Hopkins Explains Effective Use of Derivatives at Local Meeting of Society for Applied Spectroscopy

At the November meeting of the New York–New Jersey Chapter of the Society for Applied Spectroscopy an attentive audience heard David Hopkins present his work on the effective use of derivatives in spectroscopy. The meeting was held on November 9 at Horiba Scientific in Edison, New Jersey, and hosted by Fran Adar, a long-time member and friend of the chapter.  

Hopkins is an experienced analytical chemist with expertise in spectroscopy and chemometrics, with an emphasis on near-infrared (NIR) analysis. For 25 years he has been a consultant to corporations that use or manufacture instruments for NIR, UV–visible, infrared (IR), Raman, and fluorescence spectroscopy. He teaches short courses in NIR spectroscopy and chemometrics and has authored papers on derivatives, ranging from his PhD thesis to recent publications on his work with Karl Norris.

In his presentation, Hopkins described his experience using the Norris regression method developed by Karl Norris. This method uses multiple linear regression (MLR) with terms composed of quotients of gap derivatives measured at optimized wavelengths for the numerator and denominator. While Norris’s research using quotients to develop quantitative algorithms has been in private use by Norris and his followers, more recently Hopkins and Norris have developed Matlab programs to enable the reproduction of the method originally developed using Fortran programs. In the talk, Hopkins discussed some applications of using first, second, third, and fourth derivatives; the advantages of using derivatives; and the benefits of using smoothing to improve a model performance. In addition, Hopkins demonstrated how to use the standard error of calibration to optimize the amount of smoothing required to enhance model performance while maintaining spectral information. 

Spectra can often be more readily interpreted when a broad band made up of closely spaced or overlapping bands is sharpened and separated using derivatives of order 1 to 4. For measurement of oil, moisture, and protein in agricultural products such as soybeans and wheat, MLR can be used to generate models that can unravel the highly overlapping starch, protein, and oil bands and separate the moisture information. One of the key points made by Hopkins pertaining to Norris’s work is that his MLR quotient models typically only contained a few terms and that was enough to remove the multiplicative scatter effects in the spectra and yield better models for the constituents. Examples of this approach were presented by Hopkins from wheat and hemoglobin data sets.

Hopkins’s talk was both informative and entertaining, and included anecdotes to illustrate his points as well as advice for how to use the standard error of calibration to optimize the smoothing function used with derivatives. Hopkins discussed his work to create an automated Matlab algorithm that will identify the optimum set of quotients to use in the final MLR model. The method does such a good job at modeling the data for quantitiave analysis that only a few terms are needed, and it often outperforms partial least squares (PLS) methods. He demonstrated how derivatives, and in particular the fourth derivative, can be used to identify inherient moisture in the spectra that arises either from the sample or from the atmosphere when collecting the sample. One attendee mentioned he is using the fourth derivative to process differential scanning calorimetry (DSC) data.

The dinner meeting was attended by 15 professionals from various industries, including a few new members. Networking before and after the talk was plentiful and offered a way for the audience to get to know one another and discuss areas of common interest. Hopkins was presented with a speaker gift from the chapter-a custom prism. The organization welcomes volunteer speakers for future talks. More information about the chapter and the schedule of meetings can be found at www.nysas.org.

Debbie Peru is the secretary of the New York–New Jersey Chapter of the Society for Applied Spectroscopy (NYSAS).