Analytic Journal is growing – your feedback shapes the platform.Send feedback
Category: Spectroscopy | PAT | PAT | Measurement Principles
Author: Dusko Kadijevic, Publisher Analytic Journal
Last updated: May 2026
Reading time: approx. 10 minutes
Target audience: Process engineers, PAT specialists, quality managers, plant designers in pharmaceutical, chemical and food manufacturing
NIR (Near-Infrared Spectroscopy) and Raman Spectroscopy are two optical measurement principles used in PAT for qualitative and quantitative characterisation of substances. Both work without physical contact and provide molecular information — but differ fundamentally in their physical interaction with the sample, the bond types they detect, and their suitability for different process environments.
The most common source of error when choosing between NIR and Raman is the wrong starting point: many projects begin by comparing technologies before the measurement task is clearly defined. This leads to projects that fail — not because of poor technology, but because the wrong questions were asked.
The right starting question is not "NIR or Raman?" — but: "What measurement problem needs to be solved?"
Only once the target variable, sample matrix, process conditions, and requirements for speed and accuracy are clear does the technology decision become obvious.
NIR measures the absorption of light in the wavelength range of approximately 780 to 2500 nm. The light excites overtones and combination vibrations of bonds containing hydrogen atoms — primarily C-H, O-H and N-H bonds. Absorption is relatively weak, which is why NIR is well suited for measuring major components at concentrations above approximately one percent.
NIR typically operates in reflectance mode for solids and powders, and in transmission mode for liquids in flow cells.
Raman is based on a different physical effect: a laser beam strikes the sample, and a small fraction of the light is inelastically scattered — with a frequency shift characteristic of the specific molecular vibration. Raman is particularly sensitive to C=C, C≡C and S-S bonds, as well as symmetric vibrations that are infrared-inactive.
Water shows only a weak signal in the Raman spectrum — this is the decisive practical advantage over NIR in aqueous samples.
| Property | NIR | Raman |
|---|---|---|
| Physical effect | Absorption of near-infrared light | Inelastic light scattering (Raman effect) |
| Sensitive bond types | C-H, O-H, N-H | C=C, C≡C, S-S, symmetric bonds |
| Signal strength | Medium to strong | Weak (approx. 1 in 10⁷ photons) |
| Water interference | Strong — broad O-H absorption | Low — water is Raman-inactive |
| Typical concentration range | > 1% (major components) | < 1% possible (trace components) |
| Sample volume | Large (cm³ range) | Small (µm³ range) |
| Measurement through containers | Limited | Possible through glass and transparent plastics |
Selection starts at the molecular level. Which bond types and molecular properties need to be captured?
| Measurement task | NIR | Raman |
|---|---|---|
| Moisture / water content | ✅ | ⚠️ |
| Proteins, fats, carbohydrates | ✅ | ⚠️ |
| Polymorphs and crystal structures | ⚠️ | ✅ |
| Aqueous solutions, API in water | ❌ | ✅ |
| Aromatics, conjugated systems | ⚠️ | ✅ |
| Total content of major components | ✅ | ⚠️ |
| Trace components < 1% | ❌ | ✅ |
| Blend homogeneity in powders | ✅ | ⚠️ |
NIR requires a larger sample volume — an advantage with inhomogeneous materials such as powder blends, because more material is captured and statistical representativeness increases. Raman measures with microscopic precision — advantageous for stratified samples or spatially resolved measurements.
| Sample characteristic | NIR | Raman |
|---|---|---|
| Large-volume powders and bulk materials | ✅ | ⚠️ |
| Inhomogeneous solids | ✅ | ⚠️ |
| Liquids in flow cells | ✅ | ✅ |
| Measurement through glass reactor or container | ❌ | ✅ |
| Spatially resolved measurement (µm range) | ❌ | ✅ |
| Multiphase systems | ❌ | ✅ |
Note on optical spot size: NIR and Raman differ considerably in the diameter of the optical measurement point — and therefore in the actual amount of sample being analysed. This has significant consequences for sample representativeness and measurement strategy. This aspect is covered in depth in a separate Analytic Journal article.
Practical note on multiphase measurement: Raman has proven to be the superior method for multiphase systems in practice. In organic-aqueous two-phase systems, both phases can be monitored simultaneously. From a practical field report shared by a colleague in process analytical technology development at a major German chemical company, Raman was also applied successfully in three-phase systems — solid, organic liquid and aqueous phase simultaneously — tracking both reactants and products in a stirred tank reaction. NIR is not suitable for these tasks due to strong water interference and limited phase selectivity.
