Infrared Reimagined: How FT-IR Spectroscopy Learned to Read Molecules, Diagnose Disease, and See Below the Diffraction Limit

Abstract

Fourier transform infrared (FT-IR) spectroscopy is undergoing a quiet transformation. Deep learning models can now translate a raw infrared spectrum into a candidate molecular structure without consulting a spectral library, while nanoscale infrared techniques resolve chemical composition at length scales once reserved for electron microscopes. Meanwhile, FT-IR imaging is entering biopharmaceutical production lines to watch antibody formulations in real time, portable instruments are diagnosing rheumatologic and hematologic disease from blood and bloodspots, and machine learning classifiers are turning FT-IR into a frontline weapon against microplastic pollution. This article surveys these developments, drawn from research published over the past several years and covered extensively in Spectroscopy and the primary literature, and considers where the technique is headed as it converges with artificial intelligence, nanofabrication, and field-deployable instrumentation.

Introduction

Fourier transform infrared spectroscopy is often taught as a mature, almost quaint technique, useful for confirming a carbonyl stretch or matching a polymer against a reference library, but rarely described as a frontier method. That reputation is well out of date. Over the past five years, FT-IR has been pulled into nearly every major current reshaping analytical science: artificial intelligence, nanoscale chemical imaging, point-of-care diagnostics, and environmental forensics. Instead of simply matching spectra to library entries, today’s FT-IR platforms are inferring molecular structure without a database, resolving chemistry at tens of nanometers, monitoring therapeutic proteins as they elute from a purification column, and flagging disease from a few microliters of blood. This is not an incremental update to an old workhorse. It is a repositioning of FT-IR from a confirmatory bench tool into a predictive, autonomous, and field-deployable discovery platform. The sections below examine six areas where recent FT-IR research, much of it highlighted in Spectroscopy, is changing what the technique can do.

Teaching Machines to Read Infrared Fingerprints

For decades, converting an infrared spectrum into a molecular structure required either an experienced spectroscopist or an expensive quantum-chemical calculation such as density functional theory. That changed in 2025, when a transformer-based deep learning model demonstrated it could invert a raw infrared spectrum directly into a candidate chemical structure, correctly identifying the exact molecule on its first guess 63.8% of the time and capturing the right structure among its top ten guesses nearly 84% of the time.¹ The model requires no spectral database lookup; it learns the mapping between vibrational fingerprint and molecular graph directly. A complementary framework, Vib2Mol, extended this idea further, using a single encoder-decoder architecture to retrieve or generate molecular structures from infrared and Raman spectra, and even to perform peptide sequencing and predict reaction products directly from vibrational data.² Together, these systems suggest a near future in which an unknown infrared spectrum, rather than triggering a manual search through thousands of reference compounds, yields a ranked list of plausible structures within seconds, a capability with obvious value for high-throughput screening, forensic chemistry, and autonomous laboratories.

Seeing Below the Diffraction Limit With Nano-FT-IR and AFM-IR

Conventional FT-IR is blind to structure below roughly ten micrometers, a limitation imposed by the diffraction of infrared light. Nano-FTIR and its close relative AFM-IR sidestep that limit entirely by pairing an atomic force microscope tip with an infrared source, extracting chemical information from a probe volume tens of nanometers across. In 2025, researchers reported the first nano-FTIR measurements of the O-H stretching vibrations of liquid water, deuterated water, and aqueous salt solutions, a result long thought difficult to achieve because of the strong and broad absorption of water in this spectral region.³ Because stretching vibrations encode far more detail about hydrogen-bonding strength than the bending mode typically used in nanoscale water studies, this work opens the door to nanoscale hydration studies relevant to biomolecules, membranes, and electrochemical interfaces. On the instrumentation side, a new force-volume AFM-IR mode was reported that acquires simultaneous chemical and nanomechanical maps at every pixel of a sample, decoupling the infrared response from the mechanical damping that previously complicated interpretation of soft or heterogeneous materials such as polymer blends and biological tissue.⁴ Nano-FTIR is no longer a specialist curiosity; it is becoming a practical tool for chemical imaging at the scale of individual nanostructures.

Watching Biotherapeutics in Real Time With FT-IR Imaging

FT-IR spectroscopic imaging has traditionally been constrained by spatial resolution and by the difficulty of measuring concentrated biological samples under realistic process conditions. Recent work from Imperial College London describes a microfluidic attenuated total reflection (ATR) FT-IR accessory that measures antibody formulations in flow, under varying pH and temperature, effectively turning FT-IR imaging into an in-line process analytical technology for biopharmaceutical manufacturing.⁵ Unlike chromatographic or mass spectrometric methods, ATR-FT-IR imaging is not limited by protein concentration, allowing it to characterize monoclonal antibody formulations at the very high concentrations, up to roughly 200 milligrams per milliliter, used in patient self-administered injections. The same research group is now developing multi-channel imaging designs to compare several formulations simultaneously, and is exploring quantum cascade laser sources and bundled mid-infrared fiber optics to eventually connect FT-IR imaging directly to purification and manufacturing lines.⁵ The direction is unmistakable: FT-IR imaging is migrating from the microscope bench toward the production floor.

