NIR Technology Forum

Article

Near-infrared (NIR) spectroscopy is used in various application areas, including pharmaceutical analysis, medical diagnostics, and food and agrochemical analysis. Participants in this Technology Forum are Edwin Weusthof of Avantes BV, John Richmond of Bruker Optics, and Yvette Mattley of Ocean Optics.

Near-infrared (NIR) spectroscopy is used in various application areas, including pharmaceutical analysis, medical diagnostics, and food and agrochemical analysis. Participants in this Technology Forum are Edwin Weusthof of Avantes BV, John Richmond of Bruker Optics, and Yvette Mattley of Ocean Optics.

Calibration transfer is a concern in quantitative spectroscopic analysis, particularly in the NIR region but more generally in all spectral regions. How would you define “calibration transfer”? What features or capabilities are needed that are not available in current software for calibration?

Weusthof: Calibration transfer would be in a ideal situation continuosly monitoring the effect of change in wavelength due to temperature and humidity effects. This could be done by referencing with a calibration lightsource like a mercury argon calibration source in the NIR instrument.

Richmond: Calibration transfer has been a subject of debate for many years in the NIR community due to the fact that huge amounts of time, effort and money are invested in developing calibrations. Some data sets run into many thousands of spectra and it is absolutely vital that these calibrations can be transferred between spectrometers. Spectrometer design is very important and instrument standardization is a key area. In Fourier-transform (FT)-NIR, a high resolution spectrum of water vapor is collected to calibrate the frequency axis and this enables direct calibration transfer between spectrometers. In dispersive spectrometers, it is not possible to get the same degree of standardization and consequently direct calibration transfer is more difficult. At the very least, it is likely that a skew and bias will need to be applied. However, software algorithms such as piecewise direct standardization (PDS) are now being routinely used to help overcome these challenges — in fact it is now possible to use PDS to transfer spectra from a dispersive system to an FT system. Once the spectra have been transferred, it is relatively simple task to regenerate the calibration model.

Mattley: Robust, repeatable, and precise calibration is the key to the success of NIR measurements used for quality analysis or process control. The ability to transfer a robust calibration model to another instrument helps to avoid having to develop a calibration models for each new instrument. Calibration transfer is used to adjust the predictive math algorithm developed with one instrument for use with another instrument. The generation of calibration transfer constants for a new instrument is much less involved than the development of a new calibration model.

Adjusting for these parameters is a concern because even the highest performance instruments will not work for quantifying analytes if the calibration is not predicting properly. Fortunately, the adjustment of these parameters is very doable with the use of spectral processing and calibration software (for example, GRAMS, Unscrambler, MatLab). Analytical corrections like slope and bias constants are achieved with software using a few standard samples and well known procedures. These tools are quite efficient and with enough experience it is possible to get reproducible results.

There are tools available to apply corrections that enable the transfer of calibration files from one spectrometer to another with varying spectral response. This software works by “converting” spectra from a new spectrometer to match the spectra for the original spectrometer used to create the calibration model. Once the conversion factors have been calculated, they are applied to the each of the spectra in the calibration spectra database to create a new calibration model adjusted to the peak locations of the new instrument. This procedure allows the user to avoid having to run all the calibration samples again for each instrument by creating a new adjusted calibration model for the new instrument from the original calibration model.

One of the big challenges facing NIR spectroscopy analysis with calibration transfer is getting instruments that are reproducible enough to get consistent results without requiring conversion factors or very specific and complicated technical procedures.

Is process NIR likely to extend to applications other than pharmaceuticals? If so, what industries are most likely to implement it?

Weusthof: Yes. It will also be implemented in food industry, petrochemical industry, and chemistry.

Richmond: In fact, process NIR has been in use for many years in a number of industries. Back in the 1980s the chemical and fuel refining industries were routinely using process NIR for OH values of polyols and octane number of gasoline. Looking forward, as consumers demand better product quality and consistency, I see process NIR being implemented in industries such as animal feed, food, and dairy.

Mattley: NIR is a very well known technology. It is quite easy to implement with a very attractive cost-benefit model. With the right calibrations and stable technology, NIR use in process (on-line applications) is growing in popularity. In addition to pharmaceuticals, the agriculture and food industries employ NIR for solids processing to control parameters like protein, moisture, fiber, and fat. Moisture is one of the most commonly measured parameters in on-line NIR applications, being measured in dairy, meat, powders, flours, and many other samples.

Multiparametric and multipoint process technologies are less developed but also are growing because of the versatility and the economic advantages of NIR spectroscopy. In these industries, the volume of production is so big that a significant amount of product is produced in just a few minutes. A simple delay in the method used to control these large-scale processes means the user cannot make a critical decision or much worse, will make the wrong decision. These are the industries amenable to the use of on-line process control with NIR. While NIR spectroscopy does not detect every analyte, it does a great job detecting the major analytes of interest in industries like agriculture, food and biofuels. These are the largest, strongest markets that would benefit from process NIR.

What future applications of NIR imaging will be developed?

Weusthof: In art, pharmaceutics, tablets analysis, it will become the standard technique as it becomes more and more cost-effective.

Richmond: Chemical Imaging is an exciting area and a lot of research is going in this direction. NIR imaging of pharmaceutical products has been done in the laboratories for some time to determine distribution of API for example. If future technology developments could speed up analysis times significantly, I would envisage NIR Imaging being routinely used for online assessment of product quality in the pharmaceutical industry. There is also a lot of interest in using NIR imaging for medical applications such as the treatment of burns or coronary imaging.

Mattley: NIR spectroscopy has applications across a wide range of industries. With energy research in mind, applications such as ethanol process control, adulteration in fuels, and investigation of biofuels are increasingly important. Additional development of NIR techniques in the pharmaceuticals and food and beverage industries is likely, with blending control, improvements in wine processing, and determination of tannins in sorghum among the possibilities. Additional process and on-line applications are increasingly evident as the technology is further refined. The universe of applications is vast.