Plastic waste has become one of the most pressing environmental challenges of the 21st century. With global production exceeding 400 million tonnes annually and only about 9% of all plastic ever produced being recycled effectively, the need for efficient sorting technologies has never been greater. At the heart of modern recycling operations lies the ability to quickly and accurately distinguish between recyclable and non-recyclable plastics. Spectroscopic techniques have emerged as the cornerstone of this effort, offering rapid, non-destructive analysis that can be scaled to industrial volumes. By exploiting the unique ways different polymers interact with light, these methods provide a chemical fingerprint for each plastic type, enabling automated sorting systems to separate materials with precision far beyond human capability.

The Plastic Recycling Challenge: Why Sorting Matters

Not all plastics are created equal in the recycling stream. Thermoplastics such as PET (polyethylene terephthalate, resin code 1) and HDPE (high-density polyethylene, resin code 2) are widely recycled, while others like PVC (polyvinyl chloride, code 3) or mixed-material composites often contaminate batches. Even among recyclable types, different polymers require separate processing—mixing PET with HDPE in a melt stream results in poor-quality material unsuitable for new products. The presence of non-recyclable plastics, such as thermosets or black plastics that absorb infrared light, further complicates sorting. Contamination rates as low as 5% can render a batch unrecyclable, driving the need for analytical techniques that can identify plastics with over 99% accuracy.

Manual sorting, still common in many regions, relies on human visual inspection of resin identification codes (the numbers inside the chasing arrow symbol). However, these codes are often missing, worn, or covered by labels. Moreover, many plastics look identical to the naked eye—clear PET bottles and clear PVC bottles appear similar but have different chemical compositions. Spectroscopic methods overcome these limitations by probing the molecular structure directly, making them indispensable for modern recycling facilities.

Spectroscopic Techniques: Principles and Workflows

Spectroscopy measures the interaction between matter and electromagnetic radiation. When applied to plastics, different polymers absorb, reflect, or scatter light at characteristic wavelengths, creating a unique spectral signature. By comparing this signature against a database of known polymers, the material can be identified in milliseconds. This process can be deployed on conveyor belts moving at several metres per second, allowing for real-time sorting of thousands of pieces per hour. The most widely used techniques in plastic recycling include FTIR, Raman, NIR, LIBS, and hyperspectral imaging.

Fourier Transform Infrared (FTIR) Spectroscopy

FTIR spectroscopy works by shining a broad spectrum of infrared light onto a sample and measuring which wavelengths are absorbed. The absorption peaks correspond to specific molecular vibrations—for example, C-H stretching in polyethylene or carbonyl groups in PET. Each polymer has a distinct infrared spectrum, much like a fingerprint. Modern FTIR systems used in recycling can classify plastics in less than 100 milliseconds. They are particularly effective for transparent and translucent plastics and are often integrated into near-infrared (NIR) systems for broader wavelength coverage.

A critical advantage of FTIR is its ability to differentiate between chemically similar polymers, such as different types of polypropylene or nylon grades. This precision reduces cross-contamination and improves the quality of recycled pellets. According to a study published in Waste Management & Research, FTIR-based sorting achieved over 98% accuracy for common packaging plastics (source).

Raman Spectroscopy

Unlike FTIR, which measures absorbed light, Raman spectroscopy detects inelastically scattered light from a laser-illuminated sample. The scattered photons gain or lose energy corresponding to molecular vibrations, yielding a complementary spectrum. Raman is less sensitive to water interference and can analyse very small particles (down to few micrometers), making it ideal for sorting plastic flakes or microplastics. It also works well with coloured and opaque materials, including black plastics that are notoriously difficult for NIR systems.

Recent advancements have led to portable Raman devices that can be used on-site at recycling centres or even by waste inspectors. However, Raman signals can be weak and require careful laser power management to avoid burning the sample. Despite this, its ability to identify additives, fillers, and coatings—materials that often affect recyclability—makes it a valuable complement to FTIR (source).

