Element distribution within the cementitious matrix of pozzolanic roman mortars: a new approach using image processing via x-ray map analyser

A new semi-automated image processing procedure has been here used to compute the element distribution betweenand within specific phases, recognised inside a hydraulic Roman mortar sample. This new tool package, based on thesemi-automated combined use of the Principal Component Analysis (PCA) and of the supervised Maximum LikelihoodClassification (MLC) of X-ray maps, is largely centred on functions implemented in ArcGis® (Ortolano et al., 2014).These functions are able to highlight, for instance, the local reaction phenomena responsible for the compositionalvariability within the same phase, and also allow to combine different primary outputs useful to extrapolate complexreaction indexes (e.g. Hydraulicity Index - HI, Boynton, 1980).Three different micro-domains were selected within the investigated archaeological mortar, whose hydraulicproperties were reached by mixing an aerial lime with a pozzolanic aggregate from the Alban Hills. Attention has beenhere focused on the cementitious matrix in order to assess the distribution of hydraulic phases within the investigateddomains. To this aim, three different X-Ray map arrays have been acquired with a SEM-EDS device at 1024*800 pixelresolution with a magnification closely ranging from 120 to 230X.Starting from these EDS X-ray maps, the procedure included two different cycles of analysis: the first processed theentire selected domain in order to classify all the recognisable components (cementitious matrix, aggregate and pores);the second allowed to investigate in more detail the element distribution within the cementitious matrix to put intoevidence the effect of the differential hydraulic reactions that develop between pozzolanic aggregate and hydrated lime.In addition, the application of the kernel density function permitted us to obtain the density distribution of Ca, Al, Si,Mg, K and Fe within the cementitious matrix. The use of raster calculator functions allowed then to obtain furtherderived maps showing, for instance, the variability of the HI within the three selected domains.Obtained results highlight as the used tool package proved to be powerful in the study of pozzolanic mortars since itprovides a simple visualization of the newly formed phases and their distribution within a restricted but compositionallycomplex system.