Microplastics Analysis: FT-IR Imaging Down to 5 Micrometres
Ask around a water laboratory and you will hear a number: FT-IR stops somewhere around 50 µm, so microplastics analysis below that means Raman. It is a reasonable belief. It is also roughly a decade out of date, and it is costing labs the wrong instrument decision.
Bruker's own product note on the subject is blunt about where the floor actually sits. Particles of 5 µm are detected and identified with FT-IR imaging, on real environmental filters with real debris on them.
Where the 50 micrometre belief comes from
It comes from a real and much cited paper. A comparison of FTIR and Raman imaging for environmental microplastics in Analytical and Bioanalytical Chemistry reported that "FTIR imaging leads to significant underestimation (about 35 %) of microplastics compared to Raman imaging, especially in the size range <20 µm", and proposed splitting the work into 500 µm to 50 µm by FTIR imaging and 50 µm to 1 µm by Raman imaging.
That was sound advice for the instruments available when it was written. It predates the focal plane array and IR laser imaging systems that now do this work routinely, and it has outlived the hardware it described.
The physics has not changed. Mid infrared wavelengths sit in the micrometre range and diffraction still sets a floor. What changed is where that floor actually lies once you image with an array detector instead of mapping point by point.
What FT-IR imaging reaches today
Bruker Product Note M195 set out to answer exactly this question. A 32x32 focal plane array detector on a LUMOS II was used on three samples: an environmental sample full of debris, and artificial samples of polypropylene and polystyrene beads of 5 µm suspended in water and filtered onto an aluminium oxide membrane.
The note states the working figures plainly. "At a pixel resolution of 5 μm, imaging speeds of >750 spectra per second, at 8 cm-1 spectral resolution, are achieved with LUMOS II." And the conclusion: "Particles of 5 µm can be reliably detected and identified using FT-IR imaging data."
Detected and identified, not just seen. The polypropylene bead gave a correct size determination because particle and pixel coincided. The polystyrene bead landed on a single pixel and the spectral quality was still good enough for an unambiguous library identification.
For context on the optics, Bruker specifies the LUMOS II FPA imaging option at "1.25 µm spatial resolution (ATR)" and "FT-IR images at 1.6 mm² per sec". The LUMOS II ILIM, which uses a quantum cascade laser instead of a thermal source, is specified at "4.25 µm spatial resolution" with "up to 169x faster area scanning speed compared to FT-IR".
So the honest answer to "how small can infrared go" is single digit micrometres, not fifty.
What the regulation actually asks for
This matters more than the specification sheet, because it decides what a monitoring laboratory is obliged to do.
ISO 16094-2:2025 covers microplastics in drinking water and low solids water from 1 µm to 5000 µm, and it asks for four things per sample: particle size, a count, classification by size range, and chemical identification of the polymer.
For the technique, Bruker's microplastics solutions overview states it directly: "EU Decision 2024/1441 and ISO 24187:2024, both specify IR microscopy for particle-based microplastic characterization, defining results in terms of particle counts and size classes."
IR microscopy. Not Raman as an equal alternative, and not a free choice between two techniques. For particle based characterisation under the current European framework, infrared imaging is the named method, and Bruker's Microplastic Identifier is documented as usable with Commission Delegated Decision (EU) 2024/1441 and ISO 24187.
The bottleneck was never the optics
Here is the part that gets missed. Once an array detector images a whole filter, resolution stops being the constraint and data volume becomes one.
A complete 25 mm filter measured on a LUMOS II ILIM produces 22.5 million mid infrared fingerprint spectra. Nobody is comparing those against a spectral library by hand, and the old workflow of picking particles visually and measuring them one at a time collapses at that scale.
Bruker's answer is the Microplastic Identifier, MPID, embedded in OPUS. Product Note M210 describes it as using "deep-learning-based models to classify polymer types and morphologies across large hyperspectral datasets acquired on Anodisc and other IR-transparent filter materials, making manual spectral comparison obsolete". It classifies each foreground pixel into one of a fixed set of classes: ABS, EVAc, PA, PC, PE, PEEK, PET, PLA, PMMA, POM, PP, PS, PU, PVC, SAN, cellulose and calcium carbonate.
Connected pixels of the same class become a particle, each particle gets an averaged spectrum and a HIT score against a microplastics reference library, and a user set threshold decides what counts as identified. Raise the threshold and you suppress marginal hits, lower it and you catch degraded polymers at the cost of false positives. That threshold is where a method developer earns their keep, and it is a deliberate control rather than a black box.