Near-Infrared (NIR) Spectroscopy

NIR spectroscopy operates in the 780–2500 nm wavelength range, where overtones and combinations of fundamental molecular vibrations appear. It is the most common technique in industrial plastic sorting due to its speed, reliability, and relatively low cost. NIR sensors can be installed above conveyor belts and collect spectral data from every piece of material passing underneath. Because NIR penetrates slightly into the surface, it can even identify plastics beneath thin labels.

NIR systems are highly effective for sorting PET, HDPE, LDPE, PP, and PS (polystyrene). Their main limitation is difficulty with dark-coloured plastics (especially carbon-black pigmented items) that absorb most of the near-infrared light. To address this, facilities often combine NIR with other sensors or pre-treat black plastics with marking agents. NIR technology continues to evolve, with multispectral and hyperspectral NIR sensors now capable of collecting dozens of wavelength bands simultaneously for more nuanced classification (EPA on plastics recycling).

Laser-Induced Breakdown Spectroscopy (LIBS)

LIBS uses a high-energy laser pulse to create a micro-plasma on the plastic surface. The plasma emits light at wavelengths characteristic of the elemental composition of the sample. Although plastic polymers are primarily carbon and hydrogen, LIBS can detect trace elements like chlorine (in PVC), bromine (in flame-retardant plastics), or heavy metals that render a plastic non-recyclable. This makes LIBS particularly useful for identifying hazardous materials that should be removed from the recycling stream.

LIBS can analyse samples in under one second, and recent portable LIBS devices have been deployed in scrap yards and e-waste recycling centres. However, its surface sensitivity means that coatings or contamination can skew results. Additionally, LIBS is a destructive technique on a microscale, which is acceptable for sorting but not for quality control of finished products.

Hyperspectral Imaging

Hyperspectral imaging combines spectroscopy with spatial information. A camera captures an entire spectral band (e.g., from visible to short-wave infrared) for every pixel in an image, creating a three-dimensional data cube (spatial coordinates x, y and spectral dimension). This allows classification not just of the plastic type but also of its spatial distribution—for example, identifying a multilayer packaging film consisting of PET and polyethylene coexisting in a single piece.

Hyperspectral systems are increasingly used in advanced recycling plants because they can handle mixed waste streams with high accuracy. They are particularly effective for sorting post-consumer packaging where multiple polymers are laminated. The computational demands are higher than with point spectroscopy, but advances in machine learning have made real-time processing feasible. A White Paper from the Technical University of Denmark demonstrated that hyperspectral sorting could increase recycling purity by 15–20% compared to conventional NIR systems (source).

Application in Automated Sorting Systems

In a typical large-scale recycling plant, unsorted plastic waste enters a series of trommel screens and magnetic separators to remove metals and other contaminants. Plastics then pass under an array of spectroscopic sensors mounted above high-speed conveyor belts. When a piece of plastic is identified as a specific polymer (e.g., PET), a precise blast of compressed air from nozzles below the belt deflects it into the correct chute. This process can handle up to 10 tons of material per hour with accuracy exceeding 95%.

Example installations include the TOMRA AUTOSORT™ series and the MSS™ (CP Group) optical sorters, both of which rely on NIR, FTIR, and sometimes multispectral imaging. Many plants combine multiple spectroscopic methods: NIR for bulk sorting, Raman for verifying difficult fractions, and LIBS for removing hazardous plastics. This multi-sensor approach dramatically reduces the amount of non-recyclable material ending up in bales destined for reprocessors.

Advantages Over Manual Sorting

  • Speed: Spectroscopic systems identify hundreds of items per second, far exceeding human capabilities.
  • Accuracy: Machine vision coupled with spectral analysis achieves >98% purity for targeted polymers.
  • Consistency: Automated systems operate 24/7 without fatigue, maintaining high throughput.
  • Safety: Removes workers from contact with sharp, contaminated, or hazardous waste.
  • Cost-effectiveness: Although initial investment is high, reduced labour costs and higher quality recyclates yield strong ROI over time.
  • Data collection: Spectroscopic data can be logged for quality control, material flow analysis, and compliance reporting.

