“Has my active stayed in the expected form after drying? Has this hydrate dehydrated during milling? Has my amorphous phase started to recrystallise?” Three questions about how matter is arranged, not about what it is made of.
Two forms of the same molecule have the same elemental composition. What differs between them is the lattice: the way the molecules stack. And with the lattice go solubility, dissolution rate, storage stability and sometimes compressibility. A change of form does not show up in an assay.
X-ray powder diffraction remains the reference: it is what identifies a form, and its value is the one that counts. It calls for a sample, a preparation and an equipped laboratory.
What is missing between two samples is when the transformation happens: during drying, during milling, at compression, or later in the packaging.
In short
RAMAN tells two known forms apart with a library and a threshold; quantifying a proportion requires a full model. TERAHERTZ sees the lattice vibrations themselves, including through closed packaging. X-ray powder diffraction remains the reference that identifies the form.
What is actually measured
A form, not a content
The question is not “how much is there” but “which one is it”. The result is an identification, or a proportion between two known forms. That changes the nature of the validation: you demonstrate an ability to discriminate before you demonstrate accuracy.
The lattice is read at low frequencies
The vibrations that move the whole unit cell — rather than one isolated bond — sit at the low-frequency end of the spectrum. That is where two polymorphs separate most clearly, while the bands of individual bonds stay almost superposable.
Amorphous or crystalline: a matter of disorder
An amorphous phase has no ordered lattice: its bands broaden and the lattice lines disappear. The transition is therefore followed as a continuum, which is what allows a partial recrystallisation to be caught before it shows up any other way.
The criterion is a spectral difference, not a reference value to predict. To tell two known forms apart, a library and a threshold are usually enough: there is no campaign of samples to assay, because there is nothing to assay.
A model becomes necessary again as soon as you want a proportion — ten per cent of form B in form A — and it then calls for prepared mixtures of the two pure forms, in known amounts. Why some measurements call for a model and others do not.
The limit of detection of a minor form is established on your product, by trial. It depends on the spectral contrast between the two forms, and no generic value exists.
Which technology for which case
| Technology | What it sees of the lattice | Where it sits | Calibration burden | Known condition |
|---|---|---|---|---|
| RAMAN | Lattice modes, at low frequencies, and the band shifts that come with a change of packing | Immersed probe, contact-free measurement, or a portable through a vial | None to discriminate two known forms: a library and a threshold. Full for a proportion | Fluorescence of the matrix. Local heating can itself trigger a transition on a sensitive product |
| TERAHERTZ | The lattice vibrations themselves, in the band where they sit | Contact-free, including through closed packaging | A model to identify or class | Figures of merit are published, in the laboratory: quantification of polymorphism and of degree of crystallinity by terahertz time-domain spectroscopy. What remains to be established by trial is the transfer in line, on your product |
| NIR spectroscopy | Nothing directly. It sees the effect of the lattice on hydrogen bonding and on light scattering | In line, in contact or contact-free, where NIR is already installed | Full: it is a classification learned on batches of both forms | The response is indirect. A change in particle size can imitate a change of form, and that is the main trap |
| X-ray powder diffraction | The distances between crystal planes: the direct description of the lattice | In the laboratory, on a prepared sample | None to identify | Off line. It is the reference: it is what will qualify and then verify the in-line measurement |
In pharma, X-ray powder diffraction is covered by a general chapter in both pharmacopoeias: Ph. Eur. 2.9.33 and USP <941>. Among the three in-line routes, RAMAN (<858>) and NIR (<856>) have an enforceable chapter at the USP; TERAHERTZ does not.
What is worth preparing on your side
A measurement of form is prepared differently from a measurement of content. Three things decide feasibility, and they are to be gathered before the trial.
- Samples of both forms, pure. That is the first condition. Without an isolated form B there is nothing to discriminate, and no library to build. Obtaining it is sometimes the real subject of the project.
- Their characterisation by the reference method. X-ray powder diffraction establishes what those samples are. That value is what will later qualify the in-line measurement, and then verify it over time.
- The conditions that trigger the transformation. Temperature, humidity, mechanical energy, the presence of solvent. Being able to reproduce them is what makes it possible to manufacture, in trial, the cases the measurement will have to recognise in production.
If form B has never been isolated, a feasibility trial starts there rather than with the instrument.
Conditions for success, and limits
Spectral contrast decides, and it is observed
Two very close forms give very close spectra. No general rule says in advance whether the difference will be enough: it is measured on your samples, during a feasibility trial. That is the first question to settle, and it is settled by experiment.
The measurement can alter what it observes
A focused laser heats the point it lights. On a product whose transition sits close to ambient conditions, that heating can be enough to trigger what you are trying to detect. Power and integration time are then set accordingly, and it is checked.
A minor form is hard to detect at low level
Detecting one per cent of a second form is not of the same order as detecting twenty. The useful threshold is established on your prepared mixtures, and it governs the value of the project: a measurement that detects at twenty per cent often arrives after the decision.
The sample seen is not the batch
A measurement spot probes a few milligrams. On a heterogeneous transformation, one that starts at the surface or in one zone of the bed, the question of representativeness comes before that of sensitivity.
What it changes, in practice
A change of form caught in production rather than at the final check moves the decision. The batch is not finished; the step responsible is still identifiable; and the cause is still observable.
On a drying step that means being able to stop before a hydrate dehydrates rather than noting it afterwards. On milling, being able to set the energy applied rather than take the amorphisation it produces. On storage, knowing that a recrystallisation has started before it reaches the level that matters.
In a dossier it is also the difference between a form presumed stable and a form whose stability has been observed over the duration of the process. That difference defends itself in front of an auditor.
Frequently asked questions
Can X-ray diffraction be replaced by an in-line measurement?
No, and that is not the point. Diffraction remains the method that identifies a form and that counts. The in-line measurement answers a different question: when the transformation occurs, and on which setting it starts. The two work together — one qualifies the other, then verifies it over time.
How many forms do you need to know before starting?
Two are enough to begin: the one you want, and the one you fear. A library is then completed form by form, as they are isolated. What does not work is looking for a form you have no sample of: the measurement does not discover what it has not been shown.
Can it measure through the container?
In RAMAN, through a transparent or translucent vial, often yes. In TERAHERTZ, through closed packaging, including packaging opaque to the eye — that is its particularity. Metal, on the other hand, stops the measurement in both cases.
What sensitivity can be expected?
It depends entirely on the pair of forms and on the matrix, and no generic value means anything here. It is established on your samples, with mixtures prepared in known proportions. That is precisely what a feasibility trial has to produce: a figure on your product, not a promise.
Tell us which forms you watch, and at which step they switch. That is where it starts.
Forty-five minutes is enough to know whether your two forms separate, what samples the trial would call for, and at which step of the process the measurement would serve best. Where the contrast is too weak, we will say which route to look at instead.