A side-by-side ranking of the open-source software stack practitioners actually use for Raman, IR, and NIR processing in 2026 - what each does best, where it stops short, and which combinations work in production.
Recent literature on SERS substrates points at flexible cellulose, 3D-printed metals, and uniformity-engineered nanostructures. We review what is useful for industrial work and what stays in the lab.
The open-source stack that academic spectroscopy groups now teach with - mdatools, Quasar, pyMCR, rampy, pybaselines - has matured into something an industry hire arrives knowing. A field snapshot.
Four mid-May physics.chem-ph preprints worth a PAT reader's time: vibrational eigenstate computation, polaritonic spectroscopy reliability, ML for photochemical decay, and DFT-accuracy ML force fields.
Six vendors that matter for bioreactor monitoring - what each one is good at, what they are not, and the methodology we used to scope the field down from the full process Raman market.
SERS makes Raman a billion times stronger and a hundred times more fragile. For most inline measurements that is the wrong trade. The places where it pays off are narrow but real.
Six recent reviews map where hyperspectral imaging now sits as a process-analytics tool in pharma, food, recycling, and beyond - and where chemometrics, data hunger, and standards still hold it back.
Five recent papers sketch where deep-learning chemometrics sits versus PLS for process Raman: a public bioprocess benchmark, two classifier benchmarks, self-supervised pre-training, and an explainability tool for GMP.
scikit-learn, pyChemometrics, ChemometricsLib, and the rest. A practitioner's review of what works for inline spectroscopy, what doesn't, and where the gaps still are.
Recent peer-reviewed work on time-gated Raman - 512-pixel SPAD line arrays, Fourier-transform variants, and time-resolved use beyond fluorescence rejection - and what process-analytics teams can take from it.