Quantification of vitamin C using FT-NIR spectroscopy

Importance of the Topic

Ascorbic acid (vitamin C) is widely used in pharmaceuticals, dietary supplements, food preservation and polymer manufacturing. Routine, accurate quantification of ascorbic acid in formulated products (e.g., blends with starch excipients) is essential for quality control, label compliance and process monitoring. Traditional wet-chemistry assays (titrations, extractions, fluorimetry) are accurate but slow, reagent-intensive and require skilled personnel. FT-NIR spectroscopy offers a rapid, non-destructive alternative suitable for high-throughput QC and on-line/at-line process analytical technology (PAT) applications.

Objectives and Study Overview

This application note demonstrates the development and validation of a partial least-squares (PLS) chemometric model for quantifying ascorbic acid blended with starch using FT-NIR spectroscopy. The goals were to: build a robust calibration across a relevant concentration range, evaluate model statistics (calibration, cross-validation and prediction), and demonstrate that FT-NIR can accurately predict ascorbic acid content in independent validation samples with minimal sample preparation.

Methodology

Sample preparation and measurement:

• Standards: Ten calibration mixtures were prepared by weighing ascorbic acid and starch to produce concentrations from approximately 53 to 500 mg of ascorbic acid per gram of material. Separate validation samples, prepared at several concentrations within this range, were used to evaluate predictive performance.
• Sampling: Samples were placed into glass vials and mixed prior to analysis to ensure homogeneity. Spectra were acquired through the vial bottom using a diffuse reflectance probe.
• Acquisition parameters: Spectra were recorded on a Thermo Scientific Antaris II MDS FT-NIR analyzer using a SabIR diffuse reflectance probe. Each sample was measured three times, each measurement consisting of 110 scans averaged at 8 cm-1 spectral resolution; the three replicates were averaged to form the composite spectrum used for modelling.

Chemometric processing and calibration:

• Spectral preprocessing: Second derivative spectra were used to enhance spectral features and suppress baseline effects. Multiplicative signal correction (MSC) was applied as a pathlength correction.
• Spectral region: The model was developed over 4,800 to 7,600 cm-1, where the variation between standards was most pronounced.
• Modeling approach: Partial least-squares (PLS) regression was applied using Thermo Scientific TQ Analyst software. The prediction residual error sum of squares (PRESS) suggested a single latent factor provided optimal calibration performance.

Instrumentation Used

• Thermo Scientific Antaris II MDS FT-NIR Analyzer
• SabIR Diffuse Reflectance Probe (sampling through vial bottom)
• Thermo Scientific TQ Analyst chemometric software

Main Results and Discussion

The PLS model delivered strong statistical performance:

• Calibration correlation coefficient (R) = 0.99841; root mean square error of calibration (RMSEC) = 8.05 (units consistent with mg/g concentration scale).
• Cross-validation correlation coefficient = 0.99692; root mean square error of cross-validation (RMSECV) = 11.4.
• Independent validation using eight new samples returned a root mean square error of prediction (RMSEP) = 11.2, indicating good predictive ability across the tested concentration range.

Visual inspection of raw and second-derivative spectra showed clear, concentration-dependent variation in the 4,800–7,600 cm-1 interval supporting the choice of spectral window. Residual and calibration plots demonstrated close agreement between predicted and actual concentrations for both calibration and validation sets, with no evidence of systematic bias across the range studied.

Key practical observations:

• Minimal sample preparation (simple mixing and vial-based measurement) supports rapid throughput and low operator skill requirements.
• Use of derivative preprocessing and MSC effectively reduced baseline/pathlength variability from diffuse reflectance sampling.
• A single PLS factor was sufficient for this well-behaved matrix (ascorbic acid in starch), simplifying model implementation and maintenance.

Benefits and Practical Applications

FT-NIR quantification of ascorbic acid in starch blends offers several advantages relevant to QC and manufacturing:

• Speed: Rapid spectral acquisition and immediate prediction enable high sample throughput.
• Non-destructive and reagent-free: Eliminates need for corrosive or toxic chemicals and reduces waste and consumables cost.
• Ease of use: Measurement through vials and straightforward chemometric models make the approach suitable for routine QC by non-specialist operators.
• On-line/at-line potential: The diffuse reflectance probe and FT-NIR platform are compatible with PAT implementations for real-time monitoring.

Limitations and Considerations

• Model dependency: Accurate predictions require representative calibration sets that cover expected formulation and process variability (particle size, packing density, moisture, excipient variation).
• Matrix effects: Different excipients or manufacturing changes can alter spectral signatures; model transfer or re-calibration may be necessary.
• Sampling homogeneity: Proper mixing and consistent sampling geometry are critical to minimize variance from sample presentation.
• Measurement units: Errors reported (RMSEC, RMSECV, RMSEP) are on the same concentration units as standards; acceptable tolerance should be set relative to product specifications.

Future Trends and Potential Uses

Opportunities to extend and strengthen NIR-based ascorbic acid analysis include:

• Model transfer and standardization strategies to enable deployment across multiple instruments and manufacturing sites.
• Integration into PAT frameworks for continuous monitoring of blending, tableting or coating operations.
• Application of advanced chemometrics and machine-learning methods to improve robustness against matrix variability and to enable multi-component quantitation in complex formulations.
• Miniaturized or handheld NIR devices for field or point-of-receipt screening of raw materials and finished goods.
• Combining NIR with imaging or hyperspectral techniques to assess spatial homogeneity within samples or dosage forms.

Conclusions

This study demonstrates that FT-NIR spectroscopy, coupled with appropriate preprocessing (second derivative, MSC) and a simple PLS model, can accurately quantify ascorbic acid in starch matrices across a broad concentration range with minimal sample preparation. The approach provides a fast, reagent-free alternative to classical wet-chemistry assays and is well suited for routine QC and PAT applications, provided that calibration models are appropriately built and maintained for the target matrices.

Reference

Thermo Fisher Scientific. Quantification of vitamin C using FT-NIR spectroscopy. Application Note AN51633_E 05/22M. (Authors: Gabriela Budinova et al.)