Tag
11 articles tagged spectroscopy.
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.
Recap of process analytical talks, awards and sessions from PITTCON 2026 in San Antonio and the analytica conference 2026 in Munich, with verified program details from both organizers.
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.
Univariate SPC watches one number. A spectroscopic analyser produces a thousand. Multivariate SPC condenses them into two statistics that have been the standard since the early 1990s.
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.
A primer on PAT - the FDA-driven framework for measuring chemistry where it happens, why pharmaceutical and chemical manufacturers adopt it, and the tools that make it practical.
analytica 2026 closed in Munich with 35,000 visitors and 1,135 exhibitors. AI and automation drew the loudest applause; the process analytics floor was quieter than the lab side.
Both techniques measure chemical composition without sampling. The choice is rarely close once you map your process onto five questions: water, fluorescence, specificity, dynamic range, and probe access.
Spectra are not answers. The bridge between a 1,000-channel spectrum and a number you can act on is chemometrics - a small toolbox of multivariate methods, used the same way for forty years.
Both methods take a 1,000-channel spectrum and compress it into a handful of latent variables. They are not interchangeable. PCA describes; PLS predicts. The decision is older than most of the software.