Predicting Linear Operating Range for Charged Aerosol Detection Using an Inverse Power Function Framework

Importance of the topic

Charged aerosol detection (CAD) is a broadly applicable, mass‑dependent detector used in liquid chromatography for analytes lacking strong chromophores, such as lipids, surfactants, carbohydrates, and many small‑molecule impurities. Its inherent utility in pharmaceutical R&D, formulation development (including lipid nanoparticle analysis), and quality control stems from near‑universal response to nonvolatile analytes. However, CAD response is intrinsically nonlinear across wide dynamic ranges due to aerosol formation and particle‑size effects. This nonlinearity complicates quantitative method development, increases experimental burden, and may reduce robustness if detector scaling and regression strategy are chosen empirically. A predictive, data‑driven framework to define linear operating ranges and optimal detector scaling therefore has clear practical importance for accelerating method development, improving quantitative reliability, and ensuring compliance in regulated environments.

Objectives and overview of the study

This application note develops and demonstrates a model‑based method to predict linear operating ranges for CAD using calibration data acquired at a single reference condition (PFV = 1.00). The approach: (1) quantifies the detector’s intrinsic power‑law response from PFV = 1.00 calibration curves; (2) uses an inverse power function to simulate detector behavior at alternative PFV settings; (3) evaluates linearity, residual structure, and appropriate regression weighting across simulated PFV sweeps; and (4) validates predictions experimentally using representative lipid nanoparticle (LNP) components. The goal is to enable objective selection of PFV and regression strategy while minimizing extra calibration experiments.

Methodology

The core model assumes CAD signal follows a power law: signal ∝ amount^p. The reciprocal 1/p corresponds to the instrument PFV scaling parameter. Steps implemented:

• Acquire calibration data at PFV = 1.00 across a wide dynamic range.
• Estimate the intrinsic exponent p using ordinary least squares regression on the untransformed data (or appropriate transformed representation) to capture detector curvature.
• Apply an inverse power transformation (scaling by 1/p) to simulate response at alternative PFVs and linearize the curve where appropriate.
• Use Monte Carlo simulation (95% confidence) with Gaussian noise from residual standard deviation to compute confidence bands and PFV ranges that yield acceptable linearity.
• Assess linearity not only by R2 but critically by standardized residual behavior across concentrations; test common weighting schemes (unweighted, 1/x, 1/x^2) to select the weighting that yields near‑constant variance.

This workflow is implemented using commonly available tools (Empower CDS for data management and Microsoft Excel for modeling and solver-based optimization) to facilitate adoption in regulated labs.

Instrumentation used

A representative experimental setup used to validate the approach included:

• LC system: ACQUITY Premier (Waters).
• Column: Waters GTxResolve Lipid Phenyl‑Hexyl+ RP, 1.6 µm, 230 Å, 2.1 × 50 mm.
• Injection volume: 3 µL; column temperature 50 °C; flow 0.400 mL/min; sample at ambient temperature.
• Mobile phases: A = 0.1% formic acid in water; B = 50:50 MeOH:MeCN with 0.1% formic acid.
• Detection: Charged Aerosol Detector settings included 10 Hz sampling rate, ‘normal’ time constant, ion trap 20 V, evaporation temperature 35 °C. A secondary TUV (200/280 nm) was used in parallel for chromatography monitoring.
• Data systems: Empower Chromatography Data Software for acquisition/traceability; Microsoft Excel for modeling and solver workflows.

