Exploring Sampling for near Infrared Spectroscopy

Abstract

S ampling and sample preparation are two aspects of analytical work by near infrared (NIR) spectroscopy or reference methods in which, regrettably, research has usually lagged behind other developments in the technology. There are a number of things that the average NIR user needs to know about samples, sampling, sample preparation and sample presentation to the instrument. These include: 1) What is a sample? 2) What is the significance of a “truly representative” sample? 3) How should a sample be taken? 4) What affects the accuracy of sampling? An in-depth article on sampling that appeared in the Mythbuster column of the previous issue of NIR news comprehensively dealt with items 2 and 3. This article will discuss items 1 and 4, together with some other aspects of this all-important aspect of NIR technology. But first a note on the “Truly Representative Sample”. The truly representative sample should provide the material for the most accurate method of determination of the composition and functionality of the entire population. To save expense and time in testing, fully representative samples are usually thoroughly-blended composites of sub-samples that have been taken during operations such as storage of grain in elevators or silos, and other operations where the populations are too big, and the number of sub-samples too numerous to test individually. The danger with truly representative composite samples of large bulks is that they mask the variance that occurs within the sample. For example, if a composite sample is taken during storage of farmers’ deliveries of freshly-harvested grain into a 3000-tonne silo, the composite may show that the moisture content is 12.9%, a safe value for storage. But the individual deliveries may have ranged from 11% to 16% in moisture content. The delivery of, for example, 12 tonnes of grain at 16% moisture into a silo could cause spoilage within the silo that would not become apparent until the grain was later moved out of the silo for marketing. The heated grain would become dispersed throughout the grain as it moved out of the silo (Figure 1). When a composite sample is going to be prepared from a series of deliveries or sub-samples of cargo loading, the variance or the standard deviation (SD) among the samples and the range in analytical data should be determined. To obtain this important information, each individual sub-sample that goes to make up the composite should be tested before compositing. This used to be impracticable with reference testing, but it can be easily and cheaply achieved using NIR spectroscopy by scanning every subsample as it is taken.