Mid-IR photothermal sensing of liquids for trace analysis

Abstract

Photothermal Spectroscopy (PTS) is an indirect analytical technique in which the optical signal is directly proportional to the laser emission intensity. This direct dependence on the laser power means that - in contrast to more conventional transmission-absorption techniques - PTS fully benefits from the high power of novel tunable mid-infrared laser sources such as Quantum Cascade Lasers (QCLs). In particular, QCLs equipped with an external cavity (EC) allow broad tunability which can be exploited in the detection of liquids identified by broad absorption bands. To achieve high sensitivity in PTS it is also important to choose a sensitive mode of transducing photothermal signal. Among the PTS transduction techniques photothermal interferometry (i.e. the detection of the phase change resulting from sample heating) stands out due to its high sensitivity. In this work, we use an EC-QCL in a photothermal interferometry PTS setup for trace water detection. We employ a HeNe laser-based Mach-Zehnder Interferometer (MZI) with liquid flow-cells inserted in the two arms. An EC-QCL emitting in the range of 1570-1730 cm-1 is arranged co-linear to the analyte arm of the interferometer and used to target the bending mode (𝜈2 ~ 1645 cm-1) of water molecules in different matrices. Highest linearity and sensitivity are ensured by locking the MZI at its quadrature point via an active-feedback loop. Fluctuations and drifts are further minimized by means of temperature stabilization. When benchmarking the system against commercial FTIR spectrometers it is shown to be in excellent agreement with regards to band shapes, band positions and relative intensities and to compare favorably in terms of sensitivity. Achieved limits of detection (LODs) for water in chloroform and jet-fuel are in the low ppm range. Higher LODs orders of magnitude were obtained indeed for the case of water in ethanol. An analysis of the matrix influence on the PTS signal’s strength has been carried out. Results show how the choice of the matrix dramatically influences limits of detection and limits of quantification (LOQs).