Checking without opening, without destroying, without a radiation protection area. Let us say it from the outset: this is an end-of-line and goods-in need, not a process steering need. The rest of this page only makes sense on that condition.
On receipt, verifying what is inside a sealed container means opening it. Opening means breaking the integrity, starting a cleaning, sometimes destroying the unit. So little is sampled, and much is extrapolated. On batches of several tens of thousands of units.
At the end of packaging, a blister with a broken tablet or an empty pocket can be seen. When someone looks at it. Visual inspection tires, and X-ray inspection imposes a controlled area, dosimetry, trained staff. Many sites give it up not for technical reasons, but because the operating constraint is too heavy for the line concerned.
And on finished products, structure inspection remains destructive: you cut, embed, polish, observe under the microscope. The result is excellent, it arrives days later, and the unit examined no longer exists.
In short
TERAHERTZ checks the presence, absence and integrity of the contents through non-metallic packaging, without radiation protection. OCT describes the structure 100 to 200 µm below the surface, but does not go through packaging. Neither detects metals: that is the role of X-rays.
What is actually measured
Two very different families answer this need. One goes through the packaging — and reads two things there. The other looks just below the surface. They are not in opposition. They do not observe the same depth.
Going through the packaging
TERAHERTZ radiation goes through transparent, translucent and even opaque materials: cardboard, plastic, film, foam. Except metal, which reflects it. That is what makes it possible to observe the content without opening. It is also the first question to ask about your packaging. Does it contain aluminium, a metallised lidding, a metallic barrier layer?
An echo on the internal interfaces
Each change of medium sends back part of the signal. The return time of these echoes gives the position of the interfaces, and therefore layered structure information. On tablets, the demonstration exists: pulsed terahertz imaging was validated against cross-section microscopy as early as 2007, then against X-ray microtomography in 2012, and an in-line sensor measured the coating of individual tablets in real time in a pan coater in 2011. What remains to be established is the transposition through an industrial packaging, where spectral distortion changes the terms of the problem.
A structure just below the surface
Optical coherence tomography gives the internal morphology over 100 to 200 µm below the surface. It also gives a surface roughness descriptor, whose use is still being validated. The quantity comes from the physics of the measurement: no multivariate model to build or maintain. In return, it does not go through a packaging. It looks at an accessible object.
The argument for TERAHERTZ against X-rays is harmlessness, and therefore operability, never detection performance. It is the most important nuance on this page, and the one most often distorted by the market.
The figure it rests on: a photon at 1 THz carries about 4 meV, against 26 meV of thermal agitation at 300 K. Less energy than what already reigns in the material. Neither ionising nor dissociating effect, and therefore no radiation protection: no controlled area, no dosimetry, no installation constraint.
And what this does not say. TERAHERTZ does not detect metals and does not replace X-rays in detection. It is non-metals that are transparent to it. Metal, for its part, is a wall. A site looking to detect a metallic foreign body will not find its answer here, and we will not sell it one.
Which technology for which case
| Technology | What it follows | Suitable matrices | Calibration burden | Installation | Known condition |
|---|---|---|---|---|---|
| TERAHERTZ | Presence, absence, integrity of the content. Layered structure. Substance identification | Transparent, translucent or opaque non-metallic packaging, dry solids | Full for an identification. None for a presence or absence check | Contact-free, no radiation protection, no controlled area | Works on non-metals. Strong spectral distortion through packaging |
| Optical coherence tomography OCT | Internal morphology over 100 to 200 µm below the surface, roughness, layer thickness | Accessible solids, coatings | None: direct physical measurement | Contact-free, at-line or in-line | Looks at an accessible surface. Probe to sample distance held to within a few tens of microns |
| X-ray inspection (for reference) | Dense foreign bodies, metallic ones included. Internal structure in depth | All matrices, metallic packaging included | None most of the time | Controlled area, radiation protection, dosimetry, authorised staff | The operating constraint rather than the performance, and it is what sets the solution aside on many lines |
X-ray inspection appears in this table although it sits outside what we deploy. It is the only honest way to set out the arbitration. On detecting a metal it stays the answer. On a line where radiation protection is out of reach, it is not.