Limitations and Challenges

Despite their power, spectroscopic techniques are not a silver bullet. Key challenges include:

Black Plastics and Dark Pigments

Carbon black, the most common pigment, absorbs infrared and near-infrared light, making black plastics nearly invisible to NIR and FTIR sensors. Raman and LIBS can partially overcome this, but their lower throughput limits their use in high-speed sorting. Some facilities introduce fluorescent markers or use visible-light cameras in combination with deep learning to identify black items based on shape and texture.

Mixed-Material and Multilayered Plastics

Many packaging items combine multiple polymers—for example, a PET bottle with a PP cap or a PE-coated paperboard. Spectroscopic methods often read only the surface layer. Hyperspectral imaging and advanced signal processing can separate contributions from different layers, but it remains a challenge for conventional point sensors.

Equipment and Maintenance Costs

High-end spectroscopic sorters can cost hundreds of thousands of dollars, making them prohibitive for small-scale recyclers. Additionally, sensors require calibration and protection from dust and mechanical shock. The need for trained technicians to operate and maintain these systems can also be a barrier in developing regions.

Sample Preparation and Presentation

Plastics that are wet, soiled, or crushed may present unreliable spectra. Consistent belt speed, lighting, and material flow are critical for accurate identification. In some cases, a pre-wash step is necessary to remove organic contaminants before sorting.

Spectral Database Updates

New plastic formulations, biodegradable polymers, and additives appear regularly. Maintaining an up-to-date spectral library is essential for accurate classification. Cloud-based databases and automated learning from plant data are emerging solutions.

Future Directions

The next generation of spectroscopic sorting will be driven by three key trends: artificial intelligence, miniaturisation, and data integration. Deep learning models, especially convolutional neural networks (CNNs), can classify spectral data with higher accuracy and robustness to noise than traditional chemometric methods. These models can also combine spectral inputs with colour, shape, and texture data from conventional cameras, enabling more holistic sorting decisions.

Portable and handheld spectrometers are becoming more affordable and accurate. These devices allow waste auditors to verify sorting results on the spot, and can be deployed at collection points or landfill sites. For example, the SpectroSort handheld NIR device can identify 30 plastic types in under two seconds, with accuracy comparable to benchtop instruments. Such tools empower small recyclers and informal waste workers who currently rely on manual methods.

Integration with the Internet of Things (IoT) and cloud computing will enable real-time sharing of spectral data across the recycling chain. A sorting plant in one city could receive updates to its classification models based on data collected at another plant processing similar waste streams. This collaborative approach will accelerate the creation of comprehensive spectral libraries and improve sorting performance globally.

Finally, research into hybrid sensor systems that combine NIR, Raman, LIBS, and X-ray fluorescence in a single unit is underway. Such systems would provide a complete chemical and elemental profile of each item, enabling detection of non-recyclable contaminants like flame retardants or heavy metals that would otherwise go unnoticed. The goal is to achieve near-100% purity in recycled plastics, making them truly competitive with virgin materials.

Conclusion

Spectroscopic techniques have already transformed plastic recycling from a crude separation process into a precision engineering operation. FTIR, Raman, NIR, LIBS, and hyperspectral imaging each bring unique strengths to the task of differentiating recyclable from non-recyclable plastics. While challenges remain—especially with black plastics, mixed materials, and cost barriers—continuous advancements in sensor technology, machine learning, and portable devices promise to make high-accuracy sorting accessible to every recycling facility worldwide. By investing in these analytical tools, the recycling industry can dramatically reduce contamination, increase the supply of high-quality recyclate, and ultimately help close the loop on plastic waste.