The BreathSens consortium, funded under the German project code 13GW0325C, carried out a 44‑month research effort from 1 May 2019 to 31 December 2022 to develop a compact photoacoustic sensor capable of detecting acetone in human exhaled breath at sub‑ppm concentrations. The overarching goal was to enable a mobile point‑of‑care (POC) device for diagnosing metabolic disorders and monitoring ketogenic diets in children with epilepsy. The project was led by Prof. Dr. Rudolf Bierl of the University of Regensburg, with key partners including the technology provider nanoplus, the laboratory specialist Ganshorn, and the Institute of Technology at OTH Regensburg. The consortium’s work was structured around three complementary light‑source approaches—near‑infrared (NIR) interband cascade lasers (ICL), mid‑infrared (MIR) quantum cascade lasers (QCL), and ultraviolet (UV) high‑power LEDs—each targeting a distinct acetone absorption band identified through extensive literature review and spectral simulation.

Technically, the team achieved a sensor footprint of 14 × 6 × 6 cm³, well below the 14 × 9 × 9 cm³ design target, by integrating a macro‑scale metal measurement cell equipped with adjustable buffer volumes to accommodate real‑time changes in gas matrix composition. The NIR ICL, supplied and iteratively improved by nanoplus, delivered an optical power increase from 15 mW to over 30 mW and a focal shift from 3 m to 40 cm, resulting in a detection limit of approximately 14 ppbV for acetone. The MIR QCL, chosen for its high optical output and large absorption cross‑section, set a new photoacoustic world record with a detection limit of 100 pptV. The UV LED, operating near the 278 nm acetone band, provided a low‑cost alternative that, while less sensitive than the laser sources, contributed to the overall robustness of the system. Together, these sources enabled sub‑ppm detection across all three spectral regions, with the QCL surpassing the sub‑ppb requirement by more than three orders of magnitude.

To mitigate cross‑sensitivities to common breath constituents such as water vapor and CO₂, the consortium performed an exhaustive absorption‑spectra analysis and incorporated auxiliary sensors for these interferents. Machine‑learning algorithms were trained on a database of photoacoustic spectra to simultaneously quantify acetone, ethanol, water, and CO₂ from a limited set of measurement points. A two‑stage sampling protocol was developed: first, exhaled breath is captured in a sterile Tedlar bag using a Ganshorn‑designed sampler; second, the stored sample is pumped through the photoacoustic cell at a constant flow, ensuring stable mass transport and averaging over multiple breaths. This approach overcomes the vibration and low‑volume challenges that initially limited real‑time measurements.

The scientific outcomes of the project were disseminated through five peer‑reviewed journal articles, five conference presentations, and a chapter in the Springer series “Bioanalytical Reviews” titled *Breath Analysis – An approach for smart diagnostics*. The research also supported the completion of a bachelor thesis, a project thesis, and a doctoral dissertation, with an additional PhD project initiated. The consortium’s success demonstrates that a compact, highly sensitive photoacoustic sensor can be realized using a combination of advanced laser technologies, a robust measurement cell, and sophisticated data‑analysis techniques, paving the way for practical breath‑analysis diagnostics in clinical settings.