Where the check sits, and why it is end of line rather than steering
This family of measurements goes where the product is already formed, often already packed, rather than into a reactor or a dryer. It is a boundary of perimeter, and it is better known early.
| Position | What it allows |
|---|---|
| On receipt | Verifying a content with the integrity of the container intact, and so checking more units for the same effort. The case where the sampling gain is the most direct |
| End of packaging | Presence, absence and integrity checking on the line, with no controlled area. The achievable rate is the first point to establish |
| At-line | Structure or coating checking on samples, in seconds, with no preparation and the unit intact. The usual position for optical coherence tomography |
| Laboratory | Stays the reference: section, microscopy, destructive methods. It is what every non-destructive method is built and verified against |
What is worth preparing on your side
- The exact composition of the packaging, layer by layer, thickness included. An aluminium lidding or a metallised barrier layer stops TERAHERTZ radiation. It is a governing criterion, and it is verified on a data sheet.
- Real defective units, in sufficient number: broken tablet, empty cavity, seal defect. A check is set up against what it has to find, and artificially made defects do not always resemble production ones.
- The written definition of the defect to detect: minimum size, position, and above all the acceptable false positive rate. A check sensitive enough to reject 2 per cent of conforming units costs more than the defect it hunts.
- The target line rate and the surface to cover. That pair drives the architecture of the system, much more than the choice of technology.
- Time for the data processing. The available sources are explicit here: careful data preparation is essential for reliable identification, and it is a substantial part of the work.
Conditions for success and limits
TERAHERTZ passes through non-metals
Worth repeating, because the reverse circulates. No serious technical source claims metal detection by TERAHERTZ, and the physics runs the other way: it is the non-metals that are traversed. Its advantage over X-rays is harmlessness, and so the ability to deploy almost anywhere.
Water content and polymorphism: possible in principle
Those two applications are announced as achievable by the manufacturer ; the measurement range, the prediction error and the coefficient of determination are established by trial. The laboratory literature does publish them, for water in a solid as for polymorphism; what remains to be established is the transfer to a packed object. The wording used is feasibility in principle, and it stays in the conditional. That is a statement of feasibility rather than a qualified method, so we engage it as a trial.
The physical conditions are known and acknowledged
Polar liquids, water first among them, are the demanding case. The spectral lines of solids are broadened, which reduces selectivity. And through a pack there is strong spectral distortion, from water absorption and scattering. All three points are stated by the technical sources themselves, and they are why a laboratory result is transposed to a packed product by measurement rather than by assumption.
Four figures worth asking for
Lateral resolution, depth resolution, scanning speed and maximum measurable thickness depend on the geometry and the material. They are the questions to carry into a requirements specification with the real geometry and material of your product. An answer on your sample is worth the trip.
What it changes, in practice
Three pieces of evidence, and they must be read for what they are, no more, no less.
Blister inspection is demonstrated. A broken tablet and a missing tablet are clearly identified through the packaging, at 100 GHz and at 300 GHz. It is a concrete result, on a concrete application, and it is the best entry point for a first trial.
Substance separation, we established on our own trials, outside a pharmaceutical context: seven substances separated by principal component analysis. The data belong to the client and we publish no performance figure from them. What transposes is a discrimination capability, not a performance. Transposing it as it stands to an identification through a packaging would be abusive: that is precisely where the spectral distortion mentioned above comes in. The only performance that counts is the one measured on your units, in your packaging.
Going through a packaging is documented, in open access. A substance was identified inside a closed envelope, through eight and then sixteen sheets of paper. At normal incidence, internal reflections cover its lines; at the Brewster angle, the false-positive rate is lowered by a factor of about 14 for para-aminobenzoic acid and 10 for lactose monohydrate, at a correlation threshold of 0.2. What this result establishes is geometric: the packaging can be crossed, provided the angle is controlled. Paper is not a blister, and the transfer is demonstrated by a trial.
