Análisis y estimación de concentración de NaCl en solución acuosa mediante el uso de espectroscopía Raman.
No Thumbnail Available
Date
2026
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
La cuantificación de la salinidad en soluciones acuosas es un problema relevante en múltiples áreas de la ingeniería y la ciencia aplicada, debido a su impacto sobre propiedades fisicoquímicas y a la necesidad de monitoreo rápido y no destructivo en diversas condiciones operacionales. En particular, la determinación de la concentración de cloruro de sodio (NaCl) mediante espectroscopía Raman presenta un desafío fundamental: al separarse en átomos individuales que han ganado o perdido uno o más electrones, el NaCl no exhibe bandas Raman intensas en el rango espectral habitual, por lo que la estimación debe realizarse de manera indirecta a partir de la respuesta vibracional del solvente. En este contexto, la banda de estiramiento O–H del agua constituye una región espectral de alta sensibilidad, cuya forma varía sistemáticamente con la concentración salina y con la temperatura. Sin embargo, la marcada dependencia térmica de dicha banda introduce un confusor dominante que debe ser considerado explícitamente para lograr estimaciones robustas. Esta tesis desarrolla y valida una metodología cuantitativa basada en espectroscopía Raman para estimar la concentración de NaCl en soluciones acuosas a partir de descriptores efectivos extraídos de la banda O–H, incorporando la temperatura como variable del modelo. Se prepararon soluciones estándar de NaCl en el rango de 0 a 5 g/100 mL y se adquirieron espectros Raman a distintas temperaturas controladas entre 10 y 50 ◦C, conformando una base de datos de entrenamiento y un conjunto independiente de validación externa. Los espectros fueron recortados en la región 3150– 3550 cm−1 y se parametrizaron mediante una descomposición en un número reducido de funciones Gaussianas, empleada como herramienta matemática para capturar redistribuciones sistemáticas de intensidad sin atribuir necesariamente poblaciones moleculares discretas a cada componente. A partir de las amplitudes obtenidas se formuló un modelo de regresión que relaciona cada descriptor con la concentración C, la temperatura T y un término de interacción C · T. Esta formulación permite derivar un esquema para estimar C desde un espectro observado bajo una temperatura conocida, favoreciendo la interpretabilidad y la trazabilidad del proceso de estimación. El desempeño de la metodología se evaluó mediante validación interna y externa, cuantificando la exactitud a través de métricas estándar, tales como el error absoluto medio (MAE) y la raíz del error cuadrático medio (RMSE). Adicionalmente, se comparó el método propuesto con un enfoque de referencia basado en regresión por mínimos cuadrados parciales (PLS), analizando ventajas y limitaciones en términos de precisión, estabilidad frente a variaciones térmicas e interpretabilidad. En conjunto, los resultados muestran que la banda O–H del agua contiene información cuantitativa suficiente para estimar la concentración de NaCl dentro del dominio experimental estudiado, siempre que se incorpore explícitamente la temperatura. La estrategia propuesta contribuye con un marco metodológico robusto para la cuantificación indirecta de salinidad mediante Raman, estableciendo bases para futuras extensiones hacia rangos más amplios de concentración, mezclas multicomponente y aplicaciones en condiciones operacionales más complejas.
Indirect Raman quantification of dissolved salts in water is attractive because it can be performed with minimal sample preparation, yet it is challenging for electrolytes such as NaCl where concentration must be inferred from solvent-mediated spectral changes rather than from strong solute-specific bands. The Raman O–H stretching band is information-rich, but it is also strongly temperature-dependent, making temperature a dominant confounder for salinity estimation unless it is handled explicitly. Here, we develop and evaluate two distinct temperature-aware Raman routes for NaCl estimation in water (0–5 g/100 mL) using a common region of interest within the O–H band (3150– 3550 cm−1). Spectra were treated by dark subtraction and analyzed in absolute intensity (counts), without baseline correction, smoothing, or normalization. The first route builds an interpretable estimator by representing the O–H sub-band with a global constrained three-Gaussian basis, extracting non-negative amplitudes, and performing bounded concentration inversion using the measured acquisition temperature. This formulation provides a compact set of physically interpretable features and enables concentration predictions that remain within the calibrated domain by construction. The second route addresses temperature effects through Individual Contribution Standardization (ICS), standardizing spectra to a reference temperature prior to Partial Least Squares regression (LV = 3). Under both internal (in-sample) evaluation and external testing on spectra acquired on a different day at intermediate temperatures, both routes delivered consistently high agreement between predicted and reference concentrations. Beyond concentration accuracy, spectral overlay analyses supported physical fidelity of the Gaussian representation on both calibration and external examples, indicating that the model reproduces the O–H band structure rather than relying on compensation artifacts. The two pipelines exhibited complementary operational behavior: the inverse formulation enforces physically plausible outputs, whereas the ICS–PLS route follows a standard chemometric workflow and can yield occasional out-of-range predictions typical of unconstrained linear regression. This work provides a transparent framework for temperature-aware Raman estimation of NaCl concentration from the O–H band and demonstrates the value of reporting method comparisons under a shared ROI and validation protocol. The results highlight practical trade-offs between an interpretable, physically inverse model and a multivariate regression approach following explicit temperature standardization. These findings motivate temperature-aware method design and critical pipeline comparisons when indirect Raman quantification relies on water-band features.
Indirect Raman quantification of dissolved salts in water is attractive because it can be performed with minimal sample preparation, yet it is challenging for electrolytes such as NaCl where concentration must be inferred from solvent-mediated spectral changes rather than from strong solute-specific bands. The Raman O–H stretching band is information-rich, but it is also strongly temperature-dependent, making temperature a dominant confounder for salinity estimation unless it is handled explicitly. Here, we develop and evaluate two distinct temperature-aware Raman routes for NaCl estimation in water (0–5 g/100 mL) using a common region of interest within the O–H band (3150– 3550 cm−1). Spectra were treated by dark subtraction and analyzed in absolute intensity (counts), without baseline correction, smoothing, or normalization. The first route builds an interpretable estimator by representing the O–H sub-band with a global constrained three-Gaussian basis, extracting non-negative amplitudes, and performing bounded concentration inversion using the measured acquisition temperature. This formulation provides a compact set of physically interpretable features and enables concentration predictions that remain within the calibrated domain by construction. The second route addresses temperature effects through Individual Contribution Standardization (ICS), standardizing spectra to a reference temperature prior to Partial Least Squares regression (LV = 3). Under both internal (in-sample) evaluation and external testing on spectra acquired on a different day at intermediate temperatures, both routes delivered consistently high agreement between predicted and reference concentrations. Beyond concentration accuracy, spectral overlay analyses supported physical fidelity of the Gaussian representation on both calibration and external examples, indicating that the model reproduces the O–H band structure rather than relying on compensation artifacts. The two pipelines exhibited complementary operational behavior: the inverse formulation enforces physically plausible outputs, whereas the ICS–PLS route follows a standard chemometric workflow and can yield occasional out-of-range predictions typical of unconstrained linear regression. This work provides a transparent framework for temperature-aware Raman estimation of NaCl concentration from the O–H band and demonstrates the value of reporting method comparisons under a shared ROI and validation protocol. The results highlight practical trade-offs between an interpretable, physically inverse model and a multivariate regression approach following explicit temperature standardization. These findings motivate temperature-aware method design and critical pipeline comparisons when indirect Raman quantification relies on water-band features.
Description
Tesis presentada para optar al grado de Magíster en Ingeniería Eléctrica.
Keywords
Salinidad, Cloruro de sodio, Análisis espectral