Garde Casasnovas, Raúl
Resano Ezcaray, Martín (dir.) ; García Ruíz, María Esperanza (dir.)
Universidad de Zaragoza,
2022
Resumen: In this thesis three calibration approaches applied to high resolution continuum source graphite furnace atomic/molecular absorption spectrometry (HR CS GF AAS/MAS) have been studied.
¿ The advantages of multi-energy calibration for Br determination using Ca by monitoring the molecule CaBr have been assessed and compared with a similar approach developed during this study called multi energy ratios (MER). MER Data treatment does not require linear regression, and a dispersion of ratios between sample and sample+spike is calculated, this relationship is equivalent to the slope calculated in MEC. Since the basis mathematical relationship of this methodology is not linear, accuracy and precision of the method depend on the relative amounts of sample and amount of spike added. Better results were observed if the ratios/slope between signals of sample and saple+spike were near 0.5. results of MEC and MER were compared at two transitions of CaBr, ¿2¿ ¿ A2¿ (0,0) y ¿2¿ ¿ B2 ¿ (1,0), showing that for similar molecular transition sensitivities, MER is a more reliable strategy. If the number of transitions is high and with different sensitivities, MEC provides better precision. Besides, a method for the calculus of the limits of detection and quantification using MER, which allows to identify if the lines have the sensitivity needed for its application. Despite a wavy background around 625 nm, and the presence of an unknown molecule, the measurements of a certified reference material QC3060 showed accurate results for Br.
¿ The capabilities of a machine learning algorithm (ML) for the determination of halogen proportions in active pharmaceutical ingredients (API) by means of a modular simultaneous echelle spectrograph (MOSES) prototype, capable of increase the spectral window 100 times the values found in commercial continuum source instruments, coupled to a contrAA800G. Due to the univariate nature of the tree boosting algorithm, three models for F, Cl and Br each were calculated. Br model showed a bad performance, and it was decided to calculate its concentration through the remaining fraction. Validation parameters (Venetian blinds with ten splits and root mean square error of calibration, RMSEC) along with the four CRM proportions, which have variable quantities of the three halogens of interest, have shown the correct performance of the ML model. Additionally, MER was applied for the determination of F in these samples,.with errors under 5%. Afterwards, according to the proportions provided by the XGboost model and F concentration determined using MER, Br and Cl concentrations were determined, with errors under 10%.
¿ In this study, nanoparticles (NPs) and core-shell nanoparticles (SNPs) have been analyzed in order to identify between the atomization processes of NPs and ions of Cu, Pt, SiO2, Au and Ag by means of HR CS GF AAS. Thus, a parameter called tdelay (atomization delay) was used, Absorbance spectra of NPs and ions of these elements did not show differences at any of the ramps studied.
However, Au and Ag did show differences in their tedlay, calculated using a peak fit strategy. Au showed a unimodal signal for NPs following a Voigt profile,on the other hand, Au ions presented from one to two peaks, depending on the state of the tube. Ag, on the contrary, showed a complex signal for every specie analyzed, adjusting better for three Gaussian peaks. The analysis of mixtures of Au showed that only a few mixtures with high differing masses presented a double peak. The rest of the mixtures appear as a unimodal signal. In both cases, size was estimated comparing their tdelay with pure AuNPs. Results show variable errors but iut seems that mixtures with significantly differing sizes provided size values closer to the expected ones. No mixture of AgNPs showed a bimodal peak. Hence, size was compared to the average size and average volume of NPs. Ag results presented lower error, except for a mixture with -19%, all mixtures showed less than 10% error in size relative to the average size.
CSNP measurements showed that size estimated using the shell if it is comprised of Au and Ag, provided similar results to the overall size of the CSNPS. Core size characterization, on the other hand, showed errors around 20%. CSNPs smaller of 20 nm presented a high error and imprecision. Lastly, the effect of the presence of silica in the structure of CSNPs with Au was assessed. In Au@SiNPs, Au-core atomizes much later than expected according to previous experiments, with a different peak profile. Si@AuNPs, Res660, Res800 and Res980, were analyzed to calculate the size of the Au structure (shell). However, these CSNPs are larger than the standards measured, and results were calculated by extrapolation. Errors between -20 and 20% were achieved, relatively good values were achieved for extrapolated data. Finally, Si-core was estimated in Si@AuNPs through the difference of the total volume of the CSNP (estimated experimentally sizing the Au-shell) and the volume of atoms of the shell occupied (volume per concentration of Au in the sample including the particle number, provided by the manufacturer).