Both methods have characteristic interference sources that must be considered in selection:
| Interference source | NIR | Raman |
|---|---|---|
| Water | ❌ Strongly interfering | ✅ Barely interfering |
| Fluorescence | ✅ Not affected | ❌ Can completely mask signal |
| Particle size / scattering | ⚠️ Affects spectrum | ✅ Low influence |
| Temperature changes | ⚠️ Influences spectrum | ⚠️ Influences laser power |
| Ambient light | ✅ Non-critical | ⚠️ Shielding required |
| Coloured or dark samples | ⚠️ Conditionally suitable | ❌ Fluorescence likely |
Fluorescence — the Achilles heel of Raman
Fluorescence is the most significant weakness of Raman spectroscopy. When a sample fluoresces under laser irradiation, the fluorescence signal can completely overwhelm the considerably weaker Raman signal — rendering the measurement unusable. NIR is fundamentally unaffected by this problem because it is based on absorption, not on weak inelastic scattering.
A well-known workaround is the use of lasers in the NIR range (typically 1064 nm) instead of the more common visible lasers (532 nm, 785 nm). Longer-wavelength lasers excite fluorescence significantly less — making the problem manageable in many cases. The trade-off: the quantum yield of the Raman effect decreases considerably with increasing wavelength. This means weaker signals, longer acquisition times and a lower signal-to-noise ratio. It is a direct trade-off between fluorescence suppression and spectroscopic sensitivity.
In practice: before any Raman installation, the fluorescence tendency of the sample and potential interfering substances should be systematically tested — and the laser wavelength selected deliberately for the specific application.
Practical note from the field — two separate problems at a Raman installation in organic synthesis:
Problem 1 — Probe window fouling: In the early phase of the installation, Raman spectra deteriorated progressively. The cause was fouling of the probe windows by process deposits. The solution was implementing a regular cleaning cycle — an internal purge integrated into the probe design. Without this measure, the analyser would not have been viable for sustained continuous operation.
Problem 2 — Thermal oil breakthrough: Later, a thermal oil breakthrough from the heat exchanger into the process occurred. The thermal oil entered the reaction medium in trace quantities — invisible to the eye and to conventional process monitoring. The Raman system detected the breakthrough considerably faster than the operators could through manual sampling from the running reactor — identified by a characteristic fluorescence signal attributable to the thermal oil. What initially appeared to be an instrument failure was in fact an early warning of a process deviation that would otherwise have gone undetected for considerably longer.
The lesson from both cases: Probe maintenance and process integrity are just as critical in Raman applications as the spectroscopic performance of the instrument. And sometimes an unexpected signal is not a fault — it is the most valuable measurement of the day.
NIR is the faster method — measurement times below one second are achievable. Raman typically requires several seconds to one minute depending on signal strength. In slow batch processes this difference is usually irrelevant — in continuous high-speed processes it can be decisive.
| Speed requirement | NIR | Raman |
|---|---|---|
| Measurement time | 0.1 – 2 seconds typical | 2 – 60 seconds typical |
| Continuous high-speed processes | ✅ | ⚠️ |
| Batch processes on hour timescales | ✅ | ✅ |
| Endpoint detection in slow reactions | ✅ | ✅ |
Capital cost is only one part of the total cost equation. Calibration effort, maintenance, probe service life and regulatory requirements determine the actual operating costs over the lifecycle.
| Cost factor | NIR | Raman |
|---|---|---|
| Typical capital cost for process instrument | Lower | Higher |
| Laser maintenance required | No | Yes |
| Probe wear in harsh environments | Lower | Higher (windows, seals, fibres) |
| Calibration model development effort | Comparable | Comparable |
| Regulatory acceptance | Very high (20+ years track record) | Growing, but less mature |
| Spectral preprocessing required | More intensive (scatter correction) | Less intensive |
Note: Cost figures are indicative. Current pricing and configurations should be obtained directly from manufacturers.
In demanding PAT applications, NIR and Raman do not have to be mutually exclusive. Some of the most effective implementations combine both methods deliberately because they provide complementary information:
Example — API crystallisation:
Example — tablet manufacturing:
When is combined use appropriate?
