All the following experiments were carried out using surface seawater collected along the track of the AMT-12 cruise, filtered through 0.4 µm pore size filters, acidified with 1 µL Q-HCl per mL seawater stored in polycarbonate bottles, and allowed to react with the reducing reagent (sodium sulphite, 2.5 µL per mL seawater) for more than 10h in polycarbonate bottles. The iron concentration of this seawater was estimated at about 1 nM.
The problem of reproducibility was identified when several experiments showed that after a gradual increase in peak height, the CL signal for acidified filtered seawater with sulphite seemed to stabilise, but with relatively reproducibility (e.g. Figure II.11, precision = 12.3% rsd (n = 16) in this example). Several components and parameters of the system may influence reproducibility and were thus tested (Table II.3), and their influence on precision was reported when possible.
Loading pH 3.0 3.5 4.0 4.5 5.0 5.5 P eak ar ea bla n k co rre cte d 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Figure II.11: Reproducibility
experiment of 20 analytical cycles with acidified filtered seawater with sulphite (PMT
gain = 6). Atlantic surface
water with [Fe(II)] ~ 1 nM, [S(IV)] = 100 µM.
Experiments Precision (% rsd)
i) Performance of the equipment
Change valves 14.4% (n = 11)
Change PMT and flow cell 13.1% (n = 14)
Change low-voltage pumps 9.6% (n = 9)
ii) Effect of flow rates and eluent strength
Flow rates 9 – 18% (n = 5-7)
Increased eluent concentration poor
iii) Effect of the flow cell design
Change design flow cell poor
iv) Other factors: Changes in pH
No change in CL pH monitored
Table II.3: Summary of the experiments performed to improve reproducibility.
i) Performance of the equipment
Almost all mechanical components of the system were tested to check for variations in their repetitive functioning. Air bubbles were observed in the standard/sample line on using the switching valve (V2), and this and one other valve (V3) were removed from the system (Figure II.9), but did not result in any obvious amelioration in precision (14.4% (n = 11)). In order to test other components of the system, the photomultiplier tube, flow cell, and switching valve (V1) were all exchanged with spares, but these modifications did not appear to improve the precision (13.1% (n = 15) before and 13.0% (n = 14) after changing components).
Variability in the standard/sample flow rate would change the quantity of iron loaded onto the resin. Relatively high pulsing was observed with the lab-made low-voltage pumps initially used, due to their slow rotating speed. These pumps were therefore
Number of scans (10scans/sec)
0 5000 10000 15000 20000 25000 30000 35000 P M T si gn al (V) 0.0 0.1 0.2 0.3 0.4 0.5 0.6
exchanged with Ismatec pumps which showed much less pulsing as their rotation speed was much faster. Variations in the volume delivered by the pump with time were monitored and the volume of solution delivered was found to only decrease by about 1.7% over 40 analytical cycles (data not shown). Peristaltic pump tubing was changed regularly to minimise this effect. Precision was thus slightly improved (9.6% (n = 9)).
ii) Effect of flow rates
Variability in the elution efficiency was tested by changing reagent flow rates. Decreasing flow rates of the luminol reagent and eluent changed the peaks shape and intensity as the residence time in the flow cell varied, but did not seem to improve reproducibility significantly (Figure II.12). Moreover, if the elution was not complete during the elution step, a carry over effect would be expected between peaks. However, increasing the eluent strength and elution time did not change peak area (data not shown), suggesting that the strength of the eluent and elution time used previously were close to optimum.
Figure II.12: Experiments
where luminol reagent and eluent flow rates were
changed by ± 25%. Flow
rates (mL.min-1) and the
precision (% rsd) for each of
the tests are indicated.
Atlantic surface water with [Fe(II)] ~ 1 nM,
[S(IV)] = 100 µM.
iii) Effect of flow cell design
As the CL light emitting reaction is very rapid (~ 100 ms), signal loss is possible if the mixing of reagents occurs away from the PMT. Thus another design for the flow cell was tested (Figure II.13). Instead of having the luminol reagent and eluent mixing just before entering the flow cell, the reagents mixed in front of the PMT window as the critical factor is the time for mixing of reagents in front of the PMT window. Peak
Number of scans (10/sec)
0 5000 10000 15000 20000 25000 30000 PMT sign al (V) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 1.3 mL.min-1 9% 1.6 mL.min-1 11% 1.1 mL.min-1 18%
shape was similar with both designs but the response was weaker with the new design which may be due to a modification in the mixing efficiency. The first design was therefore retained in subsequent experiments.
Figure II.13:
Experiment comparing two flow
cell designs with acidified filtered
seawater. Atlantic
surface water with [Fe(II)] ~ 1 nM, [S(IV)] = 100 µM.
iv) Other factors: Changes in pH
Variability in the CL pH in the flow cell would change the efficiency of the CL reaction. However, measurements of the pH in waste showed that there was no variation in the CL pH between replicate peaks during the detection step.
Given that most of the above experiments showed little improvement on precision, it was hypothesised that poor reproducibility was due to the 8-HQ resin, which seemed to require several cycles before stabilising when starting a new experiment. This problem was investigated further with the help of S. Ussher from the University of Plymouth (see Section II.3.3.4).