# Determination of CO2 Solubility in Water by NIR Spectroscopy under Different Temperature and Pressure Conditions

## TL;DR

This paper introduces a new method using near-infrared spectroscopy to monitor CO2 solubility in water under varying temperature and pressure conditions.

## Contribution

A noninvasive, real-time method for estimating CO2 solubility in water using NIR spectroscopy and PLS regression is developed.

## Key findings

- The PLS model achieved high predictive performance (R² > 0.99) with an average external prediction error of 9.91%.
- NIR spectroscopy is sensitive to molecular interactions between CO2 and water, enabling robust predictive models.
- The method provides a reliable alternative for real-time monitoring of CO2 solubility in industrial and environmental applications.

## Abstract

Monitoring the solubility of carbon dioxide (CO2) in
water under high-pressure and -temperature conditions is critical
for industrial processes. Although the solubility behavior of CO2 is well-established in the literature, most conventional
methods rely on offline, invasive, or indirect measurements. In this
study, a methodology was developed to monitor CO2 solubility
in water using near-infrared (NIR) spectroscopy due to its sensitivity
to molecular interactions and ability to detect structural changes
in the aqueous matrix in real time. The experimental system consisted
of an apparatus similar to the static synthetic method coupled with
an NIR spectrophotometer, enabling noninvasive spectral data acquisition
during the dissolution of CO2 in ultrapure water under
varying temperature (40 to 60 °C) and pressure (0.15 to 2.35
MPa) conditions. Spectral data were preprocessed and correlated with
CO2 solubility using partial least-squares (PLS) regression.
The PLS model was calibrated using solubility values obtained from
a thermodynamic model based on the Peng–Robinson equation of
state (PR-EoS) and the NRTL activity coefficient model, which was
parametrized using experimental data from the literature. The developed
model showed high predictive performance (R
2 > 0.99) with an average external prediction error of 9.91%. In
comparison,
the thermodynamic model yielded a mean absolute deviation of 0.073%
in pressure predictions. The results demonstrate that NIR spectroscopy
is sensitive to spectral variations associated with molecular interactions
between CO2 and water, enabling the construction of robust
predictive models. This integrated approach provides a reliable and
real-time method for indirectly estimating CO2 solubility
from spectral data. It thus represents an alternative tool for industrial
and environmental monitoring of dissolved gases.