Legend: ✅ Preferred | ⚠️ Conditionally suitable | ❌ Not suitable
| Application | NIR | Raman | Note |
|---|---|---|---|
| Blend homogeneity in powders | ✅ | ⚠️ | NIR preferred — larger sample volume |
| Polymorph control | ⚠️ | ✅ | Raman superior for crystal structure |
| Moisture in granules / powders | ✅ | ❌ | NIR — O-H bond directly measurable |
| API in aqueous solution | ❌ | ✅ | Raman — water does not interfere |
| Reaction monitoring through glass reactor | ❌ | ✅ | Raman — measurement through glass possible |
| Fermentation online | ⚠️ | ✅ | Raman preferred in aqueous matrix |
| Tablet coating control | ⚠️ | ✅ | Raman — spatially resolved, through coating |
| Oil and fat analysis | ✅ | ⚠️ | NIR established, broad data base |
| Aromatics content — petrochemicals | ⚠️ | ✅ | Raman sensitive to C=C bonds |
| Raw material identification | ✅ | ✅ | Both suitable — matrix dependent |
| Trace components < 0.1% | ❌ | ✅ | Raman more sensitive at low concentrations |
| High-speed continuous measurement | ✅ | ⚠️ | NIR faster |
| Fluorescent samples | ✅ | ❌ | NIR unaffected by fluorescence |
| Multiphase systems organic/aqueous | ❌ | ✅ | Raman can capture both phases simultaneously |
| Solid/liquid multiphase systems | ❌ | ✅ | Raman documented for 3-phase systems |
Practical note: The technology decision accounts for approximately 20 percent of project success. The sampling strategy accounts for 80 percent. An expensive Raman system can fail at a poorly chosen sampling point. A straightforward NIR system can deliver positive ROI within four months when sampling is designed in accordance with the Theory of Sampling (TOS). The best probe only measures what it receives. If the sample is not representative, no technology will compensate.
Typical errors in technology selection:
In pharmaceutical manufacturing, NIR and Raman are both recognised analytical techniques under FDA guidance. The FDA PAT Guidance (2004) explicitly supports spectroscopic process monitoring as part of a science- and risk-based approach to pharmaceutical development and manufacturing.
USP <1119> governs NIR spectroscopy for pharmaceutical applications. USP <858> covers Raman spectroscopy. Both are subject to ICH Q2(R2) for method validation. For data integrity in GMP environments, 21 CFR Part 11 applies to all electronic records generated by process spectroscopic systems.
In the UK, pharmaceutical spectroscopic methods are governed by the European Pharmacopoeia (EP) standards retained under UK law: EP 2.2.40 for NIR and EP 2.2.48 for Raman. MHRA GMP guidelines apply to method validation and data integrity requirements. Post-Brexit, UK manufacturers must comply with MHRA requirements independently of EMA — though in practice EP methods remain the technical reference.
For industrial and environmental applications outside pharma, no specific UK or US regulation mandates NIR or Raman — but method validation to ISO and ASTM standards is standard practice for any process analytical measurement used for quality decisions or regulatory reporting.
| Standard / Regulation | Issuer | Content |
|---|---|---|
| ASTM E1655 | ASTM International | Test practices for multivariate infrared quantitative analysis — foundation for NIR calibration models |
| ASTM E1840 | ASTM International | Guide to Raman spectroscopy — fundamentals, instrument requirements, applications |
| USP <1119> | USP | NIR spectroscopy in pharmaceutical applications — instrument and method requirements |
| USP <858> | USP | Raman spectroscopy — pharmaceutical qualification requirements |
| EP 2.2.40 | EDQM | NIR requirements in the European Pharmacopoeia (UK retained) |
| EP 2.2.48 | EDQM | Raman requirements in the European Pharmacopoeia (UK retained) |
| ICH Q2(R2) | ICH | Validation of analytical procedures — applies to spectroscopic PAT methods |
| FDA PAT Guidance (2004) | FDA | Framework for PAT implementations in pharmaceutical manufacturing |
| 21 CFR Part 11 | FDA | Electronic records and data integrity — applies to spectroscopic GMP systems |
| MHRA GMP Guidelines | MHRA | UK equivalent to FDA GMP — method validation and data integrity |
What is the fundamental difference between NIR and Raman?
NIR measures the absorption of near-infrared light by hydrogen-containing bonds — primarily C-H, O-H and N-H. Raman measures the inelastic scattering of laser light and is particularly sensitive to C=C, C≡C and S-S bonds and symmetric molecular vibrations. Water shows strong absorption in NIR but a weak signal in Raman — this is the decisive practical difference in aqueous applications.
When is NIR better than Raman?
NIR is better suited for moisture, proteins, fats and carbohydrates in solid or liquid samples without high water content. NIR is faster, more robust against fluorescence, has lower capital costs and a broader regulatory track record. For blend homogeneity monitoring in large-volume powder mixers, NIR is often the better choice because of its larger sampling volume.
When is Raman better than NIR?
Raman is better for aqueous systems, polymorph control, measurements through glass containers, and trace components below one percent. In fermentation monitoring and API measurements in aqueous formulations, Raman has a clear advantage. Raman is also suited for spatially resolved measurements in coatings and layered structures, and for multiphase systems where NIR is severely limited.