Source of the third point: Molter et al., “Mail Inspection Based on Terahertz Time-Domain Spectroscopy”, Applied Sciences 2021, 11(3), 950, doi:10.3390/app11030950, open access. This is postal inspection, not a pharmaceutical application — hence the caution about the transfer.
OCT tomography, for its part, brings direct value. A structure over 100 to 200 µm below the surface and a roughness, with no model to build or maintain, on a product that is not destroyed. Its axial resolution, of the order of a micrometre, makes it the finest tool for thin coatings.
That does not make it “the” coating technology, and terahertz is not excluded. On tablet coating, the reference literature — often signed by the same authors for both techniques — explicitly calls them complementary, and the division is made on thickness: OCT suits coatings from 10 to 60 µm, terahertz imaging from 40 µm and beyond. The first has the fineness, with an axial resolution of 0.9 µm, but near-infrared light is scattered by the matrix and by pigments. The second has the penetration and the indifference to pigments, at the cost of an axial resolution of 30 to 40 µm. Used together on the same pan coater, they covered from 20 µm to more than 280 µm.
Lin, Zhang, Markl, Zeitler, Shen, Applied Sciences 2018, 8(12), 2700, doi:10.3390/app8122700, open access. The right question is therefore not which one is the reference, but what thickness you need to see — and, on this page, through what you need to see it: that is where terahertz remains the only candidate. The full trade-off, on the coating side.
Frequently asked questions
Does it replace X-ray inspection?
The two answer different questions. TERAHERTZ works on non-metals, and its detection capability is not presented as superior to that of X-rays. Its advantage is being non-ionising, and so deployable with no controlled area and no dosimetry. Where your problem is a metallic foreign body, X-ray inspection or a dedicated metal detection stays the answer. Where your problem is that an X-ray installation is impractical on that line, the conversation becomes interesting. In the food industry, it is your food safety plan, under Regulation (EC) No 852/2004, that sets the detection level to reach and the conduct on a deviation: no analytical chapter replaces it.
Can we check through a blister with an aluminium lidding?
Metal reflects TERAHERTZ radiation, so the content is reached another way there. It is a governing criterion, verified on the packaging data sheet before any technical discussion. On an entirely polymer or cardboard pack, the question is open again.
Can a water content be measured through the packaging?
It is announced as possible in principle, and that is where the published evidence stands: no measurement range, no prediction error, no coefficient of determination to date. So we present it as a trial to run, with an uncertain result and a perimeter defined in advance. Where you need a qualified moisture measurement, other routes are better established today. Measuring moisture content in-line.
What rate can be held at the end of a line?
Scanning speed is not among the published specifications, and we will not invent a figure. It is one of the four questions to carry into your requirements specification, with lateral resolution, depth resolution and maximum traversable thickness. The answer that counts is the one obtained on your units, in your geometry, during a trial.
Can a process be steered with these technologies?
TERAHERTZ works on a formed and often packed product, so downstream. OCT comes closer to the process and is used in-line on coating operations. For steering a blend, a drying or a reaction, the answer is in another family, and we will say which rather than stretching the perimeter of these two. The nine measurement families.
How do we know the method works on our product?
Through a trial on your real units, with your real defects, in your real packaging. A documented conclusion, that the packaging attenuates too much, that the spectral distortion limits the discrimination, that the resolution works above the defect sought, is a useful deliverable and a great deal lighter than an installation that would call for rework.
Send us the description of your packaging and of the defect to find. You will know whether it is visible.
Forty-five minutes is enough: whether your packaging can be crossed, whether the defect sought is at the scale of what the technology sees, which position in the line would suit, and which questions to ask about the unpublished specifications. Some cases call for another route — metallic lidding, a defect too fine, a metallic foreign body to detect: you will then leave with the one that fits.