Split Point Analysis and Uncertainty Quantification of Thermal-Optical Organic/Elemental Carbon Measurements

This protocol and software tool addresses key challenges in accurately quantifying and in minimizing sources of measurement uncertainty in a commonly used thermal/optical carbon analyzer. This technique considers all sources of measurement uncertainty including instrument calibration and split-point estimation and propagates these uncertainties to measured carbon masses through a comprehensive Monte Carlo method. OCECgo is currently designed to interface with a particular instrument.

However, it is our hope that this protocol and software tool will be extended to apply to all commercially available thermal/optical carbon analyzers. When first using this technique, users should perform multiple calibrations to estimate the repeatability of their individual implementation of the protocol. To install new quartz filters to prepare the organic elemental carbon analyzer for calibration, open the access panel and remove the laser shroud.

To remove the photo detector, loosen the white POM nut behind the photo detector and disconnect the metal tube fitting on the left side of the photodetector. Then, slide the photodetector housing off of the quartz insert and place the housing in the bottom of the instrument. To remove the quartz insert, loosen the white POM nut holding the quartz insert in place and slide the quartz insert out of the POM fitting.

Then, rest the quartz insert on a lint-free tissue on a flat surface. To install the filters, first use a filter-removal tool to remove and dispose of the existing quartz filters. Next, place a new large quartz filter on a lint-free tissue on a flat surface and use the filter punch tool to punch out one filter.

Using clean tweezers, remove the filter from the punch and place the filter against the POM fitting for the quartz insert, such that the textured surface of the filter is facing away from the oven. Then, use the quartz insert to slide the quartz filter into the fitting until the filter is fully seated against the oven. After punching out and installing the next filter in the same manner, punch out a third quartz filter and use clean tweezers to transfer this quarts boat filter into the end of the quartz insert.

Reintroduce the quartz insert into the instrument and loosely hand-tighten the white POM nut that secures the quartz insert in place. To replace and align the photodetector head, slide the photodetector housing onto the end of the quartz insert and reconnect the metal tube fitting on the left side of the photodetector. Than, fully hand-tighten all of the POM nuts, replace the laser shroud, and close the access panel.

To obtain a calibration point, prepare the instrument for removal of the quartz insert as demonstrated and follow the manufacturer recommended pipetting procedures to aspirate five or 10 microliters of sucrose solution. Carefully deposit the sucrose solution onto the quarts boat as close as possible to the end of the insert and reintroduce the quartz insert to the instrument. After closing the access panel, open the Run menu in the instrument software and select Dry Wet Filter.

Following the dry wet filter procedure in the Sample ID field, enter the applied sucrose volume and confirm that the desired thermal protocol par file and a suitable txt output file are selected. Confirm that the Use Sample File Times check box is unchecked, and in the Sample Minutes dropdown menu, select 0. Confirm that the Cycle check box is unchecked, and click Start Analysis, confirming that only one analysis cycle is desired.

Allow the thermal analysis to execute and to run to completion. Following the calibration data acquisition, use clean tweezers to remove the quartz boat and reinstall the quartz insert as demonstrated. To complete the calibration constant with uncertainty, load the software tool and confirm that the Calibration Tool tab is open.

In section 1 of the graphical user interface, input the nominal volume of the applied sucrose solution, the instrument reported, integrated non-dispersive infrared signal corresponding to total carbon, the instrument-reported integrated non-dispersive infrared signal during the methane loop, and a Boolean to indicate whether specific points should be used in the calibration. In section 2 of the graphical user interface, update the default uncertainty characteristics of the sucrose solution and pipette as necessary. And confirm the desired number of Monte Carlo draws.

To run the Monte Carlo analysis of the calibration data, press the Go Arrow in section 3. Use the buttons in section 3 to save the current result as the default calibration and export the results as desired. Then, update the calibration file of the instrument with the results in section 4.

To compute the carbon masses and uncertainties, acquire measurement data as instructed in the instrument manual and click to navigate to the Data Analysis Inputs tab. In section 1, subsection a, click Browse, and in the File Selection dialog, select the instrument-created txt results file to load the time-resolved instrument data. In section 1, subsection b, review the sample IDs and click to select the analysis of interest.

In subsection c, review the analysis metadata, particularly the sample start time stamp of the analysis. To define the data processing options in section 2, subsection a, select the desired laser correction procedures, and in subsection b, select the desired correction procedure for the non-dispersive infrared detector. In section 2, subsection c, confirm and update as necessary the parameters of the generalized t-distribution reported for the mass calibration constant, and the estimated calibration repeatability error.

In subsection d, press the Go Arrow to create or update the analysis thermogram and attenuation versus evolved carbon plot. In section 3, subsection a, select the desired procedure to calculate the split point and associated uncertainty. In subsection b, depending on the selected procedure to calculate the split point and uncertainty, define the nominal split point, split point uncertainty, initial laser attenuation, and/or critical attenuation decline, leveraging the attenuation versus evolved carbon plot in section 4.

In section 5, review the nominal instrument precision and the desired number of Monte Carlo draws and press the Go Arrow to run the Monte Carlo analysis. Once the Monte Carlo analysis is complete, review the results in the Data Analysis Results tab, and export the results using the Export Analysis Results button. Here, representative calibration data from the thermal optical carbon analyzer are presented.

Linear regression of the calibration data under a Monte Carlo framework reveals the two-sigma confidence interval of each of the calibration data points. The two-sigma confidence interval of the linear regression is based on the uncertain calibration data. For each Monte Carlo draw, a randomized calibration area is coupled with the uncertain linear model to obtain a Monte Carlo estimate of the methane loop carbon mass.

Monte Carlo estimates of these calibration data can then be represented in a scatter plot histogram yielding uncertainty in the calibrated carbon mass injected during the methane loop. Here, representative measurements of carbonaceous emissions from a laboratory soot generator are summarized in an analysis thermogram and in an attenuation versus evolved carbon plot. The uncertainty in this split point can be estimated using one of three approaches including a novel attenuation decline technique.

Here, key results of this example analysis, including the carbon mass statistics and the best-fitting posterior distributions are summarized. In these examples, computed uncertainty in the carbon masses are generally larger than those reported by the instrument, up to 280%in the most extreme case. Be sure to execute the same procedure when applying the sucrose standard.

This ensures a consistent bias due to uptake of ambient organics and helps to minimize calibration uncertainties. We believe that this protocol captures all major sources of measurement uncertainty for this carbon analyzer. However, standard experimental procedures should include tests of repeatability for the concerned experiment.

Users should take note of the presence of low-power laser radiation when the laser interlock is defeated and should use caution when manipulating the OEM fittings. This technique permits the robust quantification of measurement uncertainty and has recently enabled the observation of statistically significant variability and light absorption by black carbon.