What problems can fluorescence cause in Raman?
Fluorescence can completely mask the Raman signal and render the measurement unusable. It occurs particularly with coloured, dark or contaminated samples. Even trace quantities of foreign substances — such as oils from seals or lubricants — can trigger fluorescence, which may simultaneously interfere with measurement and serve as an early warning of process contamination. NIR is unaffected by fluorescence.
What is TOS and why is it critical for both methods?
TOS stands for Theory of Sampling — the scientific foundation for representative sampling. Both NIR and Raman can only deliver results as good as the sample they receive. Non-representative sampling renders even the best spectrometer worthless. In practice, PAT projects fail more often due to sampling than due to spectroscopy.
Can NIR and Raman be used in combination?
Yes — and in critical processes this is often the most effective strategy. NIR and Raman provide complementary information: NIR captures major components, moisture and physical properties of large sample volumes; Raman delivers crystal structure, trace components and spatially resolved data. In tablet manufacturing or API crystallisation, the combination of both methods can produce a complete process picture that neither method alone could provide.
How demanding is calibration for NIR and Raman?
The model development effort is comparable for both — it depends more on the complexity of the measurement task than on the method itself. NIR often requires more spectral preprocessing (scatter correction, derivatives); Raman requires fluorescence management. Both require representative calibration data under real process conditions — not laboratory conditions.
Which standards apply to NIR and Raman in pharmaceutical manufacturing?
For NIR: USP <1119> in the US and EP 2.2.40 in the UK and Europe. For Raman: USP <858> and EP 2.2.48. Both methods are subject to ICH Q2(R2) for method validation. In GMP environments, 21 CFR Part 11 (US) and MHRA GMP guidelines (UK) govern data integrity requirements for electronic records generated by spectroscopic systems.
What is Time-Gated Raman and how does it address fluorescence?
Time-Gated Raman is an instrumental approach to fluorescence suppression based on the fact that Raman scattering occurs almost instantaneously after laser excitation, while fluorescence emission is delayed by nanoseconds. By gating the detector to collect only photons coincident with the laser pulse, fluorescence is largely rejected. This enables Raman measurements of samples that would otherwise be unworkable due to fluorescence — at the cost of more complex instrumentation. It is a developing technology with growing relevance for PAT applications in fluorescent matrices.
Instrument technologies and system concepts in detail
Dispersive NIR vs. FT-NIR, as well as different Raman system concepts — including HTS-Raman (High-Throughput Screening with fast data acquisition), Time-Gated Raman as a newer instrumental approach to fluorescence rejection through temporal separation of Raman signal and fluorescence emission, and further specialised system architectures — are covered in a separate article. This instrument-level decision comes after the technology selection covered here.
Further reading: "NIR and Raman Analysers for PAT — Instrument Types, System Concepts and Selection Criteria" (forthcoming)
Chemometrics and calibration model development
The development of multivariate calibration models for NIR and Raman — PLS, PCR, machine learning — is a specialist field with its own validation requirements.
Further reading: "Calibration and Validation of Spectroscopic PAT Methods" (forthcoming)
The decision between NIR and Raman is not a question of which technology is better — there is no universal answer. It is a question of which fits the measurement task better. Anyone who knows the target variable, sample matrix and process conditions will arrive at a clear answer.
The most important lesson from practice: the technology decision is the smaller part of project success. The sampling strategy, process understanding and calibration under real conditions determine whether a PAT system delivers or fails.
And when neither technology alone is sufficient — the combination is not the exception, but in critical processes often the smarter choice.
Analytic Journal — The Portal for PAT
© 2026 Analytic Journal / Dusko Kadijevic. All rights reserved.
This article is protected by copyright. Short excerpts with source attribution and an active link to analyticjournal.com are permitted. Full or substantial reproduction — including in digital form, by AI systems or automated processes — requires the prior written consent of the publisher. Enquiries: info@analyticjournal.de
Related topics on Analytic Journal:
Editorial Note
This article has been written to the best of our knowledge based on publicly available technical information, standards, and manufacturer data. Specifications may change — for binding information, consult current manufacturer datasheets and relevant standards bodies. Analytic Journal produces editorial content independently of manufacturers and advertising partners.
Author: Dusko Kadijevic — 20 years of experience as a PAT practitioner in the process industry
Dusko Kadijevic | Publisher Analytic Journal | analyticjournal.com
Written May 2026. All content formulated in our own words, created with AI assistance.
First published on 8 June 2026 on analyticjournal.com