Detection sensitivity of field-resolved spectroscopy in the molecular fingerprint region

Summary form only given. Mid-infrared (MIR) vibrational spectroscopy is a versatile tool for identifying organic compounds (OC) in gases or liquids in a label-free manner due to their unique molecular fingerprint, promoting a large variety of applications in biology, chemistry, medicine or environmental monitoring. Although new technologies like quantum-cascade lasers are rapidly evolving, the workhorse in this field is still Fourier-transform infrared (FTIR) spectrometry based on thermal sources. Without pre-treatment of the sample and/or enrichment of the OC of interest, those techniques offer a limit of detection (LOD) for compounds in aqueous environments at ppm levels [1]. Pushing this limit to ppb levels is commonly hindered by the lack of broadband, stable and bright MIR sources as well as of low-noise, high-dynamic-range MIR detectors.Here, for the first time, we employ field-resolved spectroscopy (FRS) [2] in the molecular fingerprint region, as a time-domain method able to mitigate the above-mentioned limitations. A bright, broadband femtosecond MIR pulse is sent through the measurement cell for liquids and its resulting electric field is sampled in time with electro-optical sampling (EOS) [3, 4]. To test the performance of FRS for sensing applications, a glucose dilution series is measured and compared to the results obtained with a state-of-the-art FTIR (Bruker Vertex 70). An EOS trace of water (black) and 5000 ppm glucose solution (red) is shown in Fig 1a. The freeinduction decay (FID) of the glucose molecules can be efficiently separated in time from the main pulse propagating through the sample, which enables its detection in a background-free manner. This drastically reduces the negative influence of relative intensity noise on the LOD of the instrument. Fig. 1b,c show the Fourier transforms of the FRS signals after applying a high-pass time filter at t = 500 fs after the main pulse intensity maximum, for two different glucose concentrations. In both cases, the glucose fingerprint is clearly visible. Fig. 1d,e show the absorbance spectra obtained with FTIR for the same concentrations, revealing that at 150 ppm the glucose fingerprint is not discernible from noise anymore. Fig. 1f shows the glucose concentrations retrieved by fitting an ab-initio model to the measured data with error bars determined from multiple measurements. For a measurement time of 10 s (5 cm-1 resolution) we find an LOD of 120 ppm for the FTIR and 10 ppm for FRS (LOD defined as the concentration where the relative error of the retrieval is >100%). These measurements illustrate the potential of FRS for the detection of OC. An outlook on measurements with even higher sensitivity and a discussion of the specificity of FRS will be given at the conference.