# How Correct Sample Preparation Can Minimize Standard Deviations

## Introduction

Nowadays, even the smallest traces of sample components can be detected, as precision increases and detection limits are pushed further by using modern analytical methods. In spite of this development, sample preparation, which is performed before the actual analysis, is often neglected. When compared to the errors made during analysis, errors caused due to lack of accuracy in sample preparation have a much larger impact. Similar to an iceberg which is generally hidden under water, only a small part of the sum of errors is considered whereas the major part of potential errors is not taken into account (Figure. 1). This may be due to the fact that sample preparation and sampling have always been done in a conventional way which has become a routine over the decades and is not considered as having a major influence on the subsequent analyses.

Figure 1. Error pyramid for sample analysis. Analog to an iceberg of which only a small part is visible above the water, only a small part of the actual error sources is perceived during sample analysis.

## Sampling

The more heterogeneous a sample is, the more vital correct sample preparation becomes.

Several questions have to be answered before a sample is taken, for example, from a sand heap.

• Which is the correct sample amount so that all sample properties of the original lot are represented?
• Does it have any impact from which area of the heap the sample is taken?
• Did the sample segregate so that larger particles are mostly found in the upper part of the heap?

If the answer to the last question is yes, the subsample collected from the upper part of the heap does not represent the initial lot. In addition, other aspects may have an influence here: if the heap of sand was stored outdoors, then the material on the surface of the heap contains more moisture than the inside part. This indicates that the property “moisture” is heterogeneously distributed in the initial material.

From these simple examples, it can be seen that the analytical results are greatly affected by the sampling and the complete sample handling process. Only when the subsamples are representative, can reproducible results be achieved.

## Sample Division

Generally, the properties of a laboratory sample are heterogeneously distributed; with the help of milling, the sample can be homogenized and the properties equally distributed. Only a few milligrams or grams of the laboratory sample are needed for subsequent analysis. Hence, larger subsamples have to be divided representatively. Standardized methods such as quartering and coning or the use of sample dividers or sample splitters help to divide the subsamples, with increased automation minimizing statistic errors and increasing the representativeness of the subsample.

In RETSCH´s Rotating Sample Divider PT 100, the sample is fed into the hopper and automatically transported through a feed chute to the openings of an evenly rotating dividing head. The sample is divided into 6, 8 or 10 subsamples through this process. After the division, either one subsample can be divided further or several subsamples can be combined. If the division is repeated with similar parameters, it offers comparable results so that the analysis is reproducible. Thus, standard deviations can be reduced considerably by correct sample division. Analysis of a laboratory sample of Hyflon is shown in the following example. The thermogravimetric analyzer Thermostep from ELTRA measured the humidity of 10 samples. This instrument consists of a programmable oven including a scale and a sample carousel and measures the ash content, humidity, and volatile components of up to 19 samples fully automatically. The loss of weight of 10 samples (1-10) after correct sample division in the PT 100 was observed to be 1.3% ± 0.05%. Another 10 samples (11-20) were not divided properly but taken randomly from the laboratory sample, showing a loss of weight of 0.99% ± 0.1% (Figure 2).

The standard deviation of the randomly taken sample was 10.1% of the total value, which was minimized to 3.8% after correct sample division.

Figure 2. Standard deviations after random and automated sample division

## Homogenization of Samples by Grinding

It is rarely possible to carry out a direct analysis on a laboratory sample as it often contains large or segregated particles. Large particles can be a problem because most analytical methods will not detect all components of the particle but only the surface. The segregation effects are explained above. Grinding those samples serves to minimize the particle size so that the inner parts are available for analysis. Ideally, the properties which are originally spread heterogeneously all over the sample are homogeneously distributed after the grinding step. The following examples show that correct sample preparation is a pre-requisite to ensure increased reproducibility of analytical results and minimum standard deviations.

### Application Example Rye

The example of rye shows different results for homogeneous and inhomogeneous samples in NIR analysis (10 measurements of each sample). RETSCH`s Cyclone Mill TWISTER which is suitable for grinding rye is shown in Figure 3. Feed or corn pellets, which are soft and fibrous, are ground rapidly and efficiently with friction and impact. The optimized shape and high speed of the rotor and the grinding chamber produce an air stream that transports the sample via the integrated cyclone into the 250 ml sample bottle and cools it at the same time. In addition, the air stream helps to remove most of the sample residues. Three speeds and sieves with different aperture sizes allow for optimum adaption to an extensive range of samples. For instance, 160 g rye was ground at 14,000 min-1 to < 1 mm in 1 minute, using a 1 mm sieve. The major differences of the ground and the unground sample with respect to the fiber and protein content are shown in Table 1. The fiber content of the unground sample is higher than that of the ground sample because only the rye’s surface area was analyzed.

