Appendix 3: Method of analysis

All samples were initially tested for the presence of carbonate using an acid test, in order to confirm that the sample was sinter. Hydrochloric acid was added to a sacrificial piece broken off the sample, with the bubbling of carbon dioxide confirming the presence of carbonate.

These samples were then prepared for analysis with the help of Dr. Zhiyu Jiang and others at the Environmental and Climate Change Centre, University of Newcastle, Australia. First, fresh faces parallel to the growth axis of the samples were exposed. They were then mounted in Daystar clear casting resin, using methylethyl ketone peroxide as a hardener and a polished surface was prepared.

Laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) was used to analyze forty-three elemental concentrations along the growth axis of nineteen samples in December 2006 at the University of Melbourne, under the supervision of Dr. John Hellstrom. The lead concentrations of eighteen of those samples are presented here. Laser ablation ICP-MS uses a laser to vaporize a spot on the sample and analyze the resulting vapor in a mass spectrometer. The pulsing laser was moved along the growth axis of the sample at a constant rate. The laser ablation scan for the Pompeii Wall 2 sample is visible as a faint horizontal line just below the arrow indicating the growth axis in Figure. 9.4. These scans along the growth axis allowed the high definition investigation of variation of concentrations over the period during which the sample was laid down. Analysis was made of the entire growth axis of all the samples presented here except for those from Mura d’Arce and Ponte Tirone, where only part of the growth axis was analyzed. When choosing the location for the laser ablation scans, obvious inclusions (non-sinter, such as dirt, sand, rock particles and water, trapped within the sinter as it formed) were avoided.

The methodology followed Desmarchelier and his team (2006) except as noted below. Pulsing was set at either 5 or 10 Hz. The NBS 612 glass standard was used. Unlike the data in Desmarchelier and his team (2006), long-term temporal trends have not been removed from the data presented in this work. Areas of abnormally low calcium concentration were excluded by setting a minimum level of calcium concentration after inspecting a graph of the calcium concentrations along the growth axis. These low calcium areas (presumed to represent voids or inclusions in the sample) were excluded as the type of calibration employed will produce errors when used on material not predominantly composed of calcium carbonate. Three of the samples (Pompeii aqueduct higher wall, Praedia Euripus and Praedia Reservoir) showed much greater variability in calcium concentrations, seemingly due to greater levels of impurities. This combined with their very short growth axes made it difficult to assign the minimum calcium threshold by visual inspection, so a value of 85% of the median was used as the threshold. 85% was chosen because this was the highest percentage of the median that any of the visually assigned thresholds constituted and thus would be the most conservative option.

The lead concentration data were gathered as the total lead concentration estimated from the concentrations ofeach ofthree lead isotopes, or forms of lead, and their naturally occurring abundances. These isotopes are designated by a number preceding the chemical symbol of lead, Pb representing the varying sums of the protons and neutrons in the lead atom’s nucleus. Two lead isotopes, 206Pb and 207Pb, occur in much greater amounts in the samples, and the natural world at large, than the other, 204Pb. 206Pb and 207Pb thus provide more accurate measurements of total lead concentration than 204Pb and only the results from the first two isotopes have been presented. The spread of the data does not conform to the bell-shaped “normal” distribution, but rather has a small number of very high concentration values, or outliers. Therefore, the median, rather than the mean, of both of these lead concentrations along the growth axis has been calculated, as the median is less sensitive to such outliers, and used as a representation of the lead concentration in the sample.

As a rough method of detecting the presence of sinter coating (see above), linear regressions have

Been performed on the lead concentration data with respect to distance along the growth axis (i. e., with respect to time), using the Microsoft Excel SLOPE function.