Main results and discussion

• The CAD response from PFV = 1.00 calibration curves follows a clear power‑law relationship; estimating the intrinsic exponent p enables computation of an optimal PFV (≈1/p) to linearize response across a target range.
• Applying the inverse power function to PFV = 1.00 data and forward‑simulating detector response across PFV values produced PFV intervals predicted to provide near‑linear behavior. For the LNP components tested (SM‑102, cholesterol, DSPC, DMG‑PEG 2000), predicted PFV ranges converged on approximately 1.14–1.34.
• Monte Carlo‑derived 95% confidence bounds supplied PFV intervals that account for experimental variance, enabling objective PFV selection rather than ad hoc trial‑and‑error.
• Residual diagnostics were essential: unweighted regression showed heteroscedastic residual expansion at high concentrations; 1/x^2 weighting over‑compressed low‑concentration variance; 1/x weighting produced near‑constant standardized residuals across concentrations and therefore was selected as the most appropriate weighting for these datasets.
• Experimental calibration at PFV = 1.30 validated model predictions: regression statistics, standardized residual behavior, and percent deviations matched simulation within small margins (experimental PFVs were within ≤ 0.08 of their predicted optima). This confirms the simulator’s ability to predict key performance characteristics without additional calibration runs at each PFV.
• Integration into Empower CDS preserves audit trails and supports compliance while reducing the number of physical calibrations required for PFV optimization.

Benefits and practical applications

• Significant reduction in experimental workload by predicting usable PFV ranges from a single calibration set (PFV = 1.00).
• Objective, variance‑aware definition of linear operating ranges improves method robustness and reduces subjective decision making in method development and manufacturing transfer.
• Data‑driven selection of regression weighting (e.g., 1/x) enhances accuracy across broad dynamic ranges, aligning with ICH Q2 principles for validation of linearity and range.
• Workflow compatibility with Empower CDS enables traceable, compliant implementation suitable for regulated pharmaceutical environments.

Future trends and applications

• Automation and embedding of the inverse power function workflow directly into chromatography data systems could streamline PFV optimization and make the approach routine during method setup and technology transfer.
• Extension to multicomponent systems and automated per‑analyte PFV recommendation could support complex formulations (e.g., LNPs) where components span diverse chemistries and concentration ranges.
• Integration with laboratory information management systems (LIMS) and model‑governance frameworks would support scaling to manufacturing, enabling periodic revalidation and trend monitoring of detector behavior.
• Combining CAD predictive modeling with orthogonal detectors (MS, UV) may further reduce uncertainty by cross‑validating quantitative ranges and addressing analyte‑specific matrix effects.

Conclusions

The inverse power function framework provides a practical, statistically principled route to predict CAD linear operating ranges using only calibration data acquired at PFV = 1.00. By estimating the intrinsic detector exponent and forward‑simulating PFV effects with confidence intervals and residual diagnostics, the method enables objective PFV selection, appropriate weighting choice (commonly 1/x), and reduced experimental burden. Validation with LNP components confirmed strong agreement between simulated and experimental behavior. The approach is readily integrated into Empower CDS workflows and supports compliant, efficient method development and manufacturing deployment.

References

1. Han D., DeLaney K., Alden B., Birdsall R., Yu Y. Lipid Nanoparticle Analysis: Leveraging MS to Reduce Risk. Waters Application Note (2022), 720007716.
2. Fekete S., et al. Challenges and emerging trends in liquid chromatography–based analyses of mRNA pharmaceuticals. Journal of Pharmaceutical and Biomedical Analysis. 2023;224:115174.
3. Han D., Birdsall R., Simeone J., Fogwill M., Yu Y. Comparing ELSD and CAD Performance on Polysorbate Quantification in Infliximab Drug Products. Waters Application Note (2022), 720007501.
4. Maziarz M., Harden S., Rainville P. Determination of Fatty Acid Composition in Polysorbate 80 using HPLC with Charged Aerosol Detection. Waters Application Note (2026), 720009340.
5. Birdsall R., Du X., Bigos P., Han D., Bhiwankar N. Automating Charged Aerosol Detection (CAD) Analysis with Empower CDS Software Using a Single‑Vendor Integrated LC Platform. Waters Application Note (2026), 720009297.
6. International Council for Harmonisation (ICH). ICH Q2(R2): Validation of Analytical Procedures. ICH; 2023.