**3. Results**

While the MLA software package o ffers several options for exploring data, not all are suited to study SSA. The locking operation o ffers insights into preferential association of material phases, e.g., how a certain component forms particles with other material phases. This encompasses preferential partner species in multi-material particles as well as the degree of liberated (mono-material) particles. Another function, calculated assay, gives a rough estimate of the elemental composition of the whole sample. This is used as a relative function to see trends, in the case here, P-content of SSA (6%) dropped to 0.5% in acid residue after.

The other operation deemed applicable here is chemical composition of the material as based on EDX spectra. Unlike X-ray di ffractometry (XRD) measurements, which are only applicable to the crystalline portion of the sample, this technique yields information about the whole sample (including amorphous material). In addition, XRD measurements of this material su ffer from peak shifting due to the considerable portion of amorphous material and hardly any clearly developed peaks. Initial endeavors to fit EDX-derived material phases to XRD-derived mineral phases were only able to

determine quartz. The reflexes of known phosphorus minerals (whitlockite, apatite, etc.) are not developed in the XRD spectra. This is interpreted as one of two modes: (1) P-phases in this SSA are present as a ye<sup>t</sup> unknown mineral, or (2) the P-content of SSA is largely present within the amorphous part of the material. As Table 1 lists, phosphate-bearing components found in SSA contain more than just Ca, P and O. For P-recovery, secondary resource users are interested in phases with low complexity and as few elements as possible. During phosphoric acid production, these other elements would necessitate purification, an elaborate and costly process. As P-content rises, the complexity of phases decreases. Thus, it might be interesting to focus on high P portions of SSA in recovery in order to lower purification cost. The general appearance of this SSA prior to chemical treatment is reflected in the BSE images in Figure 1. In the following, particles are discussed. Particles in this material are either inherited—they existed prior to combustion—or a product of incineration processes.

**Figure 1.** BSE images of sewage sludge ash mixed with graphite for particle separation and stabilized with epoxy for analysis. (**A**) agglomeration of fine flakes; (**B**) inherited mineral grain, presumably from fluidized bed, bottom right of the grain shows P-rich material condensing onto grain; (**C**) general overview of SSA and (**D**) agglomeration around inherited grain.

Using the target component grouping subsequent to a generic labelling approach [16], the material composition of three samples is shown. Figure 2 compares the initial ash, digestion residue after HCl treatment, and digestion residue after thermochemical pre-treatment.

**Figure 2.** Composition of untreated SSA before and after chemical digestion by different acids (generic labelling); SSA: untreated, original sewage sludge ash, HCl: residue after SSA digestion with HCl; FSC-HCl: residue of HCl digestion of thermochemically treated SSA. Shares of material given in area-%.

When untreated and treated samples are compared, some of the material is missing. This material has been fully digested and moved to the liquid part. Analysis of the raw phosphoric acid showed other elements as well [11]. Wherever material has been removed in significant amounts, the proportion of other components changes. Upon closer investigation, three "key" spectra were identified. They represent the main P-bearing components of this SSA and are fully digested using HCl.

#### *3.1. Insights Gathered by Generic Labelling*

Using generic labelling [16], the spectra names are CaFeSiAlPO, CaPAlSiMgFeKO and PCaAlSiFeMgO. They contain 6, 12 and 16 wt% P, respectively, as determined by EDX analysis, and account for about 40 area-% of the material. By generic labelling, the first two of these would have ended up in the Ca-dominant group, and only the third as part of the P-rich material. Thus, when tracking P deportment, it is helpful to label spectra according to their P-content rather than the most abundant element.
