Here is the result of the calibration trial for Friday 18 Nov 2016. Much credit goes to Megan Onuma for all the work she did to collect sea-water, prepare samples and run the lab photometer, and David Spafford for setting us up.
We ran four sample blanks and reads through both narrow band (N) and broad band (B) photometers and got the results in the graph above. At first we thought we had only three samples because of a wild data point but it turned out the wild data point was caused by me copying down the number wrong. This added a forth point at RB value of ~1.1.
mPC had been prepared for dilution at a rate of 50 μL for a 3 ml sample for the narrow band spectrophotometer, which we’ll call NBPS or just N. Assuming that the same concentration was required for the broad band photometer we built (BBP or B) , the concentration was extrapolated for the 80 ml sample far to 1.3 ml of mPC per sample.
Figure 0. mCP. Stuff really is purple. To the left the 100 ml bottle used in the broadband device and to the right, a 3ml cuvette for the narrow band spectrophotometer.
Difference from B.Yang et.al. 1.3 ml is a lot more than the .03 ml of mPC used by B.Yang etal and something to revisit in the future. º
The SOP 6b Spectrophotometric pH chapter best practice is to use enough dye to obtain absorption readings between 0.4 and 1.0 at each absorbance peak.
It can be seen on Table 1 that the readings for the N-A2 were below 0.4. Maybe this spectrophotometer needs more dye. Next time around should also try less dye than we used in the mini photometer — which was 1.3 ml or 1.3 cm³, a lot out of that tiny bottle of mCL.
SOP 6b also suggests use of a 3rd absorption shot using 730 nm light, non-absorbing for mCP. The difference between the blank and sample shot is subtracted from the absorption values at absorbing wavelengths to correct differences caused by a change in the optical path between shots. Interesting.
We had time to take 4 samples. Of the four, sample number two was initially suspected as being an outlier and was discarded, assuming that the expected result is a linear relationship between RN and RB. After reviewing data extracted from photometer B, I found a transcription error which put a wrong number from B into the dataset that I read off to Megan to plug into her regression calc.
I have the corrected data It’s not an outlier so the data point is added back in. Also in data sets 1 and 2, I actually took two readings. The readings are close but not identical and illustrate the sampling drift present in the broadband system.
The calibration plan was to
- obtain absorption readings for prepared sea water samples on N at two wavelengths 578 and 434. The LED’s used by B are close in wavelength : 574 (green LED) and 434 (blue LED)
- compute RB and RN
- compute the linear regression between RB and RN ordered pairs.
Table 1. Trail calibration absorption and absorption ratio data. B- broadband, N – narrowband. In S1 and S2 I recorded two separate scans of the same sample using the original blank reading. These resulted in A1/A2 readings with drift in the hundredths column and reveal the performance limitations of the System B.
Figure 1a – below- (full scatter diagram of calibration data)
Figure 1b (below) – scatter detail from Figure 1a.
Figure 1. Scatter plot and regression line.
Table 1. Linear regression data and result is the calibration:
RN = 4.59225 RB – 2.5374
N RB RN pH 1 0.97079 1.90525 7.87600 2 0.98455 1.90525 7.87600 3 1.11758 2.58971 8.08000 4 1.09835 2.58971 8.08000 5 0.77541 0.96551 7.60000 6 0.69837 0.74350 7.50000 a -2.5374 b 4.59225 n=% r 0.996392811159731
The data in Figure 1a and Fig 1b shows a linear relationship between RB and RN. For reasons yet to be determined, it is not the same calibration to that obtained by B.Yang et.al. below in Fig 2.
Figure 2. Below-
We succeeded getting through the procedure four times. For me having seen a spectrophotometer for the first time, each run was a learning experience. The result does produce a calibration, and it was important to me to observe the B system produce actual absorption numbers using real sea-water and real mCP.
The SOP 6b Spectrophotometric pH (2009) protocol is for .05-.1 cm3 of dye to the sample and that the amount of dye should produce absorbance values from 0.4 to 1.0 at each absorbance peaks. Table 1 shows that the protocol we used produced A2 samples consistently less than 0.400.
RB to RN to pH Computation.
B.Yang et.al.(2014) equations 4,5,6, and 7 was the core of their RN to pH method.
Equation 1. $pH = $eqConstant + log10(( $RN2 – $e1)/(1-$RN2*$e23) );
The “Spec Sheet.xslx” spreadsheet is in use at the UH contains a similar pH equation
Equation 2. pH = P4+LOG((N4-0.0069)/(2.222-N4*0.133))
Table 2 – Below from the SpecSheet Spreadsheet.
Table 3. Using Equation 1, and the calibration equation we derived with 6 data points, I calculated the following. The corresponding SpecSheet spreadhsheet pH calculations are shown from Table 2
RB 0.69837, RN-est 0.66968, pH: 7.4358 SpecSheet pH Calculation 7.546 RB 0.77541, RN-est 1.02344, pH: 7.6306 SpecSheet pH Calculation 7.669 RB 0.97079, RN-est 1.92068, pH: 7.9297 SpecSheet pH Calculation 7.99 RB 0.98455, RN-est 1.98387, pH: 7.9455 RB 1.11758, RN-est 2.59479, pH: 8.0800 SpecSheet pH Calculation 8.144 RB 1.09835, RN-est 2.50647, pH: 8.0623
The broadband device had a noticeable drift of around ± 0.01 – .02 between readings. What causes the drift? How much does it affect the “accuracy” of the system?
Would using a 730 nm led (deep red/near infrared) or maybe 700 nm led (much more available) reading produce a usable correction to improve accuracy?