The most common error in a Process Analytical Technology presentation that includes a focused beam reflectance probe is the moment the speaker calls the output a particle size distribution. It is not. The probe returns a chord-length distribution, a different quantity, and the two are related by a geometric transform that depends on assumptions the operator rarely wants to defend in front of a review chemist. The acronym has been in the field for thirty years, the technique is a workhorse of pharmaceutical crystallization and polymer process monitoring, and the chord-length vocabulary is settled in the peer-reviewed literature. The slippage happens in slide decks, training material, and informal conversation, and it is the slippage that gets a calibration story flagged on inspection.
This explainer sets out what the probe actually measures, where the signal is genuinely useful, and where treating chord length as particle size introduces an error large enough to matter. It assumes a working familiarity with the inline, online, at-line, off-line vocabulary and with the general framing of process analytical technology.
What the probe does
A focused beam reflectance measurement probe is a window-tipped optical assembly inserted directly into a slurry, emulsion, or crystallizing solution. Inside the head, a laser is focused to a small spot and the focal point is swept across the window face in a circle at a known and constant tangential velocity, typically of order a few metres per second. Each time the focal spot crosses a particle close enough to the window to scatter coherent light back into the collection optics, the detector records a reflectance pulse. The duration of the pulse, multiplied by the known velocity, is recorded as a chord length: the straight-line distance the beam traversed across that particle during that single scan event.
The probe accumulates tens of thousands of these chord events per second and bins them into a histogram. That histogram is the chord-length distribution. It is updated continuously as the process evolves, with no sampling, dilution, or operator intervention. The instrument is at its most useful where this real-time, undiluted, in-situ signal is structurally hard to get any other way.
Why chord length is not particle size
A chord is not a diameter. A chord is the random line the laser happens to draw across a particle on a given scan, and the length of that chord depends on three things: the size of the particle, its shape, and the random offset between the beam path and the particle centre. For a perfect sphere of fixed diameter, the chord-length distribution is itself a distribution of values smaller than the diameter, with most chords shorter than the actual sphere and only a few approaching the maximum. For a population of spheres with a real size distribution, the observed chord-length distribution is the convolution of the size distribution with this offset-statistics kernel.
For non-spherical particles - the case in almost every interesting process - the convolution depends on the particle’s shape and on its orientation when the beam crosses it. A needle-shaped crystal looks short or long depending on which way it is pointing at the moment of measurement. Vendor software offers chord-to-size inversion routines that assume a shape model and a calibration to an external particle sizer, but the inversion is only as good as the shape assumption, and the assumption is the part of the procedure the review chemist will ask about.
The practical posture, settled in the chemical engineering literature and reflected in the way operators run their charts, is to use the chord-length distribution directly and to leave the inversion alone. Moments of the chord distribution - mean chord, square-weighted mean chord, counts per second above a threshold - are stable, easy to interpret, and not contingent on a shape assumption. The full chord-length distribution is the diagnostic record; the moments are the trend variables.
Where the signal is useful
In a cooling crystallization, the square-weighted mean chord rises as crystals grow and falls when a secondary nucleation event sprays the vessel with fines. The instrument resolves both phenomena clearly enough that the chord trace is a routine input to the supervisory control logic. In a fluidized-bed granulation, the chord distribution plateaus at the granulation endpoint, and an operator can read the plateau directly off the trace without offline sampling. In a flocculation step, the mean chord rises monotonically as floc structures aggregate, and the rate of rise is a reliable kinetic indicator. In an emulsion polymerization, the count rate above a chord threshold tracks the onset of coalescence.
The common thread is that all four applications are interested in change, not in the absolute size. The chord-length distribution is at its strongest as a relative process-state monitor and at its weakest when an operator tries to use it to certify a release attribute. A separate offline particle sizer - laser diffraction by ISO 13320, dynamic image analysis, or a wet-sieve fractionation - is the reference, and the chord trace is the in-process indicator that the process is or is not behaving as the reference characterised it.
Practical limitations
Window fouling is the first failure mode. A caked deposit on the sapphire window blinds the probe progressively; the chord trace looks stable while the underlying signal has decayed. Most pharmaceutical installations schedule probe cleaning at fixed intervals and supplement that with a daily verification against a reference suspension.
Refractive-index contrast is the second. Particles whose refractive index is close to that of the suspending medium scatter weakly, and the discrimination threshold of the probe may register them poorly or not at all. Organic crystals in organic solvents are the classic awkward case; aqueous slurries of inorganic crystals are easy.
Probe-tip sample heterogeneity is the third. The chord distribution describes the slurry within a few millimetres of the window. A poorly mixed vessel can have a probe-tip composition that diverges from the bulk, and the trace will then be locally honest but globally misleading. Probe placement, vessel mixing, and trace verification against grab samples are the routine countermeasures.
A fourth and underappreciated point is the lower size limit. Chord lengths below the laser spot size, of order one micrometre with the standard optics, are not resolved. Sub-micrometre populations - many nanoparticulate suspensions, very fine fines - sit below the probe’s discrimination threshold and contribute to the count rate only as a diffuse background.
What this means for the calibration story
A chord-length trace rarely sits alone in a modern PAT installation. It pairs with a spectroscopic measurement - near-infrared for moisture and end-point in granulation, Raman for polymorph identity in crystallization, attenuated total reflectance infrared for chemistry in emulsion polymerization - and the chemometric model on the spectrometer side carries the chemistry, while the chord trace carries the physical state. The two streams are independent and complementary, and the validation file documents them as such.
The regulatory framing is straightforward. ICH Q14 expects the analytical procedure file to describe what the instrument measures, with the vocabulary the standards body uses. ASTM E2891 sets out the documentation expectations for the multivariate side. The FDA PAT framework guidance leaves the choice of measurement to the operator but is firm that the choice has to be justified against what the process needs. Calling the chord-length distribution a particle-size distribution in any of those documents is the kind of phrasing that wastes the chemist’s time at the first round of review. Using chord-length vocabulary, citing the supervisory particle sizer separately, and documenting the chord-to-size inversion only where it is used - and on what assumptions - is the unglamorous procedural discipline that keeps the file clean.
Closing
The focused beam reflectance probe has earned its place in pharmaceutical crystallization and in polymer process monitoring because the chord-length distribution is a stable, undiluted, real-time signal in conditions that defeat both laser diffraction and image analysis. It is at its most useful when the operator treats it as a relative process-state monitor and writes the documentation in chord-length terms throughout. It is at its most awkward when a slide deck recasts the output as a particle-size distribution and a reviewer asks where the shape assumption comes from. The right vocabulary is in the literature; the work is to keep it in the slide deck.