From Bench to Bedside With Portable FT-IR Diagnostics

Portable FT-IR spectrometers, once a compromise on performance, are now being validated as legitimate clinical screening tools. In one study, a portable FT-IR instrument analyzed dried bloodspots from patients with fibromyalgia syndrome alongside those with lupus, osteoarthritis, and rheumatoid arthritis, and pattern-recognition modeling classified the disorders with correlation coefficients above 0.93 and no misclassifications, pointing to peptide backbone and aromatic amino acid signatures as candidate biomarkers.⁶ A separate study coupled benchtop FT-IR spectra of dried blood serum with principal components analysis and support vector machines to detect biochemical changes associated with primary myelofibrosis, a rare bone marrow cancer, demonstrating that machine learning-assisted FT-IR can distinguish disease-specific spectral signatures from serum with high classification accuracy.⁷ Neither approach requires reagents, labels, or lengthy sample preparation, and both point toward a future in which a portable FT-IR unit paired with a validated chemometric model functions as a rapid, low-cost triage tool at the point of care, provided that clinicians receive the training and regulatory frameworks needed to trust the results.

Hunting Microplastics With Machine Learning-Enhanced FT-IR

Environmental monitoring has become one of FT-IR’s fastest-growing application areas, driven largely by the microplastics crisis. Rather than relying on manual library matching of small sample sets, one research group trained four machine learning classifiers, including one-dimensional and two-dimensional convolutional neural networks, a decision tree, and a random forest, on large-scale blended microplastic datasets, achieving accuracies of 96.4% on a smaller dataset and 97.4% on a larger one directly from raw FT-IR spectra.⁸ The resulting open-source analysis tool allows environmental scientists to classify microplastic polymer types far faster than traditional spectral library searches permit. A comprehensive 2024 review further detailed how FT-IR complements Raman spectroscopy and pyrolysis-gas chromatography-mass spectrometry for microplastic identification, while cataloging the technique’s remaining challenges: detecting particles below a few micrometers, resolving mixed polymer particles, and standardizing sample preparation across laboratories.⁹ As regulatory pressure on plastic pollution intensifies, FT-IR paired with automated classifiers is positioned to become a standard environmental monitoring tool rather than a specialized research technique.

Hyperspectral and Multimodal Fusion Expand FT-IR’s Data Universe

FT-IR is also being folded into a broader hyperspectral imaging ecosystem that spans the visible, near-infrared, and short-wave infrared regions, complemented by Raman spectroscopy. A 2025 review documented the rapid expansion of hyperspectral imaging into agriculture, food quality assessment, biomedicine, environmental science, and materials characterization, driven by the marriage of pixel-wise spectral data with deep learning architectures capable of automated segmentation and classification.¹⁰ Increasingly, FT-IR chemical maps are being fused with mass spectrometry, elemental analysis, or structural imaging data to build composite, systems-level pictures of tissues, composites, and environmental samples. This multimodal approach multiplies the analytical power of any single technique, though it also multiplies the complexity of data alignment, preprocessing, and uncertainty quantification, challenges that the field is only beginning to standardize.

Summary and Conclusions

Across six very different research fronts, a common pattern emerges: FT-IR spectroscopy is being paired with computation, miniaturization, and nanofabrication to do things that pure vibrational spectroscopy could not do on its own. Deep learning models are inverting spectra into structures without reference libraries.^1,2 Nano-FTIR and AFM-IR are resolving chemistry at the nanoscale, including the stretching vibrations of liquid water itself.^3,4 FT-IR imaging is migrating into biopharmaceutical process lines.⁵ Portable instruments are screening blood and bloodspots for rheumatologic and hematologic disease.^6,7 Machine learning classifiers are turning FT-IR into a scalable weapon against microplastic pollution.^8,9 And hyperspectral, multimodal data fusion is extending FT-IR’s reach into agriculture, biomedicine, and materials science alike.¹⁰ None of these advances replaces careful spectroscopy; if anything, each raises the bar for rigorous calibration, validation, and domain expertise. But together they mark a genuine inflection point for a technique many assumed had already reached its ceiling.

