Basic knowledge about VIS NIR spectroscopy

Basic knowledge about VIS NIR spectroscopy

What is spectroscopy

Commonly, e.g. if you look up Wikipedia, spectroscopy is described as “the study of the interaction between matter and electromagnetic radiation”. More generally speaking, what you actually want to retrieve by spectroscopy is the response of a system or object to some sort of excitation.

About the similarity of guitar strings and trees

Let us explain the operation of spectroscopy with an example. Think of a guitar as the system or object of interest. When you play it, the guitar (or more specifically its strings) gets excited and delivers a response signal – the music. This signal can be detected by e.g. a microphone and furthermore analyzed by software or other analysis units to represent sound intensity as a function of frequencies contained in the recorded signal – the response or spectrum of your guitar.

The combination of a detection device (the microphone) and the analysis unit (the software) is called a spectrometer. In this respect, your ear is a spectrometer as it detects acoustical waves and (in combination with your brain) converts them into meaningful information – the sound you hear.

Another simple example for a spectrometer is your eye. It converts the interaction of visible electromagnetic radiation (e.g. the sunlight) with any object (e.g. a tree) into a meaningful picture that will be visualized by your brain. Your eye actually is a high-performance spectrometer as it analyzes millions of optical information in a second.

The meaning of wavelength in spectroscopy

However, it is limited to the visible part of the electromagnetic spectrum. What your eye “spectrometer” actually does is to split up the receiving light (which is a superposition of many colors) into the three fundamental colors red, green and blue. This allows your brain (the analysis unit) to identify the colors of the object you are looking at.

Each color is mathematically represented by a wavelength number. The wavelength number is defined as the distance between two consecutive peaks of an electromagnetic wave traveling in space (similar to the distance between peak values of a wave in a lake when you throw a stone into it). 

For example, blue color has a wavelength in the range of 450 nm, green around 600 nm and red at roughly 700 nm. Your eye can detect wavelengths in the range from ~400 to ~780 nm, also commonly called the visible range (VIS). If all wavelengths in that range exist at the same time in the radiation your eye receives, you will see white light.

How to identify the “optical fingerprint”

However, there is much more than just the VIS part of the spectrum. In the VIS you get basic information about the color of your object of interest. When you extend the observation range into e.g. the infrared (IR) region above 780 nm, you will be able to detect “fingerprints” of material composition, that means what a material is made of.

This part of the spectrum cannot be seen by your eye but can be made visible by technical spectrometer instruments. Thus, spectroscopy relies on the ability to construct technical instruments which are capable of detecting electromagnetic radiation in the wavelength range of interest with sufficient sensitivity and accuracy.

The 4 steps of spectroscopic measurements

What is chemometrics

From a purely technical point of view, our sensor hardware delivers a spectrum of each sample you measure. So, you basically get a data stream of 16 values, representing intensities at 16 wavelengths. This is – of course – not what you finally need, because you would like to identify the material or material composition and not interpret a spectrum.

To come to this final goal, the sensor needs to be trained with well-known reference samples, representing the typical range of materials you want to identify. During this training, the sensor “learns” which spectrum belongs to which material and can afterwards identify new materials based on this training. 

In technical terms, this process is called calibration and is nothing else than a piece of software, relating reference values to measured spectra. The underlying models and algorithms belong to the field of chemometrics.

The steps of developing a chemometric model

The above picture shows the typical process flow of sensor training. 

First, you select reference samples representing the variety of your materials you would like to characterize reasonably well (1). For these reference samples, you already know their reference values (2a), i.e. what material it is or of what the material is composed of. Every sample is measured using the Senorics sensor hardware (2b). 

Both information, reference values and measured spectrum for each sample, are combined in a so called chemometric model (3), relating reference values and measured spectra in an optimal way by software algorithms. The creation of this model concludes the part of sensor training or calibration.

The created chemometric model (3) now can be applied to new, unknown samples (4) to identify the material class / composition of these unknown samples. This is done by measuring the spectra of the unknown samples with Senorics hardware (5) and applying the chemometric model (3) trained before to the measured spectra. The model will look for the best fit of the new spectra to the known spectra from the sensor training and picks the associated reference value to identify the material class / composition of the new sample. 

This is, in a nutshell and very simplified, how chemometrics work, representing the intelligence of our sensor solutions to come from a spectrum to a clear and unique material identification.