*4.3. MCR-ALS Application*

From the PCA analysis of the spectral data of chalcopyrite and copper concentrates, the spectral data of CpyF and Conc. A were selected, which have discontinuous sections in which the spectral characteristics are related to copper oxides. The intensity of these spectral characteristics is the result of the temperature reached by the combustion cloud of each sample. Table 3 indicates the average combustion temperatures estimated by two-color pyrometry, and according to [23], a good selection of the sampling wavelengths to apply this method are 650 and 750 nm.


**Table 3.** Average estimated temperatures of the combustion flames.

Continuing with the analysis, the airPLS baseline extraction algorithm was applied to the two selected datasets, thereby separating the dataset to extract the continuous and discontinuous emissions. Since the spectral evidence of copper presence is associated with wavelengths of ~606 and ~616 nm, we proceeded to limit the discontinuous section in the spectral range between 540 and 650 nm, while the noise was reduced through the algorithm of Savitzky–Golay (SG) [24]. Figure 7 summarizes the discontinuities profile of CpyF and Conc. A.

Thus, to obtain a spectral profile associated with copper oxides, the spectra of the sample Conc. A were selected because they present greater irradiance intensity.

**Figure 7.** Discontinuities profile of Conc. A (orange) and CpyF (blue).

MCR-ALS can perform multiset analysis of data organized in a single matrix [25] and, therefore, for this study, we proceeded to create a matrix in the range of 540 to 650 nm that contains the discontinuous samples of the Conc. A sample and the spectral signals of sodium (Na), FeO, and Fe3O4 were those obtained by Toro et al. [16] (Figure 8). The aim of this was to make the MCR-ALS algorithm separate the signals of the known species (FeO, Fe3O4, and sodium) from those that are not, including the copper oxide profile.

**Figure 8.** Spectral profiles (**a**) associated with sodium and (**b**) associated with FeO and Fe3O4[16].

The MCR-ALS method was applied to the new matrix, as detailed in Figure 9, with 10 components and a "non-negativity" restriction on spectral profiles and concentrations. With these settings on the MCR-ALS GUI, 92.5244% of the data variance and an error of fit (PCA) of 3.0201% were achieved with 70 iterations [25]. As expected, within the 10 obtained profiles were those associated with emissions of sodium (Na), FeO, and Fe3O4, in addition to a profile whose characteristics resemble the profile reported by Knapp, as seen in Figure 10.

**Figure 9.** Data processing schemes using MCR-ALS.

**Figure 10.** Spectral emission measurements obtained via MCR-ALS and reference spectral characteristics.

On the other hand, the phase analysis of the calcine of the sample Conc. A is presented in Table 4, a majority composition of Fe and Cu oxides was observed according to what was expected, which was a product of the working conditions. The mineralogical composition of the Conc. A calcine was determined using an integrated automated mineralogy solution, QEMSCAN system (quantitative evaluation of minerals by scanning electron microscopy), produced by the Fei Company (Hillsboro, OR, USA). The Cu2S generated by the decomposition of species such as bornite or chalcopyrite can be oxidized to form CuO or Cu2O. The formation of these copper oxides in the flame follows the following transition Cu2S → Cu2O → CuO, this can be corroborated with the analysis of the stability diagram Cu-S-O in Figure 11.

**Table 4.** Calcine of the sample Conc. A mineralogical composition.


**Figure 11.** Cu-O-S stability diagram at 1500 ◦C constructed with HSC® software.

This small thermodynamic analysis ensures the formation of CuO within the working conditions, so the depicted CuOx emission profile in the combustion spectra of a copper concentrate is effectively the emission profile of copper oxides. The detection of these oxides in the process is important because it will allow having control over copper oxidation in the flame, allowing the operator to adjust process variables preventively. This would avoid greater copper losses in the slag.