Future Outlook

The trajectory of these six trends points toward an FT-IR landscape that looks substantially different by the early 2030s. Expect spectrum-to-structure AI models to be integrated directly into instrument software, turning an unknown spectrum into a ranked list of candidate structures at the push of a button rather than a manual library search. Expect nano-FTIR and AFM-IR to move from specialized nanoscience laboratories into pharmaceutical, semiconductor, and battery research groups that need chemical maps at the nanometer scale. Expect FT-IR imaging with quantum cascade laser sources and fiber-optic bundles to become a genuine in-line process analytical technology across biomanufacturing. Expect portable and eventually wearable FT-IR and related vibrational sensors to expand clinical screening beyond the hospital and into routine, real-time personal health monitoring, provided that calibration transfer, model validation, and regulatory pathways mature at the same pace as the hardware.¹¹ And expect machine learning-driven, multimodal spectral fusion to become the default analytical approach for heterogeneous samples in environmental science, food safety, and materials research. Infrared spectroscopy has always been described as a fingerprinting technique. Its next chapter is about teaching that fingerprint to talk back.

References

(1) Alberts, M.; Zipoli, F.; Laino, T. Setting New Benchmarks in AI-Driven Infrared Structure Elucidation. Digit. Discov. 2025, 4, 1936–1943. DOI: 10.1039/D5DD00131E.

(2) Lu, X.; Ma, H.; Li, H.; Li, J.; Zhu, T.; Liu, G.; Ren, B. Vib2Mol: From Vibrational Spectra to Molecular Structures—A Versatile Deep Learning Model. arXiv 2025, arXiv:2503.07014. DOI: 10.48550/arXiv.2503.07014.

(3) Kotov, N.; Keskitalo, M. M.; Johnson, C. M. Nano FTIR Spectroscopy of Liquid Water in the –OH Stretching Region. Spectrochim. Acta, Part A 2025, 330, 125640. DOI: 10.1016/j.saa.2024.125640.

(4) Wagner, M.; Hu, Q.; Hu, S.; Phillips, C.; Wang, W.; Pittenger, B.; Fali, A.; Li, C.; Mathurin, J.; Dazzi, A.; Su, C.; De Wolf, P. Force Volume Atomic Force Microscopy–Infrared for Simultaneous Nanoscale Chemical and Mechanical Spectromicroscopy. ACS Nano 2025, 19 (19), 18791–18803. DOI: 10.1021/acsnano.5c04015.

(5) Kazarian, S. G.; Byrne, B.; Wetzel, W. The Future of FT-IR Imaging. Spectroscopy, November 13, 2025. https://www.spectroscopyonline.com/view/the-future-of-ft-ir-imaging (accessed 2026-07-31).

(6) Yao, S.; Bao, H.; Nuguri, S. M.; Yu, L.; Mikulik, Z.; Osuna-Diaz, M. M.; Sebastian, K. R.; Hackshaw, K. V.; Rodriguez-Saona, L. Rapid Biomarker-Based Diagnosis of Fibromyalgia Syndrome and Related Rheumatologic Disorders by Portable FT-IR Spectroscopic Techniques. Biomedicines 2023, 11 (3), 712. DOI: 10.3390/biomedicines11030712.

(7) Guleken, Z.; Ceylan, Z.; Aday, A.; Bayrak, A. G.; Yönal Hindilerden, İ.; Nalcaci, M.; Jakubczyk, P.; Depciuch, J. Application of Fourier Transform Infrared Spectroscopy with Machine Learning, Support Vector Machine, and Principal Components Analysis to Detect Biochemical Changes in Dried Serum of Patients with Primary Myelofibrosis. Biochim. Biophys. Acta, Gen. Subj. 2023, 1867, 130438. DOI: 10.1016/j.bbagen.2023.130438.

(8) Liu, Y.; Yao, W.; Qin, F.; Zhou, L.; Zheng, Y. Spectral Classification of Large-Scale Blended (Micro)Plastics Using FT-IR Raw Spectra and Image-Based Machine Learning. Environ. Sci. Technol. 2023, 57 (16), 6656–6663. DOI: 10.1021/acs.est.2c08952.

(9) Andoh, C. N.; Attiogbe, F.; Ackerson, N. O. B.; Antwi, M.; Adu-Boahen, K. Fourier Transform Infrared Spectroscopy: An Analytical Technique for Microplastic Identification and Quantification. Infrared Phys. Technol. 2024, 136, 105070. DOI: 10.1016/j.infrared.2023.105070.

(10) Cheng, M.-F.; Mukundan, A.; Karmakar, R.; Valappil, M. A. E.; Jouhar, J.; Wang, H.-C. Modern Trends and Recent Applications of Hyperspectral Imaging: A Review. Technologies 2025, 13 (5), 170. DOI: 10.3390/technologies13050170.

(11) Workman, J., Jr. Wearable Vibrational Spectroscopy Is Here for Real-Time Sensing. Spectroscopy, November 11, 2025. https://www.spectroscopyonline.com/view/wearable-vibrational-spectroscopy-is-here-for-real-time-sensing (accessed 2026-07-31).