Resumen (otro idioma):
¿ The advantages of multi-energy calibration for Br determination using Ca by monitoring the molecule CaBr have been assessed and compared with a similar approach developed during this study called multi energy ratios (MER). MER Data treatment does not require linear regression, and a dispersion of ratios between sample and sample+spike is calculated, this relationship is equivalent to the slope calculated in MEC. Since the basis mathematical relationship of this methodology is not linear, accuracy and precision of the method depend on the relative amounts of sample and amount of spike added. Better results were observed if the ratios/slope between signals of sample and saple+spike were near 0.5. results of MEC and MER were compared at two transitions of CaBr, ¿2¿ ¿ A2¿ (0,0) y ¿2¿ ¿ B2 ¿ (1,0), showing that for similar molecular transition sensitivities, MER is a more reliable strategy. If the number of transitions is high and with different sensitivities, MEC provides better precision. Besides, a method for the calculus of the limits of detection and quantification using MER, which allows to identify if the lines have the sensitivity needed for its application. Despite a wavy background around 625 nm, and the presence of an unknown molecule, the measurements of a certified reference material QC3060 showed accurate results for Br.
¿ The capabilities of a machine learning algorithm (ML) for the determination of halogen proportions in active pharmaceutical ingredients (API) by means of a modular simultaneous echelle spectrograph (MOSES) prototype, capable of increase the spectral window 100 times the values found in commercial continuum source instruments, coupled to a contrAA800G. Due to the univariate nature of the tree boosting algorithm, three models for F, Cl and Br each were calculated. Br model showed a bad performance, and it was decided to calculate its concentration through the remaining fraction. Validation parameters (Venetian blinds with ten splits and root mean square error of calibration, RMSEC) along with the four CRM proportions, which have variable quantities of the three halogens of interest, have shown the correct performance of the ML model. Additionally, MER was applied for the determination of F in these samples,.with errors under 5%. Afterwards, according to the proportions provided by the XGboost model and F concentration determined using MER, Br and Cl concentrations were determined, with errors under 10%.
¿ In this study, nanoparticles (NPs) and core-shell nanoparticles (SNPs) have been analyzed in order to identify between the atomization processes of NPs and ions of Cu, Pt, SiO2, Au and Ag by means of HR CS GF AAS. Thus, a parameter called tdelay (atomization delay) was used, Absorbance spectra of NPs and ions of these elements did not show differences at any of the ramps studied.
However, Au and Ag did show differences in their tedlay, calculated using a peak fit strategy. Au showed a unimodal signal for NPs following a Voigt profile,on the other hand, Au ions presented from one to two peaks, depending on the state of the tube. Ag, on the contrary, showed a complex signal for every specie analyzed, adjusting better for three Gaussian peaks. The analysis of mixtures of Au showed that only a few mixtures with high differing masses presented a double peak. The rest of the mixtures appear as a unimodal signal. In both cases, size was estimated comparing their tdelay with pure AuNPs. Results show variable errors but iut seems that mixtures with significantly differing sizes provided size values closer to the expected ones. No mixture of AgNPs showed a bimodal peak. Hence, size was compared to the average size and average volume of NPs. Ag results presented lower error, except for a mixture with -19%, all mixtures showed less than 10% error in size relative to the average size.
CSNP measurements showed that size estimated using the shell if it is comprised of Au and Ag, provided similar results to the overall size of the CSNPS. Core size characterization, on the other hand, showed errors around 20%. CSNPs smaller of 20 nm presented a high error and imprecision. Lastly, the effect of the presence of silica in the structure of CSNPs with Au was assessed. In Au@SiNPs, Au-core atomizes much later than expected according to previous experiments, with a different peak profile. Si@AuNPs, Res660, Res800 and Res980, were analyzed to calculate the size of the Au structure (shell). However, these CSNPs are larger than the standards measured, and results were calculated by extrapolation. Errors between -20 and 20% were achieved, relatively good values were achieved for extrapolated data. Finally, Si-core was estimated in Si@AuNPs through the difference of the total volume of the CSNP (estimated experimentally sizing the Au-shell) and the volume of atoms of the shell occupied (volume per concentration of Au in the sample including the particle number, provided by the manufacturer).
Resumen (otro idioma):
Pal. clave: química analítica
Titulación: Programa de Doctorado en Ciencia Analítica en Química
Plan(es): Plan 487
Área de conocimiento: Ciencias
Nota: Presentado: 24 03 2022
Nota: Tesis-Univ. Zaragoza, , 2022
-
Todos los derechos reservados
Enlace permanente:
Visitas y descargas
Registro creado el 2024-04-22, última modificación el 2024-04-22
