diluted dye stock to reach a density of 1.03 $g \text{ } cm^-3$, by diluting with RO water to match the density of the seawater in the aquarium. This density-mathced, diluted dye become the new dye standard from which I measured pump capture efficiency. I ran the following test twice, named round 1 and round 2.

First, I created a calibration curve, to linearly model the Beer-Lambert Law, with the formula $A=\epsilon lC$, where A is absorbance, $\epsilon$ is the molar absorption coefficient, l is the path length of the cuvette (1 cm), and C is concentration of the sample/standard. Concentration was expressed as volume of dye per volume solution.

Since the molar absorption coefficient and path length is constant, the terms $\epsilon * l$ can be combined into the constant m, to produce the simplified equation A=mC. Standard concentrations bracketting the expected concentration of the samples were prepared and plotted against their measured absorbances to calculate the slope m.

This is done simply in R using the lm function to generate a linear model and the geom_smooth funciton of ggplot with the method=“lm” to plot the data.

For the standards, I expected the best case scenario to be 29mL dye/2500 mL solution = 1.2% and 1mL dye/72530 mL to be 0.001%, likely below the detectable limit. At the worst case, no dye would be collected and the concentration of the aquarium after the pump removed exclusively 2.5L of seawater would be 30mL dye/72530mL solution = 0.04%. So I created standard solutions in triplicate from 0.03 - 3.3 % to cover all scenarios.

calibration <- calibration %>%
  mutate(concentration = dye/(dye+RO))

calibration_lm <- lm(abs~concentration, data = calibration)

calibration %>%
  gplot(aes(concentration,abs)) +
  geom_point() +
  geom_smooth(se=F,method="lm")

Figure 1. Dye standard curve absorbance at 480nm

Figure 2. Precision increases as concentration increaes

There was high degrees of precision, with a mean coefficient of variation (CV) of 0.0504%. Nevertheless, precision decreased as concentration decreased, evidenced by the increase of coefficient of variation at low concentrations.

The linear function to determine concentration from absorbance fit the Beer-Lambert assumptions, including an intercept of approximately 0 and a high coefficient of determination ($R^2$) of 0.994.

Table 2. Round 1: Average absorbance from triplicate readings

Table 3. Round 1: Subtracted tank_init, calculate C and dye.vol

These numbers don’t make sense. While concentration is highest in effluent as expected, it’s dye vol is estimated at 1.286L, which is well above the 30mL I injected. I’m wondering if this is a problem of units? I’m using dye/solution concentration %, but Beer-Lambert is meant for molar concentrations. I’m looking into this some more.