MPID runs on LUMOS II and HYPERION II imaging data as well as on ILIM data, so the software is not tied to one instrument generation.
Two routes, and how to pick
LUMOS II with FPA imaging and MP-ID. The FT-IR route. Transmission imaging at 8 cm-1 across the mid infrared fingerprint range, 5 µm pixels, and the full spectral range available for anything unexpected on the filter. This is the route when you want a full spectrum per pixel and the flexibility to go back to the data later with a different question.
LUMOS II ILIM with the ParticlePlus workflow. The throughput route. Product Note M214 gives the numbers: a complete 25 mm filter is acquired in 13 minutes regardless of how many particles are on it, MPID processes the filter in 15 minutes or less, and the documented example ran to a total turnaround of 17 minutes for 939 particles across five polymer classes. End to end, complete results for an entire filter in under 30 minutes.
Both accept Anodisc, metal coated and silicon membrane filters, and both are designed to run without a spectroscopist standing over them.
If you measure a handful of filters a month, the FPA route is enough and gives you more spectral freedom. If microplastics monitoring is becoming a service you deliver on a schedule, the arithmetic changes fast: at under half an hour per filter, one instrument covers a monitoring programme that would otherwise need a person doing manual particle picking full time.
Where it still has limits
Infrared imaging is not the answer to everything, and pretending otherwise is how you lose a technical audience.
Below 1 µm you are in Raman territory. Nanoplastics research needs the shorter wavelengths of a Raman microscope, and Bruker describes Raman as the right choice for sub micron particle characterisation. That is a research question, not a monitoring obligation, and it is worth being clear which of the two you actually have. Bruker builds both, so pointing you at the right one costs me nothing either way.
The filter is in every spectrum. Membrane choice is the most common reason a first measurement campaign disappoints. Anodisc, silicon and metal coated filters all behave differently in the infrared, and the choice belongs in the method before the sampling starts, not after.
The HIT threshold is a decision, not a default. A number that suppresses false positives on clean water will suppress genuinely degraded particles in a surface water sample. Two laboratories running different thresholds will report different counts on the same filter, and neither of them is wrong.
Blanks are not optional. Fibres from clothing, dust from the bench and residue from reagents all end up in the count. A result reported without a blank filter is provisional.
FAQ
How small a microplastic particle can FT-IR identify?
Bruker Product Note M195 demonstrates that particles of 5 µm are reliably detected and identified with FT-IR focal plane array imaging, using a pixel resolution of 5 µm on a LUMOS II. The often quoted limit of 50 µm reflects the point mapping instruments of an earlier generation rather than current array imaging.
Does ISO 16094-2 require Raman for small particles?
No. ISO 16094-2:2025 covers 1 µm to 5000 µm using microscopy coupled with vibrational spectroscopy, and does not prescribe a single technique. Bruker states that Commission Delegated Decision (EU) 2024/1441 and ISO 24187:2024 both specify IR microscopy for particle based characterisation.
How long does a full filter take?
With the ParticlePlus workflow on a LUMOS II ILIM, Bruker documents 13 minutes for automated acquisition of a complete 25 mm filter, 15 minutes or less for MPID analysis, and complete results for an entire filter in under 30 minutes. A documented example reached 17 minutes total for 939 identified particles.
Do I need a spectroscopist to run it?
The workflows are built so that you do not. Acquisition is automated across the whole filter, and MPID classifies without manual spectral comparison. Method setup, filter choice and the identification threshold still need someone who understands what they are deciding.
Which polymers does MP-ID identify?
Product Note M210 lists ABS, EVAc, PA, PC, PE, PEEK, PET, PLA, PMMA, POM, PP, PS, PU, PVC and SAN, plus cellulose and calcium carbonate as non-polymer classes that commonly appear on environmental filters.
See it on your own filters before you decide
I am Savas Kaplan, and I represent Bruker Optics as its agent for the Netherlands and Belgium. The instrument itself comes from Bruker. I am the person here who arranges the demonstration, the measurement on your own material, and the configuration that matches what you actually have to report.
If you are working out which system your monitoring programme needs, send me a filter from a real sample. We measure it on the configuration you are considering, you get the particle list with polymer classes and sizes, and I tell you plainly which fraction fell outside what the technique can resolve. It answers the question a specification sheet cannot: does this instrument do what my reporting obligation asks, on my water.
Arrange a measurement on your own filter and tell me the size range you have to report to and which system you are considering.
Related reading: what ISO 16094-2:2025 means for water laboratories, the application note on detection and analysis of microplastics, and Bruker's guide to microplastics analysis.