Figure 3. Cyclone Mill TWISTER

Table 1. Comparison of the fiber and protein content of an unground and a ground rye sample

Application example rye
Particle size Fiber content Protein content
6 mm (unground) 6.9% ± 0.62%
(8.98%)
8.46% ± 0.45%
(5.32%)
1 mm (ground with Twister) 1.1% ± 0.05%
(4.54%)
9.02% ± 0.07%
(0.77%)

### Application Example Lignite

The Jaw Crusher BB 300 was used to crush 4 kg lignite, (extremely inhomogeneous material) to 8 mm particles. RETSCH offers different models of jaw crushers which are ideal for crushing hard and brittle samples with 35 – 130 mm initial particle size. A final particle size of 0.5 – 2 mm can be achieved based on the jaw crusher model. After crushing and successive representative sample division, a 100 g subsample was finely ground to 100 µm in 30 seconds in the Ultra Centrifugal Mill ZM 200 at 18,000 min-1, employing a 0.12 mm ring sieve. In this high-speed rotor mill, the sample passes through a hopper and hits a horizontal rotor, where centrifugal forces fling it outwards. During the process the particles hit the rotating teeth of the rotor and are eventually crushed. Additional size reduction is achieved when the particles are ground between sieve and rotor by shearing forces. The sample stays in the grinding chamber for only a very short period of time before it is collected in the cassette, so the sample properties remain unaltered during the process. The sample can be completely recovered, thanks to the patented cassette system which helps to avoid cross contamination.

The Elemental Analyzer CS-580 from ELTRA (Figure 4, Table 2) measured the sulfur and the carbon content of the 100 µm and 8 mm lignite sample. The samples are burnt at 1350 °C using the CS-580 before detecting the emitted gases with infrared cells. It takes about 90 seconds for analyzing a single sample; a user-friendly software carries out the evaluation.

Figure 4. ELTRA's Carbon/Sulfur Analyzer CS-580

Table 2. Determination of the carbon and the sulfur content of lignite with different particle sizes (10 measurements of each sample)

Application example lignite
Particle size Carbon content Sulfur content
8 mm 63.77% ± 1.21%
(1.9%)
0.048% ± 0.047%
(99.9%)
100 µm 64.61% ± 0.05%
(0.08%)

0.045% ± 0.004%
(8.9%)

These results show that the standard deviation substantially decreases if the homogeneity of the sample increases. The standard deviation of the sulfur content is reduced from nearly 100% of the initial value to less than 10%.

### Case Study Compost

Different types and fractions of soil, wood and other organic particles and even plastic parts may be present in compost. For such a heterogeneous material thorough homogenization is important to ensure reliable analysis results.

RETSCH’s Cutting Mill SM 300 (Figure 5) cut a 100 g sample into particles <4 mm at 3,000 min-1 employing the parallel section rotor and a 4 mm bottom sieve. The SM 300 is ideally suited for grinding heterogeneous materials because of the variable speed of 700 – 3,000 min-1, different collecting systems, sieves and rotors. An optional cyclone improves discharge of low density materials and creates an additional cooling effect. The sample does not get very warm during the process ensuring that properties are not altered. The mill is quickly and easily cleaned thanks to the fold-back hopper and push-fit rotor.

Figure 5. Cutting Mill SM 300

In a fine-grinding step the sample was milled two times in the Ultra Centrifugal Mill ZM 200 at 18,000 min-1, first employing a 1 mm ring sieve, then using a 0.25 mm ring sieve. The sample size was reduced to 0.25 mm after 2 minutes. The Elemental Analyzer CHS-580 from ELTRA carried out the analysis of the elements such as hydrogen, sulfur and carbon in the 0.25 mm and 4 mm sample (each sample was measured 4 times). Unlike the CS-580 employed for analyzing lignite, this analyzer features an additional infrared cell for hydrogen determination.

Figure 6. Determination of carbon, sulfur and hydrogen contents of the 4 mm and the 0.25 mm sample (4 measurements) with the elemental analyzer CHS-580 from ELTRA.

The application example of compost illustrates that adequate homogenization can have considerable effects on the analysis of elements. The standard deviation is reduced by more than a factor 10. Furthermore, much higher contents of the elements were determined in the fully homogenized samples.

## Conclusion

From the examples explained in this article, it is clear that sample homogenization and sample division can greatly influence analytical results. As the sample properties are equally distributed in the sample, standard deviation is reduced and more accurate results are achieved. The sampling, sample division and preparation processes should be handled with the same accuracy as the analysis itself in order to avoid errors leading to incorrect results.

This information has been sourced, reviewed and adapted from materials provided by Retsch GmbH.