Time-gated Raman is the standard answer to one specific problem: fluorescence so strong it buries the Raman signal in shot noise. Five or six years ago the technique was mostly a Finnish specialty - one academic group, one or two commercial spinouts, a handful of pharmaceutical-polymorph papers. The 2025-2026 literature shows that has changed.

Detector groups in Edinburgh and Strathclyde are publishing on 512-pixel SPAD line arrays, optics groups in Italy and China are demonstrating Fourier-transform variants, and the materials-physics community is now using time-gated detection for transient electron-phonon studies that have nothing to do with rejecting fluorescence at all.

For process-analytics readers the question is narrower: does any of this make time-gated Raman cheaper, faster, or more useful on the floor of a real plant? The short answer is “incrementally”, and the longer answer is in the specific numbers below.

SPAD arrays are getting wider and faster

The most concrete instrumentation paper of the period is Tye et al., published in Biomedical Optics Express in mid-2025. The Edinburgh-Heriot-Watt group reports a 512-pixel CMOS single-photon avalanche-diode line sensor with on-chip timing electronics, 50 ps timing resolution, roughly 150 ps detector jitter, and a 200 ps gate window achieved by summing four 50 ps time-bins. The group demonstrates clean spectra from paracetamol and a paracetamol-with-excipients OTC product (known in the trade for its heavy fluorescent dye load) in 30 s acquisitions over a 1 m multimode fibre. Olive and sunflower oil discrimination - a standard fluorescent-matrix benchmark - is shown alongside.

The 512-pixel count matters because earlier generations of the same architecture were either 128 or 256 wide. Wider arrays mean a larger spectral window per acquisition without scanning the grating, which is the throughput bottleneck for almost every fielded time-gated Raman instrument today. The fibre demonstration matters because in-process Raman is overwhelmingly fibre-coupled, and fibre-induced background is itself a non-trivial component of what time-gating now rejects.

A related 2021 paper from Tampere and Oulu, building on the same SPAD-line-sensor approach but pairing it with a 573 nm fibre-coupled diamond Raman laser, reports a 24 to 25-fold reduction in fluorescence-to-Raman ratio versus a continuous-wave reference and roughly 73% better fluorescence suppression than a comparable 532 nm setup on the same oil samples. The 573 nm laser - shorter wavelength than the 671-785 nm pulsed sources typical in commercial time-gated instruments - shifts the trade-off back toward higher intrinsic Raman scattering efficiency, at the cost of pulling closer to the fluorescence excitation maxima of common matrix components.

Fourier-transform time-gated imaging

Collard’s March 2026 paper in Light: Advanced Manufacturing demonstrates a different architectural choice: a SPAD array behind a scanning interferometer rather than a dispersive grating. The reported numbers are aggressive - around 100 ps temporal resolution, 0.05 cm-1 spectral resolution, and a spectral window from roughly -1000 to 10,000 cm-1.

For a process context, the FT geometry is interesting because it decouples spectral resolution from detector pixel count. A line array becomes effectively a single detector behind an interferometer; spectral resolution is set by the mirror travel. That makes high-resolution time-gated Raman more tractable on the optical engineering side, though at the cost of mechanical complexity in the moving mirror and longer total acquisition time per spectrum. The paper’s stated targets - autofluorescent biological tissue, gas-phase studies, crystallinity work - overlap with classic time-gated pharmaceutical applications (the foundational pharmaceutical-polymorph work is summarised in our earlier piece on USP <858> and <1858> at five years), so cross-pollination between the imaging and process worlds is plausible over the next two years.

Beyond rejecting fluorescence

The Reuveni et al. arXiv preprint (2603.09755) is the most interesting outlier of the recent literature. The group uses time-correlated single-photon counting with a SPAD detector and a CW probe modulated against an ultrafast pump to study transient electron-phonon coupling in lightly boron-doped silicon. The time-gated technique here is not rejecting fluorescence at all - it is resolving picosecond dynamics of the Raman peak itself, extracting carrier recombination parameters from transient phonon asymmetry.

For Spectrane’s audience the relevance is indirect, but worth noting: time-gated detection has graduated from a fluorescence-rejection trick to a general-purpose time-resolved Raman tool. That broader use base pulls more detector and laser R&D into the technique, which over a five-year horizon should mean cheaper components for the more boring industrial use cases.

What this means for process analytics

Three working assumptions for process-analytics teams considering Raman on fluorescent feedstocks:

  1. The SPAD-array side is consolidating, not exploding. The recent papers improve specific numbers - pixel count, gate width, jitter - but the underlying CMOS SPAD architecture has been broadly stable since the Henderson-group work in the mid-2010s. Plan procurement on that basis: do not wait for a generational leap, because the literature does not suggest one is imminent.
  2. Pulsed-laser cost remains the binding constraint. The published instruments rely on sub-100 ps pulse sources at non-standard wavelengths (573 nm, 671 nm, 785 nm). Off-the-shelf pulsed diodes at these wavelengths are getting more available but not meaningfully cheaper at the pulse-width spec time-gating needs.
  3. Fluorescent-matrix performance is now well-characterised in the literature. A team writing a URS for a process-Raman application on a strongly fluorescent feedstock can cite numbers from peer-reviewed work for fluorescence-to-Raman ratio improvement, rather than relying on vendor marketing. Where time-gating helps and where SERS or another approach helps is now an evidence-based discussion.

Continuous-wave Raman with sophisticated chemometric baseline correction remains the default for most fluorescent-feedstock applications because the hardware is cheaper and the chemometric stack is mature. Time-gated Raman is the answer when baseline correction is not enough - when the fluorescence is so dominant or so variable that no continuous-wave instrument and no model gives stable predictions. The 2025-2026 literature is making that boundary easier to draw with confidence.